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                         os << "\\LRE{";
1152                 } else if (running_font.language()->lang() == "farsi"
1153                            || running_font.language()->lang() == "arabic_arabi")
1154                         os << "\\textLR{" << termcmd;
1155                 else
1156                         os << "\\L{";
1157                 close = true;
1158         }
1159
1160         // FIXME: Bug: we can have an empty font change here!
1161         // if there has just been a font change, we are going to close it
1162         // right now, which means stupid latex code like \textsf{}. AFAIK,
1163         // this does not harm dvi output. A minor bug, thus (JMarc)
1164
1165         // Some insets cannot be inside a font change command.
1166         // However, even such insets *can* be placed in \L or \R
1167         // or their equivalents (for RTL language switches), so we don't
1168         // close the language in those cases.
1169         // ArabTeX, though, cannot handle this special behavior, it seems.
1170         bool arabtex = basefont.language()->lang() == "arabic_arabtex"
1171                 || running_font.language()->lang() == "arabic_arabtex";
1172         if (open_font && !inset->inheritFont()) {
1173                 bool needPar = false;
1174                 bool closeLanguage = arabtex
1175                         || basefont.isRightToLeft() == running_font.isRightToLeft();
1176                 unsigned int count = running_font.latexWriteEndChanges(os,
1177                                         bparams, runparams, basefont, basefont,
1178                                         needPar, closeLanguage);
1179                 column += count;
1180                 // if any font properties were closed, update the running_font,
1181                 // making sure, however, to leave the language as it was
1182                 if (count > 0) {
1183                         // FIXME: probably a better way to keep track of the old
1184                         // language, than copying the entire font?
1185                         Font const copy_font(running_font);
1186                         basefont = owner_->getLayoutFont(bparams, outerfont);
1187                         running_font = basefont;
1188                         if (!closeLanguage)
1189                                 running_font.setLanguage(copy_font.language());
1190                         // leave font open if language is still open
1191                         open_font = (running_font.language() == basefont.language());
1192                         if (closeLanguage)
1193                                 runparams.local_font = &basefont;
1194                 }
1195         }
1196
1197         size_t const previous_row_count = os.texrow().rows();
1198
1199         try {
1200                 runparams.lastid = id_;
1201                 runparams.lastpos = i;
1202                 inset->latex(os, runparams);
1203         } catch (EncodingException & e) {
1204                 // add location information and throw again.
1205                 e.par_id = id_;
1206                 e.pos = i;
1207                 throw(e);
1208         }
1209
1210         if (close)
1211                 os << '}';
1212
1213         if (os.texrow().rows() > previous_row_count) {
1214                 os.texrow().start(owner_->id(), i + 1);
1215                 column = 0;
1216         } else {
1217                 column += (unsigned int)(os.os().tellp() - len);
1218         }
1219
1220         if (owner_->isDeleted(i))
1221                 --runparams.inDeletedInset;
1222 }
1223
1224
1225 void Paragraph::Private::latexSpecialChar(otexstream & os,
1226                                           BufferParams const & bparams,
1227                                           OutputParams const & runparams,
1228                                           Font const & running_font,
1229                                           Change const & running_change,
1230                                           Layout const & style,
1231                                           pos_type & i,
1232                                           pos_type end_pos,
1233                                           unsigned int & column)
1234 {
1235         char_type const c = owner_->getUChar(bparams, runparams, i);
1236
1237         if (style.pass_thru || runparams.pass_thru
1238             || contains(style.pass_thru_chars, c)
1239             || contains(runparams.pass_thru_chars, c)) {
1240                 if (c != '\0') {
1241                         Encoding const * const enc = runparams.encoding;
1242                         if (enc && !enc->encodable(c))
1243                                 throw EncodingException(c);
1244                         os.put(c);
1245                 }
1246                 return;
1247         }
1248
1249         // TIPA uses its own T3 encoding
1250         if (runparams.inIPA && latexSpecialT3(c, os, i, column))
1251                 return;
1252         // If T1 font encoding is used, use the special
1253         // characters it provides.
1254         // NOTE: Some languages reset the font encoding internally to a
1255         //       non-standard font encoding. If we are using such a language,
1256         //       we do not output special T1 chars.
1257         if (!runparams.inIPA && !running_font.language()->internalFontEncoding()
1258             && !runparams.isFullUnicode() && bparams.main_font_encoding() == "T1"
1259             && latexSpecialT1(c, os, i, column))
1260                 return;
1261         // NOTE: XeTeX and LuaTeX use EU1/2 (pre 2017) or TU (as of 2017) encoding
1262         else if (!runparams.inIPA && !running_font.language()->internalFontEncoding()
1263                  && runparams.isFullUnicode() && latexSpecialTU(c, os, i, column))
1264                      return;
1265
1266         // Otherwise, we use what LaTeX provides us.
1267         switch (c) {
1268         case '\\':
1269                 os << "\\textbackslash" << termcmd;
1270                 column += 15;
1271                 break;
1272         case '<':
1273                 os << "\\textless" << termcmd;
1274                 column += 10;
1275                 break;
1276         case '>':
1277                 os << "\\textgreater" << termcmd;
1278                 column += 13;
1279                 break;
1280         case '|':
1281                 os << "\\textbar" << termcmd;
1282                 column += 9;
1283                 break;
1284         case '-':
1285                 os << '-';
1286                 if (i + 1 < static_cast<pos_type>(text_.size()) &&
1287                     (end_pos == -1 || i + 1 < end_pos) &&
1288                     text_[i+1] == '-') {
1289                         // Prevent "--" becoming an en dash and "---" an em dash.
1290                         // (Within \ttfamily, "---" is merged to en dash + hyphen.)
1291                         os << "{}";
1292                         column += 2;
1293                 }
1294                 break;
1295         case '\"':
1296                 os << "\\textquotedbl" << termcmd;
1297                 column += 14;
1298                 break;
1299
1300         case '$': case '&':
1301         case '%': case '#': case '{':
1302         case '}': case '_':
1303                 os << '\\';
1304                 os.put(c);
1305                 column += 1;
1306                 break;
1307
1308         case '~':
1309                 os << "\\textasciitilde" << termcmd;
1310                 column += 16;
1311                 break;
1312
1313         case '^':
1314                 os << "\\textasciicircum" << termcmd;
1315                 column += 17;
1316                 break;
1317
1318         case '*':
1319         case '[':
1320         case ']':
1321                 // avoid being mistaken for optional arguments
1322                 os << '{';
1323                 os.put(c);
1324                 os << '}';
1325                 column += 2;
1326                 break;
1327
1328         case ' ':
1329                 // Blanks are printed before font switching.
1330                 // Sure? I am not! (try nice-latex)
1331                 // I am sure it's correct. LyX might be smarter
1332                 // in the future, but for now, nothing wrong is
1333                 // written. (Asger)
1334                 break;
1335
1336         case 0x2013:
1337         case 0x2014:
1338                 // XeTeX's dash behaviour is determined via a global setting
1339                 if (bparams.use_dash_ligatures
1340                     && owner_->getFontSettings(bparams, i).fontInfo().family() != TYPEWRITER_FAMILY
1341                     && !runparams.inIPA
1342                         // TODO #10961: && not in inset Flex Code
1343                         // TODO #10961: && not in layout LyXCode
1344                     && (!bparams.useNonTeXFonts || runparams.flavor != OutputParams::XETEX)) {
1345                         if (c == 0x2013) {
1346                                 // en-dash
1347                                 os << "--";
1348                                 column +=2;
1349                         } else {
1350                                 // em-dash
1351                                 os << "---";
1352                                 column +=3;
1353                         }
1354                         break;
1355                 }
1356                 // fall through
1357         default:
1358                 if (c == '\0')
1359                         return;
1360
1361                 Encoding const & encoding = *(runparams.encoding);
1362                 char_type next = '\0';
1363                 if (i + 1 < int(text_.size())) {
1364                         next = text_[i + 1];
1365                         if (Encodings::isCombiningChar(next)) {
1366                                 column += latexSurrogatePair(bparams, os, c, next, runparams) - 1;
1367                                 ++i;
1368                                 break;
1369                         }
1370                 }
1371                 string script;
1372                 pair<docstring, bool> latex = encoding.latexChar(c);
1373                 docstring nextlatex;
1374                 bool nexttipas = false;
1375                 string nexttipashortcut;
1376                 if (next != '\0' && next != META_INSET && encoding.encodable(next)) {
1377                         nextlatex = encoding.latexChar(next).first;
1378                         if (runparams.inIPA) {
1379                                 nexttipashortcut = Encodings::TIPAShortcut(next);
1380                                 nexttipas = !nexttipashortcut.empty();
1381                         }
1382                 }
1383                 bool tipas = false;
1384                 if (runparams.inIPA) {
1385                         string const tipashortcut = Encodings::TIPAShortcut(c);
1386                         if (!tipashortcut.empty()) {
1387                                 latex.first = from_ascii(tipashortcut);
1388                                 latex.second = false;
1389                                 tipas = true;
1390                         }
1391                 }
1392                 string fontenc;
1393                 fontenc = running_font.language()->fontenc(bparams);
1394                 // "Script chars" need to embraced in \textcyrillic and \textgreek notwithstanding
1395                 // whether they are encodable or not (it only depends on the font encoding)
1396                 if (!runparams.isFullUnicode() && Encodings::isKnownScriptChar(c, script)) {
1397                         docstring const wrapper = from_ascii("\\" + script + "{");
1398                         docstring ltx = latex.first;
1399                         if (!prefixIs(ltx, wrapper))
1400                                 ltx = wrapper + latex.first + from_ascii("}");
1401                         column += writeScriptChars(bparams, runparams, os, ltx, running_change,
1402                                                    encoding, fontenc, i) - 1;
1403                 } else if (latex.second
1404                          && ((!prefixIs(nextlatex, '\\')
1405                                && !prefixIs(nextlatex, '{')
1406                                && !prefixIs(nextlatex, '}'))
1407                              || (nexttipas
1408                                  && !prefixIs(from_ascii(nexttipashortcut), '\\')))
1409                          && !tipas) {
1410                         // Prevent eating of a following
1411                         // space or command corruption by
1412                         // following characters
1413                         if (next == ' ' || next == '\0') {
1414                                 column += latex.first.length() + 1;
1415                                 os << latex.first << "{}";
1416                         } else {
1417                                 column += latex.first.length();
1418                                 os << latex.first << " ";
1419                         }
1420                 } else {
1421                         column += latex.first.length() - 1;
1422                         os << latex.first;
1423                 }
1424                 break;
1425         }
1426 }
1427
1428
1429 bool Paragraph::Private::latexSpecialT1(char_type const c, otexstream & os,
1430         pos_type i, unsigned int & column)
1431 {
1432         switch (c) {
1433         case '>':
1434         case '<':
1435                 os.put(c);
1436                 // In T1 encoding, these characters exist
1437                 // but we should avoid ligatures
1438                 if (i + 1 >= int(text_.size()) || text_[i + 1] != c)
1439                         return true;
1440                 os << "\\textcompwordmark" << termcmd;
1441                 column += 19;
1442                 return true;
1443         case '|':
1444                 os.put(c);
1445                 return true;
1446         case '\"':
1447                 // soul.sty breaks with \char`\"
1448                 os << "\\textquotedbl" << termcmd;
1449                 column += 14;
1450                 return true;
1451         default:
1452                 return false;
1453         }
1454 }
1455
1456
1457 bool Paragraph::Private::latexSpecialTU(char_type const c, otexstream & os,
1458         pos_type i, unsigned int & column)
1459 {
1460         // TU encoding is currently on par with T1.
1461         return latexSpecialT1(c, os, i, column);
1462 }
1463
1464
1465 bool Paragraph::Private::latexSpecialT3(char_type const c, otexstream & os,
1466         pos_type /*i*/, unsigned int & column)
1467 {
1468         switch (c) {
1469         case '*':
1470         case '[':
1471         case ']':
1472         case '\"':
1473                 os.put(c);
1474                 return true;
1475         case '|':
1476                 os << "\\textvertline" << termcmd;
1477                 column += 14;
1478                 return true;
1479         default:
1480                 return false;
1481         }
1482 }
1483
1484
1485 void Paragraph::Private::validate(LaTeXFeatures & features) const
1486 {
1487         if (layout_->inpreamble && inset_owner_) {
1488                 // FIXME: Using a string stream here circumvents the encoding
1489                 // switching machinery of odocstream. Therefore the
1490                 // output is wrong if this paragraph contains content
1491                 // that needs to switch encoding.
1492                 Buffer const & buf = inset_owner_->buffer();
1493                 otexstringstream os;
1494                 os << layout_->preamble();
1495                 size_t const length = os.length();
1496                 TeXOnePar(buf, buf.text(), buf.getParFromID(owner_->id()).pit(), os,
1497                           features.runparams(), string(), 0, -1, true);
1498                 if (os.length() > length)
1499                         features.addPreambleSnippet(os.release(), true);
1500         }
1501
1502         if (features.runparams().flavor == OutputParams::HTML
1503             && layout_->htmltitle()) {
1504                 features.setHTMLTitle(owner_->asString(AS_STR_INSETS | AS_STR_SKIPDELETE));
1505         }
1506
1507         // check the params.
1508         if (!params_.spacing().isDefault())
1509                 features.require("setspace");
1510
1511         // then the layouts
1512         features.useLayout(layout_->name());
1513
1514         // then the fonts
1515         fontlist_.validate(features);
1516
1517         // then the indentation
1518         if (!params_.leftIndent().zero())
1519                 features.require("ParagraphLeftIndent");
1520
1521         // then the insets
1522         InsetList::const_iterator icit = insetlist_.begin();
1523         InsetList::const_iterator iend = insetlist_.end();
1524         for (; icit != iend; ++icit) {
1525                 if (icit->inset) {
1526                         features.inDeletedInset(owner_->isDeleted(icit->pos));
1527                         if (icit->inset->lyxCode() == FOOT_CODE) {
1528                                 // FIXME: an item inset would make things much easier.
