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
  56 #include "mathed/InsetMathHull.h"
  57
  58 #include "support/debug.h"
  59 #include "support/docstring_list.h"
  60 #include "support/ExceptionMessage.h"
  61 #include "support/gettext.h"
  62 #include "support/lassert.h"
  63 #include "support/lstrings.h"
  64 #include "support/textutils.h"
  65
  66 #include <atomic>
  67 #include <sstream>
  68 #include <vector>
  69
  70 using namespace std;
  71 using namespace lyx::support;
  72
  73 namespace lyx {
  74
  75 namespace {
  76
  77 /// Inset identifier (above 0x10ffff, for ucs-4)
  78 char_type const META_INSET = 0x200001;
  79
  80 }
  81
  82
  83 /////////////////////////////////////////////////////////////////////
  84 //
  85 // SpellResultRange
  86 //
  87 /////////////////////////////////////////////////////////////////////
  88
  89 class SpellResultRange {
  90 public:
  91         SpellResultRange(FontSpan range, SpellChecker::Result result)
  92         : range_(range), result_(result)
  93         {}
  94         ///
  95         FontSpan const & range() const { return range_; }
  96         ///
  97         void range(FontSpan const & r) { range_ = r; }
  98         ///
  99         SpellChecker::Result result() const { return result_; }
 100         ///
 101         void result(SpellChecker::Result r) { result_ = r; }
 102         ///
 103         bool contains(pos_type pos) const { return range_.contains(pos); }
 104         ///
 105         bool covered(FontSpan const & r) const
 106         {
 107                 // 1. first of new range inside current range or
 108                 // 2. last of new range inside current range or
 109                 // 3. first of current range inside new range or
 110                 // 4. last of current range inside new range
 111                 //FIXME: is this the same as !range_.intersect(r).empty() ?
 112                 return range_.contains(r.first) || range_.contains(r.last) ||
 113                         r.contains(range_.first) || r.contains(range_.last);
 114         }
 115         ///
 116         void shift(pos_type pos, int offset)
 117         {
 118                 if (range_.first > pos) {
 119                         range_.first += offset;
 120                         range_.last += offset;
 121                 } else if (range_.last >= pos) {
 122                         range_.last += offset;
 123                 }
 124         }
 125 private:
 126         FontSpan range_ ;
 127         SpellChecker::Result result_ ;
 128 };
 129
 130
 131 /////////////////////////////////////////////////////////////////////
 132 //
 133 // SpellCheckerState
 134 //
 135 /////////////////////////////////////////////////////////////////////
 136
 137 class SpellCheckerState {
 138 public:
 139         SpellCheckerState()
 140         {
 141                 needs_refresh_ = true;
 142                 current_change_number_ = 0;
 143         }
 144
 145         void setRange(FontSpan const & fp, SpellChecker::Result state)
 146         {
 147                 Ranges result;
 148                 RangesIterator et = ranges_.end();
 149                 RangesIterator it = ranges_.begin();
 150                 for (; it != et; ++it) {
 151                         if (!it->covered(fp))
 152                                 result.push_back(SpellResultRange(it->range(), it->result()));
 153                         else if (state == SpellChecker::WORD_OK) {
 154                                 // trim or split the current misspelled range
 155                                 // store misspelled ranges only
 156                                 FontSpan range = it->range();
 157                                 if (fp.first > range.first) {
 158                                         // misspelled area in front of WORD_OK
 159                                         range.last = fp.first - 1;
 160                                         result.push_back(SpellResultRange(range, it->result()));
 161                                         range = it->range();
 162                                 }
 163                                 if (fp.last < range.last) {
 164                                         // misspelled area after WORD_OK range
 165                                         range.first = fp.last + 1;
 166                                         result.push_back(SpellResultRange(range, it->result()));
 167                                 }
 168                         }
 169                 }
 170                 ranges_ = result;
 171                 if (state != SpellChecker::WORD_OK)
 172                         ranges_.push_back(SpellResultRange(fp, state));
 173         }
 174
 175         void increasePosAfterPos(pos_type pos)
 176         {
 177                 correctRangesAfterPos(pos, 1);
 178                 needsRefresh(pos);
 179         }
 180
 181         void decreasePosAfterPos(pos_type pos)
 182         {
 183                 correctRangesAfterPos(pos, -1);
 184                 needsRefresh(pos);
 185         }
 186
 187         void refreshLast(pos_type pos)
 188         {
 189                 if (pos < refresh_.last)
 190                         refresh_.last = pos;
 191         }
 192
 193         SpellChecker::Result getState(pos_type pos) const
 194         {
 195                 SpellChecker::Result result = SpellChecker::WORD_OK;
 196                 RangesIterator et = ranges_.end();
 197                 RangesIterator it = ranges_.begin();
 198                 for (; it != et; ++it) {
 199                         if(it->contains(pos)) {
 200                                 return it->result();
 201                         }
 202                 }
 203                 return result;
 204         }
 205
 206         FontSpan const & getRange(pos_type pos) const
 207         {
 208                 /// empty span to indicate mismatch
 209                 static FontSpan empty_;
 210                 RangesIterator et = ranges_.end();
 211                 RangesIterator it = ranges_.begin();
 212                 for (; it != et; ++it) {
 213                         if(it->contains(pos)) {
 214                                 return it->range();
 215                         }
 216                 }
 217                 return empty_;
 218         }
 219
 220         bool needsRefresh() const
 221         {
 222                 return needs_refresh_;
 223         }
 224
 225         SpellChecker::ChangeNumber currentChangeNumber() const
 226         {
 227                 return current_change_number_;
 228         }
 229
 230         void refreshRange(pos_type & first, pos_type & last) const
 231         {
 232                 first = refresh_.first;
 233                 last = refresh_.last;
 234         }
 235
 236         void needsRefresh(pos_type pos)
 237         {
 238                 if (needs_refresh_ && pos != -1) {
 239                         if (pos < refresh_.first)
 240                                 refresh_.first = pos;
 241                         if (pos > refresh_.last)
 242                                 refresh_.last = pos;
 243                 } else if (pos != -1) {
 244                         // init request check for neighbour positions too
 245                         refresh_.first = pos > 0 ? pos - 1 : 0;
 246                         // no need for special end of paragraph check
 247                         refresh_.last = pos + 1;
 248                 }
 249                 needs_refresh_ = pos != -1;
 250         }
 251
 252         void needsCompleteRefresh(SpellChecker::ChangeNumber change_number)
 253         {
 254                 needs_refresh_ = true;
 255                 refresh_.first = 0;
 256                 refresh_.last = -1;
 257                 current_change_number_ = change_number;
 258         }
 259 private:
 260         typedef vector<SpellResultRange> Ranges;
 261         typedef Ranges::const_iterator RangesIterator;
 262         Ranges ranges_;
 263         /// the area of the paragraph with pending spell check
 264         FontSpan refresh_;
 265         bool needs_refresh_;
 266         /// spell state cache version number
 267         SpellChecker::ChangeNumber current_change_number_;
 268
 269
 270         void correctRangesAfterPos(pos_type pos, int offset)
 271         {
 272                 RangesIterator et = ranges_.end();
 273                 Ranges::iterator it = ranges_.begin();
 274                 for (; it != et; ++it) {
 275                         it->shift(pos, offset);
 276                 }
 277         }
 278
 279 };
 280
 281 /////////////////////////////////////////////////////////////////////
 282 //
 283 // Paragraph::Private
 284 //
 285 /////////////////////////////////////////////////////////////////////
 286
 287 class Paragraph::Private
 288 {
 289         // Enforce our own "copy" constructor
 290         Private(Private const &) = delete;
 291         Private & operator=(Private const &) = delete;
 292         // Unique ID generator
 293         static int make_id();
 294 public:
 295         ///
 296         Private(Paragraph * owner, Layout const & layout);
 297         /// "Copy constructor"
 298         Private(Private const &, Paragraph * owner);
 299         /// Copy constructor from \p beg  to \p end
 300         Private(Private const &, Paragraph * owner, pos_type beg, pos_type end);
 301
 302         ///
 303         void insertChar(pos_type pos, char_type c, Change const & change);
 304
 305         /// Output the surrogate pair formed by \p c and \p next to \p os.
 306         /// \return the number of characters written.
 307         int latexSurrogatePair(otexstream & os, char_type c, char_type next,
 308                                OutputParams const &);
 309
 310         /// Output a space in appropriate formatting (or a surrogate pair
 311         /// if the next character is a combining character).
 312         /// \return whether a surrogate pair was output.
