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 "ParagraphParameters.h"
  41 #include "SpellChecker.h"
  42 #include "sgml.h"
  43 #include "texstream.h"
  44 #include "TextClass.h"
  45 #include "TexRow.h"
  46 #include "Text.h"
  47 #include "WordLangTuple.h"
  48 #include "WordList.h"
  49
  50 #include "frontends/alert.h"
  51
  52 #include "insets/InsetBibitem.h"
  53 #include "insets/InsetLabel.h"
  54 #include "insets/InsetSpecialChar.h"
  55 #include "insets/InsetText.h"
  56
  57 #include "mathed/InsetMathHull.h"
  58
  59 #include "support/debug.h"
  60 #include "support/docstring_list.h"
  61 #include "support/ExceptionMessage.h"
  62 #include "support/gettext.h"
  63 #include "support/lassert.h"
  64 #include "support/lstrings.h"
  65 #include "support/textutils.h"
  66
  67 #include <atomic>
  68 #include <sstream>
  69 #include <vector>
  70
  71 using namespace std;
  72 using namespace lyx::support;
  73
  74 // OSX clang, gcc < 4.8.0, and msvc < 2015 do not support C++11 thread_local
  75 #if defined(__APPLE__) || (defined(__GNUC__) && __GNUC__ == 4 && __GNUC_MINOR__ < 8)
  76 #define THREAD_LOCAL_STATIC static __thread
  77 #elif defined(_MSC_VER) && (_MSC_VER < 1900)
  78 #define THREAD_LOCAL_STATIC static __declspec(thread)
  79 #else
  80 #define THREAD_LOCAL_STATIC thread_local static
  81 #endif
  82
  83 namespace lyx {
  84
  85 namespace {
  86
  87 /// Inset identifier (above 0x10ffff, for ucs-4)
  88 char_type const META_INSET = 0x200001;
  89
  90 } // namespace
  91
  92
  93 /////////////////////////////////////////////////////////////////////
  94 //
  95 // SpellResultRange
  96 //
  97 /////////////////////////////////////////////////////////////////////
  98
  99 class SpellResultRange {
 100 public:
 101         SpellResultRange(FontSpan range, SpellChecker::Result result)
 102         : range_(range), result_(result)
 103         {}
 104         ///
 105         FontSpan const & range() const { return range_; }
 106         ///
 107         void range(FontSpan const & r) { range_ = r; }
 108         ///
 109         SpellChecker::Result result() const { return result_; }
 110         ///
 111         void result(SpellChecker::Result r) { result_ = r; }
 112         ///
 113         bool contains(pos_type pos) const { return range_.contains(pos); }
 114         ///
 115         bool covered(FontSpan const & r) const
 116         {
 117                 // 1. first of new range inside current range or
 118                 // 2. last of new range inside current range or
 119                 // 3. first of current range inside new range or
 120                 // 4. last of current range inside new range
 121                 //FIXME: is this the same as !range_.intersect(r).empty() ?
 122                 return range_.contains(r.first) || range_.contains(r.last) ||
 123                         r.contains(range_.first) || r.contains(range_.last);
 124         }
 125         ///
 126         void shift(pos_type pos, int offset)
 127         {
 128                 if (range_.first > pos) {
 129                         range_.first += offset;
 130                         range_.last += offset;
 131                 } else if (range_.last >= pos) {
 132                         range_.last += offset;
 133                 }
 134         }
 135 private:
 136         FontSpan range_ ;
 137         SpellChecker::Result result_ ;
 138 };
 139
 140
 141 /////////////////////////////////////////////////////////////////////
 142 //
 143 // SpellCheckerState
 144 //
 145 /////////////////////////////////////////////////////////////////////
 146
 147 class SpellCheckerState {
 148 public:
 149         SpellCheckerState()
 150         {
 151                 needs_refresh_ = true;
 152                 current_change_number_ = 0;
 153         }
 154
 155         void setRange(FontSpan const & fp, SpellChecker::Result state)
 156         {
 157                 Ranges result;
 158                 RangesIterator et = ranges_.end();
 159                 RangesIterator it = ranges_.begin();
 160                 for (; it != et; ++it) {
 161                         if (!it->covered(fp))
 162                                 result.push_back(SpellResultRange(it->range(), it->result()));
 163                         else if (state == SpellChecker::WORD_OK) {
 164                                 // trim or split the current misspelled range
 165                                 // store misspelled ranges only
 166                                 FontSpan range = it->range();
 167                                 if (fp.first > range.first) {
 168                                         // misspelled area in front of WORD_OK
 169                                         range.last = fp.first - 1;
 170                                         result.push_back(SpellResultRange(range, it->result()));
 171                                         range = it->range();
 172                                 }
 173                                 if (fp.last < range.last) {
 174                                         // misspelled area after WORD_OK range
 175                                         range.first = fp.last + 1;
 176                                         result.push_back(SpellResultRange(range, it->result()));
 177                                 }
 178                         }
 179                 }
 180                 ranges_ = result;
 181                 if (state != SpellChecker::WORD_OK)
 182                         ranges_.push_back(SpellResultRange(fp, state));
 183         }
 184
 185         void increasePosAfterPos(pos_type pos)
 186         {
 187                 correctRangesAfterPos(pos, 1);
 188                 needsRefresh(pos);
 189         }
 190
 191         void decreasePosAfterPos(pos_type pos)
 192         {
 193                 correctRangesAfterPos(pos, -1);
 194                 needsRefresh(pos);
 195         }
 196
 197         void refreshLast(pos_type pos)
 198         {
 199                 if (pos < refresh_.last)
 200                         refresh_.last = pos;
 201         }
 202
 203         SpellChecker::Result getState(pos_type pos) const
 204         {
 205                 SpellChecker::Result result = SpellChecker::WORD_OK;
 206                 RangesIterator et = ranges_.end();
 207                 RangesIterator it = ranges_.begin();
 208                 for (; it != et; ++it) {
 209                         if(it->contains(pos)) {
 210                                 return it->result();
 211                         }
 212                 }
 213                 return result;
 214         }
 215
 216         FontSpan const & getRange(pos_type pos) const
 217         {
 218                 /// empty span to indicate mismatch
 219                 static FontSpan empty_;
 220                 RangesIterator et = ranges_.end();
 221                 RangesIterator it = ranges_.begin();
 222                 for (; it != et; ++it) {
 223                         if(it->contains(pos)) {
 224                                 return it->range();
 225                         }
 226                 }
 227                 return empty_;
 228         }
 229
 230         bool needsRefresh() const
 231         {
 232                 return needs_refresh_;
 233         }
 234
 235         SpellChecker::ChangeNumber currentChangeNumber() const
 236         {
 237                 return current_change_number_;
 238         }
 239
 240         void refreshRange(pos_type & first, pos_type & last) const
 241         {
 242                 first = refresh_.first;
 243                 last = refresh_.last;
 244         }
 245
 246         void needsRefresh(pos_type pos)
 247         {
 248                 if (needs_refresh_ && pos != -1) {
 249                         if (pos < refresh_.first)
 250                                 refresh_.first = pos;
 251                         if (pos > refresh_.last)
 252                                 refresh_.last = pos;
 253                 } else if (pos != -1) {
 254                         // init request check for neighbour positions too
 255                         refresh_.first = pos > 0 ? pos - 1 : 0;
 256                         // no need for special end of paragraph check
 257                         refresh_.last = pos + 1;
 258                 }
 259                 needs_refresh_ = pos != -1;
 260         }
 261
 262         void needsCompleteRefresh(SpellChecker::ChangeNumber change_number)
 263         {
 264                 needs_refresh_ = true;
 265                 refresh_.first = 0;
 266                 refresh_.last = -1;
 267                 current_change_number_ = change_number;
 268         }
 269 private:
 270         typedef vector<SpellResultRange> Ranges;
 271         typedef Ranges::const_iterator RangesIterator;
 272         Ranges ranges_;
 273         /// the area of the paragraph with pending spell check
 274         FontSpan refresh_;
 275         bool needs_refresh_;
 276         /// spell state cache version number
 277         SpellChecker::ChangeNumber current_change_number_;
 278
 279
 280         void correctRangesAfterPos(pos_type pos, int offset)
 281         {
 282                 RangesIterator et = ranges_.end();
 283                 Ranges::iterator it = ranges_.begin();
 284                 for (; it != et; ++it) {
 285                         it->shift(pos, offset);
 286                 }
 287         }
 288
 289 };
 290
 291 /////////////////////////////////////////////////////////////////////
 292 //
 293 // Paragraph::Private
 294 //
 295 /////////////////////////////////////////////////////////////////////
 296
 297 class Paragraph::Private
 298 {
 299         // Enforce our own "copy" constructor
 300         Private(Private const &) = delete;
 301         Private & operator=(Private const &) = delete;
 302         // Unique ID generator
 303         static int make_id();
 304 public:
 305         ///
 306         Private(Paragraph * owner, Layout const & layout);
 307         /// "Copy constructor"
 308         Private(Private const &, Paragraph * owner);
 309         /// Copy constructor from \p beg  to \p end
 310         Private(Private const &, Paragraph * owner, pos_type beg, pos_type end);
 311
 312         ///
 313         void insertChar(pos_type pos, char_type c, Change const & change);
 314
 315         /// Output the surrogate pair formed by \p c and \p next to \p os.
 316         /// \return the number of characters written.
 317         int latexSurrogatePair(BufferParams const &, otexstream & os,
 318                                char_type c, char_type next,
 319                                OutputParams const &);
 320
 321         /// Output a space in appropriate formatting (or a surrogate pair
 322         /// if the next character is a combining character).
 323         /// \return whether a surrogate pair was output.
 324         bool simpleTeXBlanks(BufferParams const &,
 325                              OutputParams const &,
 326                              otexstream &,
 327                              pos_type i,
 328                              unsigned int & column,
 329                              Font const & font,
 330                              Layout const & style);
 331
 332         /// Output consecutive unicode chars, belonging to the same script as
 333         /// specified by the latex macro \p ltx, to \p os starting from \p i.
 334         /// \return the number of characters written.
 335         int writeScriptChars(BufferParams const &, OutputParams const &,
 336                              otexstream & os,
 337                              docstring const & ltx,
 338                              Change const &, Encoding const &,
 339                              std::string const, pos_type & i);
 340
 341         /// This could go to ParagraphParameters if we want to.
 342         int startTeXParParams(BufferParams const &, otexstream &,
 343                               OutputParams const &) const;
 344
 345         /// This could go to ParagraphParameters if we want to.
 346         bool endTeXParParams(BufferParams const &, otexstream &,
 347                              OutputParams const &) const;
 348
 349         ///
 350         void latexInset(BufferParams const &,
 351                                    otexstream &,
 352                                    OutputParams &,
 353                                    Font & running_font,
 354                                    Font & basefont,
 355                                    Font const & outerfont,
 356                                    bool & open_font,
 357                                    Change & running_change,
 358                                    Layout const & style,
 359                                    pos_type & i,
 360                                    unsigned int & column);
 361
 362         ///
 363         void latexSpecialChar(
 364                                    otexstream & os,
 365                                    BufferParams const & bparams,
 366                                    OutputParams const & runparams,
 367                                    Font const & running_font,
 368                                    Change const & running_change,
 369                                    Layout const & style,
 370                                    pos_type & i,
 371                                    pos_type end_pos,
 372                                    unsigned int & column);
 373
 374         ///
 375         bool latexSpecialT1(
 376                 char_type const c,
 377                 otexstream & os,
 378                 pos_type i,
 379                 unsigned int & column);
 380         ///
 381         bool latexSpecialTU(
 382                 char_type const c,
 383                 otexstream & os,
 384                 pos_type i,
 385                 unsigned int & column);
 386         ///
 387         bool latexSpecialT3(
 388                 char_type const c,
 389                 otexstream & os,
 390                 pos_type i,
 391                 unsigned int & column);
 392
 393         ///
 394         void validate(LaTeXFeatures & features) const;
 395
 396         /// Checks if the paragraph contains only text and no inset or font change.
 397         bool onlyText(Buffer const & buf, Font const & outerfont,
 398                       pos_type initial) const;
 399
 400         /// a vector of speller skip positions
 401         typedef vector<FontSpan> SkipPositions;
 402         typedef SkipPositions::const_iterator SkipPositionsIterator;
 403
 404         void appendSkipPosition(SkipPositions & skips, pos_type const pos) const;
 405
 406         Language * getSpellLanguage(pos_type const from) const;
 407
 408         Language * locateSpellRange(pos_type & from, pos_type & to,
 409                                     SkipPositions & skips) const;
 410
 411         bool hasSpellerChange() const
 412         {
 413                 SpellChecker::ChangeNumber speller_change_number = 0;
 414                 if (theSpellChecker())
 415                         speller_change_number = theSpellChecker()->changeNumber();
 416                 return speller_change_number > speller_state_.currentChangeNumber();
 417         }
 418
 419         bool ignoreWord(docstring const & word) const ;
 420
 421         void setMisspelled(pos_type from, pos_type to, SpellChecker::Result state)
 422         {
 423                 pos_type textsize = owner_->size();
 424                 // check for sane arguments
 425                 if (to <= from || from >= textsize)
 426                         return;
 427                 FontSpan fp = FontSpan(from, to - 1);
 428                 speller_state_.setRange(fp, state);
 429         }
 430
 431         void requestSpellCheck(pos_type pos)
 432         {
 433                 if (pos == -1)
 434                         speller_state_.needsCompleteRefresh(speller_state_.currentChangeNumber());
 435                 else
 436                         speller_state_.needsRefresh(pos);
 437         }
 438
 439         void readySpellCheck()
 440         {
 441                 speller_state_.needsRefresh(-1);
 442         }
 443
 444         bool needsSpellCheck() const
 445         {
 446                 return speller_state_.needsRefresh();
 447         }
 448
 449         void rangeOfSpellCheck(pos_type & first, pos_type & last) const
 450         {
 451                 speller_state_.refreshRange(first, last);
 452                 if (last == -1) {
 453                         last = owner_->size();
 454                         return;
 455                 }
 456                 pos_type endpos = last;
 457                 owner_->locateWord(first, endpos, WHOLE_WORD);
 458                 if (endpos < last) {
 459                         endpos = last;
 460                         owner_->locateWord(last, endpos, WHOLE_WORD);
 461                 }
 462                 last = endpos;
 463         }
 464
 465         int countSkips(SkipPositionsIterator & it, SkipPositionsIterator const et,
 466                             int & start) const
 467         {
 468                 int numskips = 0;
 469                 while (it != et && it->first < start) {
 470                         int skip = it->last - it->first + 1;
 471                         start += skip;
 472                         numskips += skip;
 473                         ++it;
 474                 }
 475                 return numskips;
 476         }
 477
 478         void markMisspelledWords(pos_type const & first, pos_type const & last,
 479                                                          SpellChecker::Result result,
 480                                                          docstring const & word,
 481                                                          SkipPositions const & skips);
 482
 483         InsetCode ownerCode() const
 484         {
 485                 return inset_owner_ ? inset_owner_->lyxCode() : NO_CODE;
 486         }
 487
 488         /// Which Paragraph owns us?
 489         Paragraph * owner_;
 490
 491         /// In which Inset?
 492         Inset const * inset_owner_;
 493
 494         ///
 495         FontList fontlist_;
 496
 497         ///
 498         int id_;
 499
 500         ///
 501         ParagraphParameters params_;
 502
 503         /// for recording and looking up changes
 504         Changes changes_;
 505
 506         ///
 507         InsetList insetlist_;
 508
 509         /// end of label
 510         pos_type begin_of_body_;
 511
 512         typedef docstring TextContainer;
 513         ///
 514         TextContainer text_;
 515
 516         typedef set<docstring> Words;
 517         typedef map<string, Words> LangWordsMap;
 518         ///
 519         LangWordsMap words_;
 520         ///
 521         Layout const * layout_;
 522         ///
 523         SpellCheckerState speller_state_;
 524 };
 525
 526
 527 Paragraph::Private::Private(Paragraph * owner, Layout const & layout)
 528         : owner_(owner), inset_owner_(0), id_(-1), begin_of_body_(0), layout_(&layout)
 529 {
 530         text_.reserve(100);
 531 }
 532
 533
 534 //static
 535 int Paragraph::Private::make_id()
 536 {
 537         // The id is unique per session across buffers because it is used in
 538         // LFUN_PARAGRAPH_GOTO to switch to a different buffer, for instance in the
 539         // outliner.
 540         // (thread-safe)
 541         static atomic_uint next_id(0);
 542         return next_id++;
 543 }
 544
 545
 546 Paragraph::Private::Private(Private const & p, Paragraph * owner)
 547         : owner_(owner), inset_owner_(p.inset_owner_), fontlist_(p.fontlist_),
 548           id_(make_id()),
 549           params_(p.params_), changes_(p.changes_), insetlist_(p.insetlist_),
 550           begin_of_body_(p.begin_of_body_), text_(p.text_), words_(p.words_),
 551           layout_(p.layout_)
 552 {
 553         requestSpellCheck(p.text_.size());
 554 }
 555
 556
 557 Paragraph::Private::Private(Private const & p, Paragraph * owner,
 558         pos_type beg, pos_type end)
 559         : owner_(owner), inset_owner_(p.inset_owner_), id_(make_id()),
 560           params_(p.params_), changes_(p.changes_),
 561           insetlist_(p.insetlist_, beg, end),
 562           begin_of_body_(p.begin_of_body_), words_(p.words_),
 563           layout_(p.layout_)
 564 {
 565         if (beg >= pos_type(p.text_.size()))
 566                 return;
 567         text_ = p.text_.substr(beg, end - beg);
 568
 569         FontList::const_iterator fcit = fontlist_.begin();
 570         FontList::const_iterator fend = fontlist_.end();
 571         for (; fcit != fend; ++fcit) {
 572                 if (fcit->pos() < beg)
 573                         continue;
 574                 if (fcit->pos() >= end) {
 575                         // Add last entry in the fontlist_.
 576                         fontlist_.set(text_.size() - 1, fcit->font());
 577                         break;
 578                 }
 579                 // Add a new entry in the fontlist_.
 580                 fontlist_.set(fcit->pos() - beg, fcit->font());
 581         }
 582         requestSpellCheck(p.text_.size());
 583 }
 584
 585
 586 void Paragraph::addChangesToToc(DocIterator const & cdit, Buffer const & buf,
 587                                 bool output_active, TocBackend & backend) const
 588 {
 589         d->changes_.addToToc(cdit, buf, output_active, backend);
 590 }
 591
 592
 593 void Paragraph::addChangesToBuffer(Buffer const & buf) const
 594 {
 595         d->changes_.updateBuffer(buf);
 596 }
 597
 598
 599 bool Paragraph::isChangeUpdateRequired() const
 600 {
 601         return d->changes_.isUpdateRequired();
 602 }
 603
 604
 605 bool Paragraph::isDeleted(pos_type start, pos_type end) const
 606 {
 607         LASSERT(start >= 0 && start <= size(), return false);
 608         LASSERT(end > start && end <= size() + 1, return false);
 609
 610         return d->changes_.isDeleted(start, end);
 611 }
 612
 613
 614 bool Paragraph::isChanged(pos_type start, pos_type end) const
 615 {
 616         LASSERT(start >= 0 && start <= size(), return false);
 617         LASSERT(end > start && end <= size() + 1, return false);
 618
 619         return d->changes_.isChanged(start, end);
 620 }
 621
 622
 623 bool Paragraph::isMergedOnEndOfParDeletion(bool trackChanges) const
 624 {
 625         // keep the logic here in sync with the logic of eraseChars()
 626         if (!trackChanges)
 627                 return true;
 628
 629         Change const & change = d->changes_.lookup(size());
 630         return change.inserted() && change.currentAuthor();
 631 }
 632
 633
 634 void Paragraph::setChange(Change const & change)
 635 {
 636         // beware of the imaginary end-of-par character!