1529                                 if ((layout_->latextype == LATEX_LIST_ENVIRONMENT
1530                                      || (layout_->latextype == LATEX_ITEM_ENVIRONMENT
1531                                          && layout_->margintype == MARGIN_FIRST_DYNAMIC))
1532                                     && (icit->pos < begin_of_body_
1533                                         || (icit->pos == begin_of_body_
1534                                             && (icit->pos == 0 || text_[icit->pos - 1] != ' '))))
1535                                         features.saveNoteEnv("description");
1536                         }
1537                         icit->inset->validate(features);
1538                         features.inDeletedInset(false);
1539                         if (layout_->needprotect &&
1540                             icit->inset->lyxCode() == FOOT_CODE)
1541                                 features.require("NeedLyXFootnoteCode");
1542                 }
1543         }
1544
1545         // then the contents
1546         BufferParams const bp = features.runparams().is_child
1547                 ? features.buffer().masterParams() : features.buffer().params();
1548         for (pos_type i = 0; i < int(text_.size()) ; ++i) {
1549                 char_type c = text_[i];
1550                 CharInfo const & ci = Encodings::unicodeCharInfo(c);
1551                 if (c == 0x0022) {
1552                         if (features.runparams().isFullUnicode() && bp.useNonTeXFonts)
1553                                 features.require("textquotedblp");
1554                         else if (bp.main_font_encoding() != "T1"
1555                                  || ((&owner_->getFontSettings(bp, i))->language()->internalFontEncoding()))
1556                                 features.require("textquotedbl");
1557                 } else if (ci.textfeature() && contains(ci.textpreamble(), '=')) {
1558                         // features that depend on the font or input encoding
1559                         string feats = ci.textpreamble();
1560                         string fontenc = (&owner_->getFontSettings(bp, i))->language()->fontenc(bp);
1561                         if (fontenc.empty())
1562                                 fontenc = features.runparams().main_fontenc;
1563                         while (!feats.empty()) {
1564                                 string feat;
1565                                 feats = split(feats, feat, ',');
1566                                 if (contains(feat, "!=")) {
1567                                         // a feature that is required except for the spcified
1568                                         // font or input encodings
1569                                         string realfeature;
1570                                         string const contexts = ltrim(split(feat, realfeature, '!'), "=");
1571                                         // multiple encodings are separated by semicolon
1572                                         vector<string> context = getVectorFromString(contexts, ";");
1573                                         // require feature if the context matches neither current font
1574                                         // nor input encoding
1575                                         if (std::find(context.begin(), context.end(), fontenc) == context.end()
1576                                             && std::find(context.begin(), context.end(),
1577                                                          features.runparams().encoding->name()) == context.end())
1578                                                 features.require(realfeature);
1579                                 } else if (contains(feat, '=')) {
1580                                         // a feature that is required only for the spcified
1581                                         // font or input encodings
1582                                         string realfeature;
1583                                         string const contexts = split(feat, realfeature, '=');
1584                                         // multiple encodings are separated by semicolon
1585                                         vector<string> context = getVectorFromString(contexts, ";");
1586                                         // require feature if the context matches either current font
1587                                         // or input encoding
1588                                         if (std::find(context.begin(), context.end(), fontenc) != context.end()
1589                                             || std::find(context.begin(), context.end(),
1590                                                          features.runparams().encoding->name()) != context.end())
1591                                                 features.require(realfeature);
1592                                 }
1593                         }
1594                 } else if (!bp.use_dash_ligatures
1595                            && (c == 0x2013 || c == 0x2014)
1596                            && bp.useNonTeXFonts
1597                            && features.runparams().flavor == OutputParams::XETEX)
1598                         // XeTeX's dash behaviour is determined via a global setting
1599                         features.require("xetexdashbreakstate");
1600                 BufferEncodings::validate(c, features);
1601         }
1602 }
1603
1604 /////////////////////////////////////////////////////////////////////
1605 //
1606 // Paragraph
1607 //
1608 /////////////////////////////////////////////////////////////////////
1609
1610 namespace {
1611         Layout const emptyParagraphLayout;
1612 }
1613
1614 Paragraph::Paragraph()
1615         : d(new Paragraph::Private(this, emptyParagraphLayout))
1616 {
1617         itemdepth = 0;
1618         d->params_.clear();
1619 }
1620
1621
1622 Paragraph::Paragraph(Paragraph const & par)
1623         : itemdepth(par.itemdepth),
1624         d(new Paragraph::Private(*par.d, this))
1625 {
1626         registerWords();
1627 }
1628
1629
1630 Paragraph::Paragraph(Paragraph const & par, pos_type beg, pos_type end)
1631         : itemdepth(par.itemdepth),
1632         d(new Paragraph::Private(*par.d, this, beg, end))
1633 {
1634         registerWords();
1635 }
1636
1637
1638 Paragraph & Paragraph::operator=(Paragraph const & par)
1639 {
1640         // needed as we will destroy the private part before copying it
1641         if (&par != this) {
1642                 itemdepth = par.itemdepth;
1643
1644                 deregisterWords();
1645                 delete d;
1646                 d = new Private(*par.d, this);
1647                 registerWords();
1648         }
1649         return *this;
1650 }
1651
1652
1653 Paragraph::~Paragraph()
1654 {
1655         deregisterWords();
1656         delete d;
1657 }
1658
1659
1660 namespace {
1661
1662 // this shall be called just before every "os << ..." action.
1663 void flushString(ostream & os, docstring & s)
1664 {
1665         os << to_utf8(s);
1666         s.erase();
1667 }
1668
1669 } // namespace
1670
1671
1672 void Paragraph::write(ostream & os, BufferParams const & bparams,
1673         depth_type & dth) const
1674 {
1675         // The beginning or end of a deeper (i.e. nested) area?
1676         if (dth != d->params_.depth()) {
1677                 if (d->params_.depth() > dth) {
1678                         while (d->params_.depth() > dth) {
1679                                 os << "\n\\begin_deeper";
1680                                 ++dth;
1681                         }
1682                 } else {
1683                         while (d->params_.depth() < dth) {
1684                                 os << "\n\\end_deeper";
1685                                 --dth;
1686                         }
1687                 }
1688         }
1689
1690         // First write the layout
1691         os << "\n\\begin_layout " << to_utf8(d->layout_->name()) << '\n';
1692
1693         d->params_.write(os);
1694
1695         Font font1(inherit_font, bparams.language);
1696
1697         Change running_change = Change(Change::UNCHANGED);
1698
1699         // this string is used as a buffer to avoid repetitive calls
1700         // to to_utf8(), which turn out to be expensive (JMarc)
1701         docstring write_buffer;
1702
1703         int column = 0;
1704         for (pos_type i = 0; i <= size(); ++i) {
1705
1706                 Change const & change = lookupChange(i);
1707                 if (change != running_change)
1708                         flushString(os, write_buffer);
1709                 Changes::lyxMarkChange(os, bparams, column, running_change, change);
1710                 running_change = change;
1711
1712                 if (i == size())
1713                         break;
1714
1715                 // Write font changes
1716                 Font font2 = getFontSettings(bparams, i);
1717                 if (font2 != font1) {
1718                         flushString(os, write_buffer);
1719                         font2.lyxWriteChanges(font1, os);
1720                         column = 0;
1721                         font1 = font2;
1722                 }
1723
1724                 char_type const c = d->text_[i];
1725                 switch (c) {
1726                 case META_INSET:
1727                         if (Inset const * inset = getInset(i)) {
1728                                 flushString(os, write_buffer);
1729                                 if (inset->directWrite()) {
1730                                         // international char, let it write
1731                                         // code directly so it's shorter in
1732                                         // the file
1733                                         inset->write(os);
1734                                 } else {
1735                                         if (i)
1736                                                 os << '\n';
1737                                         os << "\\begin_inset ";
1738                                         inset->write(os);
1739                                         os << "\n\\end_inset\n\n";
1740                                         column = 0;
1741                                 }
1742                                 // FIXME This can be removed again once the mystery
1743                                 // crash has been resolved.
1744                                 os << flush;
1745                         }
1746                         break;
1747                 case '\\':
1748                         flushString(os, write_buffer);
1749                         os << "\n\\backslash\n";
1750                         column = 0;
1751                         break;
1752                 case '.':
1753                         flushString(os, write_buffer);
1754                         if (i + 1 < size() && d->text_[i + 1] == ' ') {
1755                                 os << ".\n";
1756                                 column = 0;
1757                         } else
1758                                 os << '.';
1759                         break;
1760                 default:
1761                         if ((column > 70 && c == ' ')
1762                             || column > 79) {
1763                                 flushString(os, write_buffer);
1764                                 os << '\n';
1765                                 column = 0;
1766                         }
1767                         // this check is to amend a bug. LyX sometimes
1768                         // inserts '\0' this could cause problems.
1769                         if (c != '\0')
1770                                 write_buffer.push_back(c);
1771                         else
1772                                 LYXERR0("NUL char in structure.");
1773                         ++column;
1774                         break;
1775                 }
1776         }
1777
1778         flushString(os, write_buffer);
1779         os << "\n\\end_layout\n";
1780         // FIXME This can be removed again once the mystery
1781         // crash has been resolved.
1782         os << flush;
1783 }
1784
1785
1786 void Paragraph::validate(LaTeXFeatures & features) const
1787 {
1788         d->validate(features);
1789         bool fragile = features.runparams().moving_arg;
1790         fragile |= layout().needprotect;
1791         if (needsCProtection(fragile))
1792                 features.require("cprotect");
1793 }
1794
1795
1796 void Paragraph::insert(pos_type start, docstring const & str,
1797                        Font const & font, Change const & change)
1798 {
1799         for (size_t i = 0, n = str.size(); i != n ; ++i)
1800                 insertChar(start + i, str[i], font, change);
1801 }
1802
1803
1804 void Paragraph::appendChar(char_type c, Font const & font,
1805                 Change const & change)
1806 {
1807         // track change
1808         d->changes_.insert(change, d->text_.size());
1809         // when appending characters, no need to update tables
1810         d->text_.push_back(c);
1811         setFont(d->text_.size() - 1, font);
1812         d->requestSpellCheck(d->text_.size() - 1);
1813 }
1814
1815
1816 void Paragraph::appendString(docstring const & s, Font const & font,
1817                 Change const & change)
1818 {
1819         pos_type end = s.size();
1820         size_t oldsize = d->text_.size();
1821         size_t newsize = oldsize + end;
1822         size_t capacity = d->text_.capacity();
1823         if (newsize >= capacity)
1824                 d->text_.reserve(max(capacity + 100, newsize));
1825
1826         // when appending characters, no need to update tables
1827         d->text_.append(s);
1828
1829         // FIXME: Optimize this!
1830         for (size_t i = oldsize; i != newsize; ++i) {
1831                 // track change
1832                 d->changes_.insert(change, i);
1833                 d->requestSpellCheck(i);
1834         }
1835         d->fontlist_.set(oldsize, font);
1836         d->fontlist_.set(newsize - 1, font);
1837 }
1838
1839
1840 void Paragraph::insertChar(pos_type pos, char_type c,
1841                            bool trackChanges)
1842 {
1843         d->insertChar(pos, c, Change(trackChanges ?
1844                            Change::INSERTED : Change::UNCHANGED));
1845 }
1846
1847
1848 void Paragraph::insertChar(pos_type pos, char_type c,
1849                            Font const & font, bool trackChanges)
1850 {
1851         d->insertChar(pos, c, Change(trackChanges ?
1852                            Change::INSERTED : Change::UNCHANGED));
1853         setFont(pos, font);
1854 }
1855
1856
1857 void Paragraph::insertChar(pos_type pos, char_type c,
1858                            Font const & font, Change const & change)
1859 {
1860         d->insertChar(pos, c, change);
1861         setFont(pos, font);
1862 }
1863
1864
1865 void Paragraph::resetFonts(Font const & font)
1866 {
1867         d->fontlist_.clear();
1868         d->fontlist_.set(0, font);
1869         d->fontlist_.set(d->text_.size() - 1, font);
1870 }
1871
1872 // Gets uninstantiated font setting at position.
1873 Font const & Paragraph::getFontSettings(BufferParams const & bparams,
1874                                          pos_type pos) const
1875 {
1876         if (pos > size()) {
1877                 LYXERR0("pos: " << pos << " size: " << size());
1878                 LBUFERR(false);
1879         }
1880
1881         FontList::const_iterator cit = d->fontlist_.fontIterator(pos);
1882         if (cit != d->fontlist_.end())
1883                 return cit->font();
1884
1885         if (pos == size() && !empty())
1886                 return getFontSettings(bparams, pos - 1);
1887
1888         // Optimisation: avoid a full font instantiation if there is no
1889         // language change from previous call.
1890         static Font previous_font;
1891         static Language const * previous_lang = 0;
1892         Language const * lang = getParLanguage(bparams);
1893         if (lang != previous_lang) {
1894                 previous_lang = lang;
1895                 previous_font = Font(inherit_font, lang);
1896         }
1897         return previous_font;
1898 }
1899
1900
1901 FontSpan Paragraph::fontSpan(pos_type pos) const
1902 {
1903         LBUFERR(pos <= size());
1904
1905         if (pos == size())
1906                 return FontSpan(pos, pos);
1907
1908         pos_type start = 0;
1909         FontList::const_iterator cit = d->fontlist_.begin();
1910         FontList::const_iterator end = d->fontlist_.end();
1911         for (; cit != end; ++cit) {
1912                 if (cit->pos() >= pos) {
1913                         if (pos >= beginOfBody())
1914                                 return FontSpan(max(start, beginOfBody()),
1915                                                 cit->pos());
1916                         else
1917                                 return FontSpan(start,
1918                                                 min(beginOfBody() - 1,
1919                                                          cit->pos()));
1920                 }
1921                 start = cit->pos() + 1;
1922         }
1923
1924         // This should not happen, but if so, we take no chances.
1925         LYXERR0("Paragraph::fontSpan: position not found in fontinfo table!");
1926         LASSERT(false, return FontSpan(pos, pos));
1927 }
1928
1929
1930 // Gets uninstantiated font setting at position 0
1931 Font const & Paragraph::getFirstFontSettings(BufferParams const & bparams) const
1932 {
1933         if (!empty() && !d->fontlist_.empty())
1934                 return d->fontlist_.begin()->font();
1935
1936         // Optimisation: avoid a full font instantiation if there is no
1937         // language change from previous call.
1938         static Font previous_font;
1939         static Language const * previous_lang = 0;
1940         if (bparams.language != previous_lang) {
1941                 previous_lang = bparams.language;
1942                 previous_font = Font(inherit_font, bparams.language);
1943         }
1944
1945         return previous_font;
1946 }
1947
1948
1949 // Gets the fully instantiated font at a given position in a paragraph
1950 // This is basically the same function as Text::GetFont() in text2.cpp.