 313         bool simpleTeXBlanks(OutputParams const &,
 314                              otexstream &,
 315                              pos_type i,
 316                              unsigned int & column,
 317                              Font const & font,
 318                              Layout const & style);
 319
 320         /// Output consecutive unicode chars, belonging to the same script as
 321         /// specified by the latex macro \p ltx, to \p os starting from \p i.
 322         /// \return the number of characters written.
 323         int writeScriptChars(otexstream & os, docstring const & ltx,
 324                            Change const &, Encoding const &, pos_type & i);
 325
 326         /// This could go to ParagraphParameters if we want to.
 327         int startTeXParParams(BufferParams const &, otexstream &,
 328                               OutputParams const &) const;
 329
 330         /// This could go to ParagraphParameters if we want to.
 331         bool endTeXParParams(BufferParams const &, otexstream &,
 332                              OutputParams const &) const;
 333
 334         ///
 335         void latexInset(BufferParams const &,
 336                                    otexstream &,
 337                                    OutputParams &,
 338                                    Font & running_font,
 339                                    Font & basefont,
 340                                    Font const & outerfont,
 341                                    bool & open_font,
 342                                    Change & running_change,
 343                                    Layout const & style,
 344                                    pos_type & i,
 345                                    unsigned int & column);
 346
 347         ///
 348         void latexSpecialChar(
 349                                    otexstream & os,
 350                                    BufferParams const & bparams,
 351                                    OutputParams const & runparams,
 352                                    Font const & running_font,
 353                                    Change const & running_change,
 354                                    Layout const & style,
 355                                    pos_type & i,
 356                                    pos_type end_pos,
 357                                    unsigned int & column);
 358
 359         ///
 360         bool latexSpecialT1(
 361                 char_type const c,
 362                 otexstream & os,
 363                 pos_type i,
 364                 unsigned int & column);
 365         ///
 366         bool latexSpecialT3(
 367                 char_type const c,
 368                 otexstream & os,
 369                 pos_type i,
 370                 unsigned int & column);
 371
 372         ///
 373         void validate(LaTeXFeatures & features) const;
 374
 375         /// Checks if the paragraph contains only text and no inset or font change.
 376         bool onlyText(Buffer const & buf, Font const & outerfont,
 377                       pos_type initial) const;
 378
 379         /// a vector of speller skip positions
 380         typedef vector<FontSpan> SkipPositions;
 381         typedef SkipPositions::const_iterator SkipPositionsIterator;
 382
 383         void appendSkipPosition(SkipPositions & skips, pos_type const pos) const;
 384
 385         Language * getSpellLanguage(pos_type const from) const;
 386
 387         Language * locateSpellRange(pos_type & from, pos_type & to,
 388                                     SkipPositions & skips) const;
 389
 390         bool hasSpellerChange() const
 391         {
 392                 SpellChecker::ChangeNumber speller_change_number = 0;
 393                 if (theSpellChecker())
 394                         speller_change_number = theSpellChecker()->changeNumber();
 395                 return speller_change_number > speller_state_.currentChangeNumber();
 396         }
 397
 398         bool ignoreWord(docstring const & word) const ;
 399
 400         void setMisspelled(pos_type from, pos_type to, SpellChecker::Result state)
 401         {
 402                 pos_type textsize = owner_->size();
 403                 // check for sane arguments
 404                 if (to <= from || from >= textsize)
 405                         return;
 406                 FontSpan fp = FontSpan(from, to - 1);
 407                 speller_state_.setRange(fp, state);
 408         }
 409
 410         void requestSpellCheck(pos_type pos)
 411         {
 412                 if (pos == -1)
 413                         speller_state_.needsCompleteRefresh(speller_state_.currentChangeNumber());
 414                 else
 415                         speller_state_.needsRefresh(pos);
 416         }
 417
 418         void readySpellCheck()
 419         {
 420                 speller_state_.needsRefresh(-1);
 421         }
 422
 423         bool needsSpellCheck() const
 424         {
 425                 return speller_state_.needsRefresh();
 426         }
 427
 428         void rangeOfSpellCheck(pos_type & first, pos_type & last) const
 429         {
 430                 speller_state_.refreshRange(first, last);
 431                 if (last == -1) {
 432                         last = owner_->size();
 433                         return;
 434                 }
 435                 pos_type endpos = last;
 436                 owner_->locateWord(first, endpos, WHOLE_WORD);
 437                 if (endpos < last) {
 438                         endpos = last;
 439                         owner_->locateWord(last, endpos, WHOLE_WORD);
 440                 }
 441                 last = endpos;
 442         }
 443
 444         int countSkips(SkipPositionsIterator & it, SkipPositionsIterator const et,
 445                             int & start) const
 446         {
 447                 int numskips = 0;
 448                 while (it != et && it->first < start) {
 449                         int skip = it->last - it->first + 1;
 450                         start += skip;
 451                         numskips += skip;
 452                         ++it;
 453                 }
 454                 return numskips;
 455         }
 456
 457         void markMisspelledWords(pos_type const & first, pos_type const & last,
 458                                                          SpellChecker::Result result,
 459                                                          docstring const & word,
 460                                                          SkipPositions const & skips);
 461
 462         InsetCode ownerCode() const
 463         {
 464                 return inset_owner_ ? inset_owner_->lyxCode() : NO_CODE;
 465         }
 466
 467         /// Which Paragraph owns us?
 468         Paragraph * owner_;
 469
 470         /// In which Inset?
 471         Inset const * inset_owner_;
 472
 473         ///
 474         FontList fontlist_;
 475
 476         ///
 477         int id_;
 478
 479         ///
 480         ParagraphParameters params_;
 481
 482         /// for recording and looking up changes
 483         Changes changes_;
 484
 485         ///
 486         InsetList insetlist_;
 487
 488         /// end of label
 489         pos_type begin_of_body_;
 490
 491         typedef docstring TextContainer;
 492         ///
 493         TextContainer text_;
 494
 495         typedef set<docstring> Words;
 496         typedef map<string, Words> LangWordsMap;
 497         ///
 498         LangWordsMap words_;
 499         ///
 500         Layout const * layout_;
 501         ///
 502         SpellCheckerState speller_state_;
 503 };
 504
 505
 506 Paragraph::Private::Private(Paragraph * owner, Layout const & layout)
 507         : owner_(owner), inset_owner_(0), id_(-1), begin_of_body_(0), layout_(&layout)
 508 {
 509         text_.reserve(100);
 510 }
 511
 512
 513 //static
 514 int Paragraph::Private::make_id()
 515 {
 516         // The id is unique per session across buffers because it is used in
 517         // LFUN_PARAGRAPH_GOTO to switch to a different buffer, for instance in the
 518         // outliner.
 519         // (thread-safe)
 520         static atomic_uint next_id(0);
 521         return next_id++;
 522 }
 523
 524
 525 Paragraph::Private::Private(Private const & p, Paragraph * owner)
 526         : owner_(owner), inset_owner_(p.inset_owner_), fontlist_(p.fontlist_),
 527           id_(make_id()),
 528           params_(p.params_), changes_(p.changes_), insetlist_(p.insetlist_),
 529           begin_of_body_(p.begin_of_body_), text_(p.text_), words_(p.words_),
 530           layout_(p.layout_)
 531 {
 532         requestSpellCheck(p.text_.size());
 533 }
 534
 535
 536 Paragraph::Private::Private(Private const & p, Paragraph * owner,
 537         pos_type beg, pos_type end)
 538         : owner_(owner), inset_owner_(p.inset_owner_), id_(make_id()),
 539           params_(p.params_), changes_(p.changes_),
 540           insetlist_(p.insetlist_, beg, end),
 541           begin_of_body_(p.begin_of_body_), words_(p.words_),
 542           layout_(p.layout_)
 543 {
 544         if (beg >= pos_type(p.text_.size()))
 545                 return;
 546         text_ = p.text_.substr(beg, end - beg);
 547
 548         FontList::const_iterator fcit = fontlist_.begin();
 549         FontList::const_iterator fend = fontlist_.end();
 550         for (; fcit != fend; ++fcit) {
 551                 if (fcit->pos() < beg)
 552                         continue;
 553                 if (fcit->pos() >= end) {
 554                         // Add last entry in the fontlist_.
 555                         fontlist_.set(text_.size() - 1, fcit->font());
 556                         break;
 557                 }
 558                 // Add a new entry in the fontlist_.