 637         d->changes_.set(change, 0, size() + 1);
 638
 639         /*
 640          * Propagate the change recursively - but not in case of DELETED!
 641          *
 642          * Imagine that your co-author makes changes in an existing inset. He
 643          * sends your document to you and you come to the conclusion that the
 644          * inset should go completely. If you erase it, LyX must not delete all
 645          * text within the inset. Otherwise, the change tracked insertions of
 646          * your co-author get lost and there is no way to restore them later.
 647          *
 648          * Conclusion: An inset's content should remain untouched if you delete it
 649          */
 650
 651         if (!change.deleted()) {
 652                 for (pos_type pos = 0; pos < size(); ++pos) {
 653                         if (Inset * inset = getInset(pos))
 654                                 inset->setChange(change);
 655                 }
 656         }
 657 }
 658
 659
 660 void Paragraph::setChange(pos_type pos, Change const & change)
 661 {
 662         LASSERT(pos >= 0 && pos <= size(), return);
 663         d->changes_.set(change, pos);
 664
 665         // see comment in setChange(Change const &) above
 666         if (!change.deleted() && pos < size())
 667                         if (Inset * inset = getInset(pos))
 668                                 inset->setChange(change);
 669 }
 670
 671
 672 Change const & Paragraph::lookupChange(pos_type pos) const
 673 {
 674         LBUFERR(pos >= 0 && pos <= size());
 675         return d->changes_.lookup(pos);
 676 }
 677
 678
 679 void Paragraph::acceptChanges(pos_type start, pos_type end)
 680 {
 681         LASSERT(start >= 0 && start <= size(), return);
 682         LASSERT(end > start && end <= size() + 1, return);
 683
 684         for (pos_type pos = start; pos < end; ++pos) {
 685                 switch (lookupChange(pos).type) {
 686                         case Change::UNCHANGED:
 687                                 // accept changes in nested inset
 688                                 if (Inset * inset = getInset(pos))
 689                                         inset->acceptChanges();
 690                                 break;
 691
 692                         case Change::INSERTED:
 693                                 d->changes_.set(Change(Change::UNCHANGED), pos);
 694                                 // also accept changes in nested inset
 695                                 if (Inset * inset = getInset(pos))
 696                                         inset->acceptChanges();
 697                                 break;
 698
 699                         case Change::DELETED:
 700                                 // Suppress access to non-existent
 701                                 // "end-of-paragraph char"
 702                                 if (pos < size()) {
 703                                         eraseChar(pos, false);
 704                                         --end;
 705                                         --pos;
 706                                 }
 707                                 break;
 708                 }
 709
 710         }
 711 }
 712
 713
 714 void Paragraph::rejectChanges(pos_type start, pos_type end)
 715 {
 716         LASSERT(start >= 0 && start <= size(), return);
 717         LASSERT(end > start && end <= size() + 1, return);
 718
 719         for (pos_type pos = start; pos < end; ++pos) {
 720                 switch (lookupChange(pos).type) {
 721                         case Change::UNCHANGED:
 722                                 // reject changes in nested inset
 723                                 if (Inset * inset = getInset(pos))
 724                                                 inset->rejectChanges();
 725                                 break;
 726
 727                         case Change::INSERTED:
 728                                 // Suppress access to non-existent
 729                                 // "end-of-paragraph char"
 730                                 if (pos < size()) {
 731                                         eraseChar(pos, false);
 732                                         --end;
 733                                         --pos;
 734                                 }
 735                                 break;
 736
 737                         case Change::DELETED:
 738                                 d->changes_.set(Change(Change::UNCHANGED), pos);
 739
 740                                 // Do NOT reject changes within a deleted inset!
 741                                 // There may be insertions of a co-author inside of it!
 742
 743                                 break;
 744                 }
 745         }
 746 }
 747
 748
 749 void Paragraph::Private::insertChar(pos_type pos, char_type c,
 750                 Change const & change)
 751 {
 752         LASSERT(pos >= 0 && pos <= int(text_.size()), return);
 753
 754         // track change
 755         changes_.insert(change, pos);
 756
 757         // This is actually very common when parsing buffers (and
 758         // maybe inserting ascii text)
 759         if (pos == pos_type(text_.size())) {
 760                 // when appending characters, no need to update tables
 761                 text_.push_back(c);
 762                 // but we want spell checking
 763                 requestSpellCheck(pos);
 764                 return;
 765         }
 766
 767         text_.insert(text_.begin() + pos, c);
 768
 769         // Update the font table.
 770         fontlist_.increasePosAfterPos(pos);
 771
 772         // Update the insets
 773         insetlist_.increasePosAfterPos(pos);
 774
 775         // Update list of misspelled positions
 776         speller_state_.increasePosAfterPos(pos);
 777 }
 778
 779
 780 bool Paragraph::insertInset(pos_type pos, Inset * inset,
 781                                    Font const & font, Change const & change)
 782 {
 783         LASSERT(inset, return false);
 784         LASSERT(pos >= 0 && pos <= size(), return false);
 785
 786         // Paragraph::insertInset() can be used in cut/copy/paste operation where
 787         // d->inset_owner_ is not set yet.
 788         if (d->inset_owner_ && !d->inset_owner_->insetAllowed(inset->lyxCode()))
 789                 return false;
 790
 791         d->insertChar(pos, META_INSET, change);
 792         LASSERT(d->text_[pos] == META_INSET, return false);
 793
 794         // Add a new entry in the insetlist_.
 795         d->insetlist_.insert(inset, pos);
 796
 797         // Some insets require run of spell checker
 798         requestSpellCheck(pos);
 799         setFont(pos, font);
 800         return true;
 801 }
 802
 803
 804 bool Paragraph::eraseChar(pos_type pos, bool trackChanges)
 805 {
 806         LASSERT(pos >= 0 && pos <= size(), return false);
 807
 808         // keep the logic here in sync with the logic of isMergedOnEndOfParDeletion()
 809
 810         if (trackChanges) {
 811                 Change change = d->changes_.lookup(pos);
 812
 813                 // set the character to DELETED if
 814                 //  a) it was previously unchanged or
 815                 //  b) it was inserted by a co-author
 816
 817                 if (!change.changed() ||
 818                       (change.inserted() && !change.currentAuthor())) {
 819                         setChange(pos, Change(Change::DELETED));
 820                         // request run of spell checker
 821                         requestSpellCheck(pos);
 822                         return false;
 823                 }
 824
 825                 if (change.deleted())
 826                         return false;
 827         }
 828
 829         // Don't physically access the imaginary end-of-paragraph character.
 830         // eraseChar() can only mark it as DELETED. A physical deletion of
 831         // end-of-par must be handled externally.
 832         if (pos == size()) {
 833                 return false;
 834         }
 835
 836         // track change
 837         d->changes_.erase(pos);
 838
 839         // if it is an inset, delete the inset entry
 840         if (d->text_[pos] == META_INSET)
 841                 d->insetlist_.erase(pos);
 842
 843         d->text_.erase(d->text_.begin() + pos);
 844
 845         // Update the fontlist_
 846         d->fontlist_.erase(pos);
 847
 848         // Update the insetlist_
 849         d->insetlist_.decreasePosAfterPos(pos);
 850
 851         // Update list of misspelled positions
 852         d->speller_state_.decreasePosAfterPos(pos);
 853         d->speller_state_.refreshLast(size());
 854
 855         return true;
 856 }
 857
 858
 859 int Paragraph::eraseChars(pos_type start, pos_type end, bool trackChanges)
 860 {
 861         LASSERT(start >= 0 && start <= size(), return 0);
 862         LASSERT(end >= start && end <= size() + 1, return 0);
 863
 864         pos_type i = start;
 865         for (pos_type count = end - start; count; --count) {
 866                 if (!eraseChar(i, trackChanges))
 867                         ++i;
 868         }
 869         return end - i;
 870 }
 871
 872
 873 int Paragraph::Private::latexSurrogatePair(BufferParams const & bparams,
 874                 otexstream & os, char_type c, char_type next,
 875                 OutputParams const & runparams)
 876 {
 877         // Writing next here may circumvent a possible font change between
 878         // c and next. Since next is only output if it forms a surrogate pair
 879         // with c we can ignore this:
 880         // A font change inside a surrogate pair does not make sense and is
 881         // hopefully impossible to input.
 882         // FIXME: change tracking
 883         // Is this correct WRT change tracking?
 884         Encoding const & encoding = *(runparams.encoding);
 885         docstring latex1 = encoding.latexChar(next).first;
 886         if (runparams.inIPA) {
 887                 string const tipashortcut = Encodings::TIPAShortcut(next);
 888                 if (!tipashortcut.empty()) {
 889                         latex1 = from_ascii(tipashortcut);
 890                 }
 891         }
 892         docstring latex2 = encoding.latexChar(c).first;
 893
 894         if (docstring(1, next) == latex1) {
 895                 // The encoding supports the combination:
 896                 // output as is (combining char after base char).
 897                 os << latex2 << latex1;
 898                 return latex1.length() + latex2.length();
 899         }
 900
 901         // Handle combining characters in "script" context (i.e., \textgreek and \textcyrillic)
 902         docstring::size_type const brace1 = latex2.find_first_of(from_ascii("{"));
 903         docstring::size_type const brace2 = latex2.find_last_of(from_ascii("}"));
 904         string script = to_ascii(latex2.substr(1, brace1 - 1));
 905         // "Script chars" need to embraced in \textcyrillic and \textgreek notwithstanding
 906         // whether they are encodable or not (it only depends on the font encoding)
 907         if (!runparams.isFullUnicode())
 908                 // This will get us a script value to deal with below
 909                 Encodings::isKnownScriptChar(c, script);
 910         int pos = 0;
 911         int length = brace2;
 912         string fontenc;
 913         if (runparams.local_font)
 914                 fontenc = runparams.local_font->language()->fontenc(bparams);
 915         else
 916                 fontenc = bparams.language->fontenc(bparams);
 917         docstring scriptmacro;
 918         docstring cb;
 919         if (script == "textgreek" || script == "textcyrillic") {
 920                 // We separate the script macro (\text[greek|cyr]) from the rest,
 921                 // since we need to include the combining char in it (#6463).
 922                 // This is "the rest":
 923                 pos = brace1 + 1;
 924                 length -= pos;
 925                 latex2 = latex2.substr(pos, length);
 926                 // We only need the script macro with non-native font encodings
 927                 if (Encodings::needsScriptWrapper(script, fontenc)) {
 928                         scriptmacro = from_ascii("\\" + script + "{");
 929                         cb = from_ascii("}");
 930                 }
 931         }
 932
 933         docstring lb;
 934         docstring rb;
 935         // polutonikogreek does not play nice with brackets
 936         if (!runparams.local_font
 937             || runparams.local_font->language()->lang() != "polutonikogreek") {
 938                 lb = from_ascii("{");
 939                 rb = from_ascii("}");
 940         }
 941
 942         os << scriptmacro << latex1 << lb << latex2 << rb << cb;
 943         return latex1.length() + latex2.length() + lb.length() + rb.length() + cb.length();
 944 }
 945
 946
 947 bool Paragraph::Private::simpleTeXBlanks(BufferParams const & bparams,
 948                                        OutputParams const & runparams,
 949                                        otexstream & os,
 950                                        pos_type i,
 951                                        unsigned int & column,
 952                                        Font const & font,
 953                                        Layout const & style)
 954 {
 955         if (style.pass_thru || runparams.pass_thru)
 956                 return false;
 957
 958         if (i + 1 < int(text_.size())) {
 959                 char_type next = text_[i + 1];
 960                 if (Encodings::isCombiningChar(next)) {
 961                         // This space has an accent, so we must always output it.
 962                         column += latexSurrogatePair(bparams, os, ' ', next, runparams) - 1;
 963                         return true;
 964                 }
 965         }
 966
 967         if (runparams.linelen > 0
 968             && column > runparams.linelen
 969             && i
 970             && text_[i - 1] != ' '
 971             && (i + 1 < int(text_.size()))
 972             // same in FreeSpacing mode
 973             && !owner_->isFreeSpacing()
 974             // In typewriter mode, we want to avoid
 975             // ! . ? : at the end of a line
 976             && !(font.fontInfo().family() == TYPEWRITER_FAMILY
 977                  && (text_[i - 1] == '.'
 978                      || text_[i - 1] == '?'
 979                      || text_[i - 1] == ':'
 980                      || text_[i - 1] == '!'))) {
 981                 os << '\n';
 982                 os.texrow().start(owner_->id(), i + 1);
 983                 column = 0;
 984         } else if (style.free_spacing) {
 985                 os << '~';
 986         } else {
 987                 os << ' ';
 988         }
 989         return false;
 990 }
 991
 992
 993 int Paragraph::Private::writeScriptChars(BufferParams const & bparams,
 994                                          OutputParams const & runparams,
 995                                          otexstream & os,
 996                                          docstring const & ltx,
 997                                          Change const & runningChange,
 998                                          Encoding const & encoding,
 999                                          string const fontenc,
1000                                          pos_type & i)
1001 {
1002         // FIXME: modifying i here is not very nice...
1003
1004         // We only arrive here when character text_[i] could not be translated
1005         // into the current latex encoding (or its latex translation has been forced,)
1006         // and it belongs to a known script.
1007         // Parameter ltx contains the latex translation of text_[i] as specified
1008         // in the unicodesymbols file and is something like "\textXXX{<spec>}".
1009         // The latex macro name "textXXX" specifies the script to which text_[i]
1010         // belongs and we use it in order to check whether characters from the
1011         // same script immediately follow, such that we can collect them in a
1012         // single "\textXXX" macro. So, we have to retain "\textXXX{<spec>"
1013         // for the first char but only "<spec>" for all subsequent chars.
1014         docstring::size_type const brace1 = ltx.find_first_of(from_ascii("{"));
1015         docstring::size_type const brace2 = ltx.find_last_of(from_ascii("}"));
1016         string script = to_ascii(ltx.substr(1, brace1 - 1));
1017         int pos = 0;
1018         int length = brace2;
1019         bool closing_brace = true;
1020         if (!Encodings::needsScriptWrapper(script, fontenc)) {
1021                 // Correct font encoding is being used, so we can avoid \text[greek|cyr].
1022                 pos = brace1 + 1;
1023                 length -= pos;
1024                 closing_brace = false;
1025         }
1026         os << ltx.substr(pos, length);
1027         int size = text_.size();
1028         while (i + 1 < size) {
1029                 char_type const next = text_[i + 1];
1030                 // Stop here if next character belongs to another script
1031                 // or there is a change in change tracking status.
1032                 if (!Encodings::isKnownScriptChar(next, script) ||
1033                     runningChange != owner_->lookupChange(i + 1))
1034                         break;
1035                 Font prev_font;
1036                 bool found = false;
1037                 FontList::const_iterator cit = fontlist_.begin();
1038                 FontList::const_iterator end = fontlist_.end();
1039                 for (; cit != end; ++cit) {
1040                         if (cit->pos() >= i && !found) {
1041                                 prev_font = cit->font();
1042                                 found = true;
1043                         }
1044                         if (cit->pos() >= i + 1)
1045                                 break;
1046                 }
1047                 // Stop here if there is a font attribute or encoding change.
1048                 if (found && cit != end && prev_font != cit->font())
1049                         break;
1050
1051                 // Check whether we have a combining pair
1052                 char_type next_next = '\0';
1053                 if (i + 2 < size) {
1054                         next_next = text_[i + 2];
1055                         if (Encodings::isCombiningChar(next_next)) {
1056                                 length += latexSurrogatePair(bparams, os, next, next_next, runparams) - 1;
1057                                 i += 2;
1058                                 continue;
1059                         }
1060                 }
1061
1062                 docstring const latex = encoding.latexChar(next).first;
1063                 docstring::size_type const b1 =
1064                                         latex.find_first_of(from_ascii("{"));
1065                 docstring::size_type const b2 =
1066                                         latex.find_last_of(from_ascii("}"));
1067                 int const len = b2 - b1 - 1;
1068                 os << latex.substr(b1 + 1, len);
1069                 length += len;
1070                 ++i;
1071         }
1072         if (closing_brace) {
1073                 os << '}';
1074                 ++length;
1075         }
1076         return length;
1077 }
1078
1079
1080 void Paragraph::Private::latexInset(BufferParams const & bparams,
1081                                     otexstream & os,
1082                                     OutputParams & runparams,
1083                                     Font & running_font,
1084                                     Font & basefont,
1085                                     Font const & outerfont,
1086                                     bool & open_font,
1087                                     Change & running_change,
1088                                     Layout const & style,
1089                                     pos_type & i,
1090                                     unsigned int & column)
1091 {
1092         Inset * inset = owner_->getInset(i);
1093         LBUFERR(inset);
1094
1095         if (style.pass_thru) {
1096                 odocstringstream ods;
1097                 inset->plaintext(ods, runparams);
1098                 os << ods.str();
1099                 return;
1100         }
1101
1102         // FIXME: move this to InsetNewline::latex
1103         if (inset->lyxCode() == NEWLINE_CODE || inset->lyxCode() == SEPARATOR_CODE) {
1104                 // newlines are handled differently here than
1105                 // the default in simpleTeXSpecialChars().
1106                 if (!style.newline_allowed) {
1107                         os << '\n';
1108                 } else {
1109                         if (open_font) {
1110                                 bool needPar = false;
1111                                 column += running_font.latexWriteEndChanges(
1112                                         os, bparams, runparams,
1113                                         basefont, basefont, needPar);
1114                                 open_font = false;
1115                         }
1116
1117                         if (running_font.fontInfo().family() == TYPEWRITER_FAMILY)
1118                                 os << '~';
1119
1120                         basefont = owner_->getLayoutFont(bparams, outerfont);
1121                         running_font = basefont;
1122
1123                         if (runparams.moving_arg)
1124                                 os << "\\protect ";
1125
1126                 }
1127                 os.texrow().start(owner_->id(), i + 1);
1128                 column = 0;
1129         }
1130
1131         if (owner_->isDeleted(i)) {
1132                 if( ++runparams.inDeletedInset == 1)
1133                         runparams.changeOfDeletedInset = owner_->lookupChange(i);
1134         }
1135
1136         if (inset->canTrackChanges()) {
1137                 column += Changes::latexMarkChange(os, bparams, running_change,
1138                         Change(Change::UNCHANGED), runparams);
1139                 running_change = Change(Change::UNCHANGED);
1140         }
1141
1142         bool close = false;
1143         odocstream::pos_type const len = os.os().tellp();
1144
1145         if (inset->forceLTR()
1146             && running_font.isRightToLeft()
1147             // ERT is an exception, it should be output with no
1148             // decorations at all
1149             && inset->lyxCode() != ERT_CODE) {
1150                 if (runparams.use_polyglossia) {
1151                         if (style.isCommand())
1152                                 os << "\\LR{";
1153                         os << "\\begin{LTR}";
1154                 } else if (running_font.language()->lang() == "farsi"
1155                            || running_font.language()->lang() == "arabic_arabi")
1156                         os << "\\textLR{" << termcmd;
1157                 else
1158                         os << "\\L{";
1159                 close = true;
1160         }
1161
1162         // FIXME: Bug: we can have an empty font change here!