1951 // The difference is that this one is used for generating the LaTeX file,
1952 // and thus cosmetic "improvements" are disallowed: This has to deliver
1953 // the true picture of the buffer. (Asger)
1954 Font const Paragraph::getFont(BufferParams const & bparams, pos_type pos,
1955                                  Font const & outerfont) const
1956 {
1957         LBUFERR(pos >= 0);
1958
1959         Font font = getFontSettings(bparams, pos);
1960
1961         pos_type const body_pos = beginOfBody();
1962         FontInfo & fi = font.fontInfo();
1963         if (pos < body_pos)
1964                 fi.realize(d->layout_->labelfont);
1965         else
1966                 fi.realize(d->layout_->font);
1967
1968         fi.realize(outerfont.fontInfo());
1969         fi.realize(bparams.getFont().fontInfo());
1970
1971         return font;
1972 }
1973
1974
1975 Font const Paragraph::getLabelFont
1976         (BufferParams const & bparams, Font const & outerfont) const
1977 {
1978         FontInfo tmpfont = d->layout_->labelfont;
1979         tmpfont.realize(outerfont.fontInfo());
1980         tmpfont.realize(bparams.getFont().fontInfo());
1981         return Font(tmpfont, getParLanguage(bparams));
1982 }
1983
1984
1985 Font const Paragraph::getLayoutFont
1986         (BufferParams const & bparams, Font const & outerfont) const
1987 {
1988         FontInfo tmpfont = d->layout_->font;
1989         tmpfont.realize(outerfont.fontInfo());
1990         tmpfont.realize(bparams.getFont().fontInfo());
1991         return Font(tmpfont, getParLanguage(bparams));
1992 }
1993
1994
1995 char_type Paragraph::getUChar(BufferParams const & bparams,
1996                               OutputParams const & rp,
1997                               pos_type pos) const
1998 {
1999         char_type c = d->text_[pos];
2000
2001         // Return unchanged character in LTR languages.
2002         if (!getFontSettings(bparams, pos).isRightToLeft())
2003                 return c;
2004
2005         // FIXME This is a complete mess due to all the language-specific
2006         // special cases. We need to unify this eventually, but this
2007         // requires a file format change and some thought.
2008         // We also need to unify the input of parentheses in different RTL
2009         // languages. Currently, some have their own methods (Arabic:
2010         // 18599/lyxsvn, Hebrew: e5f42f67d/lyxgit), some don't (Urdu, Syriac).
2011         // Also note that the representation in the LyX file is probably wrong
2012         // (see FIXME in TextMetrics::breakRow).
2013         // Most likely, we should simply rely on Qt's unicode handling here.
2014         string const & lang = getFontSettings(bparams, pos).language()->lang();
2015
2016         // With polyglossia, brackets and stuff need not be reversed in RTL scripts
2017         // FIXME: The special casing for Hebrew parens is due to the special
2018         // handling on input (for Hebrew in e5f42f67d/lyxgit); see #8251.
2019         char_type uc = c;
2020         if (rp.use_polyglossia) {
2021                 switch (c) {
2022                 case '(':
2023                         if (lang == "hebrew")
2024                                 uc = ')';
2025                         break;
2026                 case ')':
2027                         if (lang == "hebrew")
2028                                 uc = '(';
2029                         break;
2030                 }
2031                 return uc;
2032         }
2033
2034         // In the following languages, brackets don't need to be reversed.
2035         // Furthermore, in arabic_arabi, they are transformed to Arabic
2036         // Ornate Parentheses (dunno if this is really wanted)
2037         bool const reversebrackets = lang != "arabic_arabtex"
2038                         && lang != "arabic_arabi"
2039                         && lang != "farsi";
2040
2041         switch (c) {
2042         case '[':
2043                 if (reversebrackets)
2044                         uc = ']';
2045                 break;
2046         case ']':
2047                 if (reversebrackets)
2048                         uc = '[';
2049                 break;
2050         case '{':
2051                 uc = '}';
2052                 break;
2053         case '}':
2054                 uc = '{';
2055                 break;
2056         case '<':
2057                 uc = '>';
2058                 break;
2059         case '>':
2060                 uc = '<';
2061                 break;
2062         }
2063
2064         return uc;
2065 }
2066
2067
2068 void Paragraph::setFont(pos_type pos, Font const & font)
2069 {
2070         LASSERT(pos <= size(), return);
2071
2072         // First, reduce font against layout/label font
2073         // Update: The setCharFont() routine in text2.cpp already
2074         // reduces font, so we don't need to do that here. (Asger)
2075
2076         d->fontlist_.set(pos, font);
2077 }
2078
2079
2080 void Paragraph::makeSameLayout(Paragraph const & par)
2081 {
2082         d->layout_ = par.d->layout_;
2083         d->params_ = par.d->params_;
2084 }
2085
2086
2087 bool Paragraph::stripLeadingSpaces(bool trackChanges)
2088 {
2089         if (isFreeSpacing())
2090                 return false;
2091
2092         int pos = 0;
2093         int count = 0;
2094
2095         while (pos < size() && (isNewline(pos) || isLineSeparator(pos))) {
2096                 if (eraseChar(pos, trackChanges))
2097                         ++count;
2098                 else
2099                         ++pos;
2100         }
2101
2102         return count > 0 || pos > 0;
2103 }
2104
2105
2106 bool Paragraph::hasSameLayout(Paragraph const & par) const
2107 {
2108         return par.d->layout_ == d->layout_
2109                 && d->params_.sameLayout(par.d->params_);
2110 }
2111
2112
2113 depth_type Paragraph::getDepth() const
2114 {
2115         return d->params_.depth();
2116 }
2117
2118
2119 depth_type Paragraph::getMaxDepthAfter() const
2120 {
2121         if (d->layout_->isEnvironment())
2122                 return d->params_.depth() + 1;
2123         else
2124                 return d->params_.depth();
2125 }
2126
2127
2128 LyXAlignment Paragraph::getAlign() const
2129 {
2130         if (d->params_.align() == LYX_ALIGN_LAYOUT)
2131                 return d->layout_->align;
2132         else
2133                 return d->params_.align();
2134 }
2135
2136
2137 docstring const & Paragraph::labelString() const
2138 {
2139         return d->params_.labelString();
2140 }
2141
2142
2143 // the next two functions are for the manual labels
2144 docstring const Paragraph::getLabelWidthString() const
2145 {
2146         if (d->layout_->margintype == MARGIN_MANUAL
2147             || d->layout_->latextype == LATEX_BIB_ENVIRONMENT)
2148                 return d->params_.labelWidthString();
2149         else
2150                 return _("Senseless with this layout!");
2151 }
2152
2153
2154 void Paragraph::setLabelWidthString(docstring const & s)
2155 {
2156         d->params_.labelWidthString(s);
2157 }
2158
2159
2160 docstring Paragraph::expandLabel(Layout const & layout,
2161                 BufferParams const & bparams) const
2162 {
2163         return expandParagraphLabel(layout, bparams, true);
2164 }
2165
2166
2167 docstring Paragraph::expandDocBookLabel(Layout const & layout,
2168                 BufferParams const & bparams) const
2169 {
2170         return expandParagraphLabel(layout, bparams, false);
2171 }
2172
2173
2174 docstring Paragraph::expandParagraphLabel(Layout const & layout,
2175                 BufferParams const & bparams, bool process_appendix) const
2176 {
2177         DocumentClass const & tclass = bparams.documentClass();
2178         string const & lang = getParLanguage(bparams)->code();
2179         bool const in_appendix = process_appendix && d->params_.appendix();
2180         docstring fmt = translateIfPossible(layout.labelstring(in_appendix), lang);
2181
2182         if (fmt.empty() && !layout.counter.empty())
2183                 return tclass.counters().theCounter(layout.counter, lang);
2184
2185         // handle 'inherited level parts' in 'fmt',
2186         // i.e. the stuff between '@' in   '@Section@.\arabic{subsection}'
2187         size_t const i = fmt.find('@', 0);
2188         if (i != docstring::npos) {
2189                 size_t const j = fmt.find('@', i + 1);
2190                 if (j != docstring::npos) {
2191                         docstring parent(fmt, i + 1, j - i - 1);
2192                         docstring label = from_ascii("??");
2193                         if (tclass.hasLayout(parent))
2194                                 label = expandParagraphLabel(tclass[parent], bparams,
2195                                                       process_appendix);
2196                         fmt = docstring(fmt, 0, i) + label
2197                                 + docstring(fmt, j + 1, docstring::npos);
2198                 }
2199         }
2200
2201         return tclass.counters().counterLabel(fmt, lang);
2202 }
2203
2204
2205 void Paragraph::applyLayout(Layout const & new_layout)
2206 {
2207         d->layout_ = &new_layout;
2208         LyXAlignment const oldAlign = d->params_.align();
2209
2210         if (!(oldAlign & d->layout_->alignpossible)) {
2211                 frontend::Alert::warning(_("Alignment not permitted"),
2212                         _("The new layout does not permit the alignment previously used.\nSetting to default."));
2213                 d->params_.align(LYX_ALIGN_LAYOUT);
2214         }
2215 }
2216
2217
2218 pos_type Paragraph::beginOfBody() const
2219 {
2220         return d->begin_of_body_;
2221 }
2222
2223
2224 void Paragraph::setBeginOfBody()
2225 {
2226         if (d->layout_->labeltype != LABEL_MANUAL) {
2227                 d->begin_of_body_ = 0;
2228                 return;
2229         }
2230
2231         // Unroll the first two cycles of the loop
2232         // and remember the previous character to
2233         // remove unnecessary getChar() calls
2234         pos_type i = 0;
2235         pos_type end = size();
2236         if (i < end && !(isNewline(i) || isEnvSeparator(i))) {
2237                 ++i;
2238                 if (i < end) {
2239                         char_type previous_char = d->text_[i];
2240                         if (!(isNewline(i) || isEnvSeparator(i))) {
2241                                 ++i;
2242                                 while (i < end && previous_char != ' ') {
2243                                         char_type temp = d->text_[i];
2244                                         if (isNewline(i) || isEnvSeparator(i))
2245                                                 break;
2246                                         ++i;
2247                                         previous_char = temp;
2248                                 }
2249                         }
2250                 }
2251         }
2252
2253         d->begin_of_body_ = i;
2254 }
2255
2256
2257 bool Paragraph::allowParagraphCustomization() const
2258 {
2259         return inInset().allowParagraphCustomization();
2260 }
2261
2262
2263 bool Paragraph::usePlainLayout() const
2264 {
2265         return inInset().usePlainLayout();
2266 }
2267
2268
2269 bool Paragraph::isPassThru() const
2270 {
2271         return inInset().isPassThru() || d->layout_->pass_thru;
2272 }
2273
2274 namespace {
2275
2276 // paragraphs inside floats need different alignment tags to avoid
2277 // unwanted space
2278
2279 bool noTrivlistCentering(InsetCode code)
2280 {
2281         return code == FLOAT_CODE
2282                || code == WRAP_CODE
2283                || code == CELL_CODE;
2284 }
2285
2286
2287 string correction(string const & orig)
2288 {
2289         if (orig == "flushleft")
2290                 return "raggedright";
2291         if (orig == "flushright")
2292                 return "raggedleft";
2293         if (orig == "center")
2294                 return "centering";
2295         return orig;
2296 }
2297
2298
2299 bool corrected_env(otexstream & os, string const & suffix, string const & env,
2300         InsetCode code, bool const lastpar, int & col)
2301 {
2302         string macro = suffix + "{";
2303         if (noTrivlistCentering(code)) {
2304                 if (lastpar) {
2305                         // the last paragraph in non-trivlist-aligned
2306                         // context is special (to avoid unwanted whitespace)
2307                         if (suffix == "\\begin") {
2308                                 macro = "\\" + correction(env) + "{}";
2309                                 os << from_ascii(macro);
2310                                 col += macro.size();
2311                                 return true;
2312                         }
2313                         return false;
2314                 }
2315                 macro += correction(env);
2316         } else
2317                 macro += env;
2318         macro += "}";
2319         if (suffix == "\\par\\end") {
2320                 os << breakln;
2321                 col = 0;
2322         }
2323         os << from_ascii(macro);
2324         col += macro.size();
2325         if (suffix == "\\begin") {
2326                 os << breakln;
2327                 col = 0;
2328         }
2329         return true;
2330 }
2331
2332 } // namespace
2333
2334
2335 int Paragraph::Private::startTeXParParams(BufferParams const & bparams,
2336                         otexstream & os, OutputParams const & runparams) const
2337 {
2338         int column = 0;
2339
2340         bool canindent =
2341                 (bparams.paragraph_separation == BufferParams::ParagraphIndentSeparation) ?
2342                         (layout_->toggle_indent != ITOGGLE_NEVER) :
2343                         (layout_->toggle_indent == ITOGGLE_ALWAYS);
2344
2345         if (canindent && params_.noindent() && !layout_->pass_thru) {
2346                 os << "\\noindent ";
2347                 column += 10;
2348         }
2349
2350         LyXAlignment const curAlign = params_.align();
2351
2352         if (curAlign == layout_->align)
2353                 return column;
2354
2355         switch (curAlign) {
2356         case LYX_ALIGN_NONE:
2357         case LYX_ALIGN_BLOCK:
2358         case LYX_ALIGN_LAYOUT:
2359         case LYX_ALIGN_SPECIAL:
2360         case LYX_ALIGN_DECIMAL:
2361                 break;
2362         case LYX_ALIGN_LEFT:
2363         case LYX_ALIGN_RIGHT:
2364         case LYX_ALIGN_CENTER:
2365                 if (runparams.moving_arg) {
2366                         os << "\\protect";
2367                         column += 8;
2368                 }
2369                 break;
2370         }
2371
2372         string const begin_tag = "\\begin";
2373         InsetCode code = ownerCode();
2374         bool const lastpar = runparams.isLastPar;
2375
2376         switch (curAlign) {
2377         case LYX_ALIGN_NONE:
2378         case LYX_ALIGN_BLOCK:
2379         case LYX_ALIGN_LAYOUT:
2380         case LYX_ALIGN_SPECIAL:
2381         case LYX_ALIGN_DECIMAL:
2382                 break;
2383         case LYX_ALIGN_LEFT: {
2384                 if (!owner_->getParLanguage(bparams)->rightToLeft())
2385                         corrected_env(os, begin_tag, "flushleft", code, lastpar, column);
2386                 else
2387                         corrected_env(os, begin_tag, "flushright", code, lastpar, column);
2388                 break;
2389         } case LYX_ALIGN_RIGHT: {
2390                 if (!owner_->getParLanguage(bparams)->rightToLeft())
2391                         corrected_env(os, begin_tag, "flushright", code, lastpar, column);
2392                 else
2393                         corrected_env(os, begin_tag, "flushleft", code, lastpar, column);
2394                 break;
2395         } case LYX_ALIGN_CENTER: {
2396                 corrected_env(os, begin_tag, "center", code, lastpar, column);
2397                 break;
2398         }
2399         }
2400
2401         return column;
2402 }
2403
2404
2405 bool Paragraph::Private::endTeXParParams(BufferParams const & bparams,
2406                         otexstream & os, OutputParams const & runparams) const
2407 {
2408         LyXAlignment const curAlign = params_.align();
2409
2410         if (curAlign == layout_->align)
2411                 return false;
2412
2413         switch (curAlign) {
2414         case LYX_ALIGN_NONE:
2415         case LYX_ALIGN_BLOCK:
2416         case LYX_ALIGN_LAYOUT:
2417         case LYX_ALIGN_SPECIAL:
2418         case LYX_ALIGN_DECIMAL:
2419                 break;
2420         case LYX_ALIGN_LEFT:
2421         case LYX_ALIGN_RIGHT:
2422         case LYX_ALIGN_CENTER:
2423                 if (runparams.moving_arg)
2424                         os << "\\protect";
2425                 break;
2426         }
2427
2428         bool output = false;
2429         int col = 0;
2430         string const end_tag = "\\par\\end";
2431         InsetCode code = ownerCode();
2432         bool const lastpar = runparams.isLastPar;
2433
2434         switch (curAlign) {
2435         case LYX_ALIGN_NONE:
2436         case LYX_ALIGN_BLOCK:
2437         case LYX_ALIGN_LAYOUT:
2438         case LYX_ALIGN_SPECIAL:
2439         case LYX_ALIGN_DECIMAL:
2440                 break;
2441         case LYX_ALIGN_LEFT: {
2442                 if (!owner_->getParLanguage(bparams)->rightToLeft())
2443                         output = corrected_env(os, end_tag, "flushleft", code, lastpar, col);
2444                 else
2445                         output = corrected_env(os, end_tag, "flushright", code, lastpar, col);
2446                 break;
2447         } case LYX_ALIGN_RIGHT: {
2448                 if (!owner_->getParLanguage(bparams)->rightToLeft())
2449                         output = corrected_env(os, end_tag, "flushright", code, lastpar, col);
2450                 else
2451                         output = corrected_env(os, end_tag, "flushleft", code, lastpar, col);
2452                 break;
2453         } case LYX_ALIGN_CENTER: {
2454                 corrected_env(os, end_tag, "center", code, lastpar, col);
2455                 break;
2456         }
2457         }
2458
2459         return output || lastpar;
2460 }
2461
2462
2463 // This one spits out the text of the paragraph
2464 void Paragraph::latex(BufferParams const & bparams,
2465         Font const & outerfont,
2466         otexstream & os,
2467         OutputParams const & runparams,
2468         int start_pos, int end_pos, bool force) const
2469 {
2470         LYXERR(Debug::LATEX, "Paragraph::latex...     " << this);
2471
2472         // FIXME This check should not be needed. Perhaps issue an
2473         // error if it triggers.