 559                 fontlist_.set(fcit->pos() - beg, fcit->font());
 560         }
 561         requestSpellCheck(p.text_.size());
 562 }
 563
 564
 565 void Paragraph::addChangesToToc(DocIterator const & cdit, Buffer const & buf,
 566                                 bool output_active, TocBackend & backend) const
 567 {
 568         d->changes_.addToToc(cdit, buf, output_active, backend);
 569 }
 570
 571
 572 void Paragraph::addChangesToBuffer(Buffer const & buf) const
 573 {
 574         d->changes_.updateBuffer(buf);
 575 }
 576
 577
 578 bool Paragraph::isChangeUpdateRequired() const
 579 {
 580         return d->changes_.isUpdateRequired();
 581 }
 582
 583
 584 bool Paragraph::isDeleted(pos_type start, pos_type end) const
 585 {
 586         LASSERT(start >= 0 && start <= size(), return false);
 587         LASSERT(end > start && end <= size() + 1, return false);
 588
 589         return d->changes_.isDeleted(start, end);
 590 }
 591
 592
 593 bool Paragraph::isChanged(pos_type start, pos_type end) const
 594 {
 595         LASSERT(start >= 0 && start <= size(), return false);
 596         LASSERT(end > start && end <= size() + 1, return false);
 597
 598         return d->changes_.isChanged(start, end);
 599 }
 600
 601
 602 bool Paragraph::isMergedOnEndOfParDeletion(bool trackChanges) const
 603 {
 604         // keep the logic here in sync with the logic of eraseChars()
 605         if (!trackChanges)
 606                 return true;
 607
 608         Change const & change = d->changes_.lookup(size());
 609         return change.inserted() && change.currentAuthor();
 610 }
 611
 612
 613 void Paragraph::setChange(Change const & change)
 614 {
 615         // beware of the imaginary end-of-par character!
 616         d->changes_.set(change, 0, size() + 1);
 617
 618         /*
 619          * Propagate the change recursively - but not in case of DELETED!
 620          *
 621          * Imagine that your co-author makes changes in an existing inset. He
 622          * sends your document to you and you come to the conclusion that the
 623          * inset should go completely. If you erase it, LyX must not delete all
 624          * text within the inset. Otherwise, the change tracked insertions of
 625          * your co-author get lost and there is no way to restore them later.
 626          *
 627          * Conclusion: An inset's content should remain untouched if you delete it
 628          */
 629
 630         if (!change.deleted()) {
 631                 for (pos_type pos = 0; pos < size(); ++pos) {
 632                         if (Inset * inset = getInset(pos))
 633                                 inset->setChange(change);
 634                 }
 635         }
 636 }
 637
 638
 639 void Paragraph::setChange(pos_type pos, Change const & change)
 640 {
 641         LASSERT(pos >= 0 && pos <= size(), return);
 642         d->changes_.set(change, pos);
 643
 644         // see comment in setChange(Change const &) above
 645         if (!change.deleted() && pos < size())
 646                         if (Inset * inset = getInset(pos))
 647                                 inset->setChange(change);
 648 }
 649
 650
 651 Change const & Paragraph::lookupChange(pos_type pos) const
 652 {
 653         LBUFERR(pos >= 0 && pos <= size());
 654         return d->changes_.lookup(pos);
 655 }
 656
 657
 658 void Paragraph::acceptChanges(pos_type start, pos_type end)
 659 {
 660         LASSERT(start >= 0 && start <= size(), return);
 661         LASSERT(end > start && end <= size() + 1, return);
 662
 663         for (pos_type pos = start; pos < end; ++pos) {
 664                 switch (lookupChange(pos).type) {
 665                         case Change::UNCHANGED:
 666                                 // accept changes in nested inset
 667                                 if (Inset * inset = getInset(pos))
 668                                         inset->acceptChanges();
 669                                 break;
 670
 671                         case Change::INSERTED:
 672                                 d->changes_.set(Change(Change::UNCHANGED), pos);
 673                                 // also accept changes in nested inset
 674                                 if (Inset * inset = getInset(pos))
 675                                         inset->acceptChanges();
 676                                 break;
 677
 678                         case Change::DELETED:
 679                                 // Suppress access to non-existent
 680                                 // "end-of-paragraph char"
 681                                 if (pos < size()) {
 682                                         eraseChar(pos, false);
 683                                         --end;
 684                                         --pos;
 685                                 }
 686                                 break;
 687                 }
 688
 689         }
 690 }
 691
 692
 693 void Paragraph::rejectChanges(pos_type start, pos_type end)
 694 {
 695         LASSERT(start >= 0 && start <= size(), return);
 696         LASSERT(end > start && end <= size() + 1, return);
 697
 698         for (pos_type pos = start; pos < end; ++pos) {
 699                 switch (lookupChange(pos).type) {
 700                         case Change::UNCHANGED:
 701                                 // reject changes in nested inset
 702                                 if (Inset * inset = getInset(pos))
 703                                                 inset->rejectChanges();
 704                                 break;
 705
 706                         case Change::INSERTED:
 707                                 // Suppress access to non-existent
 708                                 // "end-of-paragraph char"
 709                                 if (pos < size()) {
 710                                         eraseChar(pos, false);
 711                                         --end;
 712                                         --pos;
 713                                 }
 714                                 break;
 715
 716                         case Change::DELETED:
 717                                 d->changes_.set(Change(Change::UNCHANGED), pos);
 718
 719                                 // Do NOT reject changes within a deleted inset!
 720                                 // There may be insertions of a co-author inside of it!
 721
 722                                 break;
 723                 }
 724         }
 725 }
 726
 727
 728 void Paragraph::Private::insertChar(pos_type pos, char_type c,
 729                 Change const & change)
 730 {
 731         LASSERT(pos >= 0 && pos <= int(text_.size()), return);
 732
 733         // track change
 734         changes_.insert(change, pos);
 735
 736         // This is actually very common when parsing buffers (and
 737         // maybe inserting ascii text)
 738         if (pos == pos_type(text_.size())) {
 739                 // when appending characters, no need to update tables
 740                 text_.push_back(c);
 741                 // but we want spell checking
 742                 requestSpellCheck(pos);
 743                 return;
 744         }
 745
 746         text_.insert(text_.begin() + pos, c);
 747
 748         // Update the font table.
 749         fontlist_.increasePosAfterPos(pos);
 750
 751         // Update the insets
 752         insetlist_.increasePosAfterPos(pos);
 753
 754         // Update list of misspelled positions
 755         speller_state_.increasePosAfterPos(pos);
 756 }
 757
 758
 759 bool Paragraph::insertInset(pos_type pos, Inset * inset,
 760                                    Font const & font, Change const & change)
 761 {
 762         LASSERT(inset, return false);
 763         LASSERT(pos >= 0 && pos <= size(), return false);
 764
 765         // Paragraph::insertInset() can be used in cut/copy/paste operation where
 766         // d->inset_owner_ is not set yet.
 767         if (d->inset_owner_ && !d->inset_owner_->insetAllowed(inset->lyxCode()))
 768                 return false;
 769
 770         d->insertChar(pos, META_INSET, change);
 771         LASSERT(d->text_[pos] == META_INSET, return false);
 772
 773         // Add a new entry in the insetlist_.
 774         d->insetlist_.insert(inset, pos);
 775
 776         // Some insets require run of spell checker
 777         requestSpellCheck(pos);
 778         setFont(pos, font);
 779         return true;
 780 }
 781
 782
 783 bool Paragraph::eraseChar(pos_type pos, bool trackChanges)
 784 {
 785         LASSERT(pos >= 0 && pos <= size(), return false);
 786
 787         // keep the logic here in sync with the logic of isMergedOnEndOfParDeletion()
 788
 789         if (trackChanges) {
 790                 Change change = d->changes_.lookup(pos);
 791
 792                 // set the character to DELETED if
 793                 //  a) it was previously unchanged or
 794                 //  b) it was inserted by a co-author
 795
 796                 if (!change.changed() ||
 797                       (change.inserted() && !change.currentAuthor())) {
 798                         setChange(pos, Change(Change::DELETED));
 799                         // request run of spell checker
 800                         requestSpellCheck(pos);
 801                         return false;
 802                 }
 803
 804                 if (change.deleted())
 805                         return false;
 806         }
 807
 808         // Don't physically access the imaginary end-of-paragraph character.
 809         // eraseChar() can only mark it as DELETED. A physical deletion of
 810         // end-of-par must be handled externally.