1163         // if there has just been a font change, we are going to close it
1164         // right now, which means stupid latex code like \textsf{}. AFAIK,
1165         // this does not harm dvi output. A minor bug, thus (JMarc)
1166
1167         // Some insets cannot be inside a font change command.
1168         // However, even such insets *can* be placed in \L or \R
1169         // or their equivalents (for RTL language switches), so we don't
1170         // close the language in those cases.
1171         // ArabTeX, though, cannot handle this special behavior, it seems.
1172         bool arabtex = basefont.language()->lang() == "arabic_arabtex"
1173                 || running_font.language()->lang() == "arabic_arabtex";
1174         if (open_font && !inset->inheritFont()) {
1175                 bool needPar = false;
1176                 bool closeLanguage = arabtex
1177                         || basefont.isRightToLeft() == running_font.isRightToLeft();
1178                 unsigned int count = running_font.latexWriteEndChanges(os,
1179                                         bparams, runparams, basefont, basefont,
1180                                         needPar, closeLanguage);
1181                 column += count;
1182                 // if any font properties were closed, update the running_font,
1183                 // making sure, however, to leave the language as it was
1184                 if (count > 0) {
1185                         // FIXME: probably a better way to keep track of the old
1186                         // language, than copying the entire font?
1187                         Font const copy_font(running_font);
1188                         basefont = owner_->getLayoutFont(bparams, outerfont);
1189                         running_font = basefont;
1190                         if (!closeLanguage)
1191                                 running_font.setLanguage(copy_font.language());
1192                         // leave font open if language is still open
1193                         open_font = (running_font.language() == basefont.language());
1194                         if (closeLanguage)
1195                                 runparams.local_font = &basefont;
1196                 }
1197         }
1198
1199         size_t const previous_row_count = os.texrow().rows();
1200
1201         try {
1202                 runparams.lastid = id_;
1203                 runparams.lastpos = i;
1204                 inset->latex(os, runparams);
1205         } catch (EncodingException & e) {
1206                 // add location information and throw again.
1207                 e.par_id = id_;
1208                 e.pos = i;
1209                 throw(e);
1210         }
1211
1212         if (close) {
1213                 if (runparams.use_polyglossia && !style.isCommand())
1214                         os << "\\end{LTR}";
1215                 else
1216                         os << '}';
1217         }
1218
1219         if (os.texrow().rows() > previous_row_count) {
1220                 os.texrow().start(owner_->id(), i + 1);
1221                 column = 0;
1222         } else {
1223                 column += (unsigned int)(os.os().tellp() - len);
1224         }
1225
1226         if (owner_->isDeleted(i))
1227                 --runparams.inDeletedInset;
1228 }
1229
1230
1231 void Paragraph::Private::latexSpecialChar(otexstream & os,
1232                                           BufferParams const & bparams,
1233                                           OutputParams const & runparams,
1234                                           Font const & running_font,
1235                                           Change const & running_change,
1236                                           Layout const & style,
1237                                           pos_type & i,
1238                                           pos_type end_pos,
1239                                           unsigned int & column)
1240 {
1241         char_type const c = owner_->getUChar(bparams, runparams, i);
1242
1243         if (style.pass_thru || runparams.pass_thru
1244             || contains(style.pass_thru_chars, c)
1245             || contains(runparams.pass_thru_chars, c)) {
1246                 if (c != '\0') {
1247                         Encoding const * const enc = runparams.encoding;
1248                         if (enc && !enc->encodable(c))
1249                                 throw EncodingException(c);
1250                         os.put(c);
1251                 }
1252                 return;
1253         }
1254
1255         // TIPA uses its own T3 encoding
1256         if (runparams.inIPA && latexSpecialT3(c, os, i, column))
1257                 return;
1258         // If T1 font encoding is used, use the special
1259         // characters it provides.
1260         // NOTE: Some languages reset the font encoding internally to a
1261         //       non-standard font encoding. If we are using such a language,
1262         //       we do not output special T1 chars.
1263         if (!runparams.inIPA && !running_font.language()->internalFontEncoding()
1264             && !runparams.isFullUnicode() && bparams.main_font_encoding() == "T1"
1265             && latexSpecialT1(c, os, i, column))
1266                 return;
1267         // NOTE: XeTeX and LuaTeX use EU1/2 (pre 2017) or TU (as of 2017) encoding
1268         else if (!runparams.inIPA && !running_font.language()->internalFontEncoding()
1269                  && runparams.isFullUnicode() && latexSpecialTU(c, os, i, column))
1270                      return;
1271
1272         // Otherwise, we use what LaTeX provides us.
1273         switch (c) {
1274         case '\\':
1275                 os << "\\textbackslash" << termcmd;
1276                 column += 15;
1277                 break;
1278         case '<':
1279                 os << "\\textless" << termcmd;
1280                 column += 10;
1281                 break;
1282         case '>':
1283                 os << "\\textgreater" << termcmd;
1284                 column += 13;
1285                 break;
1286         case '|':
1287                 os << "\\textbar" << termcmd;
1288                 column += 9;
1289                 break;
1290         case '-':
1291                 os << '-';
1292                 if (i + 1 < static_cast<pos_type>(text_.size()) &&
1293                     (end_pos == -1 || i + 1 < end_pos) &&
1294                     text_[i+1] == '-') {
1295                         // Prevent "--" becoming an en dash and "---" an em dash.
1296                         // (Within \ttfamily, "---" is merged to en dash + hyphen.)
1297                         os << "{}";
1298                         column += 2;
1299                 }
1300                 break;
1301         case '\"':
1302                 os << "\\textquotedbl" << termcmd;
1303                 column += 14;
1304                 break;
1305
1306         case '$': case '&':
1307         case '%': case '#': case '{':
1308         case '}': case '_':
1309                 os << '\\';
1310                 os.put(c);
1311                 column += 1;
1312                 break;
1313
1314         case '~':
1315                 os << "\\textasciitilde" << termcmd;
1316                 column += 16;
1317                 break;
1318
1319         case '^':
1320                 os << "\\textasciicircum" << termcmd;
1321                 column += 17;
1322                 break;
1323
1324         case '*':
1325         case '[':
1326         case ']':
1327                 // avoid being mistaken for optional arguments
1328                 os << '{';
1329                 os.put(c);
1330                 os << '}';
1331                 column += 2;
1332                 break;
1333
1334         case ' ':
1335                 // Blanks are printed before font switching.
1336                 // Sure? I am not! (try nice-latex)
1337                 // I am sure it's correct. LyX might be smarter
1338                 // in the future, but for now, nothing wrong is
1339                 // written. (Asger)
1340                 break;
1341
1342         case 0x2013:
1343         case 0x2014:
1344                 // XeTeX's dash behaviour is determined via a global setting
1345                 if (bparams.use_dash_ligatures
1346                     && owner_->getFontSettings(bparams, i).fontInfo().family() != TYPEWRITER_FAMILY
1347                     && !runparams.inIPA
1348                         // TODO #10961: && not in inset Flex Code
1349                         // TODO #10961: && not in layout LyXCode
1350                     && (!bparams.useNonTeXFonts || runparams.flavor != OutputParams::XETEX)) {
1351                         if (c == 0x2013) {
1352                                 // en-dash
1353                                 os << "--";
1354                                 column +=2;
1355                         } else {
1356                                 // em-dash
1357                                 os << "---";
1358                                 column +=3;
1359                         }
1360                         break;
1361                 }
1362                 // fall through
1363         default:
1364                 if (c == '\0')
1365                         return;
1366
1367                 Encoding const & encoding = *(runparams.encoding);
1368                 char_type next = '\0';
1369                 if (i + 1 < int(text_.size())) {
1370                         next = text_[i + 1];
1371                         if (Encodings::isCombiningChar(next)) {
1372                                 column += latexSurrogatePair(bparams, os, c, next, runparams) - 1;
1373                                 ++i;
1374                                 break;
1375                         }
1376                 }
1377                 string script;
1378                 pair<docstring, bool> latex = encoding.latexChar(c);
1379                 docstring nextlatex;
1380                 bool nexttipas = false;
1381                 string nexttipashortcut;
1382                 if (next != '\0' && next != META_INSET && encoding.encodable(next)) {
1383                         nextlatex = encoding.latexChar(next).first;
1384                         if (runparams.inIPA) {
1385                                 nexttipashortcut = Encodings::TIPAShortcut(next);
1386                                 nexttipas = !nexttipashortcut.empty();
1387                         }
1388                 }
1389                 bool tipas = false;
1390                 if (runparams.inIPA) {
1391                         string const tipashortcut = Encodings::TIPAShortcut(c);
1392                         if (!tipashortcut.empty()) {
1393                                 latex.first = from_ascii(tipashortcut);
1394                                 latex.second = false;
1395                                 tipas = true;
1396                         }
1397                 }
1398                 string fontenc;
1399                 fontenc = running_font.language()->fontenc(bparams);
1400                 // "Script chars" need to embraced in \textcyrillic and \textgreek notwithstanding
1401                 // whether they are encodable or not (it only depends on the font encoding)
1402                 if (!runparams.isFullUnicode() && Encodings::isKnownScriptChar(c, script)) {
1403                         docstring const wrapper = from_ascii("\\" + script + "{");
1404                         docstring ltx = latex.first;
1405                         if (!prefixIs(ltx, wrapper))
1406                                 ltx = wrapper + latex.first + from_ascii("}");
1407                         column += writeScriptChars(bparams, runparams, os, ltx, running_change,
1408                                                    encoding, fontenc, i) - 1;
1409                 } else if (latex.second
1410                          && ((!prefixIs(nextlatex, '\\')
1411                                && !prefixIs(nextlatex, '{')
1412                                && !prefixIs(nextlatex, '}'))
1413                              || (nexttipas
1414                                  && !prefixIs(from_ascii(nexttipashortcut), '\\')))
1415                          && !tipas) {
1416                         // Prevent eating of a following
1417                         // space or command corruption by
1418                         // following characters
1419                         if (next == ' ' || next == '\0') {
1420                                 column += latex.first.length() + 1;
1421                                 os << latex.first << "{}";
1422                         } else {
1423                                 column += latex.first.length();
1424                                 os << latex.first << " ";
1425                         }
1426                 } else {
1427                         column += latex.first.length() - 1;
1428                         os << latex.first;
1429                 }
1430                 break;
1431         }
1432 }
1433
1434
1435 bool Paragraph::Private::latexSpecialT1(char_type const c, otexstream & os,
1436         pos_type i, unsigned int & column)
1437 {
1438         switch (c) {
1439         case '>':
1440         case '<':
1441                 os.put(c);
1442                 // In T1 encoding, these characters exist
1443                 // but we should avoid ligatures
1444                 if (i + 1 >= int(text_.size()) || text_[i + 1] != c)
1445                         return true;
1446                 os << "\\textcompwordmark" << termcmd;
1447                 column += 19;
1448                 return true;
1449         case '|':
1450                 os.put(c);
1451                 return true;
1452         case '\"':
1453                 // soul.sty breaks with \char`\"
1454                 os << "\\textquotedbl" << termcmd;
1455                 column += 14;
1456                 return true;
1457         default:
1458                 return false;
1459         }
1460 }
1461
1462
1463 bool Paragraph::Private::latexSpecialTU(char_type const c, otexstream & os,
1464         pos_type i, unsigned int & column)
1465 {
1466         // TU encoding is currently on par with T1.
1467         return latexSpecialT1(c, os, i, column);
1468 }
1469
1470
1471 bool Paragraph::Private::latexSpecialT3(char_type const c, otexstream & os,
1472         pos_type /*i*/, unsigned int & column)
1473 {
1474         switch (c) {
1475         case '*':
1476         case '[':
1477         case ']':
1478         case '\"':
1479                 os.put(c);
1480                 return true;
1481         case '|':
1482                 os << "\\textvertline" << termcmd;
1483                 column += 14;
1484                 return true;
1485         default:
1486                 return false;
1487         }
1488 }
1489
1490
1491 void Paragraph::Private::validate(LaTeXFeatures & features) const
1492 {
1493         if (layout_->inpreamble && inset_owner_) {
1494                 // FIXME: Using a string stream here circumvents the encoding
1495                 // switching machinery of odocstream. Therefore the
1496                 // output is wrong if this paragraph contains content
1497                 // that needs to switch encoding.
1498                 Buffer const & buf = inset_owner_->buffer();
1499                 otexstringstream os;
1500                 os << layout_->preamble();
1501                 size_t const length = os.length();
1502                 TeXOnePar(buf, buf.text(), buf.getParFromID(owner_->id()).pit(), os,
1503                           features.runparams(), string(), 0, -1, true);
1504                 if (os.length() > length)
1505                         features.addPreambleSnippet(os.release(), true);
1506         }
1507
1508         if (features.runparams().flavor == OutputParams::HTML
1509             && layout_->htmltitle()) {
1510                 features.setHTMLTitle(owner_->asString(AS_STR_INSETS | AS_STR_SKIPDELETE));
1511         }
1512
1513         // check the params.
1514         if (!params_.spacing().isDefault())
1515                 features.require("setspace");
1516
1517         // then the layouts
1518         features.useLayout(layout_->name());
1519
1520         // then the fonts
1521         fontlist_.validate(features);
1522
1523         // then the indentation
1524         if (!params_.leftIndent().zero())
1525                 features.require("ParagraphLeftIndent");
1526
1527         // then the insets
1528         InsetList::const_iterator icit = insetlist_.begin();
1529         InsetList::const_iterator iend = insetlist_.end();
1530         for (; icit != iend; ++icit) {
1531                 if (icit->inset) {
1532                         features.inDeletedInset(owner_->isDeleted(icit->pos));
1533                         if (icit->inset->lyxCode() == FOOT_CODE) {
1534                                 // FIXME: an item inset would make things much easier.
1535                                 if ((layout_->latextype == LATEX_LIST_ENVIRONMENT
1536                                      || (layout_->latextype == LATEX_ITEM_ENVIRONMENT
1537                                          && layout_->margintype == MARGIN_FIRST_DYNAMIC))
1538                                     && (icit->pos < begin_of_body_
1539                                         || (icit->pos == begin_of_body_
1540                                             && (icit->pos == 0 || text_[icit->pos - 1] != ' '))))
1541                                         features.saveNoteEnv("description");
1542                         }
1543                         icit->inset->validate(features);
1544                         features.inDeletedInset(false);
1545                         if (layout_->needprotect &&
1546                             icit->inset->lyxCode() == FOOT_CODE)
1547                                 features.require("NeedLyXFootnoteCode");
1548                 }
1549         }
1550
1551         // then the contents
1552         BufferParams const bp = features.runparams().is_child
1553                 ? features.buffer().masterParams() : features.buffer().params();
1554         for (pos_type i = 0; i < int(text_.size()) ; ++i) {
1555                 char_type c = text_[i];
1556                 CharInfo const & ci = Encodings::unicodeCharInfo(c);
1557                 if (c == 0x0022) {
1558                         if (features.runparams().isFullUnicode() && bp.useNonTeXFonts)
1559                                 features.require("textquotedblp");
1560                         else if (bp.main_font_encoding() != "T1"
1561                                  || ((&owner_->getFontSettings(bp, i))->language()->internalFontEncoding()))
1562                                 features.require("textquotedbl");
1563                 } else if (ci.textfeature() && contains(ci.textpreamble(), '=')) {
1564                         // features that depend on the font or input encoding
1565                         string feats = ci.textpreamble();
1566                         string fontenc = (&owner_->getFontSettings(bp, i))->language()->fontenc(bp);
1567                         if (fontenc.empty())
1568                                 fontenc = features.runparams().main_fontenc;
1569                         while (!feats.empty()) {
1570                                 string feat;
1571                                 feats = split(feats, feat, ',');
1572                                 if (contains(feat, "!=")) {
1573                                         // a feature that is required except for the spcified
1574                                         // font or input encodings
1575                                         string realfeature;
1576                                         string const contexts = ltrim(split(feat, realfeature, '!'), "=");
1577                                         // multiple encodings are separated by semicolon
1578                                         vector<string> context = getVectorFromString(contexts, ";");
1579                                         // require feature if the context matches neither current font
1580                                         // nor input encoding
1581                                         if (std::find(context.begin(), context.end(), fontenc) == context.end()
1582                                             && std::find(context.begin(), context.end(),
1583                                                          features.runparams().encoding->name()) == context.end())
1584                                                 features.require(realfeature);
1585                                 } else if (contains(feat, '=')) {
1586                                         // a feature that is required only for the spcified
1587                                         // font or input encodings
1588                                         string realfeature;
1589                                         string const contexts = split(feat, realfeature, '=');
1590                                         // multiple encodings are separated by semicolon
1591                                         vector<string> context = getVectorFromString(contexts, ";");
1592                                         // require feature if the context matches either current font
1593                                         // or input encoding
1594                                         if (std::find(context.begin(), context.end(), fontenc) != context.end()
1595                                             || std::find(context.begin(), context.end(),
1596                                                          features.runparams().encoding->name()) != context.end())
1597                                                 features.require(realfeature);
1598                                 }
1599                         }
1600                 } else if (!bp.use_dash_ligatures
1601                            && (c == 0x2013 || c == 0x2014)
1602                            && bp.useNonTeXFonts
1603                            && features.runparams().flavor == OutputParams::XETEX)
1604                         // XeTeX's dash behaviour is determined via a global setting
1605                         features.require("xetexdashbreakstate");
1606                 BufferEncodings::validate(c, features);
1607         }
1608 }
1609
1610 /////////////////////////////////////////////////////////////////////
1611 //
1612 // Paragraph
1613 //
1614 /////////////////////////////////////////////////////////////////////
1615
1616 namespace {
1617         Layout const emptyParagraphLayout;
1618 }
1619
1620 Paragraph::Paragraph()
1621         : d(new Paragraph::Private(this, emptyParagraphLayout))
1622 {
1623         itemdepth = 0;
1624         d->params_.clear();
1625 }
1626
1627
1628 Paragraph::Paragraph(Paragraph const & par)
1629         : itemdepth(par.itemdepth),
1630         d(new Paragraph::Private(*par.d, this))
1631 {
1632         registerWords();
1633 }
1634
1635
1636 Paragraph::Paragraph(Paragraph const & par, pos_type beg, pos_type end)
1637         : itemdepth(par.itemdepth),
1638         d(new Paragraph::Private(*par.d, this, beg, end))
1639 {
1640         registerWords();
1641 }
1642
1643
1644 Paragraph & Paragraph::operator=(Paragraph const & par)
1645 {
1646         // needed as we will destroy the private part before copying it
1647         if (&par != this) {
1648                 itemdepth = par.itemdepth;
1649
1650                 deregisterWords();
1651                 delete d;
1652                 d = new Private(*par.d, this);
1653                 registerWords();
1654         }
1655         return *this;
1656 }
1657
1658
1659 Paragraph::~Paragraph()
1660 {
1661         deregisterWords();
1662         delete d;
1663 }
1664
1665
1666 namespace {
1667
1668 // this shall be called just before every "os << ..." action.
1669 void flushString(ostream & os, docstring & s)
1670 {
1671         os << to_utf8(s);
1672         s.erase();
1673 }
1674
1675 } // namespace
1676
1677
1678 void Paragraph::write(ostream & os, BufferParams const & bparams,
1679         depth_type & dth) const
1680 {
1681         // The beginning or end of a deeper (i.e. nested) area?