2474         Layout const & style = inInset().forcePlainLayout() ?
2475                 bparams.documentClass().plainLayout() : *d->layout_;
2476
2477         if (!force && style.inpreamble)
2478                 return;
2479
2480         bool const allowcust = allowParagraphCustomization();
2481
2482         // Current base font for all inherited font changes, without any
2483         // change caused by an individual character, except for the language:
2484         // It is set to the language of the first character.
2485         // As long as we are in the label, this font is the base font of the
2486         // label. Before the first body character it is set to the base font
2487         // of the body.
2488         Font basefont;
2489
2490         // Maybe we have to create a optional argument.
2491         pos_type body_pos = beginOfBody();
2492         unsigned int column = 0;
2493
2494         if (body_pos > 0) {
2495                 // the optional argument is kept in curly brackets in
2496                 // case it contains a ']'
2497                 // This is not strictly needed, but if this is changed it
2498                 // would be a file format change, and tex2lyx would need
2499                 // to be adjusted, since it unconditionally removes the
2500                 // braces when it parses \item.
2501                 os << "[{";
2502                 column += 2;
2503                 basefont = getLabelFont(bparams, outerfont);
2504         } else {
2505                 basefont = getLayoutFont(bparams, outerfont);
2506         }
2507
2508         // Which font is currently active?
2509         Font running_font(basefont);
2510         // Do we have an open font change?
2511         bool open_font = false;
2512
2513         Change runningChange = Change(Change::UNCHANGED);
2514
2515         Encoding const * const prev_encoding = runparams.encoding;
2516
2517         os.texrow().start(id(), 0);
2518
2519         // if the paragraph is empty, the loop will not be entered at all
2520         if (empty()) {
2521                 // For InTitle commands, we have already opened a group
2522                 // in output_latex::TeXOnePar.
2523                 if (style.isCommand() && !style.intitle) {
2524                         os << '{';
2525                         ++column;
2526                 }
2527                 if (!style.leftdelim().empty()) {
2528                         os << style.leftdelim();
2529                         column += style.leftdelim().size();
2530                 }
2531                 if (allowcust)
2532                         column += d->startTeXParParams(bparams, os, runparams);
2533         }
2534
2535         // Whether a \par can be issued for insets typeset inline with text.
2536         // Yes if greater than 0. This has to be static.
2537         THREAD_LOCAL_STATIC int parInline = 0;
2538
2539         for (pos_type i = 0; i < size(); ++i) {
2540                 // First char in paragraph or after label?
2541                 if (i == body_pos) {
2542                         if (body_pos > 0) {
2543                                 if (open_font) {
2544                                         bool needPar = false;
2545                                         column += running_font.latexWriteEndChanges(
2546                                                 os, bparams, runparams,
2547                                                 basefont, basefont, needPar);
2548                                         open_font = false;
2549                                 }
2550                                 basefont = getLayoutFont(bparams, outerfont);
2551                                 running_font = basefont;
2552
2553                                 column += Changes::latexMarkChange(os, bparams,
2554                                                 runningChange, Change(Change::UNCHANGED),
2555                                                 runparams);
2556                                 runningChange = Change(Change::UNCHANGED);
2557
2558                                 os << "}] ";
2559                                 column +=3;
2560                         }
2561                         // For InTitle commands, we have already opened a group
2562                         // in output_latex::TeXOnePar.
2563                         if (style.isCommand() && !style.intitle) {
2564                                 os << '{';
2565                                 ++column;
2566                         }
2567
2568                         if (!style.leftdelim().empty()) {
2569                                 os << style.leftdelim();
2570                                 column += style.leftdelim().size();
2571                         }
2572
2573                         if (allowcust)
2574                                 column += d->startTeXParParams(bparams, os,
2575                                                             runparams);
2576                 }
2577
2578                 runparams.wasDisplayMath = runparams.inDisplayMath;
2579                 runparams.inDisplayMath = false;
2580                 bool deleted_display_math = false;
2581                 Change const & change = runparams.inDeletedInset
2582                         ? runparams.changeOfDeletedInset : lookupChange(i);
2583
2584                 // Check whether a display math inset follows
2585                 if (d->text_[i] == META_INSET
2586                     && i >= start_pos && (end_pos == -1 || i < end_pos)) {
2587                         InsetMath const * im = getInset(i)->asInsetMath();
2588                         if (im && im->asHullInset()
2589                             && im->asHullInset()->outerDisplay()) {
2590                                 runparams.inDisplayMath = true;
2591                                 // runparams.inDeletedInset will be set by
2592                                 // latexInset later, but we need this info
2593                                 // before it is called. On the other hand, we
2594                                 // cannot set it here because it is a counter.
2595                                 deleted_display_math = isDeleted(i);
2596                         }
2597                         if (bparams.output_changes && deleted_display_math
2598                             && runningChange == change
2599                             && change.type == Change::DELETED
2600                             && !os.afterParbreak()) {
2601                                 // A display math in the same paragraph follows.
2602                                 // We have to close and then reopen \lyxdeleted,
2603                                 // otherwise the math will be shifted up.
2604                                 OutputParams rp = runparams;
2605                                 if (open_font) {
2606                                         bool needPar = false;
2607                                         column += running_font.latexWriteEndChanges(
2608                                                 os, bparams, rp, basefont,
2609                                                 basefont, needPar);
2610                                         open_font = false;
2611                                 }
2612                                 basefont = getLayoutFont(bparams, outerfont);
2613                                 running_font = basefont;
2614                                 column += Changes::latexMarkChange(os, bparams,
2615                                         Change(Change::INSERTED), change, rp);
2616                         }
2617                 }
2618
2619                 if (bparams.output_changes && runningChange != change) {
2620                         if (open_font) {
2621                                 bool needPar = false;
2622                                 column += running_font.latexWriteEndChanges(
2623                                                 os, bparams, runparams,
2624                                                 basefont, basefont, needPar);
2625                                 open_font = false;
2626                         }
2627                         basefont = getLayoutFont(bparams, outerfont);
2628                         running_font = basefont;
2629                         column += Changes::latexMarkChange(os, bparams, runningChange,
2630                                                            change, runparams);
2631                         runningChange = change;
2632                 }
2633
2634                 // do not output text which is marked deleted
2635                 // if change tracking output is disabled
2636                 if (!bparams.output_changes && change.deleted()) {
2637                         continue;
2638                 }
2639
2640                 ++column;
2641
2642                 // Fully instantiated font
2643                 Font const current_font = getFont(bparams, i, outerfont);
2644
2645                 Font const last_font = running_font;
2646
2647                 // Do we need to close the previous font?
2648                 if (open_font &&
2649                     (current_font != running_font ||
2650                      current_font.language() != running_font.language()))
2651                 {
2652                         bool needPar = false;
2653                         column += running_font.latexWriteEndChanges(
2654                                     os, bparams, runparams, basefont,
2655                                     (i == body_pos-1) ? basefont : current_font,
2656                                     needPar);
2657                         running_font = basefont;
2658                         open_font = false;
2659                 }
2660
2661                 string const running_lang = runparams.use_polyglossia ?
2662                         running_font.language()->polyglossia() : running_font.language()->babel();
2663                 // close babel's font environment before opening CJK.
2664                 string const lang_end_command = runparams.use_polyglossia ?
2665                         "\\end{$$lang}" : lyxrc.language_command_end;
2666                 bool const using_begin_end = runparams.use_polyglossia ||
2667                                                 !lang_end_command.empty();
2668                 if (!running_lang.empty() &&
2669                     (!using_begin_end || running_lang == openLanguageName()) &&
2670                     current_font.language()->encoding()->package() == Encoding::CJK) {
2671                                 string end_tag = subst(lang_end_command,
2672                                                         "$$lang",
2673                                                         running_lang);
2674                                 os << from_ascii(end_tag);
2675                                 column += end_tag.length();
2676                                 if (using_begin_end)
2677                                         popLanguageName();
2678                 }
2679
2680                 // Switch file encoding if necessary (and allowed)
2681                 if (!runparams.pass_thru && !style.pass_thru &&
2682                     runparams.encoding->package() != Encoding::none &&
2683                     current_font.language()->encoding()->package() != Encoding::none) {
2684                         pair<bool, int> const enc_switch =
2685                                 switchEncoding(os.os(), bparams, runparams,
2686                                         *(current_font.language()->encoding()));
2687                         if (enc_switch.first) {
2688                                 column += enc_switch.second;
2689                                 runparams.encoding = current_font.language()->encoding();
2690                         }
2691                 }
2692
2693                 char_type const c = d->text_[i];
2694
2695                 // A display math inset inside an ulem command will be output
2696                 // as a box of width \linewidth, so we have to either disable
2697                 // indentation if the inset starts a paragraph, or start a new
2698                 // line to accommodate such box. This has to be done before
2699                 // writing any font changing commands.
2700                 if (runparams.inDisplayMath && !deleted_display_math
2701                     && runparams.inulemcmd) {
2702                         if (os.afterParbreak())
2703                                 os << "\\noindent";
2704                         else
2705                                 os << "\\\\\n";
2706                 }
2707
2708                 // Do we need to change font?
2709                 if ((current_font != running_font ||
2710                      current_font.language() != running_font.language()) &&
2711                         i != body_pos - 1)
2712                 {
2713                         odocstringstream ods;
2714                         column += current_font.latexWriteStartChanges(ods, bparams,
2715                                                               runparams, basefont,
2716                                                               last_font);
2717                         // Check again for display math in ulem commands as a
2718                         // font change may also occur just before a math inset.
2719                         if (runparams.inDisplayMath && !deleted_display_math
2720                             && runparams.inulemcmd) {
2721                                 if (os.afterParbreak())
2722                                         os << "\\noindent";
2723                                 else
2724                                         os << "\\\\\n";
2725                         }
2726                         running_font = current_font;
2727                         open_font = true;
2728                         docstring fontchange = ods.str();
2729                         // check whether the fontchange ends with a \\textcolor
2730                         // modifier and the text starts with a space (bug 4473)
2731                         docstring const last_modifier = rsplit(fontchange, '\\');
2732                         if (prefixIs(last_modifier, from_ascii("textcolor")) && c == ' ')
2733                                 os << fontchange << from_ascii("{}");
2734                         // check if the fontchange ends with a trailing blank
2735                         // (like "\small " (see bug 3382)
2736                         else if (suffixIs(fontchange, ' ') && c == ' ')
2737                                 os << fontchange.substr(0, fontchange.size() - 1)
2738                                    << from_ascii("{}");
2739                         else
2740                                 os << fontchange;
2741                 }
2742
2743                 // FIXME: think about end_pos implementation...
2744                 if (c == ' ' && i >= start_pos && (end_pos == -1 || i < end_pos)) {
2745                         // FIXME: integrate this case in latexSpecialChar
2746                         // Do not print the separation of the optional argument
2747                         // if style.pass_thru is false. This works because
2748                         // latexSpecialChar ignores spaces if
2749                         // style.pass_thru is false.
2750                         if (i != body_pos - 1) {
2751                                 if (d->simpleTeXBlanks(bparams, runparams, os,
2752                                                 i, column, current_font, style)) {
2753                                         // A surrogate pair was output. We
2754                                         // must not call latexSpecialChar
2755                                         // in this iteration, since it would output
2756                                         // the combining character again.
2757                                         ++i;
2758                                         continue;
2759                                 }
2760                         }
2761                 }
2762
2763                 OutputParams rp = runparams;
2764                 rp.free_spacing = style.free_spacing;
2765                 rp.local_font = &current_font;
2766                 rp.intitle = style.intitle;
2767
2768                 // Two major modes:  LaTeX or plain
2769                 // Handle here those cases common to both modes
2770                 // and then split to handle the two modes separately.
2771                 if (c == META_INSET) {
2772                         if (i >= start_pos && (end_pos == -1 || i < end_pos)) {
2773                                 // Greyedout notes and, in general, all insets
2774                                 // with InsetLayout::isDisplay() == false,
2775                                 // are typeset inline with the text. So, we
2776                                 // can add a \par to the last paragraph of
2777                                 // such insets only if nothing else follows.