 811         if (pos == size()) {
 812                 return false;
 813         }
 814
 815         // track change
 816         d->changes_.erase(pos);
 817
 818         // if it is an inset, delete the inset entry
 819         if (d->text_[pos] == META_INSET)
 820                 d->insetlist_.erase(pos);
 821
 822         d->text_.erase(d->text_.begin() + pos);
 823
 824         // Update the fontlist_
 825         d->fontlist_.erase(pos);
 826
 827         // Update the insetlist_
 828         d->insetlist_.decreasePosAfterPos(pos);
 829
 830         // Update list of misspelled positions
 831         d->speller_state_.decreasePosAfterPos(pos);
 832         d->speller_state_.refreshLast(size());
 833
 834         return true;
 835 }
 836
 837
 838 int Paragraph::eraseChars(pos_type start, pos_type end, bool trackChanges)
 839 {
 840         LASSERT(start >= 0 && start <= size(), return 0);
 841         LASSERT(end >= start && end <= size() + 1, return 0);
 842
 843         pos_type i = start;
 844         for (pos_type count = end - start; count; --count) {
 845                 if (!eraseChar(i, trackChanges))
 846                         ++i;
 847         }
 848         return end - i;
 849 }
 850
 851
 852 int Paragraph::Private::latexSurrogatePair(otexstream & os, char_type c,
 853                 char_type next, OutputParams const & runparams)
 854 {
 855         // Writing next here may circumvent a possible font change between
 856         // c and next. Since next is only output if it forms a surrogate pair
 857         // with c we can ignore this:
 858         // A font change inside a surrogate pair does not make sense and is
 859         // hopefully impossible to input.
 860         // FIXME: change tracking
 861         // Is this correct WRT change tracking?
 862         Encoding const & encoding = *(runparams.encoding);
 863         docstring latex1 = encoding.latexChar(next).first;
 864         if (runparams.inIPA) {
 865                 string const tipashortcut = Encodings::TIPAShortcut(next);
 866                 if (!tipashortcut.empty()) {
 867                         latex1 = from_ascii(tipashortcut);
 868                 }
 869         }
 870         docstring const latex2 = encoding.latexChar(c).first;
 871         if (docstring(1, next) == latex1) {
 872                 // the encoding supports the combination
 873                 os << latex2 << latex1;
 874                 return latex1.length() + latex2.length();
 875         } else if (runparams.local_font &&
 876                    runparams.local_font->language()->lang() == "polutonikogreek") {
 877                 // polutonikogreek only works without the brackets
 878                 os << latex1 << latex2;
 879                 return latex1.length() + latex2.length();
 880         } else
 881                 os << latex1 << '{' << latex2 << '}';
 882         return latex1.length() + latex2.length() + 2;
 883 }
 884
 885
 886 bool Paragraph::Private::simpleTeXBlanks(OutputParams const & runparams,
 887                                        otexstream & os,
 888                                        pos_type i,
 889                                        unsigned int & column,
 890                                        Font const & font,
 891                                        Layout const & style)
 892 {
 893         if (style.pass_thru || runparams.pass_thru)
 894                 return false;
 895
 896         if (i + 1 < int(text_.size())) {
 897                 char_type next = text_[i + 1];
 898                 if (Encodings::isCombiningChar(next)) {
 899                         // This space has an accent, so we must always output it.
 900                         column += latexSurrogatePair(os, ' ', next, runparams) - 1;
 901                         return true;
 902                 }
 903         }
 904
 905         if (runparams.linelen > 0
 906             && column > runparams.linelen
 907             && i
 908             && text_[i - 1] != ' '
 909             && (i + 1 < int(text_.size()))
 910             // same in FreeSpacing mode
 911             && !owner_->isFreeSpacing()
 912             // In typewriter mode, we want to avoid
 913             // ! . ? : at the end of a line
 914             && !(font.fontInfo().family() == TYPEWRITER_FAMILY
 915                  && (text_[i - 1] == '.'
 916                      || text_[i - 1] == '?'
 917                      || text_[i - 1] == ':'
 918                      || text_[i - 1] == '!'))) {
 919                 os << '\n';
 920                 os.texrow().start(owner_->id(), i + 1);
 921                 column = 0;
 922         } else if (style.free_spacing) {
 923                 os << '~';
 924         } else {
 925                 os << ' ';
 926         }
 927         return false;
 928 }
 929
 930
 931 int Paragraph::Private::writeScriptChars(otexstream & os,
 932                                          docstring const & ltx,
 933                                          Change const & runningChange,
 934                                          Encoding const & encoding,
 935                                          pos_type & i)
 936 {
 937         // FIXME: modifying i here is not very nice...
 938
 939         // We only arrive here when character text_[i] could not be translated
 940         // into the current latex encoding (or its latex translation has been forced,)
 941         // and it belongs to a known script.
 942         // TODO: We need \textcyr and \textgreek wrappers also for characters
 943         //       that can be encoded in the "LaTeX encoding" but not in the
 944         //       current *font encoding*.
 945         //       (See #9681 for details and test)
 946         // Parameter ltx contains the latex translation of text_[i] as specified
 947         // in the unicodesymbols file and is something like "\textXXX{<spec>}".
 948         // The latex macro name "textXXX" specifies the script to which text_[i]
 949         // belongs and we use it in order to check whether characters from the
 950         // same script immediately follow, such that we can collect them in a
 951         // single "\textXXX" macro. So, we have to retain "\textXXX{<spec>"
 952         // for the first char but only "<spec>" for all subsequent chars.
 953         docstring::size_type const brace1 = ltx.find_first_of(from_ascii("{"));
 954         docstring::size_type const brace2 = ltx.find_last_of(from_ascii("}"));
 955         string script = to_ascii(ltx.substr(1, brace1 - 1));
 956         int pos = 0;
 957         int length = brace2;
 958         bool closing_brace = true;
 959         if (script == "textgreek" && encoding.latexName() == "iso-8859-7") {
 960                 // Correct encoding is being used, so we can avoid \textgreek.
 961                 // TODO: wrong test: we need to check the *font encoding*
 962                 //       (i.e. the active language and its FontEncoding tag)
 963                 //       instead of the LaTeX *input encoding*!
 964                 //       See #9637 for details and test-cases.
 965                 pos = brace1 + 1;
 966                 length -= pos;
 967                 closing_brace = false;
 968         }
 969         os << ltx.substr(pos, length);
 970         int size = text_.size();
 971         while (i + 1 < size) {
 972                 char_type const next = text_[i + 1];
 973                 // Stop here if next character belongs to another script
 974                 // or there is a change in change tracking status.
 975                 if (!Encodings::isKnownScriptChar(next, script) ||
 976                     runningChange != owner_->lookupChange(i + 1))
 977                         break;
 978                 Font prev_font;
 979                 bool found = false;
 980                 FontList::const_iterator cit = fontlist_.begin();
 981                 FontList::const_iterator end = fontlist_.end();
 982                 for (; cit != end; ++cit) {
 983                         if (cit->pos() >= i && !found) {
 984                                 prev_font = cit->font();
 985                                 found = true;
 986                         }
 987                         if (cit->pos() >= i + 1)
 988                                 break;
 989                 }
 990                 // Stop here if there is a font attribute or encoding change.
 991                 if (found && cit != end && prev_font != cit->font())
 992                         break;
 993                 docstring const latex = encoding.latexChar(next).first;
 994                 docstring::size_type const b1 =
 995                                         latex.find_first_of(from_ascii("{"));
 996                 docstring::size_type const b2 =
 997                                         latex.find_last_of(from_ascii("}"));
 998                 int const len = b2 - b1 - 1;
 999                 os << latex.substr(b1 + 1, len);
1000                 length += len;
1001                 ++i;
1002         }
1003         if (closing_brace) {
1004                 os << '}';
1005                 ++length;
1006         }
1007         return length;
1008 }
1009
1010
1011 void Paragraph::Private::latexInset(BufferParams const & bparams,
1012                                     otexstream & os,
1013                                     OutputParams & runparams,
1014                                     Font & running_font,
1015                                     Font & basefont,
1016                                     Font const & outerfont,
1017                                     bool & open_font,
1018                                     Change & running_change,
1019                                     Layout const & style,
1020                                     pos_type & i,
1021                                     unsigned int & column)
1022 {
1023         Inset * inset = owner_->getInset(i);
1024         LBUFERR(inset);
1025
1026         if (style.pass_thru) {
1027                 odocstringstream ods;
1028                 inset->plaintext(ods, runparams);
1029                 os << ods.str();
1030                 return;
1031         }
1032
1033         // FIXME: move this to InsetNewline::latex
1034         if (inset->lyxCode() == NEWLINE_CODE || inset->lyxCode() == SEPARATOR_CODE) {
1035                 // newlines are handled differently here than
1036                 // the default in simpleTeXSpecialChars().