1682         if (dth != d->params_.depth()) {
1683                 if (d->params_.depth() > dth) {
1684                         while (d->params_.depth() > dth) {
1685                                 os << "\n\\begin_deeper";
1686                                 ++dth;
1687                         }
1688                 } else {
1689                         while (d->params_.depth() < dth) {
1690                                 os << "\n\\end_deeper";
1691                                 --dth;
1692                         }
1693                 }
1694         }
1695
1696         // First write the layout
1697         os << "\n\\begin_layout " << to_utf8(d->layout_->name()) << '\n';
1698
1699         d->params_.write(os);
1700
1701         Font font1(inherit_font, bparams.language);
1702
1703         Change running_change = Change(Change::UNCHANGED);
1704
1705         // this string is used as a buffer to avoid repetitive calls
1706         // to to_utf8(), which turn out to be expensive (JMarc)
1707         docstring write_buffer;
1708
1709         int column = 0;
1710         for (pos_type i = 0; i <= size(); ++i) {
1711
1712                 Change const & change = lookupChange(i);
1713                 if (change != running_change)
1714                         flushString(os, write_buffer);
1715                 Changes::lyxMarkChange(os, bparams, column, running_change, change);
1716                 running_change = change;
1717
1718                 if (i == size())
1719                         break;
1720
1721                 // Write font changes
1722                 Font font2 = getFontSettings(bparams, i);
1723                 if (font2 != font1) {
1724                         flushString(os, write_buffer);
1725                         font2.lyxWriteChanges(font1, os);
1726                         column = 0;
1727                         font1 = font2;
1728                 }
1729
1730                 char_type const c = d->text_[i];
1731                 switch (c) {
1732                 case META_INSET:
1733                         if (Inset const * inset = getInset(i)) {
1734                                 flushString(os, write_buffer);
1735                                 if (inset->directWrite()) {
1736                                         // international char, let it write
1737                                         // code directly so it's shorter in
1738                                         // the file
1739                                         inset->write(os);
1740                                 } else {
1741                                         if (i)
1742                                                 os << '\n';
1743                                         os << "\\begin_inset ";
1744                                         inset->write(os);
1745                                         os << "\n\\end_inset\n\n";
1746                                         column = 0;
1747                                 }
1748                                 // FIXME This can be removed again once the mystery
1749                                 // crash has been resolved.
1750                                 os << flush;
1751                         }
1752                         break;
1753                 case '\\':
1754                         flushString(os, write_buffer);
1755                         os << "\n\\backslash\n";
1756                         column = 0;
1757                         break;
1758                 case '.':
1759                         flushString(os, write_buffer);
1760                         if (i + 1 < size() && d->text_[i + 1] == ' ') {
1761                                 os << ".\n";
1762                                 column = 0;
1763                         } else
1764                                 os << '.';
1765                         break;
1766                 default:
1767                         if ((column > 70 && c == ' ')
1768                             || column > 79) {
1769                                 flushString(os, write_buffer);
1770                                 os << '\n';
1771                                 column = 0;
1772                         }
1773                         // this check is to amend a bug. LyX sometimes
1774                         // inserts '\0' this could cause problems.
1775                         if (c != '\0')
1776                                 write_buffer.push_back(c);
1777                         else
1778                                 LYXERR0("NUL char in structure.");
1779                         ++column;
1780                         break;
1781                 }
1782         }
1783
1784         flushString(os, write_buffer);
1785         os << "\n\\end_layout\n";
1786         // FIXME This can be removed again once the mystery
1787         // crash has been resolved.
1788         os << flush;
1789 }
1790
1791
1792 void Paragraph::validate(LaTeXFeatures & features) const
1793 {
1794         d->validate(features);
1795         bool fragile = features.runparams().moving_arg;
1796         fragile |= layout().needprotect;
1797         if (needsCProtection(fragile))
1798                 features.require("cprotect");
1799 }
1800
1801
1802 void Paragraph::insert(pos_type start, docstring const & str,
1803                        Font const & font, Change const & change)
1804 {
1805         for (size_t i = 0, n = str.size(); i != n ; ++i)
1806                 insertChar(start + i, str[i], font, change);
1807 }
1808
1809
1810 void Paragraph::appendChar(char_type c, Font const & font,
1811                 Change const & change)
1812 {
1813         // track change
1814         d->changes_.insert(change, d->text_.size());
1815         // when appending characters, no need to update tables
1816         d->text_.push_back(c);
1817         setFont(d->text_.size() - 1, font);
1818         d->requestSpellCheck(d->text_.size() - 1);
1819 }
1820
1821
1822 void Paragraph::appendString(docstring const & s, Font const & font,
1823                 Change const & change)
1824 {
1825         pos_type end = s.size();
1826         size_t oldsize = d->text_.size();
1827         size_t newsize = oldsize + end;
1828         size_t capacity = d->text_.capacity();
1829         if (newsize >= capacity)
1830                 d->text_.reserve(max(capacity + 100, newsize));
1831
1832         // when appending characters, no need to update tables
1833         d->text_.append(s);
1834
1835         // FIXME: Optimize this!
1836         for (size_t i = oldsize; i != newsize; ++i) {
1837                 // track change
1838                 d->changes_.insert(change, i);
1839                 d->requestSpellCheck(i);
1840         }
1841         d->fontlist_.set(oldsize, font);
1842         d->fontlist_.set(newsize - 1, font);
1843 }
1844
1845
1846 void Paragraph::insertChar(pos_type pos, char_type c,
1847                            bool trackChanges)
1848 {
1849         d->insertChar(pos, c, Change(trackChanges ?
1850                            Change::INSERTED : Change::UNCHANGED));
1851 }
1852
1853
1854 void Paragraph::insertChar(pos_type pos, char_type c,
1855                            Font const & font, bool trackChanges)
1856 {
1857         d->insertChar(pos, c, Change(trackChanges ?
1858                            Change::INSERTED : Change::UNCHANGED));
1859         setFont(pos, font);
1860 }
1861
1862
1863 void Paragraph::insertChar(pos_type pos, char_type c,
1864                            Font const & font, Change const & change)
1865 {
1866         d->insertChar(pos, c, change);
1867         setFont(pos, font);
1868 }
1869
1870
1871 void Paragraph::resetFonts(Font const & font)
1872 {
1873         d->fontlist_.clear();
1874         d->fontlist_.set(0, font);
1875         d->fontlist_.set(d->text_.size() - 1, font);
1876 }
1877
1878 // Gets uninstantiated font setting at position.
1879 Font const & Paragraph::getFontSettings(BufferParams const & bparams,
1880                                          pos_type pos) const
1881 {
1882         if (pos > size()) {
1883                 LYXERR0("pos: " << pos << " size: " << size());
1884                 LBUFERR(false);
1885         }
1886
1887         FontList::const_iterator cit = d->fontlist_.fontIterator(pos);
1888         if (cit != d->fontlist_.end())
1889                 return cit->font();
1890
1891         if (pos == size() && !empty())
1892                 return getFontSettings(bparams, pos - 1);
1893
1894         // Optimisation: avoid a full font instantiation if there is no
1895         // language change from previous call.
1896         static Font previous_font;
1897         static Language const * previous_lang = 0;
1898         Language const * lang = getParLanguage(bparams);
1899         if (lang != previous_lang) {
1900                 previous_lang = lang;
1901                 previous_font = Font(inherit_font, lang);
1902         }
1903         return previous_font;
1904 }
1905
1906
1907 FontSpan Paragraph::fontSpan(pos_type pos) const
1908 {
1909         LBUFERR(pos <= size());
1910
1911         if (pos == size())
1912                 return FontSpan(pos, pos);
1913
1914         pos_type start = 0;
1915         FontList::const_iterator cit = d->fontlist_.begin();
1916         FontList::const_iterator end = d->fontlist_.end();
1917         for (; cit != end; ++cit) {
1918                 if (cit->pos() >= pos) {
1919                         if (pos >= beginOfBody())
1920                                 return FontSpan(max(start, beginOfBody()),
1921                                                 cit->pos());
1922                         else
1923                                 return FontSpan(start,
1924                                                 min(beginOfBody() - 1,
1925                                                          cit->pos()));
1926                 }
1927                 start = cit->pos() + 1;
1928         }
1929
1930         // This should not happen, but if so, we take no chances.
1931         LYXERR0("Paragraph::fontSpan: position not found in fontinfo table!");
1932         LASSERT(false, return FontSpan(pos, pos));
1933 }
1934
1935
1936 // Gets uninstantiated font setting at position 0
1937 Font const & Paragraph::getFirstFontSettings(BufferParams const & bparams) const
1938 {
1939         if (!empty() && !d->fontlist_.empty())
1940                 return d->fontlist_.begin()->font();
1941
1942         // Optimisation: avoid a full font instantiation if there is no
1943         // language change from previous call.
1944         static Font previous_font;
1945         static Language const * previous_lang = 0;
1946         if (bparams.language != previous_lang) {
1947                 previous_lang = bparams.language;
1948                 previous_font = Font(inherit_font, bparams.language);
1949         }
1950
1951         return previous_font;
1952 }
1953
1954
1955 // Gets the fully instantiated font at a given position in a paragraph
1956 // This is basically the same function as Text::GetFont() in text2.cpp.
1957 // The difference is that this one is used for generating the LaTeX file,
1958 // and thus cosmetic "improvements" are disallowed: This has to deliver
1959 // the true picture of the buffer. (Asger)
1960 Font const Paragraph::getFont(BufferParams const & bparams, pos_type pos,
1961                                  Font const & outerfont) const
1962 {
1963         LBUFERR(pos >= 0);
1964
1965         Font font = getFontSettings(bparams, pos);
1966
1967         pos_type const body_pos = beginOfBody();
1968         FontInfo & fi = font.fontInfo();
1969         if (pos < body_pos)
1970                 fi.realize(d->layout_->labelfont);
1971         else
1972                 fi.realize(d->layout_->font);
1973
1974         fi.realize(outerfont.fontInfo());
1975         fi.realize(bparams.getFont().fontInfo());
1976
1977         return font;
1978 }
1979
1980
1981 Font const Paragraph::getLabelFont
1982         (BufferParams const & bparams, Font const & outerfont) const
1983 {
1984         FontInfo tmpfont = d->layout_->labelfont;
1985         tmpfont.realize(outerfont.fontInfo());
1986         tmpfont.realize(bparams.getFont().fontInfo());
1987         return Font(tmpfont, getParLanguage(bparams));
1988 }
1989
1990
1991 Font const Paragraph::getLayoutFont
1992         (BufferParams const & bparams, Font const & outerfont) const
1993 {
1994         FontInfo tmpfont = d->layout_->font;
1995         tmpfont.realize(outerfont.fontInfo());
1996         tmpfont.realize(bparams.getFont().fontInfo());
1997         return Font(tmpfont, getParLanguage(bparams));
1998 }
1999
2000
2001 char_type Paragraph::getUChar(BufferParams const & bparams,
2002                               OutputParams const & rp,
2003                               pos_type pos) const
2004 {
2005         char_type c = d->text_[pos];
2006
2007         // Return unchanged character in LTR languages.
2008         if (!getFontSettings(bparams, pos).isRightToLeft())
2009                 return c;
2010
2011         // FIXME This is a complete mess due to all the language-specific
2012         // special cases. We need to unify this eventually, but this
2013         // requires a file format change and some thought.
2014         // We also need to unify the input of parentheses in different RTL
2015         // languages. Currently, some have their own methods (Arabic:
2016         // 18599/lyxsvn, Hebrew: e5f42f67d/lyxgit), some don't (Urdu, Syriac).
2017         // Also note that the representation in the LyX file is probably wrong
2018         // (see FIXME in TextMetrics::breakRow).
2019         // Most likely, we should simply rely on Qt's unicode handling here.
2020         string const & lang = getFontSettings(bparams, pos).language()->lang();
2021
2022         // With polyglossia, brackets and stuff need not be reversed in RTL scripts
2023         // FIXME: The special casing for Hebrew parens is due to the special
2024         // handling on input (for Hebrew in e5f42f67d/lyxgit); see #8251.
2025         char_type uc = c;
2026         if (rp.use_polyglossia) {
2027                 switch (c) {
2028                 case '(':
2029                         if (lang == "hebrew")
2030                                 uc = ')';
2031                         break;
2032                 case ')':
2033                         if (lang == "hebrew")
2034                                 uc = '(';
2035                         break;
2036                 }
2037                 return uc;
2038         }
2039
2040         // In the following languages, brackets don't need to be reversed.
2041         // Furthermore, in arabic_arabi, they are transformed to Arabic
2042         // Ornate Parentheses (dunno if this is really wanted)
2043         bool const reversebrackets = lang != "arabic_arabtex"
2044                         && lang != "arabic_arabi"
2045                         && lang != "farsi";
2046
2047         switch (c) {
2048         case '[':
2049                 if (reversebrackets)
2050                         uc = ']';
2051                 break;
2052         case ']':
2053                 if (reversebrackets)
2054                         uc = '[';
2055                 break;
2056         case '{':
2057                 uc = '}';
2058                 break;
2059         case '}':
2060                 uc = '{';
2061                 break;
2062         case '<':
2063                 uc = '>';
2064                 break;
2065         case '>':
2066                 uc = '<';
2067                 break;
2068         }
2069
2070         return uc;
2071 }
2072
2073
2074 void Paragraph::setFont(pos_type pos, Font const & font)
2075 {
2076         LASSERT(pos <= size(), return);
2077
2078         // First, reduce font against layout/label font
2079         // Update: The setCharFont() routine in text2.cpp already
2080         // reduces font, so we don't need to do that here. (Asger)
2081
2082         d->fontlist_.set(pos, font);
2083 }
2084
2085
2086 void Paragraph::makeSameLayout(Paragraph const & par)
2087 {
2088         d->layout_ = par.d->layout_;
2089         d->params_ = par.d->params_;
2090 }
2091
2092
2093 bool Paragraph::stripLeadingSpaces(bool trackChanges)
2094 {
2095         if (isFreeSpacing())
2096                 return false;
2097
2098         int pos = 0;
2099         int count = 0;
2100
2101         while (pos < size() && (isNewline(pos) || isLineSeparator(pos))) {
2102                 if (eraseChar(pos, trackChanges))
2103                         ++count;
2104                 else
2105                         ++pos;
2106         }
2107
2108         return count > 0 || pos > 0;
2109 }
2110
2111
2112 bool Paragraph::hasSameLayout(Paragraph const & par) const
2113 {
2114         return par.d->layout_ == d->layout_
2115                 && d->params_.sameLayout(par.d->params_);
2116 }
2117
2118
2119 depth_type Paragraph::getDepth() const
2120 {
2121         return d->params_.depth();
2122 }
2123
2124
2125 depth_type Paragraph::getMaxDepthAfter() const
2126 {
2127         if (d->layout_->isEnvironment())
2128                 return d->params_.depth() + 1;
2129         else
2130                 return d->params_.depth();
2131 }
2132
2133
2134 LyXAlignment Paragraph::getAlign() const
2135 {
2136         if (d->params_.align() == LYX_ALIGN_LAYOUT)
2137                 return d->layout_->align;
2138         else
2139                 return d->params_.align();
2140 }
2141
2142
2143 docstring const & Paragraph::labelString() const
2144 {
2145         return d->params_.labelString();
2146 }
2147
2148
2149 // the next two functions are for the manual labels
2150 docstring const Paragraph::getLabelWidthString() const
2151 {
2152         if (d->layout_->margintype == MARGIN_MANUAL
2153             || d->layout_->latextype == LATEX_BIB_ENVIRONMENT)
2154                 return d->params_.labelWidthString();
2155         else
2156                 return _("Senseless with this layout!");
2157 }
2158
2159
2160 void Paragraph::setLabelWidthString(docstring const & s)
2161 {
2162         d->params_.labelWidthString(s);
2163 }
2164
2165
2166 docstring Paragraph::expandLabel(Layout const & layout,
2167                 BufferParams const & bparams) const
2168 {
2169         return expandParagraphLabel(layout, bparams, true);
2170 }
2171
2172
2173 docstring Paragraph::expandDocBookLabel(Layout const & layout,
2174                 BufferParams const & bparams) const
2175 {
2176         return expandParagraphLabel(layout, bparams, false);
2177 }
2178
2179
2180 docstring Paragraph::expandParagraphLabel(Layout const & layout,
2181                 BufferParams const & bparams, bool process_appendix) const
2182 {
2183         DocumentClass const & tclass = bparams.documentClass();
2184         string const & lang = getParLanguage(bparams)->code();
2185         bool const in_appendix = process_appendix && d->params_.appendix();
2186         docstring fmt = translateIfPossible(layout.labelstring(in_appendix), lang);
2187
2188         if (fmt.empty() && !layout.counter.empty())
2189                 return tclass.counters().theCounter(layout.counter, lang);
2190
2191         // handle 'inherited level parts' in 'fmt',
2192         // i.e. the stuff between '@' in   '@Section@.\arabic{subsection}'
2193         size_t const i = fmt.find('@', 0);
2194         if (i != docstring::npos) {
2195                 size_t const j = fmt.find('@', i + 1);
2196                 if (j != docstring::npos) {
2197                         docstring parent(fmt, i + 1, j - i - 1);
2198                         docstring label = from_ascii("??");
2199                         if (tclass.hasLayout(parent))
2200                                 label = expandParagraphLabel(tclass[parent], bparams,
2201                                                       process_appendix);
2202                         fmt = docstring(fmt, 0, i) + label
2203                                 + docstring(fmt, j + 1, docstring::npos);
2204                 }
2205         }
2206
2207         return tclass.counters().counterLabel(fmt, lang);
2208 }
2209
2210
2211 void Paragraph::applyLayout(Layout const & new_layout)
2212 {
2213         d->layout_ = &new_layout;
2214         LyXAlignment const oldAlign = d->params_.align();
2215
2216         if (!(oldAlign & d->layout_->alignpossible)) {
2217                 frontend::Alert::warning(_("Alignment not permitted"),
2218                         _("The new layout does not permit the alignment previously used.\nSetting to default."));
2219                 d->params_.align(LYX_ALIGN_LAYOUT);
2220         }
2221 }
2222
2223
2224 pos_type Paragraph::beginOfBody() const
2225 {
2226         return d->begin_of_body_;
2227 }
2228
2229
2230 void Paragraph::setBeginOfBody()
2231 {
2232         if (d->layout_->labeltype != LABEL_MANUAL) {
2233                 d->begin_of_body_ = 0;
2234                 return;
2235         }
2236
2237         // Unroll the first two cycles of the loop
2238         // and remember the previous character to
2239         // remove unnecessary getChar() calls
2240         pos_type i = 0;
2241         pos_type end = size();
2242         if (i < end && !(isNewline(i) || isEnvSeparator(i))) {
2243                 ++i;
2244                 if (i < end) {
2245                         char_type previous_char = d->text_[i];
2246                         if (!(isNewline(i) || isEnvSeparator(i))) {
2247                                 ++i;
2248                                 while (i < end && previous_char != ' ') {
2249                                         char_type temp = d->text_[i];
2250                                         if (isNewline(i) || isEnvSeparator(i))
2251                                                 break;
2252                                         ++i;
2253                                         previous_char = temp;
2254                                 }
2255                         }
2256                 }
2257         }
2258
2259         d->begin_of_body_ = i;
2260 }
2261
2262
2263 bool Paragraph::allowParagraphCustomization() const
2264 {
2265         return inInset().allowParagraphCustomization();
2266 }
2267
2268
2269 bool Paragraph::usePlainLayout() const
2270 {
2271         return inInset().usePlainLayout();
2272 }
2273
2274
2275 bool Paragraph::isPassThru() const
2276 {
2277         return inInset().isPassThru() || d->layout_->pass_thru;
2278 }
2279
2280 namespace {
2281
2282 // paragraphs inside floats need different alignment tags to avoid
2283 // unwanted space
2284
2285 bool noTrivlistCentering(InsetCode code)
2286 {
2287         return code == FLOAT_CODE
2288                || code == WRAP_CODE
2289                || code == CELL_CODE;
2290 }
2291
2292
2293 string correction(string const & orig)
2294 {
2295         if (orig == "flushleft")
2296                 return "raggedright";
2297         if (orig == "flushright")
2298                 return "raggedleft";
2299         if (orig == "center")
2300                 return "centering";
2301         return orig;
2302 }
2303
2304
2305 bool corrected_env(otexstream & os, string const & suffix, string const & env,
2306         InsetCode code, bool const lastpar, int & col)
2307 {
2308         string macro = suffix + "{";
2309         if (noTrivlistCentering(code)) {
2310                 if (lastpar) {
2311                         // the last paragraph in non-trivlist-aligned
2312                         // context is special (to avoid unwanted whitespace)
2313                         if (suffix == "\\begin") {
2314                                 macro = "\\" + correction(env) + "{}";
2315                                 os << from_ascii(macro);
2316                                 col += macro.size();
2317                                 return true;
2318                         }
2319                         return false;
2320                 }
2321                 macro += correction(env);
2322         } else
2323                 macro += env;
2324         macro += "}";
2325         if (suffix == "\\par\\end") {
2326                 os << breakln;
2327                 col = 0;
2328         }
2329         os << from_ascii(macro);
2330         col += macro.size();
2331         if (suffix == "\\begin") {
2332                 os << breakln;
2333                 col = 0;
2334         }
2335         return true;
2336 }
2337
2338 } // namespace
2339
2340
2341 int Paragraph::Private::startTeXParParams(BufferParams const & bparams,
2342                         otexstream & os, OutputParams const & runparams) const
2343 {
2344         int column = 0;
2345
2346         bool canindent =
2347                 (bparams.paragraph_separation == BufferParams::ParagraphIndentSeparation) ?