2778                                 bool incremented = false;
2779                                 Inset const * inset = getInset(i);
2780                                 InsetText const * textinset = inset
2781                                                         ? inset->asInsetText()
2782                                                         : 0;
2783                                 if (i + 1 == size() && textinset
2784                                     && !inset->getLayout().isDisplay()) {
2785                                         ParagraphList const & pars =
2786                                                 textinset->text().paragraphs();
2787                                         pit_type const pit = pars.size() - 1;
2788                                         Font const lastfont =
2789                                                 pit < 0 || pars[pit].empty()
2790                                                 ? pars[pit].getLayoutFont(
2791                                                                 bparams,
2792                                                                 outerfont)
2793                                                 : pars[pit].getFont(bparams,
2794                                                         pars[pit].size() - 1,
2795                                                         outerfont);
2796                                         if (lastfont.fontInfo().size() !=
2797                                             basefont.fontInfo().size()) {
2798                                                 ++parInline;
2799                                                 incremented = true;
2800                                         }
2801                                 }
2802                                 d->latexInset(bparams, os, rp, running_font,
2803                                                 basefont, outerfont, open_font,
2804                                                 runningChange, style, i, column);
2805                                 if (incremented)
2806                                         --parInline;
2807                         }
2808                 } else if (i >= start_pos && (end_pos == -1 || i < end_pos)) {
2809                         try {
2810                                 d->latexSpecialChar(os, bparams, rp,
2811                                                     running_font, runningChange,
2812                                                     style, i, end_pos, column);
2813                         } catch (EncodingException & e) {
2814                                 if (runparams.dryrun) {
2815                                         os << "<" << _("LyX Warning: ")
2816                                            << _("uncodable character") << " '";
2817                                         os.put(c);
2818                                         os << "'>";
2819                                 } else {
2820                                         // add location information and throw again.
2821                                         e.par_id = id();
2822                                         e.pos = i;
2823                                         throw(e);
2824                                 }
2825                         }
2826                 }
2827
2828                 // Set the encoding to that returned from latexSpecialChar (see
2829                 // comment for encoding member in OutputParams.h)
2830                 runparams.encoding = rp.encoding;
2831
2832                 // Also carry on the info on a closed ulem command for insets
2833                 // such as Note that do not produce any output, so that no
2834                 // command is ever executed but its opening was recorded.
2835                 runparams.inulemcmd = rp.inulemcmd;
2836         }
2837
2838         // If we have an open font definition, we have to close it
2839         if (open_font) {
2840                 // Make sure that \\par is done with the font of the last
2841                 // character if this has another size as the default.
2842                 // This is necessary because LaTeX (and LyX on the screen)
2843                 // calculates the space between the baselines according
2844                 // to this font. (Matthias)
2845                 //
2846                 // We must not change the font for the last paragraph
2847                 // of non-multipar insets, tabular cells or commands,
2848                 // since this produces unwanted whitespace.
2849
2850                 Font const font = empty()
2851                         ? getLayoutFont(bparams, outerfont)
2852                         : getFont(bparams, size() - 1, outerfont);
2853
2854                 InsetText const * textinset = inInset().asInsetText();
2855
2856                 bool const maintext = textinset
2857                         ? textinset->text().isMainText()
2858                         : false;
2859
2860                 size_t const numpars = textinset
2861                         ? textinset->text().paragraphs().size()
2862                         : 0;
2863
2864                 bool needPar = false;
2865
2866                 if (style.resfont.size() != font.fontInfo().size()
2867                     && (!runparams.isLastPar || maintext
2868                         || (numpars > 1 && d->ownerCode() != CELL_CODE
2869                             && (inInset().getLayout().isDisplay()
2870                                 || parInline)))
2871                     && !style.isCommand()) {
2872                         needPar = true;
2873                 }
2874 #ifdef FIXED_LANGUAGE_END_DETECTION
2875                 if (next_) {
2876                         running_font.latexWriteEndChanges(os, bparams,
2877                                         runparams, basefont,
2878                                         next_->getFont(bparams, 0, outerfont),
2879                                                        needPar);
2880                 } else {
2881                         running_font.latexWriteEndChanges(os, bparams,
2882                                         runparams, basefont, basefont, needPar);
2883                 }
2884 #else
2885 //FIXME: For now we ALWAYS have to close the foreign font settings if they are
2886 //FIXME: there as we start another \selectlanguage with the next paragraph if
2887 //FIXME: we are in need of this. This should be fixed sometime (Jug)
2888                 running_font.latexWriteEndChanges(os, bparams, runparams,
2889                                 basefont, basefont, needPar);
2890 #endif
2891                 if (needPar) {
2892                         // The \par could not be inserted at the same nesting
2893                         // level of the font size change, so do it now.
2894                         os << "{\\" << font.latexSize() << "\\par}";
2895                 }
2896         }
2897
2898         column += Changes::latexMarkChange(os, bparams, runningChange,
2899                                            Change(Change::UNCHANGED), runparams);
2900
2901         // Needed if there is an optional argument but no contents.
2902         if (body_pos > 0 && body_pos == size()) {
2903                 os << "}]~";
2904         }
2905
2906         if (!style.rightdelim().empty()) {
2907                 os << style.rightdelim();
2908                 column += style.rightdelim().size();
2909         }
2910
2911         if (allowcust && d->endTeXParParams(bparams, os, runparams)
2912             && runparams.encoding != prev_encoding) {
2913                 runparams.encoding = prev_encoding;
2914                 os << setEncoding(prev_encoding->iconvName());
2915         }
2916
2917         LYXERR(Debug::LATEX, "Paragraph::latex... done " << this);
2918 }
2919
2920
2921 bool Paragraph::emptyTag() const
2922 {
2923         for (pos_type i = 0; i < size(); ++i) {
2924                 if (Inset const * inset = getInset(i)) {
2925                         InsetCode lyx_code = inset->lyxCode();
2926                         // FIXME testing like that is wrong. What is
2927                         // the intent?
2928                         if (lyx_code != TOC_CODE &&
2929                             lyx_code != INCLUDE_CODE &&
2930                             lyx_code != GRAPHICS_CODE &&
2931                             lyx_code != ERT_CODE &&
2932                             lyx_code != LISTINGS_CODE &&
2933                             lyx_code != FLOAT_CODE &&
2934                             lyx_code != TABULAR_CODE) {
2935                                 return false;
2936                         }
2937                 } else {
2938                         char_type c = d->text_[i];
2939                         if (c != ' ' && c != '\t')
2940                                 return false;
2941                 }
2942         }
2943         return true;
2944 }
2945
2946
2947 string Paragraph::getID(Buffer const & buf, OutputParams const & runparams)
2948         const
2949 {
2950         for (pos_type i = 0; i < size(); ++i) {
2951                 if (Inset const * inset = getInset(i)) {
2952                         InsetCode lyx_code = inset->lyxCode();
2953                         if (lyx_code == LABEL_CODE) {
2954                                 InsetLabel const * const il = static_cast<InsetLabel const *>(inset);
2955                                 docstring const & id = il->getParam("name");
2956                                 return "id='" + to_utf8(sgml::cleanID(buf, runparams, id)) + "'";
2957                         }
2958                 }
2959         }
2960         return string();
2961 }
2962
2963
2964 pos_type Paragraph::firstWordDocBook(odocstream & os, OutputParams const & runparams)
2965         const
2966 {
2967         pos_type i;
2968         for (i = 0; i < size(); ++i) {
2969                 if (Inset const * inset = getInset(i)) {
2970                         inset->docbook(os, runparams);
2971                 } else {
2972                         char_type c = d->text_[i];
2973                         if (c == ' ')
2974                                 break;
2975                         os << sgml::escapeChar(c);
2976                 }
2977         }
2978         return i;
2979 }
2980
2981
2982 pos_type Paragraph::firstWordLyXHTML(XHTMLStream & xs, OutputParams const & runparams)
2983         const
2984 {
2985         pos_type i;
2986         for (i = 0; i < size(); ++i) {
2987                 if (Inset const * inset = getInset(i)) {
2988                         inset->xhtml(xs, runparams);
2989                 } else {
2990                         char_type c = d->text_[i];
2991                         if (c == ' ')
2992                                 break;
2993                         xs << c;
2994                 }
2995         }
2996         return i;
2997 }
2998
2999
3000 bool Paragraph::Private::onlyText(Buffer const & buf, Font const & outerfont, pos_type initial) const
3001 {
3002         Font font_old;
3003         pos_type size = text_.size();
3004         for (pos_type i = initial; i < size; ++i) {
3005                 Font font = owner_->getFont(buf.params(), i, outerfont);
3006                 if (text_[i] == META_INSET)
3007                         return false;
3008                 if (i != initial && font != font_old)
3009                         return false;
3010                 font_old = font;
3011         }
3012
3013         return true;
3014 }
3015
3016
3017 void Paragraph::simpleDocBookOnePar(Buffer const & buf,
3018                                     odocstream & os,
3019                                     OutputParams const & runparams,
3020                                     Font const & outerfont,
3021                                     pos_type initial) const
3022 {
3023         bool emph_flag = false;
3024
3025         Layout const & style = *d->layout_;
3026         FontInfo font_old =
3027                 style.labeltype == LABEL_MANUAL ? style.labelfont : style.font;
3028
3029         if (style.pass_thru && !d->onlyText(buf, outerfont, initial))
3030                 os << "]]>";
3031
3032         // parsing main loop
3033         for (pos_type i = initial; i < size(); ++i) {
3034                 Font font = getFont(buf.params(), i, outerfont);
3035
3036                 // handle <emphasis> tag
3037                 if (font_old.emph() != font.fontInfo().emph()) {
3038                         if (font.fontInfo().emph() == FONT_ON) {
3039                                 os << "<emphasis>";
3040                                 emph_flag = true;
3041                         } else if (i != initial) {
3042                                 os << "</emphasis>";
3043                                 emph_flag = false;
3044                         }
3045                 }
3046
3047                 if (Inset const * inset = getInset(i)) {
3048                         inset->docbook(os, runparams);
3049                 } else {
3050                         char_type c = d->text_[i];
3051
3052                         if (style.pass_thru)
3053                                 os.put(c);
3054                         else
3055                                 os << sgml::escapeChar(c);
3056                 }
3057                 font_old = font.fontInfo();
3058         }
3059
3060         if (emph_flag) {
3061                 os << "</emphasis>";
3062         }
3063
3064         if (style.free_spacing)
3065                 os << '\n';
3066         if (style.pass_thru && !d->onlyText(buf, outerfont, initial))
3067                 os << "<![CDATA[";
3068 }
3069
3070
3071 namespace {
3072 void doFontSwitch(vector<html::FontTag> & tagsToOpen,
3073                   vector<html::EndFontTag> & tagsToClose,
3074                   bool & flag, FontState curstate, html::FontTypes type)
3075 {
3076         if (curstate == FONT_ON) {
3077                 tagsToOpen.push_back(html::FontTag(type));
3078                 flag = true;
3079         } else if (flag) {
3080                 tagsToClose.push_back(html::EndFontTag(type));
3081                 flag = false;
3082         }
3083 }
3084 } // namespace
3085
3086
3087 docstring Paragraph::simpleLyXHTMLOnePar(Buffer const & buf,
3088                                     XHTMLStream & xs,
3089                                     OutputParams const & runparams,
3090                                     Font const & outerfont,
3091                                     bool start_paragraph, bool close_paragraph,
3092                                     pos_type initial) const
3093 {
3094         docstring retval;
3095
3096         // track whether we have opened these tags
3097         bool emph_flag = false;
3098         bool bold_flag = false;
3099         bool noun_flag = false;
3100         bool ubar_flag = false;
3101         bool dbar_flag = false;
3102         bool sout_flag = false;
3103         bool xout_flag = false;
3104         bool wave_flag = false;
3105         // shape tags
3106         bool shap_flag = false;
3107         // family tags
3108         bool faml_flag = false;
3109         // size tags
3110         bool size_flag = false;
3111
3112         Layout const & style = *d->layout_;
3113
3114         if (start_paragraph)
3115                 xs.startDivision(allowEmpty());
3116
3117         FontInfo font_old =
3118                 style.labeltype == LABEL_MANUAL ? style.labelfont : style.font;
3119
3120         FontShape  curr_fs   = INHERIT_SHAPE;
3121         FontFamily curr_fam  = INHERIT_FAMILY;
3122         FontSize   curr_size = FONT_SIZE_INHERIT;
3123
3124         string const default_family =
3125                 buf.masterBuffer()->params().fonts_default_family;
3126
3127         vector<html::FontTag> tagsToOpen;
3128         vector<html::EndFontTag> tagsToClose;
3129
3130         // parsing main loop
3131         for (pos_type i = initial; i < size(); ++i) {
3132                 // let's not show deleted material in the output
3133                 if (isDeleted(i))
3134                         continue;
3135
3136                 Font const font = getFont(buf.masterBuffer()->params(), i, outerfont);
3137
3138                 // emphasis
3139                 FontState curstate = font.fontInfo().emph();
3140                 if (font_old.emph() != curstate)
3141                         doFontSwitch(tagsToOpen, tagsToClose, emph_flag, curstate, html::FT_EMPH);
3142
3143                 // noun
3144                 curstate = font.fontInfo().noun();
3145                 if (font_old.noun() != curstate)
3146                         doFontSwitch(tagsToOpen, tagsToClose, noun_flag, curstate, html::FT_NOUN);
3147
3148                 // underbar
3149                 curstate = font.fontInfo().underbar();
3150                 if (font_old.underbar() != curstate)
3151                         doFontSwitch(tagsToOpen, tagsToClose, ubar_flag, curstate, html::FT_UBAR);
3152
3153                 // strikeout
3154                 curstate = font.fontInfo().strikeout();
3155                 if (font_old.strikeout() != curstate)
3156                         doFontSwitch(tagsToOpen, tagsToClose, sout_flag, curstate, html::FT_SOUT);
3157
3158                 // xout
3159                 curstate = font.fontInfo().xout();
3160                 if (font_old.xout() != curstate)
3161                         doFontSwitch(tagsToOpen, tagsToClose, xout_flag, curstate, html::FT_XOUT);
3162
3163                 // double underbar
3164                 curstate = font.fontInfo().uuline();
3165                 if (font_old.uuline() != curstate)
3166                         doFontSwitch(tagsToOpen, tagsToClose, dbar_flag, curstate, html::FT_DBAR);
3167
3168                 // wavy line
3169                 curstate = font.fontInfo().uwave();
3170                 if (font_old.uwave() != curstate)
3171                         doFontSwitch(tagsToOpen, tagsToClose, wave_flag, curstate, html::FT_WAVE);
3172
3173                 // bold
3174                 // a little hackish, but allows us to reuse what we have.