1037                 if (!style.newline_allowed) {
1038                         os << '\n';
1039                 } else {
1040                         if (open_font) {
1041                                 column += running_font.latexWriteEndChanges(
1042                                         os, bparams, runparams,
1043                                         basefont, basefont);
1044                                 open_font = false;
1045                         }
1046
1047                         if (running_font.fontInfo().family() == TYPEWRITER_FAMILY)
1048                                 os << '~';
1049
1050                         basefont = owner_->getLayoutFont(bparams, outerfont);
1051                         running_font = basefont;
1052
1053                         if (runparams.moving_arg)
1054                                 os << "\\protect ";
1055
1056                 }
1057                 os.texrow().start(owner_->id(), i + 1);
1058                 column = 0;
1059         }
1060
1061         if (owner_->isDeleted(i)) {
1062                 if( ++runparams.inDeletedInset == 1)
1063                         runparams.changeOfDeletedInset = owner_->lookupChange(i);
1064         }
1065
1066         if (inset->canTrackChanges()) {
1067                 column += Changes::latexMarkChange(os, bparams, running_change,
1068                         Change(Change::UNCHANGED), runparams);
1069                 running_change = Change(Change::UNCHANGED);
1070         }
1071
1072         bool close = false;
1073         odocstream::pos_type const len = os.os().tellp();
1074
1075         if (inset->forceLTR()
1076             && !runparams.use_polyglossia
1077             && running_font.isRightToLeft()
1078             // ERT is an exception, it should be output with no
1079             // decorations at all
1080             && inset->lyxCode() != ERT_CODE) {
1081                 if (running_font.language()->lang() == "farsi")
1082                         os << "\\beginL" << termcmd;
1083                 else
1084                         os << "\\L{";
1085                 close = true;
1086         }
1087
1088         // FIXME: Bug: we can have an empty font change here!
1089         // if there has just been a font change, we are going to close it
1090         // right now, which means stupid latex code like \textsf{}. AFAIK,
1091         // this does not harm dvi output. A minor bug, thus (JMarc)
1092
1093         // Some insets cannot be inside a font change command.
1094         // However, even such insets *can* be placed in \L or \R
1095         // or their equivalents (for RTL language switches), so we don't
1096         // close the language in those cases.
1097         // ArabTeX, though, cannot handle this special behavior, it seems.
1098         bool arabtex = basefont.language()->lang() == "arabic_arabtex"
1099                 || running_font.language()->lang() == "arabic_arabtex";
1100         if (open_font && !inset->inheritFont()) {
1101                 bool closeLanguage = arabtex
1102                         || basefont.isRightToLeft() == running_font.isRightToLeft();
1103                 unsigned int count = running_font.latexWriteEndChanges(os,
1104                         bparams, runparams, basefont, basefont, closeLanguage);
1105                 column += count;
1106                 // if any font properties were closed, update the running_font,
1107                 // making sure, however, to leave the language as it was
1108                 if (count > 0) {
1109                         // FIXME: probably a better way to keep track of the old
1110                         // language, than copying the entire font?
1111                         Font const copy_font(running_font);
1112                         basefont = owner_->getLayoutFont(bparams, outerfont);
1113                         running_font = basefont;
1114                         if (!closeLanguage)
1115                                 running_font.setLanguage(copy_font.language());
1116                         // leave font open if language is still open
1117                         open_font = (running_font.language() == basefont.language());
1118                         if (closeLanguage)
1119                                 runparams.local_font = &basefont;
1120                 }
1121         }
1122
1123         size_t const previous_row_count = os.texrow().rows();
1124
1125         try {
1126                 runparams.lastid = id_;
1127                 runparams.lastpos = i;
1128                 inset->latex(os, runparams);
1129         } catch (EncodingException & e) {
1130                 // add location information and throw again.
1131                 e.par_id = id_;
1132                 e.pos = i;
1133                 throw(e);
1134         }
1135
1136         if (close) {
1137                 if (running_font.language()->lang() == "farsi")
1138                                 os << "\\endL" << termcmd;
1139                         else
1140                                 os << '}';
1141         }
1142
1143         if (os.texrow().rows() > previous_row_count) {
1144                 os.texrow().start(owner_->id(), i + 1);
1145                 column = 0;
1146         } else {
1147                 column += (unsigned int)(os.os().tellp() - len);
1148         }
1149
1150         if (owner_->isDeleted(i))
1151                 --runparams.inDeletedInset;
1152 }
1153
1154
1155 void Paragraph::Private::latexSpecialChar(otexstream & os,
1156                                           BufferParams const & bparams,
1157                                           OutputParams const & runparams,
1158                                           Font const & running_font,
1159                                           Change const & running_change,
1160                                           Layout const & style,
1161                                           pos_type & i,
1162                                           pos_type end_pos,
1163                                           unsigned int & column)
1164 {
1165         // With polyglossia, brackets and stuff need not be reversed
1166         // in RTL scripts (see bug #8251)
1167         char_type const c = (runparams.use_polyglossia) ?
1168                 owner_->getUChar(bparams, i) : text_[i];
1169
1170         if (style.pass_thru || runparams.pass_thru
1171             || contains(style.pass_thru_chars, c)
1172             || contains(runparams.pass_thru_chars, c)) {
1173                 if (c != '\0') {
1174                         Encoding const * const enc = runparams.encoding;
1175                         if (enc && !enc->encodable(c))
1176                                 throw EncodingException(c);
1177                         os.put(c);
1178                 }
1179                 return;
1180         }
1181
1182         // TIPA uses its own T3 encoding
1183         if (runparams.inIPA && latexSpecialT3(c, os, i, column))
1184                 return;
1185         // If T1 font encoding is used, use the special
1186         // characters it provides.
1187         // NOTE: Some languages reset the font encoding internally to a
1188         //       non-standard font encoding. If we are using such a language,
1189         //       we do not output special T1 chars.
1190         // NOTE: XeTeX and LuaTeX use OT1 (pre 2017) or TU (as of 2017) encoding
1191         if (!runparams.inIPA && !running_font.language()->internalFontEncoding()
1192             && !runparams.isFullUnicode() && bparams.font_encoding() == "T1"
1193             && latexSpecialT1(c, os, i, column))
1194                 return;
1195
1196         // Otherwise, we use what LaTeX provides us.
1197         switch (c) {
1198         case '\\':
1199                 os << "\\textbackslash" << termcmd;
1200                 column += 15;
1201                 break;
1202         case '<':
1203                 os << "\\textless" << termcmd;
1204                 column += 10;
1205                 break;
1206         case '>':
1207                 os << "\\textgreater" << termcmd;
1208                 column += 13;
1209                 break;
1210         case '|':
1211                 os << "\\textbar" << termcmd;
1212                 column += 9;
1213                 break;
1214         case '-':
1215                 os << '-';
1216                 if (i + 1 < static_cast<pos_type>(text_.size()) &&
1217                     (end_pos == -1 || i + 1 < end_pos) &&
1218                     text_[i+1] == '-') {
1219                         // Prevent "--" becoming an endash and "---" becoming
1220                         // an emdash.
1221                         // Within \ttfamily, "--" is merged to "-" (no endash)
1222                         // so we avoid this rather irritating ligature as well
1223                         os << "{}";
1224                         column += 2;
1225                 }
1226                 break;
1227         case '\"':
1228                 os << "\\char`\\\"" << termcmd;
1229                 column += 9;
1230                 break;
1231
1232         case '$': case '&':
1233         case '%': case '#': case '{':
1234         case '}': case '_':
1235                 os << '\\';
1236                 os.put(c);
1237                 column += 1;
1238                 break;
1239
1240         case '~':
1241                 os << "\\textasciitilde" << termcmd;
1242                 column += 16;
1243                 break;
1244
1245         case '^':
1246                 os << "\\textasciicircum" << termcmd;
1247                 column += 17;
1248                 break;
1249
1250         case '*':
1251         case '[':
1252         case ']':
1253                 // avoid being mistaken for optional arguments
1254                 os << '{';
1255                 os.put(c);
1256                 os << '}';
1257                 column += 2;
1258                 break;
1259
1260         case ' ':
1261                 // Blanks are printed before font switching.