2348                         (layout_->toggle_indent != ITOGGLE_NEVER) :
2349                         (layout_->toggle_indent == ITOGGLE_ALWAYS);
2350
2351         if (canindent && params_.noindent() && !layout_->pass_thru) {
2352                 os << "\\noindent ";
2353                 column += 10;
2354         }
2355
2356         LyXAlignment const curAlign = params_.align();
2357
2358         if (curAlign == layout_->align)
2359                 return column;
2360
2361         switch (curAlign) {
2362         case LYX_ALIGN_NONE:
2363         case LYX_ALIGN_BLOCK:
2364         case LYX_ALIGN_LAYOUT:
2365         case LYX_ALIGN_SPECIAL:
2366         case LYX_ALIGN_DECIMAL:
2367                 break;
2368         case LYX_ALIGN_LEFT:
2369         case LYX_ALIGN_RIGHT:
2370         case LYX_ALIGN_CENTER:
2371                 if (runparams.moving_arg) {
2372                         os << "\\protect";
2373                         column += 8;
2374                 }
2375                 break;
2376         }
2377
2378         string const begin_tag = "\\begin";
2379         InsetCode code = ownerCode();
2380         bool const lastpar = runparams.isLastPar;
2381
2382         switch (curAlign) {
2383         case LYX_ALIGN_NONE:
2384         case LYX_ALIGN_BLOCK:
2385         case LYX_ALIGN_LAYOUT:
2386         case LYX_ALIGN_SPECIAL:
2387         case LYX_ALIGN_DECIMAL:
2388                 break;
2389         case LYX_ALIGN_LEFT: {
2390                 if (!owner_->getParLanguage(bparams)->rightToLeft())
2391                         corrected_env(os, begin_tag, "flushleft", code, lastpar, column);
2392                 else
2393                         corrected_env(os, begin_tag, "flushright", code, lastpar, column);
2394                 break;
2395         } case LYX_ALIGN_RIGHT: {
2396                 if (!owner_->getParLanguage(bparams)->rightToLeft())
2397                         corrected_env(os, begin_tag, "flushright", code, lastpar, column);
2398                 else
2399                         corrected_env(os, begin_tag, "flushleft", code, lastpar, column);
2400                 break;
2401         } case LYX_ALIGN_CENTER: {
2402                 corrected_env(os, begin_tag, "center", code, lastpar, column);
2403                 break;
2404         }
2405         }
2406
2407         return column;
2408 }
2409
2410
2411 bool Paragraph::Private::endTeXParParams(BufferParams const & bparams,
2412                         otexstream & os, OutputParams const & runparams) const
2413 {
2414         LyXAlignment const curAlign = params_.align();
2415
2416         if (curAlign == layout_->align)
2417                 return false;
2418
2419         switch (curAlign) {
2420         case LYX_ALIGN_NONE:
2421         case LYX_ALIGN_BLOCK:
2422         case LYX_ALIGN_LAYOUT:
2423         case LYX_ALIGN_SPECIAL:
2424         case LYX_ALIGN_DECIMAL:
2425                 break;
2426         case LYX_ALIGN_LEFT:
2427         case LYX_ALIGN_RIGHT:
2428         case LYX_ALIGN_CENTER:
2429                 if (runparams.moving_arg)
2430                         os << "\\protect";
2431                 break;
2432         }
2433
2434         bool output = false;
2435         int col = 0;
2436         string const end_tag = "\\par\\end";
2437         InsetCode code = ownerCode();
2438         bool const lastpar = runparams.isLastPar;
2439
2440         switch (curAlign) {
2441         case LYX_ALIGN_NONE:
2442         case LYX_ALIGN_BLOCK:
2443         case LYX_ALIGN_LAYOUT:
2444         case LYX_ALIGN_SPECIAL:
2445         case LYX_ALIGN_DECIMAL:
2446                 break;
2447         case LYX_ALIGN_LEFT: {
2448                 if (!owner_->getParLanguage(bparams)->rightToLeft())
2449                         output = corrected_env(os, end_tag, "flushleft", code, lastpar, col);
2450                 else
2451                         output = corrected_env(os, end_tag, "flushright", code, lastpar, col);
2452                 break;
2453         } case LYX_ALIGN_RIGHT: {
2454                 if (!owner_->getParLanguage(bparams)->rightToLeft())
2455                         output = corrected_env(os, end_tag, "flushright", code, lastpar, col);
2456                 else
2457                         output = corrected_env(os, end_tag, "flushleft", code, lastpar, col);
2458                 break;
2459         } case LYX_ALIGN_CENTER: {
2460                 corrected_env(os, end_tag, "center", code, lastpar, col);
2461                 break;
2462         }
2463         }
2464
2465         return output || lastpar;
2466 }
2467
2468
2469 // This one spits out the text of the paragraph
2470 void Paragraph::latex(BufferParams const & bparams,
2471         Font const & outerfont,
2472         otexstream & os,
2473         OutputParams const & runparams,
2474         int start_pos, int end_pos, bool force) const
2475 {
2476         LYXERR(Debug::LATEX, "Paragraph::latex...     " << this);
2477
2478         // FIXME This check should not be needed. Perhaps issue an
2479         // error if it triggers.
2480         Layout const & style = inInset().forcePlainLayout() ?
2481                 bparams.documentClass().plainLayout() : *d->layout_;
2482
2483         if (!force && style.inpreamble)
2484                 return;
2485
2486         bool const allowcust = allowParagraphCustomization();
2487
2488         // Current base font for all inherited font changes, without any
2489         // change caused by an individual character, except for the language:
2490         // It is set to the language of the first character.
2491         // As long as we are in the label, this font is the base font of the
2492         // label. Before the first body character it is set to the base font
2493         // of the body.
2494         Font basefont;
2495
2496         // Maybe we have to create a optional argument.
2497         pos_type body_pos = beginOfBody();
2498         unsigned int column = 0;
2499
2500         if (body_pos > 0) {
2501                 // the optional argument is kept in curly brackets in
2502                 // case it contains a ']'
2503                 // This is not strictly needed, but if this is changed it
2504                 // would be a file format change, and tex2lyx would need
2505                 // to be adjusted, since it unconditionally removes the
2506                 // braces when it parses \item.
2507                 os << "[{";
2508                 column += 2;
2509                 basefont = getLabelFont(bparams, outerfont);
2510         } else {
2511                 basefont = getLayoutFont(bparams, outerfont);
2512         }
2513
2514         // Which font is currently active?
2515         Font running_font(basefont);
2516         // Do we have an open font change?
2517         bool open_font = false;
2518
2519         Change runningChange = Change(Change::UNCHANGED);
2520
2521         Encoding const * const prev_encoding = runparams.encoding;
2522
2523         os.texrow().start(id(), 0);
2524
2525         // if the paragraph is empty, the loop will not be entered at all
2526         if (empty()) {
2527                 // For InTitle commands, we have already opened a group
2528                 // in output_latex::TeXOnePar.
2529                 if (style.isCommand() && !style.intitle) {
2530                         os << '{';
2531                         ++column;
2532                 }
2533                 if (!style.leftdelim().empty()) {
2534                         os << style.leftdelim();
2535                         column += style.leftdelim().size();
2536                 }
2537                 if (allowcust)
2538                         column += d->startTeXParParams(bparams, os, runparams);
2539         }
2540
2541         // Whether a \par can be issued for insets typeset inline with text.
2542         // Yes if greater than 0. This has to be static.
2543         THREAD_LOCAL_STATIC int parInline = 0;
2544
2545         for (pos_type i = 0; i < size(); ++i) {
2546                 // First char in paragraph or after label?
2547                 if (i == body_pos) {
2548                         if (body_pos > 0) {
2549                                 if (open_font) {
2550                                         bool needPar = false;
2551                                         column += running_font.latexWriteEndChanges(
2552                                                 os, bparams, runparams,
2553                                                 basefont, basefont, needPar);
2554                                         open_font = false;
2555                                 }
2556                                 basefont = getLayoutFont(bparams, outerfont);
2557                                 running_font = basefont;
2558
2559                                 column += Changes::latexMarkChange(os, bparams,
2560                                                 runningChange, Change(Change::UNCHANGED),
2561                                                 runparams);
2562                                 runningChange = Change(Change::UNCHANGED);
2563
2564                                 os << "}] ";
2565                                 column +=3;
2566                         }
2567                         // For InTitle commands, we have already opened a group
2568                         // in output_latex::TeXOnePar.
2569                         if (style.isCommand() && !style.intitle) {
2570                                 os << '{';
2571                                 ++column;
2572                         }
2573
2574                         if (!style.leftdelim().empty()) {
2575                                 os << style.leftdelim();
2576                                 column += style.leftdelim().size();
2577                         }
2578
2579                         if (allowcust)
2580                                 column += d->startTeXParParams(bparams, os,
2581                                                             runparams);
2582                 }
2583
2584                 runparams.wasDisplayMath = runparams.inDisplayMath;
2585                 runparams.inDisplayMath = false;
2586                 bool deleted_display_math = false;
2587                 Change const & change = runparams.inDeletedInset
2588                         ? runparams.changeOfDeletedInset : lookupChange(i);
2589
2590                 // Check whether a display math inset follows
2591                 if (d->text_[i] == META_INSET
2592                     && i >= start_pos && (end_pos == -1 || i < end_pos)) {
2593                         InsetMath const * im = getInset(i)->asInsetMath();
2594                         if (im && im->asHullInset()
2595                             && im->asHullInset()->outerDisplay()) {
2596                                 runparams.inDisplayMath = true;
2597                                 // runparams.inDeletedInset will be set by
2598                                 // latexInset later, but we need this info
2599                                 // before it is called. On the other hand, we
2600                                 // cannot set it here because it is a counter.
2601                                 deleted_display_math = isDeleted(i);
2602                         }
2603                         if (bparams.output_changes && deleted_display_math
2604                             && runningChange == change
2605                             && change.type == Change::DELETED
2606                             && !os.afterParbreak()) {
2607                                 // A display math in the same paragraph follows.
2608                                 // We have to close and then reopen \lyxdeleted,
2609                                 // otherwise the math will be shifted up.
2610                                 OutputParams rp = runparams;
2611                                 if (open_font) {
2612                                         bool needPar = false;
2613                                         column += running_font.latexWriteEndChanges(
2614                                                 os, bparams, rp, basefont,
2615                                                 basefont, needPar);
2616                                         open_font = false;
2617                                 }
2618                                 basefont = getLayoutFont(bparams, outerfont);
2619                                 running_font = basefont;
2620                                 column += Changes::latexMarkChange(os, bparams,
2621                                         Change(Change::INSERTED), change, rp);
2622                         }
2623                 }
2624
2625                 if (bparams.output_changes && runningChange != change) {
2626                         if (open_font) {
2627                                 bool needPar = false;
2628                                 column += running_font.latexWriteEndChanges(
2629                                                 os, bparams, runparams,
2630                                                 basefont, basefont, needPar);
2631                                 open_font = false;
2632                         }
2633                         basefont = getLayoutFont(bparams, outerfont);
2634                         running_font = basefont;
2635                         column += Changes::latexMarkChange(os, bparams, runningChange,
2636                                                            change, runparams);
2637                         runningChange = change;
2638                 }
2639
2640                 // do not output text which is marked deleted
2641                 // if change tracking output is disabled
2642                 if (!bparams.output_changes && change.deleted()) {
2643                         continue;
2644                 }
2645
2646                 ++column;
2647
2648                 // Fully instantiated font
2649                 Font const current_font = getFont(bparams, i, outerfont);
2650
2651                 Font const last_font = running_font;
2652
2653                 // Do we need to close the previous font?
2654                 if (open_font &&
2655                     (current_font != running_font ||
2656                      current_font.language() != running_font.language()))
2657                 {
2658                         bool needPar = false;
2659                         column += running_font.latexWriteEndChanges(
2660                                     os, bparams, runparams, basefont,
2661                                     (i == body_pos-1) ? basefont : current_font,
2662                                     needPar);
2663                         running_font = basefont;
2664                         open_font = false;
2665                 }
2666
2667                 string const running_lang = runparams.use_polyglossia ?
2668                         running_font.language()->polyglossia() : running_font.language()->babel();
2669                 // close babel's font environment before opening CJK.
2670                 string const lang_end_command = runparams.use_polyglossia ?
2671                         "\\end{$$lang}" : lyxrc.language_command_end;
2672                 bool const using_begin_end = runparams.use_polyglossia ||
2673                                                 !lang_end_command.empty();
2674                 if (!running_lang.empty() &&
2675                     (!using_begin_end || running_lang == openLanguageName()) &&
2676                     current_font.language()->encoding()->package() == Encoding::CJK) {
2677                                 string end_tag = subst(lang_end_command,
2678                                                         "$$lang",
2679                                                         running_lang);
2680                                 os << from_ascii(end_tag);
2681                                 column += end_tag.length();
2682                                 if (using_begin_end)
2683                                         popLanguageName();
2684                 }
2685
2686                 // Switch file encoding if necessary (and allowed)
2687                 if (!runparams.pass_thru && !style.pass_thru &&
2688                     runparams.encoding->package() != Encoding::none &&
2689                     current_font.language()->encoding()->package() != Encoding::none) {
2690                         pair<bool, int> const enc_switch =
2691                                 switchEncoding(os.os(), bparams, runparams,
2692                                         *(current_font.language()->encoding()));
2693                         if (enc_switch.first) {
2694                                 column += enc_switch.second;
2695                                 runparams.encoding = current_font.language()->encoding();
2696                         }
2697                 }
2698
2699                 char_type const c = d->text_[i];
2700
2701                 // A display math inset inside an ulem command will be output
2702                 // as a box of width \linewidth, so we have to either disable
2703                 // indentation if the inset starts a paragraph, or start a new
2704                 // line to accommodate such box. This has to be done before
2705                 // writing any font changing commands.
2706                 if (runparams.inDisplayMath && !deleted_display_math
2707                     && runparams.inulemcmd) {
2708                         if (os.afterParbreak())
2709                                 os << "\\noindent";
2710                         else
2711                                 os << "\\\\\n";
2712                 }
2713
2714                 // Do we need to change font?
2715                 if ((current_font != running_font ||
2716                      current_font.language() != running_font.language()) &&
2717                         i != body_pos - 1)
2718                 {
2719                         odocstringstream ods;
2720                         column += current_font.latexWriteStartChanges(ods, bparams,
2721                                                               runparams, basefont,
2722                                                               last_font);
2723                         // Check again for display math in ulem commands as a
2724                         // font change may also occur just before a math inset.
2725                         if (runparams.inDisplayMath && !deleted_display_math
2726                             && runparams.inulemcmd) {
2727                                 if (os.afterParbreak())
2728                                         os << "\\noindent";
2729                                 else
2730                                         os << "\\\\\n";
2731                         }
2732                         running_font = current_font;
2733                         open_font = true;
2734                         docstring fontchange = ods.str();
2735                         // check whether the fontchange ends with a \\textcolor
2736                         // modifier and the text starts with a space (bug 4473)
2737                         docstring const last_modifier = rsplit(fontchange, '\\');
2738                         if (prefixIs(last_modifier, from_ascii("textcolor")) && c == ' ')
2739                                 os << fontchange << from_ascii("{}");
2740                         // check if the fontchange ends with a trailing blank
2741                         // (like "\small " (see bug 3382)
2742                         else if (suffixIs(fontchange, ' ') && c == ' ')
2743                                 os << fontchange.substr(0, fontchange.size() - 1)
2744                                    << from_ascii("{}");
2745                         else
2746                                 os << fontchange;
2747                 }
2748
2749                 // FIXME: think about end_pos implementation...
2750                 if (c == ' ' && i >= start_pos && (end_pos == -1 || i < end_pos)) {
2751                         // FIXME: integrate this case in latexSpecialChar
2752                         // Do not print the separation of the optional argument
2753                         // if style.pass_thru is false. This works because
2754                         // latexSpecialChar ignores spaces if
2755                         // style.pass_thru is false.
2756                         if (i != body_pos - 1) {
2757                                 if (d->simpleTeXBlanks(bparams, runparams, os,
2758                                                 i, column, current_font, style)) {
2759                                         // A surrogate pair was output. We
2760                                         // must not call latexSpecialChar
2761                                         // in this iteration, since it would output
2762                                         // the combining character again.
2763                                         ++i;
2764                                         continue;
2765                                 }
2766                         }
2767                 }
2768
2769                 OutputParams rp = runparams;
2770                 rp.free_spacing = style.free_spacing;
2771                 rp.local_font = &current_font;
2772                 rp.intitle = style.intitle;
2773
2774                 // Two major modes:  LaTeX or plain
2775                 // Handle here those cases common to both modes
2776                 // and then split to handle the two modes separately.
2777                 if (c == META_INSET) {
2778                         if (i >= start_pos && (end_pos == -1 || i < end_pos)) {
2779                                 // Greyedout notes and, in general, all insets
2780                                 // with InsetLayout::isDisplay() == false,
2781                                 // are typeset inline with the text. So, we
2782                                 // can add a \par to the last paragraph of
2783                                 // such insets only if nothing else follows.