3175                 curstate = (font.fontInfo().series() == BOLD_SERIES ? FONT_ON : FONT_OFF);
3176                 if (font_old.series() != font.fontInfo().series())
3177                         doFontSwitch(tagsToOpen, tagsToClose, bold_flag, curstate, html::FT_BOLD);
3178
3179                 // Font shape
3180                 curr_fs = font.fontInfo().shape();
3181                 FontShape old_fs = font_old.shape();
3182                 if (old_fs != curr_fs) {
3183                         if (shap_flag) {
3184                                 switch (old_fs) {
3185                                 case ITALIC_SHAPE:
3186                                         tagsToClose.push_back(html::EndFontTag(html::FT_ITALIC));
3187                                         break;
3188                                 case SLANTED_SHAPE:
3189                                         tagsToClose.push_back(html::EndFontTag(html::FT_SLANTED));
3190                                         break;
3191                                 case SMALLCAPS_SHAPE:
3192                                         tagsToClose.push_back(html::EndFontTag(html::FT_SMALLCAPS));
3193                                         break;
3194                                 case UP_SHAPE:
3195                                 case INHERIT_SHAPE:
3196                                         break;
3197                                 default:
3198                                         // the other tags are for internal use
3199                                         LATTEST(false);
3200                                         break;
3201                                 }
3202                                 shap_flag = false;
3203                         }
3204                         switch (curr_fs) {
3205                         case ITALIC_SHAPE:
3206                                 tagsToOpen.push_back(html::FontTag(html::FT_ITALIC));
3207                                 shap_flag = true;
3208                                 break;
3209                         case SLANTED_SHAPE:
3210                                 tagsToOpen.push_back(html::FontTag(html::FT_SLANTED));
3211                                 shap_flag = true;
3212                                 break;
3213                         case SMALLCAPS_SHAPE:
3214                                 tagsToOpen.push_back(html::FontTag(html::FT_SMALLCAPS));
3215                                 shap_flag = true;
3216                                 break;
3217                         case UP_SHAPE:
3218                         case INHERIT_SHAPE:
3219                                 break;
3220                         default:
3221                                 // the other tags are for internal use
3222                                 LATTEST(false);
3223                                 break;
3224                         }
3225                 }
3226
3227                 // Font family
3228                 curr_fam = font.fontInfo().family();
3229                 FontFamily old_fam = font_old.family();
3230                 if (old_fam != curr_fam) {
3231                         if (faml_flag) {
3232                                 switch (old_fam) {
3233                                 case ROMAN_FAMILY:
3234                                         tagsToClose.push_back(html::EndFontTag(html::FT_ROMAN));
3235                                         break;
3236                                 case SANS_FAMILY:
3237                                         tagsToClose.push_back(html::EndFontTag(html::FT_SANS));
3238                                         break;
3239                                 case TYPEWRITER_FAMILY:
3240                                         tagsToClose.push_back(html::EndFontTag(html::FT_TYPE));
3241                                         break;
3242                                 case INHERIT_FAMILY:
3243                                         break;
3244                                 default:
3245                                         // the other tags are for internal use
3246                                         LATTEST(false);
3247                                         break;
3248                                 }
3249                                 faml_flag = false;
3250                         }
3251                         switch (curr_fam) {
3252                         case ROMAN_FAMILY:
3253                                 // we will treat a "default" font family as roman, since we have
3254                                 // no other idea what to do.
3255                                 if (default_family != "rmdefault" && default_family != "default") {
3256                                         tagsToOpen.push_back(html::FontTag(html::FT_ROMAN));
3257                                         faml_flag = true;
3258                                 }
3259                                 break;
3260                         case SANS_FAMILY:
3261                                 if (default_family != "sfdefault") {
3262                                         tagsToOpen.push_back(html::FontTag(html::FT_SANS));
3263                                         faml_flag = true;
3264                                 }
3265                                 break;
3266                         case TYPEWRITER_FAMILY:
3267                                 if (default_family != "ttdefault") {
3268                                         tagsToOpen.push_back(html::FontTag(html::FT_TYPE));
3269                                         faml_flag = true;
3270                                 }
3271                                 break;
3272                         case INHERIT_FAMILY:
3273                                 break;
3274                         default:
3275                                 // the other tags are for internal use
3276                                 LATTEST(false);
3277                                 break;
3278                         }
3279                 }
3280
3281                 // Font size
3282                 curr_size = font.fontInfo().size();
3283                 FontSize old_size = font_old.size();
3284                 if (old_size != curr_size) {
3285                         if (size_flag) {
3286                                 switch (old_size) {
3287                                 case FONT_SIZE_TINY:
3288                                         tagsToClose.push_back(html::EndFontTag(html::FT_SIZE_TINY));
3289                                         break;
3290                                 case FONT_SIZE_SCRIPT:
3291                                         tagsToClose.push_back(html::EndFontTag(html::FT_SIZE_SCRIPT));
3292                                         break;
3293                                 case FONT_SIZE_FOOTNOTE:
3294                                         tagsToClose.push_back(html::EndFontTag(html::FT_SIZE_FOOTNOTE));
3295                                         break;
3296                                 case FONT_SIZE_SMALL:
3297                                         tagsToClose.push_back(html::EndFontTag(html::FT_SIZE_SMALL));
3298                                         break;
3299                                 case FONT_SIZE_LARGE:
3300                                         tagsToClose.push_back(html::EndFontTag(html::FT_SIZE_LARGE));
3301                                         break;
3302                                 case FONT_SIZE_LARGER:
3303                                         tagsToClose.push_back(html::EndFontTag(html::FT_SIZE_LARGER));
3304                                         break;
3305                                 case FONT_SIZE_LARGEST:
3306                                         tagsToClose.push_back(html::EndFontTag(html::FT_SIZE_LARGEST));
3307                                         break;
3308                                 case FONT_SIZE_HUGE:
3309                                         tagsToClose.push_back(html::EndFontTag(html::FT_SIZE_HUGE));
3310                                         break;
3311                                 case FONT_SIZE_HUGER:
3312                                         tagsToClose.push_back(html::EndFontTag(html::FT_SIZE_HUGER));
3313                                         break;
3314                                 case FONT_SIZE_INCREASE:
3315                                         tagsToClose.push_back(html::EndFontTag(html::FT_SIZE_INCREASE));
3316                                         break;
3317                                 case FONT_SIZE_DECREASE:
3318                                         tagsToClose.push_back(html::EndFontTag(html::FT_SIZE_DECREASE));
3319                                         break;
3320                                 case FONT_SIZE_INHERIT:
3321                                 case FONT_SIZE_NORMAL:
3322                                         break;
3323                                 default:
3324                                         // the other tags are for internal use
3325                                         LATTEST(false);
3326                                         break;
3327                                 }
3328                                 size_flag = false;
3329                         }
3330                         switch (curr_size) {
3331                         case FONT_SIZE_TINY:
3332                                 tagsToOpen.push_back(html::FontTag(html::FT_SIZE_TINY));
3333                                 size_flag = true;
3334                                 break;
3335                         case FONT_SIZE_SCRIPT:
3336                                 tagsToOpen.push_back(html::FontTag(html::FT_SIZE_SCRIPT));
3337                                 size_flag = true;
3338                                 break;
3339                         case FONT_SIZE_FOOTNOTE:
3340                                 tagsToOpen.push_back(html::FontTag(html::FT_SIZE_FOOTNOTE));
3341                                 size_flag = true;
3342                                 break;
3343                         case FONT_SIZE_SMALL:
3344                                 tagsToOpen.push_back(html::FontTag(html::FT_SIZE_SMALL));
3345                                 size_flag = true;
3346                                 break;
3347                         case FONT_SIZE_LARGE:
3348                                 tagsToOpen.push_back(html::FontTag(html::FT_SIZE_LARGE));
3349                                 size_flag = true;
3350                                 break;
3351                         case FONT_SIZE_LARGER:
3352                                 tagsToOpen.push_back(html::FontTag(html::FT_SIZE_LARGER));
3353                                 size_flag = true;
3354                                 break;
3355                         case FONT_SIZE_LARGEST:
3356                                 tagsToOpen.push_back(html::FontTag(html::FT_SIZE_LARGEST));
3357                                 size_flag = true;
3358                                 break;
3359                         case FONT_SIZE_HUGE:
3360                                 tagsToOpen.push_back(html::FontTag(html::FT_SIZE_HUGE));
3361                                 size_flag = true;
3362                                 break;
3363                         case FONT_SIZE_HUGER:
3364                                 tagsToOpen.push_back(html::FontTag(html::FT_SIZE_HUGER));
3365                                 size_flag = true;
3366                                 break;
3367                         case FONT_SIZE_INCREASE:
3368                                 tagsToOpen.push_back(html::FontTag(html::FT_SIZE_INCREASE));
3369                                 size_flag = true;
3370                                 break;
3371                         case FONT_SIZE_DECREASE:
3372                                 tagsToOpen.push_back(html::FontTag(html::FT_SIZE_DECREASE));
3373                                 size_flag = true;
3374                                 break;
3375                         case FONT_SIZE_NORMAL:
3376                         case FONT_SIZE_INHERIT:
3377                                 break;
3378                         default:
3379                                 // the other tags are for internal use
3380                                 LATTEST(false);
3381                                 break;
3382                         }
3383                 }
3384
3385                 // FIXME XHTML
3386                 // Other such tags? What about the other text ranges?
3387
3388                 vector<html::EndFontTag>::const_iterator cit = tagsToClose.begin();
3389                 vector<html::EndFontTag>::const_iterator cen = tagsToClose.end();
3390                 for (; cit != cen; ++cit)
3391                         xs << *cit;
3392
3393                 vector<html::FontTag>::const_iterator sit = tagsToOpen.begin();
3394                 vector<html::FontTag>::const_iterator sen = tagsToOpen.end();
3395                 for (; sit != sen; ++sit)
3396                         xs << *sit;
3397
3398                 tagsToClose.clear();
3399                 tagsToOpen.clear();
3400
3401                 Inset const * inset = getInset(i);
3402                 if (inset) {
3403                         if (!runparams.for_toc || inset->isInToc()) {
3404                                 OutputParams np = runparams;
3405                                 np.local_font = &font;
3406                                 // If the paragraph has size 1, then we are in the "special
3407                                 // case" where we do not output the containing paragraph info
3408                                 if (!inset->getLayout().htmlisblock() && size() != 1)
3409                                         np.html_in_par = true;
3410                                 retval += inset->xhtml(xs, np);
3411                         }
3412                 } else {
3413                         char_type c = getUChar(buf.masterBuffer()->params(),
3414                                                runparams, i);
3415                         xs << c;
3416                 }
3417                 font_old = font.fontInfo();
3418         }
3419
3420         // FIXME XHTML
3421         // I'm worried about what happens if a branch, say, is itself
3422         // wrapped in some font stuff. I think that will not work.
3423         xs.closeFontTags();
3424         if (close_paragraph)
3425                 xs.endDivision();
3426
3427         return retval;
3428 }
3429
3430
3431 bool Paragraph::isHfill(pos_type pos) const
3432 {
3433         Inset const * inset = getInset(pos);
3434         return inset && inset->isHfill();
3435 }
3436
3437
3438 bool Paragraph::isNewline(pos_type pos) const
3439 {
3440         // U+2028 LINE SEPARATOR
3441         // U+2029 PARAGRAPH SEPARATOR
3442         char_type const c = d->text_[pos];
3443         if (c == 0x2028 || c == 0x2029)
3444                 return true;
3445         Inset const * inset = getInset(pos);
3446         return inset && inset->lyxCode() == NEWLINE_CODE;
3447 }
3448
3449
3450 bool Paragraph::isEnvSeparator(pos_type pos) const
3451 {
3452         Inset const * inset = getInset(pos);
3453         return inset && inset->lyxCode() == SEPARATOR_CODE;
3454 }
3455
3456
3457 bool Paragraph::isLineSeparator(pos_type pos) const
3458 {
3459         char_type const c = d->text_[pos];
3460         if (isLineSeparatorChar(c))
3461                 return true;
3462         Inset const * inset = getInset(pos);
3463         return inset && inset->isLineSeparator();
3464 }
3465
3466
3467 bool Paragraph::isWordSeparator(pos_type pos) const
3468 {
3469         if (pos == size())
3470                 return true;
3471         if (Inset const * inset = getInset(pos))
3472                 return !inset->isLetter();
3473         // if we have a hard hyphen (no en- or emdash) or apostrophe
3474         // we pass this to the spell checker
3475         // FIXME: this method is subject to change, visit
3476         // https://bugzilla.mozilla.org/show_bug.cgi?id=355178
3477         // to get an impression how complex this is.