1262                 // Sure? I am not! (try nice-latex)
1263                 // I am sure it's correct. LyX might be smarter
1264                 // in the future, but for now, nothing wrong is
1265                 // written. (Asger)
1266                 break;
1267
1268         default:
1269                 if (c == '\0')
1270                         return;
1271
1272                 Encoding const & encoding = *(runparams.encoding);
1273                 char_type next = '\0';
1274                 if (i + 1 < int(text_.size())) {
1275                         next = text_[i + 1];
1276                         if (Encodings::isCombiningChar(next)) {
1277                                 column += latexSurrogatePair(os, c, next, runparams) - 1;
1278                                 ++i;
1279                                 break;
1280                         }
1281                 }
1282                 string script;
1283                 pair<docstring, bool> latex = encoding.latexChar(c);
1284                 docstring nextlatex;
1285                 bool nexttipas = false;
1286                 string nexttipashortcut;
1287                 if (next != '\0' && next != META_INSET && encoding.encodable(next)) {
1288                         nextlatex = encoding.latexChar(next).first;
1289                         if (runparams.inIPA) {
1290                                 nexttipashortcut = Encodings::TIPAShortcut(next);
1291                                 nexttipas = !nexttipashortcut.empty();
1292                         }
1293                 }
1294                 bool tipas = false;
1295                 if (runparams.inIPA) {
1296                         string const tipashortcut = Encodings::TIPAShortcut(c);
1297                         if (!tipashortcut.empty()) {
1298                                 latex.first = from_ascii(tipashortcut);
1299                                 latex.second = false;
1300                                 tipas = true;
1301                         }
1302                 }
1303                 if (Encodings::isKnownScriptChar(c, script)
1304                     && prefixIs(latex.first, from_ascii("\\" + script)))
1305                         column += writeScriptChars(os, latex.first,
1306                                         running_change, encoding, i) - 1;
1307                 else if (latex.second
1308                          && ((!prefixIs(nextlatex, '\\')
1309                                && !prefixIs(nextlatex, '{')
1310                                && !prefixIs(nextlatex, '}'))
1311                              || (nexttipas
1312                                  && !prefixIs(from_ascii(nexttipashortcut), '\\')))
1313                          && !tipas) {
1314                         // Prevent eating of a following
1315                         // space or command corruption by
1316                         // following characters
1317                         if (next == ' ' || next == '\0') {
1318                                 column += latex.first.length() + 1;
1319                                 os << latex.first << "{}";
1320                         } else {
1321                                 column += latex.first.length();
1322                                 os << latex.first << " ";
1323                         }
1324                 } else {
1325                         column += latex.first.length() - 1;
1326                         os << latex.first;
1327                 }
1328                 break;
1329         }
1330 }
1331
1332
1333 bool Paragraph::Private::latexSpecialT1(char_type const c, otexstream & os,
1334         pos_type i, unsigned int & column)
1335 {
1336         switch (c) {
1337         case '>':
1338         case '<':
1339                 os.put(c);
1340                 // In T1 encoding, these characters exist
1341                 // but we should avoid ligatures
1342                 if (i + 1 >= int(text_.size()) || text_[i + 1] != c)
1343                         return true;
1344                 os << "\\textcompwordmark" << termcmd;
1345                 column += 19;
1346                 return true;
1347         case '|':
1348                 os.put(c);
1349                 return true;
1350         case '\"':
1351                 // soul.sty breaks with \char`\"
1352                 os << "\\textquotedbl" << termcmd;
1353                 column += 14;
1354                 return true;
1355         default:
1356                 return false;
1357         }
1358 }
1359
1360
1361 bool Paragraph::Private::latexSpecialT3(char_type const c, otexstream & os,
1362         pos_type /*i*/, unsigned int & column)
1363 {
1364         switch (c) {
1365         case '*':
1366         case '[':
1367         case ']':
1368         case '\"':
1369                 os.put(c);
1370                 return true;
1371         case '|':
1372                 os << "\\textvertline" << termcmd;
1373                 column += 14;
1374                 return true;
1375         default:
1376                 return false;
1377         }
1378 }
1379
1380
1381 void Paragraph::Private::validate(LaTeXFeatures & features) const
1382 {
1383         if (layout_->inpreamble && inset_owner_) {
1384                 bool const is_command = layout_->latextype == LATEX_COMMAND;
1385                 Buffer const & buf = inset_owner_->buffer();
1386                 BufferParams const & bp = features.runparams().is_child
1387                         ? buf.masterParams() : buf.params();
1388                 Font f;
1389                 // Using a string stream here circumvents the encoding
1390                 // switching machinery of odocstream. Therefore the
1391                 // output is wrong if this paragraph contains content
1392                 // that needs to switch encoding.
1393                 otexstringstream os;
1394                 if (is_command) {
1395                         os << '\\' << from_ascii(layout_->latexname());
1396                         // we have to provide all the optional arguments here, even though
1397                         // the last one is the only one we care about.
1398                         // Separate handling of optional argument inset.
1399                         if (!layout_->latexargs().empty()) {
1400                                 OutputParams rp = features.runparams();
1401                                 rp.local_font = &owner_->getFirstFontSettings(bp);
1402                                 latexArgInsets(*owner_, os, rp, layout_->latexargs());
1403                         }
1404                         os << from_ascii(layout_->latexparam());
1405                 }
1406                 size_t const length = os.length();
1407                 // this will output "{" at the beginning, but not at the end
1408                 owner_->latex(bp, f, os, features.runparams(), 0, -1, true);
1409                 if (os.length() > length) {
1410                         if (is_command) {
1411                                 os << '}';
1412                                 if (!layout_->postcommandargs().empty()) {
1413                                         OutputParams rp = features.runparams();
1414                                         rp.local_font = &owner_->getFirstFontSettings(bp);
1415                                         latexArgInsets(*owner_, os, rp, layout_->postcommandargs(), "post:");
1416                                 }
1417                         }
1418                         features.addPreambleSnippet(os.release(), true);
1419                 }
1420         }
1421
1422         if (features.runparams().flavor == OutputParams::HTML
1423             && layout_->htmltitle()) {
1424                 features.setHTMLTitle(owner_->asString(AS_STR_INSETS | AS_STR_SKIPDELETE));
1425         }
1426
1427         // check the params.
1428         if (!params_.spacing().isDefault())
1429                 features.require("setspace");
1430
1431         // then the layouts
1432         features.useLayout(layout_->name());
1433
1434         // then the fonts
1435         fontlist_.validate(features);
1436
1437         // then the indentation
1438         if (!params_.leftIndent().zero())
1439                 features.require("ParagraphLeftIndent");
1440
1441         // then the insets
1442         InsetList::const_iterator icit = insetlist_.begin();
1443         InsetList::const_iterator iend = insetlist_.end();
1444         for (; icit != iend; ++icit) {
1445                 if (icit->inset) {
1446                         features.inDeletedInset(owner_->isDeleted(icit->pos));
1447                         icit->inset->validate(features);
1448                         features.inDeletedInset(false);
1449                         if (layout_->needprotect &&
1450                             icit->inset->lyxCode() == FOOT_CODE)
1451                                 features.require("NeedLyXFootnoteCode");
1452                 }
1453         }
1454
1455         // then the contents
1456         for (pos_type i = 0; i < int(text_.size()) ; ++i) {
1457                 BufferEncodings::validate(text_[i], features);
1458         }
1459 }
1460
1461 /////////////////////////////////////////////////////////////////////
1462 //
1463 // Paragraph
1464 //
1465 /////////////////////////////////////////////////////////////////////
1466
1467 namespace {
1468         Layout const emptyParagraphLayout;
1469 }
1470
1471 Paragraph::Paragraph()
1472         : d(new Paragraph::Private(this, emptyParagraphLayout))
1473 {
1474         itemdepth = 0;
1475         d->params_.clear();
1476 }
1477
1478
1479 Paragraph::Paragraph(Paragraph const & par)
1480         : itemdepth(par.itemdepth),
1481         d(new Paragraph::Private(*par.d, this))
1482 {
1483         registerWords();
1484 }
1485
1486
1487 Paragraph::Paragraph(Paragraph const & par, pos_type beg, pos_type end)
1488         : itemdepth(par.itemdepth),
1489         d(new Paragraph::Private(*par.d, this, beg, end))
1490 {
1491         registerWords();
1492 }
1493
1494
1495 Paragraph & Paragraph::operator=(Paragraph const & par)
1496 {
1497         // needed as we will destroy the private part before copying it
1498         if (&par != this) {
1499                 itemdepth = par.itemdepth;
1500
1501                 deregisterWords();
1502                 delete d;
1503                 d = new Private(*par.d, this);
1504                 registerWords();
1505         }
1506         return *this;
1507 }
1508
1509
1510 Paragraph::~Paragraph()
1511 {
1512         deregisterWords();
1513         delete d;
1514 }
1515
1516
1517 namespace {
1518
1519 // this shall be called just before every "os << ..." action.
1520 void flushString(ostream & os, docstring & s)
1521 {
1522         os << to_utf8(s);
1523         s.erase();
1524 }
1525
1526 }
1527
1528
1529 void Paragraph::write(ostream & os, BufferParams const & bparams,
1530         depth_type & dth) const
1531 {
1532         // The beginning or end of a deeper (i.e. nested) area?