2784                                 bool incremented = false;
2785                                 Inset const * inset = getInset(i);
2786                                 InsetText const * textinset = inset
2787                                                         ? inset->asInsetText()
2788                                                         : 0;
2789                                 if (i + 1 == size() && textinset
2790                                     && !inset->getLayout().isDisplay()) {
2791                                         ParagraphList const & pars =
2792                                                 textinset->text().paragraphs();
2793                                         pit_type const pit = pars.size() - 1;
2794                                         Font const lastfont =
2795                                                 pit < 0 || pars[pit].empty()
2796                                                 ? pars[pit].getLayoutFont(
2797                                                                 bparams,
2798                                                                 outerfont)
2799                                                 : pars[pit].getFont(bparams,
2800                                                         pars[pit].size() - 1,
2801                                                         outerfont);
2802                                         if (lastfont.fontInfo().size() !=
2803                                             basefont.fontInfo().size()) {
2804                                                 ++parInline;
2805                                                 incremented = true;
2806                                         }
2807                                 }
2808                                 d->latexInset(bparams, os, rp, running_font,
2809                                                 basefont, outerfont, open_font,
2810                                                 runningChange, style, i, column);
2811                                 if (incremented)
2812                                         --parInline;
2813                         }
2814                 } else if (i >= start_pos && (end_pos == -1 || i < end_pos)) {
2815                         try {
2816                                 d->latexSpecialChar(os, bparams, rp,
2817                                                     running_font, runningChange,
2818                                                     style, i, end_pos, column);
2819                         } catch (EncodingException & e) {
2820                                 if (runparams.dryrun) {
2821                                         os << "<" << _("LyX Warning: ")
2822                                            << _("uncodable character") << " '";
2823                                         os.put(c);
2824                                         os << "'>";
2825                                 } else {
2826                                         // add location information and throw again.
2827                                         e.par_id = id();
2828                                         e.pos = i;
2829                                         throw(e);
2830                                 }
2831                         }
2832                 }
2833
2834                 // Set the encoding to that returned from latexSpecialChar (see
2835                 // comment for encoding member in OutputParams.h)
2836                 runparams.encoding = rp.encoding;
2837
2838                 // Also carry on the info on a closed ulem command for insets
2839                 // such as Note that do not produce any output, so that no
2840                 // command is ever executed but its opening was recorded.
2841                 runparams.inulemcmd = rp.inulemcmd;
2842         }
2843
2844         // If we have an open font definition, we have to close it
2845         if (open_font) {
2846                 // Make sure that \\par is done with the font of the last
2847                 // character if this has another size as the default.
2848                 // This is necessary because LaTeX (and LyX on the screen)
2849                 // calculates the space between the baselines according
2850                 // to this font. (Matthias)
2851                 //
2852                 // We must not change the font for the last paragraph
2853                 // of non-multipar insets, tabular cells or commands,
2854                 // since this produces unwanted whitespace.
2855
2856                 Font const font = empty()
2857                         ? getLayoutFont(bparams, outerfont)
2858                         : getFont(bparams, size() - 1, outerfont);
2859
2860                 InsetText const * textinset = inInset().asInsetText();
2861
2862                 bool const maintext = textinset
2863                         ? textinset->text().isMainText()
2864                         : false;
2865
2866                 size_t const numpars = textinset
2867                         ? textinset->text().paragraphs().size()
2868                         : 0;
2869
2870                 bool needPar = false;
2871
2872                 if (style.resfont.size() != font.fontInfo().size()
2873                     && (!runparams.isLastPar || maintext
2874                         || (numpars > 1 && d->ownerCode() != CELL_CODE
2875                             && (inInset().getLayout().isDisplay()
2876                                 || parInline)))
2877                     && !style.isCommand()) {
2878                         needPar = true;
2879                 }
2880 #ifdef FIXED_LANGUAGE_END_DETECTION
2881                 if (next_) {
2882                         running_font.latexWriteEndChanges(os, bparams,
2883                                         runparams, basefont,
2884                                         next_->getFont(bparams, 0, outerfont),
2885                                                        needPar);
2886                 } else {
2887                         running_font.latexWriteEndChanges(os, bparams,
2888                                         runparams, basefont, basefont, needPar);
2889                 }
2890 #else
2891 //FIXME: For now we ALWAYS have to close the foreign font settings if they are
2892 //FIXME: there as we start another \selectlanguage with the next paragraph if
2893 //FIXME: we are in need of this. This should be fixed sometime (Jug)
2894                 running_font.latexWriteEndChanges(os, bparams, runparams,
2895                                 basefont, basefont, needPar);
2896 #endif
2897                 if (needPar) {
2898                         // The \par could not be inserted at the same nesting
2899                         // level of the font size change, so do it now.
2900                         os << "{\\" << font.latexSize() << "\\par}";
2901                 }
2902         }
2903
2904         column += Changes::latexMarkChange(os, bparams, runningChange,
2905                                            Change(Change::UNCHANGED), runparams);
2906
2907         // Needed if there is an optional argument but no contents.
2908         if (body_pos > 0 && body_pos == size()) {
2909                 os << "}]~";
2910         }
2911
2912         if (!style.rightdelim().empty()) {
2913                 os << style.rightdelim();
2914                 column += style.rightdelim().size();
2915         }
2916
2917         if (allowcust && d->endTeXParParams(bparams, os, runparams)
2918             && runparams.encoding != prev_encoding) {
2919                 runparams.encoding = prev_encoding;
2920                 os << setEncoding(prev_encoding->iconvName());
2921         }
2922
2923         LYXERR(Debug::LATEX, "Paragraph::latex... done " << this);
2924 }
2925
2926
2927 bool Paragraph::emptyTag() const
2928 {
2929         for (pos_type i = 0; i < size(); ++i) {
2930                 if (Inset const * inset = getInset(i)) {
2931                         InsetCode lyx_code = inset->lyxCode();
2932                         // FIXME testing like that is wrong. What is
2933                         // the intent?
2934                         if (lyx_code != TOC_CODE &&
2935                             lyx_code != INCLUDE_CODE &&
2936                             lyx_code != GRAPHICS_CODE &&
2937                             lyx_code != ERT_CODE &&
2938                             lyx_code != LISTINGS_CODE &&
2939                             lyx_code != FLOAT_CODE &&
2940                             lyx_code != TABULAR_CODE) {
2941                                 return false;
2942                         }
2943                 } else {
2944                         char_type c = d->text_[i];
2945                         if (c != ' ' && c != '\t')
2946                                 return false;
2947                 }
2948         }
2949         return true;
2950 }
2951
2952
2953 string Paragraph::getID(Buffer const & buf, OutputParams const & runparams)
2954         const
2955 {
2956         for (pos_type i = 0; i < size(); ++i) {
2957                 if (Inset const * inset = getInset(i)) {
2958                         InsetCode lyx_code = inset->lyxCode();
2959                         if (lyx_code == LABEL_CODE) {
2960                                 InsetLabel const * const il = static_cast<InsetLabel const *>(inset);
2961                                 docstring const & id = il->getParam("name");
2962                                 return "id='" + to_utf8(sgml::cleanID(buf, runparams, id)) + "'";
2963                         }
2964                 }
2965         }
2966         return string();
2967 }
2968
2969
2970 pos_type Paragraph::firstWordDocBook(odocstream & os, OutputParams const & runparams)
2971         const
2972 {
2973         pos_type i;
2974         for (i = 0; i < size(); ++i) {
2975                 if (Inset const * inset = getInset(i)) {
2976                         inset->docbook(os, runparams);
2977                 } else {
2978                         char_type c = d->text_[i];
2979                         if (c == ' ')
2980                                 break;
2981                         os << sgml::escapeChar(c);
2982                 }
2983         }
2984         return i;
2985 }
2986
2987
2988 pos_type Paragraph::firstWordLyXHTML(XHTMLStream & xs, OutputParams const & runparams)
2989         const
2990 {
2991         pos_type i;
2992         for (i = 0; i < size(); ++i) {
2993                 if (Inset const * inset = getInset(i)) {
2994                         inset->xhtml(xs, runparams);
2995                 } else {
2996                         char_type c = d->text_[i];
2997                         if (c == ' ')
2998                                 break;
2999                         xs << c;
3000                 }
3001         }
3002         return i;
3003 }
3004
3005
3006 bool Paragraph::Private::onlyText(Buffer const & buf, Font const & outerfont, pos_type initial) const
3007 {
3008         Font font_old;
3009         pos_type size = text_.size();
3010         for (pos_type i = initial; i < size; ++i) {
3011                 Font font = owner_->getFont(buf.params(), i, outerfont);
3012                 if (text_[i] == META_INSET)
3013                         return false;
3014                 if (i != initial && font != font_old)
3015                         return false;
3016                 font_old = font;
3017         }
3018
3019         return true;
3020 }
3021
3022
3023 void Paragraph::simpleDocBookOnePar(Buffer const & buf,
3024                                     odocstream & os,
3025                                     OutputParams const & runparams,
3026                                     Font const & outerfont,
3027                                     pos_type initial) const
3028 {
3029         bool emph_flag = false;
3030
3031         Layout const & style = *d->layout_;
3032         FontInfo font_old =
3033                 style.labeltype == LABEL_MANUAL ? style.labelfont : style.font;
3034
3035         if (style.pass_thru && !d->onlyText(buf, outerfont, initial))
3036                 os << "]]>";
3037
3038         // parsing main loop
3039         for (pos_type i = initial; i < size(); ++i) {
3040                 Font font = getFont(buf.params(), i, outerfont);
3041
3042                 // handle <emphasis> tag
3043                 if (font_old.emph() != font.fontInfo().emph()) {
3044                         if (font.fontInfo().emph() == FONT_ON) {
3045                                 os << "<emphasis>";
3046                                 emph_flag = true;
3047                         } else if (i != initial) {
3048                                 os << "</emphasis>";
3049                                 emph_flag = false;
3050                         }
3051                 }
3052
3053                 if (Inset const * inset = getInset(i)) {
3054                         inset->docbook(os, runparams);
3055                 } else {
3056                         char_type c = d->text_[i];
3057
3058                         if (style.pass_thru)
3059                                 os.put(c);
3060                         else
3061                                 os << sgml::escapeChar(c);
3062                 }
3063                 font_old = font.fontInfo();
3064         }
3065
3066         if (emph_flag) {
3067                 os << "</emphasis>";
3068         }
3069
3070         if (style.free_spacing)
3071                 os << '\n';
3072         if (style.pass_thru && !d->onlyText(buf, outerfont, initial))
3073                 os << "<![CDATA[";
3074 }
3075
3076
3077 namespace {
3078 void doFontSwitch(vector<html::FontTag> & tagsToOpen,
3079                   vector<html::EndFontTag> & tagsToClose,
3080                   bool & flag, FontState curstate, html::FontTypes type)
3081 {
3082         if (curstate == FONT_ON) {
3083                 tagsToOpen.push_back(html::FontTag(type));
3084                 flag = true;
3085         } else if (flag) {
3086                 tagsToClose.push_back(html::EndFontTag(type));
3087                 flag = false;
3088         }
3089 }
3090 } // namespace
3091
3092
3093 docstring Paragraph::simpleLyXHTMLOnePar(Buffer const & buf,
3094                                     XHTMLStream & xs,
3095                                     OutputParams const & runparams,
3096                                     Font const & outerfont,
3097                                     bool start_paragraph, bool close_paragraph,
3098                                     pos_type initial) const
3099 {
3100         docstring retval;
3101
3102         // track whether we have opened these tags
3103         bool emph_flag = false;
3104         bool bold_flag = false;
3105         bool noun_flag = false;
3106         bool ubar_flag = false;
3107         bool dbar_flag = false;
3108         bool sout_flag = false;
3109         bool xout_flag = false;
3110         bool wave_flag = false;
3111         // shape tags
3112         bool shap_flag = false;
3113         // family tags
3114         bool faml_flag = false;
3115         // size tags
3116         bool size_flag = false;
3117
3118         Layout const & style = *d->layout_;
3119
3120         if (start_paragraph)
3121                 xs.startDivision(allowEmpty());
3122
3123         FontInfo font_old =
3124                 style.labeltype == LABEL_MANUAL ? style.labelfont : style.font;
3125
3126         FontShape  curr_fs   = INHERIT_SHAPE;
3127         FontFamily curr_fam  = INHERIT_FAMILY;
3128         FontSize   curr_size = FONT_SIZE_INHERIT;
3129
3130         string const default_family =
3131                 buf.masterBuffer()->params().fonts_default_family;
3132
3133         vector<html::FontTag> tagsToOpen;
3134         vector<html::EndFontTag> tagsToClose;
3135
3136         // parsing main loop
3137         for (pos_type i = initial; i < size(); ++i) {
3138                 // let's not show deleted material in the output
3139                 if (isDeleted(i))
3140                         continue;
3141
3142                 Font const font = getFont(buf.masterBuffer()->params(), i, outerfont);
3143
3144                 // emphasis
3145                 FontState curstate = font.fontInfo().emph();
3146                 if (font_old.emph() != curstate)
3147                         doFontSwitch(tagsToOpen, tagsToClose, emph_flag, curstate, html::FT_EMPH);
3148
3149                 // noun
3150                 curstate = font.fontInfo().noun();
3151                 if (font_old.noun() != curstate)
3152                         doFontSwitch(tagsToOpen, tagsToClose, noun_flag, curstate, html::FT_NOUN);
3153
3154                 // underbar
3155                 curstate = font.fontInfo().underbar();
3156                 if (font_old.underbar() != curstate)
3157                         doFontSwitch(tagsToOpen, tagsToClose, ubar_flag, curstate, html::FT_UBAR);
3158
3159                 // strikeout
3160                 curstate = font.fontInfo().strikeout();
3161                 if (font_old.strikeout() != curstate)
3162                         doFontSwitch(tagsToOpen, tagsToClose, sout_flag, curstate, html::FT_SOUT);
3163
3164                 // xout
3165                 curstate = font.fontInfo().xout();
3166                 if (font_old.xout() != curstate)
3167                         doFontSwitch(tagsToOpen, tagsToClose, xout_flag, curstate, html::FT_XOUT);
3168
3169                 // double underbar
3170                 curstate = font.fontInfo().uuline();
3171                 if (font_old.uuline() != curstate)
3172                         doFontSwitch(tagsToOpen, tagsToClose, dbar_flag, curstate, html::FT_DBAR);
3173
3174                 // wavy line
3175                 curstate = font.fontInfo().uwave();
3176                 if (font_old.uwave() != curstate)
3177                         doFontSwitch(tagsToOpen, tagsToClose, wave_flag, curstate, html::FT_WAVE);
3178
3179                 // bold
3180                 // a little hackish, but allows us to reuse what we have.
3181                 curstate = (font.fontInfo().series() == BOLD_SERIES ? FONT_ON : FONT_OFF);
3182                 if (font_old.series() != font.fontInfo().series())
3183                         doFontSwitch(tagsToOpen, tagsToClose, bold_flag, curstate, html::FT_BOLD);
3184
3185                 // Font shape
3186                 curr_fs = font.fontInfo().shape();
3187                 FontShape old_fs = font_old.shape();
3188                 if (old_fs != curr_fs) {
3189                         if (shap_flag) {
3190                                 switch (old_fs) {
3191                                 case ITALIC_SHAPE:
3192                                         tagsToClose.push_back(html::EndFontTag(html::FT_ITALIC));
3193                                         break;
3194                                 case SLANTED_SHAPE:
3195                                         tagsToClose.push_back(html::EndFontTag(html::FT_SLANTED));
3196                                         break;
3197                                 case SMALLCAPS_SHAPE:
3198                                         tagsToClose.push_back(html::EndFontTag(html::FT_SMALLCAPS));
3199                                         break;
3200                                 case UP_SHAPE:
3201                                 case INHERIT_SHAPE:
3202                                         break;
3203                                 default:
3204                                         // the other tags are for internal use
3205                                         LATTEST(false);
3206                                         break;
3207                                 }
3208                                 shap_flag = false;
3209                         }
3210                         switch (curr_fs) {
3211                         case ITALIC_SHAPE:
3212                                 tagsToOpen.push_back(html::FontTag(html::FT_ITALIC));
3213                                 shap_flag = true;
3214                                 break;
3215                         case SLANTED_SHAPE:
3216                                 tagsToOpen.push_back(html::FontTag(html::FT_SLANTED));
3217                                 shap_flag = true;
3218                                 break;
3219                         case SMALLCAPS_SHAPE:
3220                                 tagsToOpen.push_back(html::FontTag(html::FT_SMALLCAPS));
3221                                 shap_flag = true;
3222                                 break;
3223                         case UP_SHAPE:
3224                         case INHERIT_SHAPE:
3225                                 break;
3226                         default:
3227                                 // the other tags are for internal use
3228                                 LATTEST(false);
3229                                 break;
3230                         }
3231                 }
3232
3233                 // Font family
3234                 curr_fam = font.fontInfo().family();
3235                 FontFamily old_fam = font_old.family();
3236                 if (old_fam != curr_fam) {
3237                         if (faml_flag) {
3238                                 switch (old_fam) {
3239                                 case ROMAN_FAMILY:
3240                                         tagsToClose.push_back(html::EndFontTag(html::FT_ROMAN));
3241                                         break;
3242                                 case SANS_FAMILY:
3243                                         tagsToClose.push_back(html::EndFontTag(html::FT_SANS));
3244                                         break;
3245                                 case TYPEWRITER_FAMILY:
3246                                         tagsToClose.push_back(html::EndFontTag(html::FT_TYPE));
3247                                         break;
3248                                 case INHERIT_FAMILY:
3249                                         break;
3250                                 default:
3251                                         // the other tags are for internal use
3252                                         LATTEST(false);
3253                                         break;
3254                                 }
3255                                 faml_flag = false;
3256                         }
3257                         switch (curr_fam) {
3258                         case ROMAN_FAMILY:
3259                                 // we will treat a "default" font family as roman, since we have
3260                                 // no other idea what to do.