3478         if (isHardHyphenOrApostrophe(pos))
3479                 return false;
3480         char_type const c = d->text_[pos];
3481         // We want to pass the escape chars to the spellchecker
3482         docstring const escape_chars = from_utf8(lyxrc.spellchecker_esc_chars);
3483         return !isLetterChar(c) && !isDigitASCII(c) && !contains(escape_chars, c);
3484 }
3485
3486
3487 bool Paragraph::isHardHyphenOrApostrophe(pos_type pos) const
3488 {
3489         pos_type const psize = size();
3490         if (pos >= psize)
3491                 return false;
3492         char_type const c = d->text_[pos];
3493         if (c != '-' && c != '\'')
3494                 return false;
3495         int nextpos = pos + 1;
3496         int prevpos = pos > 0 ? pos - 1 : 0;
3497         if ((nextpos == psize || isSpace(nextpos))
3498                 && (pos == 0 || isSpace(prevpos)))
3499                 return false;
3500         return true;
3501 }
3502
3503
3504 bool Paragraph::needsCProtection(bool const fragile) const
3505 {
3506         // first check the layout of the paragraph, but only in insets
3507         InsetText const * textinset = inInset().asInsetText();
3508         bool const maintext = textinset
3509                 ? textinset->text().isMainText()
3510                 : false;
3511
3512         if (!maintext && layout().needcprotect) {
3513                 // Environments need cprotection regardless the content
3514                 if (layout().latextype == LATEX_ENVIRONMENT)
3515                         return true;
3516
3517                 // Commands need cprotection if they contain specific chars
3518                 int const nchars_escape = 9;
3519                 static char_type const chars_escape[nchars_escape] = {
3520                         '&', '_', '$', '%', '#', '^', '{', '}', '\\'};
3521
3522                 docstring const pars = asString();
3523                 for (int k = 0; k < nchars_escape; k++) {
3524                         if (contains(pars, chars_escape[k]))
3525                                 return true;
3526                 }
3527         }
3528
3529         // now check whether we have insets that need cprotection
3530         pos_type size = d->text_.size();
3531         for (pos_type i = 0; i < size; ++i)
3532                 if (isInset(i) && getInset(i)->needsCProtection(maintext, fragile))
3533                         return true;
3534
3535         return false;
3536 }
3537
3538
3539 FontSpan const & Paragraph::getSpellRange(pos_type pos) const
3540 {
3541         return d->speller_state_.getRange(pos);
3542 }
3543
3544
3545 bool Paragraph::isChar(pos_type pos) const
3546 {
3547         if (Inset const * inset = getInset(pos))
3548                 return inset->isChar();
3549         char_type const c = d->text_[pos];
3550         return !isLetterChar(c) && !isDigitASCII(c) && !lyx::isSpace(c);
3551 }
3552
3553
3554 bool Paragraph::isSpace(pos_type pos) const
3555 {
3556         if (Inset const * inset = getInset(pos))
3557                 return inset->isSpace();
3558         char_type const c = d->text_[pos];
3559         return lyx::isSpace(c);
3560 }
3561
3562
3563 Language const *
3564 Paragraph::getParLanguage(BufferParams const & bparams) const
3565 {
3566         if (!empty())
3567                 return getFirstFontSettings(bparams).language();
3568         // FIXME: we should check the prev par as well (Lgb)
3569         return bparams.language;
3570 }
3571
3572
3573 bool Paragraph::isRTL(BufferParams const & bparams) const
3574 {
3575         return getParLanguage(bparams)->rightToLeft()
3576                 && !inInset().getLayout().forceLTR();
3577 }
3578
3579
3580 void Paragraph::changeLanguage(BufferParams const & bparams,
3581                                Language const * from, Language const * to)
3582 {
3583         // change language including dummy font change at the end
3584         for (pos_type i = 0; i <= size(); ++i) {
3585                 Font font = getFontSettings(bparams, i);
3586                 if (font.language() == from) {
3587                         font.setLanguage(to);
3588                         setFont(i, font);
3589                         d->requestSpellCheck(i);
3590                 }
3591         }
3592 }
3593
3594
3595 bool Paragraph::isMultiLingual(BufferParams const & bparams) const
3596 {
3597         Language const * doc_language = bparams.language;
3598         for (auto const & f : d->fontlist_)
3599                 if (f.font().language() != ignore_language &&
3600                     f.font().language() != latex_language &&
3601                     f.font().language() != doc_language)
3602                         return true;
3603         return false;
3604 }
3605
3606
3607 void Paragraph::getLanguages(std::set<Language const *> & langs) const
3608 {
3609         for (auto const & f : d->fontlist_) {
3610                 Language const * lang = f.font().language();
3611                 if (lang != ignore_language &&
3612                     lang != latex_language)
3613                         langs.insert(lang);
3614         }
3615 }
3616
3617
3618 docstring Paragraph::asString(int options) const
3619 {
3620         return asString(0, size(), options);
3621 }
3622
3623
3624 docstring Paragraph::asString(pos_type beg, pos_type end, int options, const OutputParams *runparams) const
3625 {
3626         odocstringstream os;
3627
3628         if (beg == 0
3629             && options & AS_STR_LABEL
3630             && !d->params_.labelString().empty())
3631                 os << d->params_.labelString() << ' ';
3632
3633         for (pos_type i = beg; i < end; ++i) {
3634                 if ((options & AS_STR_SKIPDELETE) && isDeleted(i))
3635                         continue;
3636                 char_type const c = d->text_[i];
3637                 if (isPrintable(c) || c == '\t'
3638                     || (c == '\n' && (options & AS_STR_NEWLINES)))
3639                         os.put(c);
3640                 else if (c == META_INSET && (options & AS_STR_INSETS)) {
3641                         if (c == META_INSET && (options & AS_STR_PLAINTEXT)) {
3642                                 LASSERT(runparams != 0, return docstring());
3643                                 getInset(i)->plaintext(os, *runparams);
3644                         } else {
3645                                 getInset(i)->toString(os);
3646                         }
3647                 }
3648         }
3649
3650         return os.str();
3651 }
3652
3653
3654 void Paragraph::forOutliner(docstring & os, size_t const maxlen,
3655                             bool const shorten, bool const label) const
3656 {
3657         size_t tmplen = shorten ? maxlen + 1 : maxlen;
3658         if (label && !labelString().empty())
3659                 os += labelString() + ' ';
3660         if (!layout().isTocCaption())
3661                 return;
3662         for (pos_type i = 0; i < size() && os.length() < tmplen; ++i) {
3663                 if (isDeleted(i))
3664                         continue;
3665                 char_type const c = d->text_[i];
3666                 if (isPrintable(c))
3667                         os += c;
3668                 else if (c == META_INSET)
3669                         getInset(i)->forOutliner(os, tmplen, false);
3670         }
3671         if (shorten)
3672                 Text::shortenForOutliner(os, maxlen);
3673 }
3674
3675
3676 void Paragraph::setInsetOwner(Inset const * inset)
3677 {
3678         d->inset_owner_ = inset;
3679 }
3680
3681
3682 int Paragraph::id() const
3683 {
3684         return d->id_;
3685 }
3686
3687
3688 void Paragraph::setId(int id)
3689 {
3690         d->id_ = id;
3691 }
3692
3693
3694 Layout const & Paragraph::layout() const
3695 {
3696         return *d->layout_;
3697 }
3698
3699
3700 void Paragraph::setLayout(Layout const & layout)
3701 {
3702         d->layout_ = &layout;
3703 }
3704
3705
3706 void Paragraph::setDefaultLayout(DocumentClass const & tc)
3707 {
3708         setLayout(tc.defaultLayout());
3709 }
3710
3711
3712 void Paragraph::setPlainLayout(DocumentClass const & tc)
3713 {
3714         setLayout(tc.plainLayout());
3715 }
3716
3717
3718 void Paragraph::setPlainOrDefaultLayout(DocumentClass const & tclass)
3719 {
3720         if (usePlainLayout())
3721                 setPlainLayout(tclass);
3722         else
3723                 setDefaultLayout(tclass);
3724 }
3725
3726
3727 Inset const & Paragraph::inInset() const
3728 {
3729         LBUFERR(d->inset_owner_);
3730         return *d->inset_owner_;
3731 }
3732
3733
3734 ParagraphParameters & Paragraph::params()
3735 {
3736         return d->params_;
3737 }
3738
3739
3740 ParagraphParameters const & Paragraph::params() const
3741 {
3742         return d->params_;
3743 }
3744
3745
3746 bool Paragraph::isFreeSpacing() const
3747 {
3748         if (d->layout_->free_spacing)
3749                 return true;
3750         return d->inset_owner_ && d->inset_owner_->isFreeSpacing();
3751 }
3752
3753
3754 bool Paragraph::allowEmpty() const
3755 {
3756         if (d->layout_->keepempty)
3757                 return true;
3758         return d->inset_owner_ && d->inset_owner_->allowEmpty();
3759 }
3760
3761
3762 bool Paragraph::brokenBiblio() const
3763 {
3764         // there is a problem if there is no bibitem at position 0 or
3765         // if there is another bibitem in the paragraph.
3766         return d->layout_->labeltype == LABEL_BIBLIO
3767                 && (d->insetlist_.find(BIBITEM_CODE) != 0
3768                     || d->insetlist_.find(BIBITEM_CODE, 1) > 0);
3769 }
3770
3771
3772 int Paragraph::fixBiblio(Buffer const & buffer)
3773 {
3774         // FIXME: What about the case where paragraph is not BIBLIO
3775         // but there is an InsetBibitem?
3776         // FIXME: when there was already an inset at 0, the return value is 1,
3777         // which does not tell whether another inset has been remove; the
3778         // cursor cannot be correctly updated.
3779
3780         if (d->layout_->labeltype != LABEL_BIBLIO)
3781                 return 0;
3782
3783         bool const track_changes = buffer.params().track_changes;
3784         int bibitem_pos = d->insetlist_.find(BIBITEM_CODE);
3785         bool const hasbibitem0 = bibitem_pos == 0;
3786
3787         if (hasbibitem0) {
3788                 bibitem_pos = d->insetlist_.find(BIBITEM_CODE, 1);
3789                 // There was an InsetBibitem at pos 0, and no other one => OK
3790                 if (bibitem_pos == -1)
3791                         return 0;
3792                 // there is a bibitem at the 0 position, but since
3793                 // there is a second one, we copy the second on the
3794                 // first. We're assuming there are at most two of
3795                 // these, which there should be.
3796                 // FIXME: why does it make sense to do that rather
3797                 // than keep the first? (JMarc)
3798                 Inset * inset = releaseInset(bibitem_pos);
3799                 d->insetlist_.begin()->inset = inset;
3800                 return -bibitem_pos;
3801         }
3802
3803         // We need to create an inset at the beginning
3804         Inset * inset = 0;
3805         if (bibitem_pos > 0) {
3806                 // there was one somewhere in the paragraph, let's move it
3807                 inset = d->insetlist_.release(bibitem_pos);
3808                 eraseChar(bibitem_pos, track_changes);
3809         } else
3810                 // make a fresh one
3811                 inset = new InsetBibitem(const_cast<Buffer *>(&buffer),
3812                                          InsetCommandParams(BIBITEM_CODE));
3813
3814         Font font(inherit_font, buffer.params().language);
3815         insertInset(0, inset, font, Change(track_changes ? Change::INSERTED
3816                                                    : Change::UNCHANGED));
3817
3818         return 1;
3819 }
3820
3821
3822 void Paragraph::checkAuthors(AuthorList const & authorList)
3823 {
3824         d->changes_.checkAuthors(authorList);
3825 }
3826
3827
3828 bool Paragraph::isChanged(pos_type pos) const
3829 {
3830         return lookupChange(pos).changed();
3831 }
3832
3833
3834 bool Paragraph::isInserted(pos_type pos) const
3835 {
3836         return lookupChange(pos).inserted();
3837 }
3838
3839
3840 bool Paragraph::isDeleted(pos_type pos) const
3841 {
3842         return lookupChange(pos).deleted();
3843 }
3844
3845
3846 InsetList const & Paragraph::insetList() const
3847 {
3848         return d->insetlist_;
3849 }
3850
3851
3852 void Paragraph::setInsetBuffers(Buffer & b)
3853 {
3854         d->insetlist_.setBuffer(b);
3855 }
3856
3857
3858 void Paragraph::resetBuffer()
3859 {
3860         d->insetlist_.resetBuffer();
3861 }
3862
3863
3864 Inset * Paragraph::releaseInset(pos_type pos)
3865 {
3866         Inset * inset = d->insetlist_.release(pos);
3867         /// does not honour change tracking!
3868         eraseChar(pos, false);
3869         return inset;
3870 }
3871
3872
3873 Inset * Paragraph::getInset(pos_type pos)
3874 {
3875         return (pos < pos_type(d->text_.size()) && d->text_[pos] == META_INSET)
3876                  ? d->insetlist_.get(pos) : 0;
3877 }
3878
3879
3880 Inset const * Paragraph::getInset(pos_type pos) const
3881 {
3882         return (pos < pos_type(d->text_.size()) && d->text_[pos] == META_INSET)
3883                  ? d->insetlist_.get(pos) : 0;
3884 }
3885
3886
3887 void Paragraph::changeCase(BufferParams const & bparams, pos_type pos,
3888                 pos_type & right, TextCase action)
3889 {
3890         // process sequences of modified characters; in change
3891         // tracking mode, this approach results in much better
3892         // usability than changing case on a char-by-char basis
3893         // We also need to track the current font, since font
3894         // changes within sequences can occur.
3895         vector<pair<char_type, Font> > changes;
3896
3897         bool const trackChanges = bparams.track_changes;
3898
3899         bool capitalize = true;
3900
3901         for (; pos < right; ++pos) {
3902                 char_type oldChar = d->text_[pos];
3903                 char_type newChar = oldChar;
3904
3905                 // ignore insets and don't play with deleted text!
3906                 if (oldChar != META_INSET && !isDeleted(pos)) {
3907                         switch (action) {
3908                                 case text_lowercase:
3909                                         newChar = lowercase(oldChar);
3910                                         break;
3911                                 case text_capitalization:
3912                                         if (capitalize) {
3913                                                 newChar = uppercase(oldChar);
3914                                                 capitalize = false;
3915                                         }
3916                                         break;
3917                                 case text_uppercase:
3918                                         newChar = uppercase(oldChar);
3919                                         break;
3920                         }
3921                 }
3922
3923                 if (isWordSeparator(pos) || isDeleted(pos)) {
3924                         // permit capitalization again
3925                         capitalize = true;
3926                 }
3927
3928                 if (oldChar != newChar) {
3929                         changes.push_back(make_pair(newChar, getFontSettings(bparams, pos)));
3930                         if (pos != right - 1)
3931                                 continue;
3932                         // step behind the changing area
3933                         pos++;
3934                 }
3935
3936                 int erasePos = pos - changes.size();
3937                 for (size_t i = 0; i < changes.size(); i++) {
3938                         insertChar(pos, changes[i].first,
3939                                    changes[i].second,
3940                                    trackChanges);
3941                         if (!eraseChar(erasePos, trackChanges)) {
3942                                 ++erasePos;
3943                                 ++pos; // advance
3944                                 ++right; // expand selection
3945                         }
3946                 }
3947                 changes.clear();
3948         }
3949 }
3950
3951
3952 int Paragraph::find(docstring const & str, bool cs, bool mw,
3953                 pos_type start_pos, bool del) const
3954 {
3955         pos_type pos = start_pos;
3956         int const strsize = str.length();
3957         int i = 0;
3958         pos_type const parsize = d->text_.size();
3959         for (i = 0; i < strsize && pos < parsize; ++i, ++pos) {
3960                 // Ignore "invisible" letters such as ligature breaks
3961                 // and hyphenation chars while searching
3962                 while (pos < parsize - 1 && isInset(pos)) {
3963                         odocstringstream os;
3964                         getInset(pos)->toString(os);
3965                         if (!getInset(pos)->isLetter() || !os.str().empty())
3966                                 break;
3967                         pos++;
3968                 }
3969                 if (cs && str[i] != d->text_[pos])
3970                         break;
3971                 if (!cs && uppercase(str[i]) != uppercase(d->text_[pos]))
3972                         break;
3973                 if (!del && isDeleted(pos))
3974                         break;
3975         }
3976
3977         if (i != strsize)
3978                 return 0;
3979
3980         // if necessary, check whether string matches word
3981         if (mw) {
3982                 if (start_pos > 0 && !isWordSeparator(start_pos - 1))
3983                         return 0;
3984                 if (pos < parsize
3985                         && !isWordSeparator(pos))
3986                         return 0;
3987         }
3988
3989         return pos - start_pos;
3990 }
3991
3992
3993 char_type Paragraph::getChar(pos_type pos) const
3994 {
3995         return d->text_[pos];
3996 }
3997
3998
3999 pos_type Paragraph::size() const
4000 {
4001         return d->text_.size();
4002 }
4003
4004
4005 bool Paragraph::empty() const
4006 {
4007         return d->text_.empty();
4008 }
4009
4010
4011 bool Paragraph::isInset(pos_type pos) const
4012 {
4013         return d->text_[pos] == META_INSET;
4014 }
4015
4016
4017 bool Paragraph::isSeparator(pos_type pos) const
4018 {
4019         //FIXME: Are we sure this can be the only separator?