1533         if (dth != d->params_.depth()) {
1534                 if (d->params_.depth() > dth) {
1535                         while (d->params_.depth() > dth) {
1536                                 os << "\n\\begin_deeper";
1537                                 ++dth;
1538                         }
1539                 } else {
1540                         while (d->params_.depth() < dth) {
1541                                 os << "\n\\end_deeper";
1542                                 --dth;
1543                         }
1544                 }
1545         }
1546
1547         // First write the layout
1548         os << "\n\\begin_layout " << to_utf8(d->layout_->name()) << '\n';
1549
1550         d->params_.write(os);
1551
1552         Font font1(inherit_font, bparams.language);
1553
1554         Change running_change = Change(Change::UNCHANGED);
1555
1556         // this string is used as a buffer to avoid repetitive calls
1557         // to to_utf8(), which turn out to be expensive (JMarc)
1558         docstring write_buffer;
1559
1560         int column = 0;
1561         for (pos_type i = 0; i <= size(); ++i) {
1562
1563                 Change const & change = lookupChange(i);
1564                 if (change != running_change)
1565                         flushString(os, write_buffer);
1566                 Changes::lyxMarkChange(os, bparams, column, running_change, change);
1567                 running_change = change;
1568
1569                 if (i == size())
1570                         break;
1571
1572                 // Write font changes
1573                 Font font2 = getFontSettings(bparams, i);
1574                 if (font2 != font1) {
1575                         flushString(os, write_buffer);
1576                         font2.lyxWriteChanges(font1, os);
1577                         column = 0;
1578                         font1 = font2;
1579                 }
1580
1581                 char_type const c = d->text_[i];
1582                 switch (c) {
1583                 case META_INSET:
1584                         if (Inset const * inset = getInset(i)) {
1585                                 flushString(os, write_buffer);
1586                                 if (inset->directWrite()) {
1587                                         // international char, let it write
1588                                         // code directly so it's shorter in
1589                                         // the file
1590                                         inset->write(os);
1591                                 } else {
1592                                         if (i)
1593                                                 os << '\n';
1594                                         os << "\\begin_inset ";
1595                                         inset->write(os);
1596                                         os << "\n\\end_inset\n\n";
1597                                         column = 0;
1598                                 }
1599                                 // FIXME This can be removed again once the mystery
1600                                 // crash has been resolved.
1601                                 os << flush;
1602                         }
1603                         break;
1604                 case '\\':
1605                         flushString(os, write_buffer);
1606                         os << "\n\\backslash\n";
1607                         column = 0;
1608                         break;
1609                 case '.':
1610                         flushString(os, write_buffer);
1611                         if (i + 1 < size() && d->text_[i + 1] == ' ') {
1612                                 os << ".\n";
1613                                 column = 0;
1614                         } else
1615                                 os << '.';
1616                         break;
1617                 default:
1618                         if ((column > 70 && c == ' ')
1619                             || column > 79) {
1620                                 flushString(os, write_buffer);
1621                                 os << '\n';
1622                                 column = 0;
1623                         }
1624                         // this check is to amend a bug. LyX sometimes
1625                         // inserts '\0' this could cause problems.
1626                         if (c != '\0')
1627                                 write_buffer.push_back(c);
1628                         else
1629                                 LYXERR0("NUL char in structure.");
1630                         ++column;
1631                         break;
1632                 }
1633         }
1634
1635         flushString(os, write_buffer);
1636         os << "\n\\end_layout\n";
1637         // FIXME This can be removed again once the mystery
1638         // crash has been resolved.
1639         os << flush;
1640 }
1641
1642
1643 void Paragraph::validate(LaTeXFeatures & features) const
1644 {
1645         d->validate(features);
1646 }
1647
1648
1649 void Paragraph::insert(pos_type start, docstring const & str,
1650                        Font const & font, Change const & change)
1651 {
1652         for (size_t i = 0, n = str.size(); i != n ; ++i)
1653                 insertChar(start + i, str[i], font, change);
1654 }
1655
1656
1657 void Paragraph::appendChar(char_type c, Font const & font,
1658                 Change const & change)
1659 {
1660         // track change
1661         d->changes_.insert(change, d->text_.size());
1662         // when appending characters, no need to update tables
1663         d->text_.push_back(c);
1664         setFont(d->text_.size() - 1, font);
1665         d->requestSpellCheck(d->text_.size() - 1);
1666 }
1667
1668
1669 void Paragraph::appendString(docstring const & s, Font const & font,
1670                 Change const & change)
1671 {
1672         pos_type end = s.size();
1673         size_t oldsize = d->text_.size();
1674         size_t newsize = oldsize + end;
1675         size_t capacity = d->text_.capacity();
1676         if (newsize >= capacity)
1677                 d->text_.reserve(max(capacity + 100, newsize));
1678
1679         // when appending characters, no need to update tables
1680         d->text_.append(s);
1681
1682         // FIXME: Optimize this!
1683         for (size_t i = oldsize; i != newsize; ++i) {
1684                 // track change
1685                 d->changes_.insert(change, i);
1686                 d->requestSpellCheck(i);
1687         }
1688         d->fontlist_.set(oldsize, font);
1689         d->fontlist_.set(newsize - 1, font);
1690 }
1691
1692
1693 void Paragraph::insertChar(pos_type pos, char_type c,
1694                            bool trackChanges)
1695 {
1696         d->insertChar(pos, c, Change(trackChanges ?
1697                            Change::INSERTED : Change::UNCHANGED));
1698 }
1699
1700
1701 void Paragraph::insertChar(pos_type pos, char_type c,
1702                            Font const & font, bool trackChanges)
1703 {
1704         d->insertChar(pos, c, Change(trackChanges ?
1705                            Change::INSERTED : Change::UNCHANGED));
1706         setFont(pos, font);
1707 }
1708
1709
1710 void Paragraph::insertChar(pos_type pos, char_type c,
1711                            Font const & font, Change const & change)
1712 {
1713         d->insertChar(pos, c, change);
1714         setFont(pos, font);
1715 }
1716
1717
1718 void Paragraph::resetFonts(Font const & font)
1719 {
1720         d->fontlist_.clear();
1721         d->fontlist_.set(0, font);
1722         d->fontlist_.set(d->text_.size() - 1, font);
1723 }
1724
1725 // Gets uninstantiated font setting at position.
1726 Font const & Paragraph::getFontSettings(BufferParams const & bparams,
1727                                          pos_type pos) const
1728 {
1729         if (pos > size()) {
1730                 LYXERR0("pos: " << pos << " size: " << size());
1731                 LBUFERR(false);
1732         }
1733
1734         FontList::const_iterator cit = d->fontlist_.fontIterator(pos);
1735         if (cit != d->fontlist_.end())
1736                 return cit->font();
1737
1738         if (pos == size() && !empty())
1739                 return getFontSettings(bparams, pos - 1);
1740
1741         // Optimisation: avoid a full font instantiation if there is no
1742         // language change from previous call.
1743         static Font previous_font;
1744         static Language const * previous_lang = 0;
1745         Language const * lang = getParLanguage(bparams);
1746         if (lang != previous_lang) {
1747                 previous_lang = lang;
1748                 previous_font = Font(inherit_font, lang);
1749         }
1750         return previous_font;
1751 }
1752
1753
1754 FontSpan Paragraph::fontSpan(pos_type pos) const
1755 {
1756         LBUFERR(pos <= size());
1757
1758         if (pos == size())
1759                 return FontSpan(pos, pos);
1760
1761         pos_type start = 0;
1762         FontList::const_iterator cit = d->fontlist_.begin();
1763         FontList::const_iterator end = d->fontlist_.end();
1764         for (; cit != end; ++cit) {
1765                 if (cit->pos() >= pos) {
1766                         if (pos >= beginOfBody())
1767                                 return FontSpan(max(start, beginOfBody()),
1768                                                 cit->pos());
1769                         else
1770                                 return FontSpan(start,
1771                                                 min(beginOfBody() - 1,
1772                                                          cit->pos()));
1773                 }
1774                 start = cit->pos() + 1;
1775         }
1776
1777         // This should not happen, but if so, we take no chances.
1778         LYXERR0("Paragraph::fontSpan: position not found in fontinfo table!");
1779         LASSERT(false, return FontSpan(pos, pos));
1780 }
1781
1782
1783 // Gets uninstantiated font setting at position 0
1784 Font const & Paragraph::getFirstFontSettings(BufferParams const & bparams) const
1785 {
1786         if (!empty() && !d->fontlist_.empty())
1787                 return d->fontlist_.begin()->font();
1788
1789         // Optimisation: avoid a full font instantiation if there is no
1790         // language change from previous call.