3261                                 if (default_family != "rmdefault" && default_family != "default") {
3262                                         tagsToOpen.push_back(html::FontTag(html::FT_ROMAN));
3263                                         faml_flag = true;
3264                                 }
3265                                 break;
3266                         case SANS_FAMILY:
3267                                 if (default_family != "sfdefault") {
3268                                         tagsToOpen.push_back(html::FontTag(html::FT_SANS));
3269                                         faml_flag = true;
3270                                 }
3271                                 break;
3272                         case TYPEWRITER_FAMILY:
3273                                 if (default_family != "ttdefault") {
3274                                         tagsToOpen.push_back(html::FontTag(html::FT_TYPE));
3275                                         faml_flag = true;
3276                                 }
3277                                 break;
3278                         case INHERIT_FAMILY:
3279                                 break;
3280                         default:
3281                                 // the other tags are for internal use
3282                                 LATTEST(false);
3283                                 break;
3284                         }
3285                 }
3286
3287                 // Font size
3288                 curr_size = font.fontInfo().size();
3289                 FontSize old_size = font_old.size();
3290                 if (old_size != curr_size) {
3291                         if (size_flag) {
3292                                 switch (old_size) {
3293                                 case FONT_SIZE_TINY:
3294                                         tagsToClose.push_back(html::EndFontTag(html::FT_SIZE_TINY));
3295                                         break;
3296                                 case FONT_SIZE_SCRIPT:
3297                                         tagsToClose.push_back(html::EndFontTag(html::FT_SIZE_SCRIPT));
3298                                         break;
3299                                 case FONT_SIZE_FOOTNOTE:
3300                                         tagsToClose.push_back(html::EndFontTag(html::FT_SIZE_FOOTNOTE));
3301                                         break;
3302                                 case FONT_SIZE_SMALL:
3303                                         tagsToClose.push_back(html::EndFontTag(html::FT_SIZE_SMALL));
3304                                         break;
3305                                 case FONT_SIZE_LARGE:
3306                                         tagsToClose.push_back(html::EndFontTag(html::FT_SIZE_LARGE));
3307                                         break;
3308                                 case FONT_SIZE_LARGER:
3309                                         tagsToClose.push_back(html::EndFontTag(html::FT_SIZE_LARGER));
3310                                         break;
3311                                 case FONT_SIZE_LARGEST:
3312                                         tagsToClose.push_back(html::EndFontTag(html::FT_SIZE_LARGEST));
3313                                         break;
3314                                 case FONT_SIZE_HUGE:
3315                                         tagsToClose.push_back(html::EndFontTag(html::FT_SIZE_HUGE));
3316                                         break;
3317                                 case FONT_SIZE_HUGER:
3318                                         tagsToClose.push_back(html::EndFontTag(html::FT_SIZE_HUGER));
3319                                         break;
3320                                 case FONT_SIZE_INCREASE:
3321                                         tagsToClose.push_back(html::EndFontTag(html::FT_SIZE_INCREASE));
3322                                         break;
3323                                 case FONT_SIZE_DECREASE:
3324                                         tagsToClose.push_back(html::EndFontTag(html::FT_SIZE_DECREASE));
3325                                         break;
3326                                 case FONT_SIZE_INHERIT:
3327                                 case FONT_SIZE_NORMAL:
3328                                         break;
3329                                 default:
3330                                         // the other tags are for internal use
3331                                         LATTEST(false);
3332                                         break;
3333                                 }
3334                                 size_flag = false;
3335                         }
3336                         switch (curr_size) {
3337                         case FONT_SIZE_TINY:
3338                                 tagsToOpen.push_back(html::FontTag(html::FT_SIZE_TINY));
3339                                 size_flag = true;
3340                                 break;
3341                         case FONT_SIZE_SCRIPT:
3342                                 tagsToOpen.push_back(html::FontTag(html::FT_SIZE_SCRIPT));
3343                                 size_flag = true;
3344                                 break;
3345                         case FONT_SIZE_FOOTNOTE:
3346                                 tagsToOpen.push_back(html::FontTag(html::FT_SIZE_FOOTNOTE));
3347                                 size_flag = true;
3348                                 break;
3349                         case FONT_SIZE_SMALL:
3350                                 tagsToOpen.push_back(html::FontTag(html::FT_SIZE_SMALL));
3351                                 size_flag = true;
3352                                 break;
3353                         case FONT_SIZE_LARGE:
3354                                 tagsToOpen.push_back(html::FontTag(html::FT_SIZE_LARGE));
3355                                 size_flag = true;
3356                                 break;
3357                         case FONT_SIZE_LARGER:
3358                                 tagsToOpen.push_back(html::FontTag(html::FT_SIZE_LARGER));
3359                                 size_flag = true;
3360                                 break;
3361                         case FONT_SIZE_LARGEST:
3362                                 tagsToOpen.push_back(html::FontTag(html::FT_SIZE_LARGEST));
3363                                 size_flag = true;
3364                                 break;
3365                         case FONT_SIZE_HUGE:
3366                                 tagsToOpen.push_back(html::FontTag(html::FT_SIZE_HUGE));
3367                                 size_flag = true;
3368                                 break;
3369                         case FONT_SIZE_HUGER:
3370                                 tagsToOpen.push_back(html::FontTag(html::FT_SIZE_HUGER));
3371                                 size_flag = true;
3372                                 break;
3373                         case FONT_SIZE_INCREASE:
3374                                 tagsToOpen.push_back(html::FontTag(html::FT_SIZE_INCREASE));
3375                                 size_flag = true;
3376                                 break;
3377                         case FONT_SIZE_DECREASE:
3378                                 tagsToOpen.push_back(html::FontTag(html::FT_SIZE_DECREASE));
3379                                 size_flag = true;
3380                                 break;
3381                         case FONT_SIZE_NORMAL:
3382                         case FONT_SIZE_INHERIT:
3383                                 break;
3384                         default:
3385                                 // the other tags are for internal use
3386                                 LATTEST(false);
3387                                 break;
3388                         }
3389                 }
3390
3391                 // FIXME XHTML
3392                 // Other such tags? What about the other text ranges?
3393
3394                 vector<html::EndFontTag>::const_iterator cit = tagsToClose.begin();
3395                 vector<html::EndFontTag>::const_iterator cen = tagsToClose.end();
3396                 for (; cit != cen; ++cit)
3397                         xs << *cit;
3398
3399                 vector<html::FontTag>::const_iterator sit = tagsToOpen.begin();
3400                 vector<html::FontTag>::const_iterator sen = tagsToOpen.end();
3401                 for (; sit != sen; ++sit)
3402                         xs << *sit;
3403
3404                 tagsToClose.clear();
3405                 tagsToOpen.clear();
3406
3407                 Inset const * inset = getInset(i);
3408                 if (inset) {
3409                         if (!runparams.for_toc || inset->isInToc()) {
3410                                 OutputParams np = runparams;
3411                                 np.local_font = &font;
3412                                 // If the paragraph has size 1, then we are in the "special
3413                                 // case" where we do not output the containing paragraph info
3414                                 if (!inset->getLayout().htmlisblock() && size() != 1)
3415                                         np.html_in_par = true;
3416                                 retval += inset->xhtml(xs, np);
3417                         }
3418                 } else {
3419                         char_type c = getUChar(buf.masterBuffer()->params(),
3420                                                runparams, i);
3421                         xs << c;
3422                 }
3423                 font_old = font.fontInfo();
3424         }
3425
3426         // FIXME XHTML
3427         // I'm worried about what happens if a branch, say, is itself
3428         // wrapped in some font stuff. I think that will not work.
3429         xs.closeFontTags();
3430         if (close_paragraph)
3431                 xs.endDivision();
3432
3433         return retval;
3434 }
3435
3436
3437 bool Paragraph::isHfill(pos_type pos) const
3438 {
3439         Inset const * inset = getInset(pos);
3440         return inset && inset->isHfill();
3441 }
3442
3443
3444 bool Paragraph::isNewline(pos_type pos) const
3445 {
3446         // U+2028 LINE SEPARATOR
3447         // U+2029 PARAGRAPH SEPARATOR
3448         char_type const c = d->text_[pos];
3449         if (c == 0x2028 || c == 0x2029)
3450                 return true;
3451         Inset const * inset = getInset(pos);
3452         return inset && inset->lyxCode() == NEWLINE_CODE;
3453 }
3454
3455
3456 bool Paragraph::isEnvSeparator(pos_type pos) const
3457 {
3458         Inset const * inset = getInset(pos);
3459         return inset && inset->lyxCode() == SEPARATOR_CODE;
3460 }
3461
3462
3463 bool Paragraph::isLineSeparator(pos_type pos) const
3464 {
3465         char_type const c = d->text_[pos];
3466         if (isLineSeparatorChar(c))
3467                 return true;
3468         Inset const * inset = getInset(pos);
3469         return inset && inset->isLineSeparator();
3470 }
3471
3472
3473 bool Paragraph::isWordSeparator(pos_type pos) const
3474 {
3475         if (pos == size())
3476                 return true;
3477         if (Inset const * inset = getInset(pos))
3478                 return !inset->isLetter();
3479         // if we have a hard hyphen (no en- or emdash) or apostrophe
3480         // we pass this to the spell checker
3481         // FIXME: this method is subject to change, visit
3482         // https://bugzilla.mozilla.org/show_bug.cgi?id=355178
3483         // to get an impression how complex this is.
3484         if (isHardHyphenOrApostrophe(pos))
3485                 return false;
3486         char_type const c = d->text_[pos];
3487         // We want to pass the escape chars to the spellchecker
3488         docstring const escape_chars = from_utf8(lyxrc.spellchecker_esc_chars);
3489         return !isLetterChar(c) && !isDigitASCII(c) && !contains(escape_chars, c);
3490 }
3491
3492
3493 bool Paragraph::isHardHyphenOrApostrophe(pos_type pos) const
3494 {
3495         pos_type const psize = size();
3496         if (pos >= psize)
3497                 return false;
3498         char_type const c = d->text_[pos];
3499         if (c != '-' && c != '\'')
3500                 return false;
3501         int nextpos = pos + 1;
3502         int prevpos = pos > 0 ? pos - 1 : 0;
3503         if ((nextpos == psize || isSpace(nextpos))
3504                 && (pos == 0 || isSpace(prevpos)))
3505                 return false;
3506         return true;
3507 }
3508
3509
3510 bool Paragraph::needsCProtection(bool const fragile) const
3511 {
3512         // first check the layout of the paragraph, but only in insets
3513         InsetText const * textinset = inInset().asInsetText();
3514         bool const maintext = textinset
3515                 ? textinset->text().isMainText()
3516                 : false;
3517
3518         if (!maintext && layout().needcprotect) {
3519                 // Environments need cprotection regardless the content
3520                 if (layout().latextype == LATEX_ENVIRONMENT)
3521                         return true;
3522
3523                 // Commands need cprotection if they contain specific chars
3524                 int const nchars_escape = 9;
3525                 static char_type const chars_escape[nchars_escape] = {
3526                         '&', '_', '$', '%', '#', '^', '{', '}', '\\'};
3527
3528                 docstring const pars = asString();
3529                 for (int k = 0; k < nchars_escape; k++) {
3530                         if (contains(pars, chars_escape[k]))
3531                                 return true;
3532                 }
3533         }
3534
3535         // now check whether we have insets that need cprotection
3536         pos_type size = d->text_.size();
3537         for (pos_type i = 0; i < size; ++i)
3538                 if (isInset(i) && getInset(i)->needsCProtection(maintext, fragile))
3539                         return true;
3540
3541         return false;
3542 }
3543
3544
3545 FontSpan const & Paragraph::getSpellRange(pos_type pos) const
3546 {
3547         return d->speller_state_.getRange(pos);
3548 }
3549
3550
3551 bool Paragraph::isChar(pos_type pos) const
3552 {
3553         if (Inset const * inset = getInset(pos))
3554                 return inset->isChar();
3555         char_type const c = d->text_[pos];
3556         return !isLetterChar(c) && !isDigitASCII(c) && !lyx::isSpace(c);
3557 }
3558
3559
3560 bool Paragraph::isSpace(pos_type pos) const
3561 {
3562         if (Inset const * inset = getInset(pos))
3563                 return inset->isSpace();
3564         char_type const c = d->text_[pos];
3565         return lyx::isSpace(c);
3566 }
3567
3568
3569 Language const *
3570 Paragraph::getParLanguage(BufferParams const & bparams) const
3571 {
3572         if (!empty())
3573                 return getFirstFontSettings(bparams).language();
3574         // FIXME: we should check the prev par as well (Lgb)
3575         return bparams.language;
3576 }
3577
3578
3579 bool Paragraph::isRTL(BufferParams const & bparams) const
3580 {
3581         return getParLanguage(bparams)->rightToLeft()
3582                 && !inInset().getLayout().forceLTR();
3583 }
3584
3585
3586 void Paragraph::changeLanguage(BufferParams const & bparams,
3587                                Language const * from, Language const * to)
3588 {
3589         // change language including dummy font change at the end
3590         for (pos_type i = 0; i <= size(); ++i) {
3591                 Font font = getFontSettings(bparams, i);
3592                 if (font.language() == from) {
3593                         font.setLanguage(to);
3594                         setFont(i, font);
3595                         d->requestSpellCheck(i);
3596                 }
3597         }
3598 }
3599
3600
3601 bool Paragraph::isMultiLingual(BufferParams const & bparams) const
3602 {
3603         Language const * doc_language = bparams.language;
3604         for (auto const & f : d->fontlist_)
3605                 if (f.font().language() != ignore_language &&
3606                     f.font().language() != latex_language &&
3607                     f.font().language() != doc_language)
3608                         return true;
3609         return false;
3610 }
3611
3612
3613 void Paragraph::getLanguages(std::set<Language const *> & langs) const
3614 {
3615         for (auto const & f : d->fontlist_) {
3616                 Language const * lang = f.font().language();
3617                 if (lang != ignore_language &&
3618                     lang != latex_language)
3619                         langs.insert(lang);
3620         }
3621 }
3622
3623
3624 docstring Paragraph::asString(int options) const
3625 {
3626         return asString(0, size(), options);
3627 }
3628
3629
3630 docstring Paragraph::asString(pos_type beg, pos_type end, int options, const OutputParams *runparams) const
3631 {
3632         odocstringstream os;
3633
3634         if (beg == 0
3635             && options & AS_STR_LABEL
3636             && !d->params_.labelString().empty())
3637                 os << d->params_.labelString() << ' ';
3638
3639         for (pos_type i = beg; i < end; ++i) {
3640                 if ((options & AS_STR_SKIPDELETE) && isDeleted(i))
3641                         continue;
3642                 char_type const c = d->text_[i];
3643                 if (isPrintable(c) || c == '\t'
3644                     || (c == '\n' && (options & AS_STR_NEWLINES)))
3645                         os.put(c);
3646                 else if (c == META_INSET && (options & AS_STR_INSETS)) {
3647                         if (c == META_INSET && (options & AS_STR_PLAINTEXT)) {
3648                                 LASSERT(runparams != 0, return docstring());
3649                                 getInset(i)->plaintext(os, *runparams);
3650                         } else {
3651                                 getInset(i)->toString(os);
3652                         }
3653                 }
3654         }
3655
3656         return os.str();
3657 }
3658
3659
3660 void Paragraph::forOutliner(docstring & os, size_t const maxlen,
3661                             bool const shorten, bool const label) const
3662 {
3663         size_t tmplen = shorten ? maxlen + 1 : maxlen;
3664         if (label && !labelString().empty())
3665                 os += labelString() + ' ';
3666         if (!layout().isTocCaption())
3667                 return;
3668         for (pos_type i = 0; i < size() && os.length() < tmplen; ++i) {
3669                 if (isDeleted(i))
3670                         continue;
3671                 char_type const c = d->text_[i];
3672                 if (isPrintable(c))
3673                         os += c;
3674                 else if (c == META_INSET)
3675                         getInset(i)->forOutliner(os, tmplen, false);
3676         }
3677         if (shorten)
3678                 Text::shortenForOutliner(os, maxlen);
3679 }
3680
3681
3682 void Paragraph::setInsetOwner(Inset const * inset)
3683 {
3684         d->inset_owner_ = inset;
3685 }
3686
3687
3688 int Paragraph::id() const
3689 {
3690         return d->id_;
3691 }
3692
3693
3694 void Paragraph::setId(int id)
3695 {
3696         d->id_ = id;
3697 }
3698
3699
3700 Layout const & Paragraph::layout() const
3701 {
3702         return *d->layout_;
3703 }
3704
3705
3706 void Paragraph::setLayout(Layout const & layout)
3707 {
3708         d->layout_ = &layout;
3709 }
3710
3711
3712 void Paragraph::setDefaultLayout(DocumentClass const & tc)
3713 {
3714         setLayout(tc.defaultLayout());
3715 }
3716
3717
3718 void Paragraph::setPlainLayout(DocumentClass const & tc)
3719 {
3720         setLayout(tc.plainLayout());
3721 }
3722
3723
3724 void Paragraph::setPlainOrDefaultLayout(DocumentClass const & tclass)
3725 {
3726         if (usePlainLayout())
3727                 setPlainLayout(tclass);
3728         else
3729                 setDefaultLayout(tclass);
3730 }
3731
3732
3733 Inset const & Paragraph::inInset() const
3734 {
3735         LBUFERR(d->inset_owner_);
3736         return *d->inset_owner_;
3737 }
3738
3739
3740 ParagraphParameters & Paragraph::params()
3741 {
3742         return d->params_;
3743 }
3744
3745
3746 ParagraphParameters const & Paragraph::params() const
3747 {
3748         return d->params_;
3749 }
3750
3751
3752 bool Paragraph::isFreeSpacing() const
3753 {
3754         if (d->layout_->free_spacing)
3755                 return true;
3756         return d->inset_owner_ && d->inset_owner_->isFreeSpacing();
3757 }
3758
3759
3760 bool Paragraph::allowEmpty() const
3761 {
3762         if (d->layout_->keepempty)
3763                 return true;
3764         return d->inset_owner_ && d->inset_owner_->allowEmpty();
3765 }
3766
3767
3768 bool Paragraph::brokenBiblio() const
3769 {
3770         // there is a problem if there is no bibitem at position 0 or
3771         // if there is another bibitem in the paragraph.
3772         return d->layout_->labeltype == LABEL_BIBLIO
3773                 && (d->insetlist_.find(BIBITEM_CODE) != 0
3774                     || d->insetlist_.find(BIBITEM_CODE, 1) > 0);
3775 }
3776
3777
3778 int Paragraph::fixBiblio(Buffer const & buffer)
3779 {
3780         // FIXME: What about the case where paragraph is not BIBLIO
3781         // but there is an InsetBibitem?
3782         // FIXME: when there was already an inset at 0, the return value is 1,
3783         // which does not tell whether another inset has been remove; the
3784         // cursor cannot be correctly updated.
3785
3786         if (d->layout_->labeltype != LABEL_BIBLIO)
3787                 return 0;
3788
3789         bool const track_changes = buffer.params().track_changes;
3790         int bibitem_pos = d->insetlist_.find(BIBITEM_CODE);
3791         bool const hasbibitem0 = bibitem_pos == 0;
3792
3793         if (hasbibitem0) {
3794                 bibitem_pos = d->insetlist_.find(BIBITEM_CODE, 1);
3795                 // There was an InsetBibitem at pos 0, and no other one => OK
3796                 if (bibitem_pos == -1)
3797                         return 0;
3798                 // there is a bibitem at the 0 position, but since
3799                 // there is a second one, we copy the second on the
3800                 // first. We're assuming there are at most two of
3801                 // these, which there should be.
3802                 // FIXME: why does it make sense to do that rather
3803                 // than keep the first? (JMarc)
3804                 Inset * inset = releaseInset(bibitem_pos);
3805                 d->insetlist_.begin()->inset = inset;
3806                 return -bibitem_pos;
3807         }
3808
3809         // We need to create an inset at the beginning
3810         Inset * inset = 0;
3811         if (bibitem_pos > 0) {
3812                 // there was one somewhere in the paragraph, let's move it
3813                 inset = d->insetlist_.release(bibitem_pos);
3814                 eraseChar(bibitem_pos, track_changes);
3815         } else
3816                 // make a fresh one
3817                 inset = new InsetBibitem(const_cast<Buffer *>(&buffer),
3818                                          InsetCommandParams(BIBITEM_CODE));
3819
3820         Font font(inherit_font, buffer.params().language);
3821         insertInset(0, inset, font, Change(track_changes ? Change::INSERTED
3822                                                    : Change::UNCHANGED));
3823
3824         return 1;
3825 }
3826
3827
3828 void Paragraph::checkAuthors(AuthorList const & authorList)
3829 {
3830         d->changes_.checkAuthors(authorList);
3831 }
3832
3833
3834 bool Paragraph::isChanged(pos_type pos) const
3835 {
3836         return lookupChange(pos).changed();
3837 }
3838
3839
3840 bool Paragraph::isInserted(pos_type pos) const
3841 {
3842         return lookupChange(pos).inserted();
3843 }
3844
3845
3846 bool Paragraph::isDeleted(pos_type pos) const
3847 {
3848         return lookupChange(pos).deleted();
3849 }
3850
3851
3852 InsetList const & Paragraph::insetList() const
3853 {
3854         return d->insetlist_;
3855 }
3856
3857
3858 void Paragraph::setInsetBuffers(Buffer & b)
3859 {
3860         d->insetlist_.setBuffer(b);
3861 }
3862
3863
3864 void Paragraph::resetBuffer()
3865 {
3866         d->insetlist_.resetBuffer();
3867 }
3868
3869
3870 Inset * Paragraph::releaseInset(pos_type pos)
3871 {
3872         Inset * inset = d->insetlist_.release(pos);
3873         /// does not honour change tracking!