4020         return d->text_[pos] == ' ';
4021 }
4022
4023
4024 void Paragraph::deregisterWords()
4025 {
4026         Private::LangWordsMap::const_iterator itl = d->words_.begin();
4027         Private::LangWordsMap::const_iterator ite = d->words_.end();
4028         for (; itl != ite; ++itl) {
4029                 WordList & wl = theWordList(itl->first);
4030                 Private::Words::const_iterator it = (itl->second).begin();
4031                 Private::Words::const_iterator et = (itl->second).end();
4032                 for (; it != et; ++it)
4033                         wl.remove(*it);
4034         }
4035         d->words_.clear();
4036 }
4037
4038
4039 void Paragraph::locateWord(pos_type & from, pos_type & to,
4040         word_location const loc) const
4041 {
4042         switch (loc) {
4043         case WHOLE_WORD_STRICT:
4044                 if (from == 0 || from == size()
4045                     || isWordSeparator(from)
4046                     || isWordSeparator(from - 1)) {
4047                         to = from;
4048                         return;
4049                 }
4050                 // fall through
4051
4052         case WHOLE_WORD:
4053                 // If we are already at the beginning of a word, do nothing
4054                 if (!from || isWordSeparator(from - 1))
4055                         break;
4056                 // fall through
4057
4058         case PREVIOUS_WORD:
4059                 // always move the cursor to the beginning of previous word
4060                 while (from && !isWordSeparator(from - 1))
4061                         --from;
4062                 break;
4063         case NEXT_WORD:
4064                 LYXERR0("Paragraph::locateWord: NEXT_WORD not implemented yet");
4065                 break;
4066         case PARTIAL_WORD:
4067                 // no need to move the 'from' cursor
4068                 break;
4069         }
4070         to = from;
4071         while (to < size() && !isWordSeparator(to))
4072                 ++to;
4073 }
4074
4075
4076 void Paragraph::collectWords()
4077 {
4078         for (pos_type pos = 0; pos < size(); ++pos) {
4079                 if (isWordSeparator(pos))
4080                         continue;
4081                 pos_type from = pos;
4082                 locateWord(from, pos, WHOLE_WORD);
4083                 // Work around MSVC warning: The statement
4084                 // if (pos < from + lyxrc.completion_minlength)
4085                 // triggers a signed vs. unsigned warning.
4086                 // I don't know why this happens, it could be a MSVC bug, or
4087                 // related to LLP64 (windows) vs. LP64 (unix) programming
4088                 // model, or the C++ standard might be ambigous in the section
4089                 // defining the "usual arithmetic conversions". However, using
4090                 // a temporary variable is safe and works on all compilers.
4091                 pos_type const endpos = from + lyxrc.completion_minlength;
4092                 if (pos < endpos)
4093                         continue;
4094                 FontList::const_iterator cit = d->fontlist_.fontIterator(from);
4095                 if (cit == d->fontlist_.end())
4096                         return;
4097                 Language const * lang = cit->font().language();
4098                 docstring const word = asString(from, pos, AS_STR_NONE);
4099                 d->words_[lang->lang()].insert(word);
4100         }
4101 }
4102
4103
4104 void Paragraph::registerWords()
4105 {
4106         Private::LangWordsMap::const_iterator itl = d->words_.begin();
4107         Private::LangWordsMap::const_iterator ite = d->words_.end();
4108         for (; itl != ite; ++itl) {
4109                 WordList & wl = theWordList(itl->first);
4110                 Private::Words::const_iterator it = (itl->second).begin();
4111                 Private::Words::const_iterator et = (itl->second).end();
4112                 for (; it != et; ++it)
4113                         wl.insert(*it);
4114         }
4115 }
4116
4117
4118 void Paragraph::updateWords()
4119 {
4120         deregisterWords();
4121         collectWords();
4122         registerWords();
4123 }
4124
4125
4126 void Paragraph::Private::appendSkipPosition(SkipPositions & skips, pos_type const pos) const
4127 {
4128         SkipPositionsIterator begin = skips.begin();
4129         SkipPositions::iterator end = skips.end();
4130         if (pos > 0 && begin < end) {
4131                 --end;
4132                 if (end->last == pos - 1) {
4133                         end->last = pos;
4134                         return;
4135                 }
4136         }
4137         skips.insert(end, FontSpan(pos, pos));
4138 }
4139
4140
4141 Language * Paragraph::Private::locateSpellRange(
4142         pos_type & from, pos_type & to,
4143         SkipPositions & skips) const
4144 {
4145         // skip leading white space
4146         while (from < to && owner_->isWordSeparator(from))
4147                 ++from;
4148         // don't check empty range
4149         if (from >= to)
4150                 return 0;
4151         // get current language
4152         Language * lang = getSpellLanguage(from);
4153         pos_type last = from;
4154         bool samelang = true;
4155         bool sameinset = true;
4156         while (last < to && samelang && sameinset) {
4157                 // hop to end of word
4158                 while (last < to && !owner_->isWordSeparator(last)) {
4159                         if (owner_->getInset(last)) {
4160                                 appendSkipPosition(skips, last);
4161                         } else if (owner_->isDeleted(last)) {
4162                                 appendSkipPosition(skips, last);
4163                         }
4164                         ++last;
4165                 }
4166                 // hop to next word while checking for insets
4167                 while (sameinset && last < to && owner_->isWordSeparator(last)) {
4168                         if (Inset const * inset = owner_->getInset(last))
4169                                 sameinset = inset->isChar() && inset->isLetter();
4170                         if (sameinset && owner_->isDeleted(last)) {
4171                                 appendSkipPosition(skips, last);
4172                         }
4173                         if (sameinset)
4174                                 last++;
4175                 }
4176                 if (sameinset && last < to) {
4177                         // now check for language change
4178                         samelang = lang == getSpellLanguage(last);
4179                 }
4180         }
4181         // if language change detected backstep is needed
4182         if (!samelang)
4183                 --last;
4184         to = last;
4185         return lang;
4186 }
4187
4188
4189 Language * Paragraph::Private::getSpellLanguage(pos_type const from) const
4190 {
4191         Language * lang =
4192                 const_cast<Language *>(owner_->getFontSettings(
4193                         inset_owner_->buffer().params(), from).language());
4194         if (lang == inset_owner_->buffer().params().language
4195                 && !lyxrc.spellchecker_alt_lang.empty()) {
4196                 string lang_code;
4197                 string const lang_variety =
4198                         split(lyxrc.spellchecker_alt_lang, lang_code, '-');
4199                 lang->setCode(lang_code);
4200                 lang->setVariety(lang_variety);
4201         }
4202         return lang;
4203 }
4204
4205
4206 void Paragraph::requestSpellCheck(pos_type pos)
4207 {
4208         d->requestSpellCheck(pos);
4209 }
4210
4211
4212 bool Paragraph::needsSpellCheck() const
4213 {
4214         SpellChecker::ChangeNumber speller_change_number = 0;
4215         if (theSpellChecker())
4216                 speller_change_number = theSpellChecker()->changeNumber();
4217         if (speller_change_number > d->speller_state_.currentChangeNumber()) {
4218                 d->speller_state_.needsCompleteRefresh(speller_change_number);
4219         }
4220         return d->needsSpellCheck();
4221 }
4222
4223
4224 bool Paragraph::Private::ignoreWord(docstring const & word) const
4225 {
4226         // Ignore words with digits
4227         // FIXME: make this customizable
4228         // (note that some checkers ignore words with digits by default)
4229         docstring::const_iterator cit = word.begin();
4230         docstring::const_iterator const end = word.end();
4231         for (; cit != end; ++cit) {
4232                 if (isNumber((*cit)))
4233                         return true;
4234         }
4235         return false;
4236 }
4237
4238
4239 SpellChecker::Result Paragraph::spellCheck(pos_type & from, pos_type & to,
4240         WordLangTuple & wl, docstring_list & suggestions,
4241         bool do_suggestion, bool check_learned) const
4242 {
4243         SpellChecker::Result result = SpellChecker::WORD_OK;
4244         SpellChecker * speller = theSpellChecker();
4245         if (!speller)
4246                 return result;
4247
4248         if (!d->layout_->spellcheck || !inInset().allowSpellCheck())
4249                 return result;
4250
4251         locateWord(from, to, WHOLE_WORD);
4252         if (from == to || from >= size())
4253                 return result;
4254
4255         docstring word = asString(from, to, AS_STR_INSETS | AS_STR_SKIPDELETE);
4256         Language * lang = d->getSpellLanguage(from);
4257
4258         if (getFontSettings(d->inset_owner_->buffer().params(), from).fontInfo().nospellcheck() == FONT_ON)
4259                 return result;
4260
4261         wl = WordLangTuple(word, lang);
4262
4263         if (word.empty())
4264                 return result;
4265
4266         if (needsSpellCheck() || check_learned) {
4267                 pos_type end = to;
4268                 if (!d->ignoreWord(word)) {
4269                         bool const trailing_dot = to < size() && d->text_[to] == '.';
4270                         result = speller->check(wl);
4271                         if (SpellChecker::misspelled(result) && trailing_dot) {
4272                                 wl = WordLangTuple(word.append(from_ascii(".")), lang);
4273                                 result = speller->check(wl);
4274                                 if (!SpellChecker::misspelled(result)) {
4275                                         LYXERR(Debug::GUI, "misspelled word is correct with dot: \"" <<
4276                                            word << "\" [" <<
4277                                            from << ".." << to << "]");
4278                                 } else {
4279                                         // spell check with dot appended failed too
4280                                         // restore original word/lang value
4281                                         word = asString(from, to, AS_STR_INSETS | AS_STR_SKIPDELETE);
4282                                         wl = WordLangTuple(word, lang);
4283                                 }
4284                         }
4285                 }
4286                 if (!SpellChecker::misspelled(result)) {
4287                         // area up to the begin of the next word is not misspelled
4288                         while (end < size() && isWordSeparator(end))
4289                                 ++end;
4290                 }
4291                 d->setMisspelled(from, end, result);
4292         } else {
4293                 result = d->speller_state_.getState(from);
4294         }
4295
4296         if (do_suggestion)
4297                 suggestions.clear();
4298
4299         if (SpellChecker::misspelled(result)) {
4300                 LYXERR(Debug::GUI, "misspelled word: \"" <<
4301                            word << "\" [" <<
4302                            from << ".." << to << "]");
4303                 if (do_suggestion)
4304                         speller->suggest(wl, suggestions);
4305         }
4306         return result;
4307 }
4308
4309
4310 void Paragraph::anonymize()
4311 {
4312         // This is a very crude anonymization for now
4313         for (char_type & c : d->text_)
4314                 if (isLetterChar(c) || isNumber(c))
4315                         c = 'a';
4316 }
4317
4318
4319 void Paragraph::Private::markMisspelledWords(
4320         pos_type const & first, pos_type const & last,
4321         SpellChecker::Result result,
4322         docstring const & word,
4323         SkipPositions const & skips)
4324 {
4325         if (!SpellChecker::misspelled(result)) {
4326                 setMisspelled(first, last, SpellChecker::WORD_OK);
4327                 return;
4328         }
4329         int snext = first;
4330         SpellChecker * speller = theSpellChecker();
4331         // locate and enumerate the error positions
4332         int nerrors = speller->numMisspelledWords();
4333         int numskipped = 0;
4334         SkipPositionsIterator it = skips.begin();
4335         SkipPositionsIterator et = skips.end();
4336         for (int index = 0; index < nerrors; ++index) {
4337                 int wstart;
4338                 int wlen = 0;
4339                 speller->misspelledWord(index, wstart, wlen);
4340                 /// should not happen if speller supports range checks
4341                 if (!wlen) continue;
4342                 docstring const misspelled = word.substr(wstart, wlen);
4343                 wstart += first + numskipped;
4344                 if (snext < wstart) {
4345                         /// mark the range of correct spelling
4346                         numskipped += countSkips(it, et, wstart);
4347                         setMisspelled(snext,
4348                                 wstart - 1, SpellChecker::WORD_OK);
4349                 }
4350                 snext = wstart + wlen;
4351                 numskipped += countSkips(it, et, snext);
4352                 /// mark the range of misspelling
4353                 setMisspelled(wstart, snext, result);
4354                 LYXERR(Debug::GUI, "misspelled word: \"" <<
4355                            misspelled << "\" [" <<
4356                            wstart << ".." << (snext-1) << "]");
4357                 ++snext;
4358         }
4359         if (snext <= last) {
4360                 /// mark the range of correct spelling at end
4361                 setMisspelled(snext, last, SpellChecker::WORD_OK);
4362         }
4363 }
4364
4365
4366 void Paragraph::spellCheck() const
4367 {
4368         SpellChecker * speller = theSpellChecker();
4369         if (!speller || empty() ||!needsSpellCheck())
4370                 return;
4371         pos_type start;
4372         pos_type endpos;
4373         d->rangeOfSpellCheck(start, endpos);
4374         if (speller->canCheckParagraph()) {
4375                 // loop until we leave the range
4376                 for (pos_type first = start; first < endpos; ) {
4377                         pos_type last = endpos;
4378                         Private::SkipPositions skips;
4379                         Language * lang = d->locateSpellRange(first, last, skips);
4380                         if (first >= endpos)
4381                                 break;
4382                         // start the spell checker on the unit of meaning
4383                         docstring word = asString(first, last, AS_STR_INSETS + AS_STR_SKIPDELETE);
4384                         WordLangTuple wl = WordLangTuple(word, lang);
4385                         SpellChecker::Result result = word.size() ?
4386                                 speller->check(wl) : SpellChecker::WORD_OK;
4387                         d->markMisspelledWords(first, last, result, word, skips);
4388                         first = ++last;
4389                 }
4390         } else {
4391                 static docstring_list suggestions;
4392                 pos_type to = endpos;
4393                 while (start < endpos) {
4394                         WordLangTuple wl;
4395                         spellCheck(start, to, wl, suggestions, false);
4396                         start = to + 1;
4397                 }
4398         }
4399         d->readySpellCheck();
4400 }
4401
4402
4403 bool Paragraph::isMisspelled(pos_type pos, bool check_boundary) const
4404 {
4405         bool result = SpellChecker::misspelled(d->speller_state_.getState(pos));
4406         if (result || pos <= 0 || pos > size())
4407                 return result;
4408         if (check_boundary && (pos == size() || isWordSeparator(pos)))
4409                 result = SpellChecker::misspelled(d->speller_state_.getState(pos - 1));
4410         return result;
4411 }
4412
4413
4414 string Paragraph::magicLabel() const
4415 {
4416         stringstream ss;
4417         ss << "magicparlabel-" << id();
4418         return ss.str();
4419 }
4420
4421
4422 } // namespace lyx