3874         eraseChar(pos, false);
3875         return inset;
3876 }
3877
3878
3879 Inset * Paragraph::getInset(pos_type pos)
3880 {
3881         return (pos < pos_type(d->text_.size()) && d->text_[pos] == META_INSET)
3882                  ? d->insetlist_.get(pos) : 0;
3883 }
3884
3885
3886 Inset const * Paragraph::getInset(pos_type pos) const
3887 {
3888         return (pos < pos_type(d->text_.size()) && d->text_[pos] == META_INSET)
3889                  ? d->insetlist_.get(pos) : 0;
3890 }
3891
3892
3893 void Paragraph::changeCase(BufferParams const & bparams, pos_type pos,
3894                 pos_type & right, TextCase action)
3895 {
3896         // process sequences of modified characters; in change
3897         // tracking mode, this approach results in much better
3898         // usability than changing case on a char-by-char basis
3899         // We also need to track the current font, since font
3900         // changes within sequences can occur.
3901         vector<pair<char_type, Font> > changes;
3902
3903         bool const trackChanges = bparams.track_changes;
3904
3905         bool capitalize = true;
3906
3907         for (; pos < right; ++pos) {
3908                 char_type oldChar = d->text_[pos];
3909                 char_type newChar = oldChar;
3910
3911                 // ignore insets and don't play with deleted text!
3912                 if (oldChar != META_INSET && !isDeleted(pos)) {
3913                         switch (action) {
3914                                 case text_lowercase:
3915                                         newChar = lowercase(oldChar);
3916                                         break;
3917                                 case text_capitalization:
3918                                         if (capitalize) {
3919                                                 newChar = uppercase(oldChar);
3920                                                 capitalize = false;
3921                                         }
3922                                         break;
3923                                 case text_uppercase:
3924                                         newChar = uppercase(oldChar);
3925                                         break;
3926                         }
3927                 }
3928
3929                 if (isWordSeparator(pos) || isDeleted(pos)) {
3930                         // permit capitalization again
3931                         capitalize = true;
3932                 }
3933
3934                 if (oldChar != newChar) {
3935                         changes.push_back(make_pair(newChar, getFontSettings(bparams, pos)));
3936                         if (pos != right - 1)
3937                                 continue;
3938                         // step behind the changing area
3939                         pos++;
3940                 }
3941
3942                 int erasePos = pos - changes.size();
3943                 for (size_t i = 0; i < changes.size(); i++) {
3944                         insertChar(pos, changes[i].first,
3945                                    changes[i].second,
3946                                    trackChanges);
3947                         if (!eraseChar(erasePos, trackChanges)) {
3948                                 ++erasePos;
3949                                 ++pos; // advance
3950                                 ++right; // expand selection
3951                         }
3952                 }
3953                 changes.clear();
3954         }
3955 }
3956
3957
3958 int Paragraph::find(docstring const & str, bool cs, bool mw,
3959                 pos_type start_pos, bool del) const
3960 {
3961         pos_type pos = start_pos;
3962         int const strsize = str.length();
3963         int i = 0;
3964         pos_type const parsize = d->text_.size();
3965         for (i = 0; i < strsize && pos < parsize; ++i, ++pos) {
3966                 // Ignore "invisible" letters such as ligature breaks
3967                 // and hyphenation chars while searching
3968                 while (pos < parsize - 1 && isInset(pos)) {
3969                         odocstringstream os;
3970                         getInset(pos)->toString(os);
3971                         if (!getInset(pos)->isLetter() || !os.str().empty())
3972                                 break;
3973                         pos++;
3974                 }
3975                 if (cs && str[i] != d->text_[pos])
3976                         break;
3977                 if (!cs && uppercase(str[i]) != uppercase(d->text_[pos]))
3978                         break;
3979                 if (!del && isDeleted(pos))
3980                         break;
3981         }
3982
3983         if (i != strsize)
3984                 return 0;
3985
3986         // if necessary, check whether string matches word
3987         if (mw) {
3988                 if (start_pos > 0 && !isWordSeparator(start_pos - 1))
3989                         return 0;
3990                 if (pos < parsize
3991                         && !isWordSeparator(pos))
3992                         return 0;
3993         }
3994
3995         return pos - start_pos;
3996 }
3997
3998
3999 char_type Paragraph::getChar(pos_type pos) const
4000 {
4001         return d->text_[pos];
4002 }
4003
4004
4005 pos_type Paragraph::size() const
4006 {
4007         return d->text_.size();
4008 }
4009
4010
4011 bool Paragraph::empty() const
4012 {
4013         return d->text_.empty();
4014 }
4015
4016
4017 bool Paragraph::isInset(pos_type pos) const
4018 {
4019         return d->text_[pos] == META_INSET;
4020 }
4021
4022
4023 bool Paragraph::isSeparator(pos_type pos) const
4024 {
4025         //FIXME: Are we sure this can be the only separator?
4026         return d->text_[pos] == ' ';
4027 }
4028
4029
4030 void Paragraph::deregisterWords()
4031 {
4032         Private::LangWordsMap::const_iterator itl = d->words_.begin();
4033         Private::LangWordsMap::const_iterator ite = d->words_.end();
4034         for (; itl != ite; ++itl) {
4035                 WordList & wl = theWordList(itl->first);
4036                 Private::Words::const_iterator it = (itl->second).begin();
4037                 Private::Words::const_iterator et = (itl->second).end();
4038                 for (; it != et; ++it)
4039                         wl.remove(*it);
4040         }
4041         d->words_.clear();
4042 }
4043
4044
4045 void Paragraph::locateWord(pos_type & from, pos_type & to,
4046         word_location const loc) const
4047 {
4048         switch (loc) {
4049         case WHOLE_WORD_STRICT:
4050                 if (from == 0 || from == size()
4051                     || isWordSeparator(from)
4052                     || isWordSeparator(from - 1)) {
4053                         to = from;
4054                         return;
4055                 }
4056                 // fall through
4057
4058         case WHOLE_WORD:
4059                 // If we are already at the beginning of a word, do nothing
4060                 if (!from || isWordSeparator(from - 1))
4061                         break;
4062                 // fall through
4063
4064         case PREVIOUS_WORD:
4065                 // always move the cursor to the beginning of previous word
4066                 while (from && !isWordSeparator(from - 1))
4067                         --from;
4068                 break;
4069         case NEXT_WORD:
4070                 LYXERR0("Paragraph::locateWord: NEXT_WORD not implemented yet");
4071                 break;
4072         case PARTIAL_WORD:
4073                 // no need to move the 'from' cursor
4074                 break;
4075         }
4076         to = from;
4077         while (to < size() && !isWordSeparator(to))
4078                 ++to;
4079 }
4080
4081
4082 void Paragraph::collectWords()
4083 {
4084         for (pos_type pos = 0; pos < size(); ++pos) {
4085                 if (isWordSeparator(pos))
4086                         continue;
4087                 pos_type from = pos;
4088                 locateWord(from, pos, WHOLE_WORD);
4089                 // Work around MSVC warning: The statement
4090                 // if (pos < from + lyxrc.completion_minlength)
4091                 // triggers a signed vs. unsigned warning.
4092                 // I don't know why this happens, it could be a MSVC bug, or
4093                 // related to LLP64 (windows) vs. LP64 (unix) programming
4094                 // model, or the C++ standard might be ambigous in the section
4095                 // defining the "usual arithmetic conversions". However, using
4096                 // a temporary variable is safe and works on all compilers.
4097                 pos_type const endpos = from + lyxrc.completion_minlength;
4098                 if (pos < endpos)
4099                         continue;
4100                 FontList::const_iterator cit = d->fontlist_.fontIterator(from);
4101                 if (cit == d->fontlist_.end())
4102                         return;
4103                 Language const * lang = cit->font().language();
4104                 docstring const word = asString(from, pos, AS_STR_NONE);
4105                 d->words_[lang->lang()].insert(word);
4106         }
4107 }
4108
4109
4110 void Paragraph::registerWords()
4111 {
4112         Private::LangWordsMap::const_iterator itl = d->words_.begin();
4113         Private::LangWordsMap::const_iterator ite = d->words_.end();
4114         for (; itl != ite; ++itl) {
4115                 WordList & wl = theWordList(itl->first);
4116                 Private::Words::const_iterator it = (itl->second).begin();
4117                 Private::Words::const_iterator et = (itl->second).end();
4118                 for (; it != et; ++it)
4119                         wl.insert(*it);
4120         }
4121 }
4122
4123
4124 void Paragraph::updateWords()
4125 {
4126         deregisterWords();
4127         collectWords();
4128         registerWords();
4129 }
4130
4131
4132 void Paragraph::Private::appendSkipPosition(SkipPositions & skips, pos_type const pos) const
4133 {
4134         SkipPositionsIterator begin = skips.begin();
4135         SkipPositions::iterator end = skips.end();
4136         if (pos > 0 && begin < end) {
4137                 --end;
4138                 if (end->last == pos - 1) {
4139                         end->last = pos;
4140                         return;
4141                 }
4142         }
4143         skips.insert(end, FontSpan(pos, pos));
4144 }
4145
4146
4147 Language * Paragraph::Private::locateSpellRange(
4148         pos_type & from, pos_type & to,
4149         SkipPositions & skips) const
4150 {
4151         // skip leading white space
4152         while (from < to && owner_->isWordSeparator(from))
4153                 ++from;
4154         // don't check empty range
4155         if (from >= to)
4156                 return 0;
4157         // get current language
4158         Language * lang = getSpellLanguage(from);
4159         pos_type last = from;
4160         bool samelang = true;
4161         bool sameinset = true;
4162         while (last < to && samelang && sameinset) {
4163                 // hop to end of word
4164                 while (last < to && !owner_->isWordSeparator(last)) {
4165                         if (owner_->getInset(last)) {
4166                                 appendSkipPosition(skips, last);
4167                         } else if (owner_->isDeleted(last)) {
4168                                 appendSkipPosition(skips, last);
4169                         }
4170                         ++last;
4171                 }
4172                 // hop to next word while checking for insets
4173                 while (sameinset && last < to && owner_->isWordSeparator(last)) {
4174                         if (Inset const * inset = owner_->getInset(last))
4175                                 sameinset = inset->isChar() && inset->isLetter();
4176                         if (sameinset && owner_->isDeleted(last)) {
4177                                 appendSkipPosition(skips, last);
4178                         }
4179                         if (sameinset)
4180                                 last++;
4181                 }
4182                 if (sameinset && last < to) {
4183                         // now check for language change
4184                         samelang = lang == getSpellLanguage(last);
4185                 }
4186         }
4187         // if language change detected backstep is needed
4188         if (!samelang)
4189                 --last;
4190         to = last;
4191         return lang;
4192 }
4193
4194
4195 Language * Paragraph::Private::getSpellLanguage(pos_type const from) const
4196 {
4197         Language * lang =
4198                 const_cast<Language *>(owner_->getFontSettings(
4199                         inset_owner_->buffer().params(), from).language());
4200         if (lang == inset_owner_->buffer().params().language
4201                 && !lyxrc.spellchecker_alt_lang.empty()) {
4202                 string lang_code;
4203                 string const lang_variety =
4204                         split(lyxrc.spellchecker_alt_lang, lang_code, '-');
4205                 lang->setCode(lang_code);
4206                 lang->setVariety(lang_variety);
4207         }
4208         return lang;
4209 }
4210
4211
4212 void Paragraph::requestSpellCheck(pos_type pos)
4213 {
4214         d->requestSpellCheck(pos);
4215 }
4216
4217
4218 bool Paragraph::needsSpellCheck() const
4219 {
4220         SpellChecker::ChangeNumber speller_change_number = 0;
4221         if (theSpellChecker())
4222                 speller_change_number = theSpellChecker()->changeNumber();
4223         if (speller_change_number > d->speller_state_.currentChangeNumber()) {
4224                 d->speller_state_.needsCompleteRefresh(speller_change_number);
4225         }
4226         return d->needsSpellCheck();
4227 }
4228
4229
4230 bool Paragraph::Private::ignoreWord(docstring const & word) const
4231 {
4232         // Ignore words with digits
4233         // FIXME: make this customizable
4234         // (note that some checkers ignore words with digits by default)
4235         docstring::const_iterator cit = word.begin();
4236         docstring::const_iterator const end = word.end();
4237         for (; cit != end; ++cit) {
4238                 if (isNumber((*cit)))
4239                         return true;
4240         }
4241         return false;
4242 }
4243
4244
4245 SpellChecker::Result Paragraph::spellCheck(pos_type & from, pos_type & to,
4246         WordLangTuple & wl, docstring_list & suggestions,
4247         bool do_suggestion, bool check_learned) const
4248 {
4249         SpellChecker::Result result = SpellChecker::WORD_OK;
4250         SpellChecker * speller = theSpellChecker();
4251         if (!speller)
4252                 return result;
4253
4254         if (!d->layout_->spellcheck || !inInset().allowSpellCheck())
4255                 return result;
4256
4257         locateWord(from, to, WHOLE_WORD);
4258         if (from == to || from >= size())
4259                 return result;
4260
4261         docstring word = asString(from, to, AS_STR_INSETS | AS_STR_SKIPDELETE);
4262         Language * lang = d->getSpellLanguage(from);
4263
4264         if (getFontSettings(d->inset_owner_->buffer().params(), from).fontInfo().nospellcheck() == FONT_ON)
4265                 return result;
4266
4267         wl = WordLangTuple(word, lang);
4268
4269         if (word.empty())
4270                 return result;
4271
4272         if (needsSpellCheck() || check_learned) {
4273                 pos_type end = to;
4274                 if (!d->ignoreWord(word)) {
4275                         bool const trailing_dot = to < size() && d->text_[to] == '.';
4276                         result = speller->check(wl);
4277                         if (SpellChecker::misspelled(result) && trailing_dot) {
4278                                 wl = WordLangTuple(word.append(from_ascii(".")), lang);
4279                                 result = speller->check(wl);
4280                                 if (!SpellChecker::misspelled(result)) {
4281                                         LYXERR(Debug::GUI, "misspelled word is correct with dot: \"" <<
4282                                            word << "\" [" <<
4283                                            from << ".." << to << "]");
4284                                 } else {
4285                                         // spell check with dot appended failed too
4286                                         // restore original word/lang value
4287                                         word = asString(from, to, AS_STR_INSETS | AS_STR_SKIPDELETE);
4288                                         wl = WordLangTuple(word, lang);
4289                                 }
4290                         }
4291                 }
4292                 if (!SpellChecker::misspelled(result)) {
4293                         // area up to the begin of the next word is not misspelled
4294                         while (end < size() && isWordSeparator(end))
4295                                 ++end;
4296                 }
4297                 d->setMisspelled(from, end, result);
4298         } else {
4299                 result = d->speller_state_.getState(from);
4300         }
4301
4302         if (do_suggestion)
4303                 suggestions.clear();
4304
4305         if (SpellChecker::misspelled(result)) {
4306                 LYXERR(Debug::GUI, "misspelled word: \"" <<
4307                            word << "\" [" <<
4308                            from << ".." << to << "]");
4309                 if (do_suggestion)
4310                         speller->suggest(wl, suggestions);
4311         }
4312         return result;
4313 }
4314
4315
4316 void Paragraph::anonymize()
4317 {
4318         // This is a very crude anonymization for now
4319         for (char_type & c : d->text_)
4320                 if (isLetterChar(c) || isNumber(c))
4321                         c = 'a';
4322 }
4323
4324
4325 void Paragraph::Private::markMisspelledWords(
4326         pos_type const & first, pos_type const & last,
4327         SpellChecker::Result result,
4328         docstring const & word,
4329         SkipPositions const & skips)
4330 {
4331         if (!SpellChecker::misspelled(result)) {
4332                 setMisspelled(first, last, SpellChecker::WORD_OK);
4333                 return;
4334         }
4335         int snext = first;
4336         SpellChecker * speller = theSpellChecker();
4337         // locate and enumerate the error positions
4338         int nerrors = speller->numMisspelledWords();
4339         int numskipped = 0;
4340         SkipPositionsIterator it = skips.begin();
4341         SkipPositionsIterator et = skips.end();
4342         for (int index = 0; index < nerrors; ++index) {
4343                 int wstart;
4344                 int wlen = 0;
4345                 speller->misspelledWord(index, wstart, wlen);
4346                 /// should not happen if speller supports range checks
4347                 if (!wlen) continue;
4348                 docstring const misspelled = word.substr(wstart, wlen);
4349                 wstart += first + numskipped;
4350                 if (snext < wstart) {
4351                         /// mark the range of correct spelling
4352                         numskipped += countSkips(it, et, wstart);
4353                         setMisspelled(snext,
4354                                 wstart - 1, SpellChecker::WORD_OK);
4355                 }
4356                 snext = wstart + wlen;
4357                 numskipped += countSkips(it, et, snext);
4358                 /// mark the range of misspelling
4359                 setMisspelled(wstart, snext, result);
4360                 LYXERR(Debug::GUI, "misspelled word: \"" <<
4361                            misspelled << "\" [" <<
4362                            wstart << ".." << (snext-1) << "]");
4363                 ++snext;
4364         }
4365         if (snext <= last) {
4366                 /// mark the range of correct spelling at end
4367                 setMisspelled(snext, last, SpellChecker::WORD_OK);
4368         }
4369 }
4370
4371
4372 void Paragraph::spellCheck() const
4373 {
4374         SpellChecker * speller = theSpellChecker();
4375         if (!speller || empty() ||!needsSpellCheck())
4376                 return;
4377         pos_type start;
4378         pos_type endpos;
4379         d->rangeOfSpellCheck(start, endpos);
4380         if (speller->canCheckParagraph()) {
4381                 // loop until we leave the range
4382                 for (pos_type first = start; first < endpos; ) {
4383                         pos_type last = endpos;
4384                         Private::SkipPositions skips;
4385                         Language * lang = d->locateSpellRange(first, last, skips);
4386                         if (first >= endpos)
4387                                 break;
4388                         // start the spell checker on the unit of meaning
4389                         docstring word = asString(first, last, AS_STR_INSETS + AS_STR_SKIPDELETE);
4390                         WordLangTuple wl = WordLangTuple(word, lang);
4391                         SpellChecker::Result result = word.size() ?
4392                                 speller->check(wl) : SpellChecker::WORD_OK;
4393                         d->markMisspelledWords(first, last, result, word, skips);
4394                         first = ++last;
4395                 }
4396         } else {
4397                 static docstring_list suggestions;
4398                 pos_type to = endpos;
4399                 while (start < endpos) {
4400                         WordLangTuple wl;
4401                         spellCheck(start, to, wl, suggestions, false);
4402                         start = to + 1;
4403                 }
4404         }
4405         d->readySpellCheck();
4406 }
4407
4408
4409 bool Paragraph::isMisspelled(pos_type pos, bool check_boundary) const
4410 {
4411         bool result = SpellChecker::misspelled(d->speller_state_.getState(pos));
4412         if (result || pos <= 0 || pos > size())
4413                 return result;
4414         if (check_boundary && (pos == size() || isWordSeparator(pos)))
4415                 result = SpellChecker::misspelled(d->speller_state_.getState(pos - 1));
4416         return result;
4417 }
4418
4419
4420 string Paragraph::magicLabel() const
4421 {
4422         stringstream ss;
4423         ss << "magicparlabel-" << id();
4424         return ss.str();
4425 }
4426
4427
4428 } // namespace lyx