1791         static Font previous_font;
1792         static Language const * previous_lang = 0;
1793         if (bparams.language != previous_lang) {
1794                 previous_lang = bparams.language;
1795                 previous_font = Font(inherit_font, bparams.language);
1796         }
1797
1798         return previous_font;
1799 }
1800
1801
1802 // Gets the fully instantiated font at a given position in a paragraph
1803 // This is basically the same function as Text::GetFont() in text2.cpp.
1804 // The difference is that this one is used for generating the LaTeX file,
1805 // and thus cosmetic "improvements" are disallowed: This has to deliver
1806 // the true picture of the buffer. (Asger)
1807 Font const Paragraph::getFont(BufferParams const & bparams, pos_type pos,
1808                                  Font const & outerfont) const
1809 {
1810         LBUFERR(pos >= 0);
1811
1812         Font font = getFontSettings(bparams, pos);
1813
1814         pos_type const body_pos = beginOfBody();
1815         FontInfo & fi = font.fontInfo();
1816         if (pos < body_pos)
1817                 fi.realize(d->layout_->labelfont);
1818         else
1819                 fi.realize(d->layout_->font);
1820
1821         fi.realize(outerfont.fontInfo());
1822         fi.realize(bparams.getFont().fontInfo());
1823
1824         return font;
1825 }
1826
1827
1828 Font const Paragraph::getLabelFont
1829         (BufferParams const & bparams, Font const & outerfont) const
1830 {
1831         FontInfo tmpfont = d->layout_->labelfont;
1832         tmpfont.realize(outerfont.fontInfo());
1833         tmpfont.realize(bparams.getFont().fontInfo());
1834         return Font(tmpfont, getParLanguage(bparams));
1835 }
1836
1837
1838 Font const Paragraph::getLayoutFont
1839         (BufferParams const & bparams, Font const & outerfont) const
1840 {
1841         FontInfo tmpfont = d->layout_->font;
1842         tmpfont.realize(outerfont.fontInfo());
1843         tmpfont.realize(bparams.getFont().fontInfo());
1844         return Font(tmpfont, getParLanguage(bparams));
1845 }
1846
1847
1848 char_type Paragraph::getUChar(BufferParams const & bparams, pos_type pos) const
1849 {
1850         char_type c = d->text_[pos];
1851         if (!getFontSettings(bparams, pos).isRightToLeft())
1852                 return c;
1853
1854         // FIXME: The arabic special casing is due to the difference of arabic
1855         // round brackets input introduced in r18599. Check if this should be
1856         // unified with Hebrew or at least if all bracket types should be
1857         // handled the same (file format change in either case).
1858         string const & lang = getFontSettings(bparams, pos).language()->lang();
1859         bool const arabic = lang == "arabic_arabtex" || lang == "arabic_arabi"
1860                 || lang == "farsi";
1861         char_type uc = c;
1862         switch (c) {
1863         case '(':
1864                 uc = arabic ? c : ')';
1865                 break;
1866         case ')':
1867                 uc = arabic ? c : '(';
1868                 break;
1869         case '[':
1870                 uc = ']';
1871                 break;
1872         case ']':
1873                 uc = '[';
1874                 break;
1875         case '{':
1876                 uc = '}';
1877                 break;
1878         case '}':
1879                 uc = '{';
1880                 break;
1881         case '<':
1882                 uc = '>';
1883                 break;
1884         case '>':
1885                 uc = '<';
1886                 break;
1887         }
1888
1889         return uc;
1890 }
1891
1892
1893 void Paragraph::setFont(pos_type pos, Font const & font)
1894 {
1895         LASSERT(pos <= size(), return);
1896
1897         // First, reduce font against layout/label font
1898         // Update: The setCharFont() routine in text2.cpp already
1899         // reduces font, so we don't need to do that here. (Asger)
1900
1901         d->fontlist_.set(pos, font);
1902 }
1903
1904
1905 void Paragraph::makeSameLayout(Paragraph const & par)
1906 {
1907         d->layout_ = par.d->layout_;
1908         d->params_ = par.d->params_;
1909 }
1910
1911
1912 bool Paragraph::stripLeadingSpaces(bool trackChanges)
1913 {
1914         if (isFreeSpacing())
1915                 return false;
1916
1917         int pos = 0;
1918         int count = 0;
1919
1920         while (pos < size() && (isNewline(pos) || isLineSeparator(pos))) {
1921                 if (eraseChar(pos, trackChanges))
1922                         ++count;
1923                 else
1924                         ++pos;
1925         }
1926
1927         return count > 0 || pos > 0;
1928 }
1929
1930
1931 bool Paragraph::hasSameLayout(Paragraph const & par) const
1932 {
1933         return par.d->layout_ == d->layout_
1934                 && d->params_.sameLayout(par.d->params_);
1935 }
1936
1937
1938 depth_type Paragraph::getDepth() const
1939 {
1940         return d->params_.depth();
1941 }
1942
1943
1944 depth_type Paragraph::getMaxDepthAfter() const
1945 {
1946         if (d->layout_->isEnvironment())
1947                 return d->params_.depth() + 1;
1948         else
1949                 return d->params_.depth();
1950 }
1951
1952
1953 LyXAlignment Paragraph::getAlign() const
1954 {
1955         if (d->params_.align() == LYX_ALIGN_LAYOUT)
1956                 return d->layout_->align;
1957         else
1958                 return d->params_.align();
1959 }
1960
1961
1962 docstring const & Paragraph::labelString() const
1963 {
1964         return d->params_.labelString();
1965 }
1966
1967
1968 // the next two functions are for the manual labels
1969 docstring const Paragraph::getLabelWidthString() const
1970 {
1971         if (d->layout_->margintype == MARGIN_MANUAL
1972             || d->layout_->latextype == LATEX_BIB_ENVIRONMENT)
1973                 return d->params_.labelWidthString();
1974         else
1975                 return _("Senseless with this layout!");
1976 }
1977
1978
1979 void Paragraph::setLabelWidthString(docstring const & s)
1980 {
1981         d->params_.labelWidthString(s);
1982 }
1983
1984
1985 docstring Paragraph::expandLabel(Layout const & layout,
1986                 BufferParams const & bparams) const
1987 {
1988         return expandParagraphLabel(layout, bparams, true);
1989 }
1990
1991
1992 docstring Paragraph::expandDocBookLabel(Layout const & layout,
1993                 BufferParams const & bparams) const
1994 {
1995         return expandParagraphLabel(layout, bparams, false);
1996 }
1997
1998
1999 docstring Paragraph::expandParagraphLabel(Layout const & layout,
2000                 BufferParams const & bparams, bool process_appendix) const
2001 {
2002         DocumentClass const & tclass = bparams.documentClass();
2003         string const & lang = getParLanguage(bparams)->code();
2004         bool const in_appendix = process_appendix && d->params_.appendix();
2005         docstring fmt = translateIfPossible(layout.labelstring(in_appendix), lang);
2006
2007         if (fmt.empty() && !layout.counter.empty())
2008                 return tclass.counters().theCounter(layout.counter, lang);
2009
2010         // handle 'inherited level parts' in 'fmt',
2011         // i.e. the stuff between '@' in   '@Section@.\arabic{subsection}'
2012         size_t const i = fmt.find('@', 0);
2013         if (i != docstring::npos) {
2014                 size_t const j = fmt.find('@', i + 1);
2015                 if (j != docstring::npos) {
2016                         docstring parent(fmt, i + 1, j - i - 1);
2017                         docstring label = from_ascii("??");
2018                         if (tclass.hasLayout(parent))
2019                                 label = expandParagraphLabel(tclass[parent], bparams,
2020                                                       process_appendix);
2021                         fmt = docstring(fmt, 0, i) + label
2022                                 + docstring(fmt, j + 1, docstring::npos);
2023                 }
2024         }
2025
2026         return tclass.counters().counterLabel(fmt, lang);
2027 }
2028
2029
2030 void Paragraph::applyLayout(Layout const & new_layout)
2031 {
2032         d->layout_ = &new_layout;
2033         LyXAlignment const oldAlign = d->params_.align();
2034
2035         if (!(oldAlign & d->layout_->alignpossible)) {
2036                 frontend::Alert::warning(_("Alignment not permitted"),
2037                         _("The new layout does not permit the alignment previously used.\nSetting to default."));
2038                 d->params_.align(LYX_ALIGN_LAYOUT);
2039         }
2040 }
2041
2042
2043 pos_type Paragraph::beginOfBody() const
2044 {
2045         return d->begin_of_body_;
2046 }
2047
2048
2049 void Paragraph::setBeginOfBody()
2050 {
2051         if (d->layout_->labeltype != LABEL_MANUAL) {
2052                 d->begin_of_body_ = 0;
2053                 return;
2054         }
2055
2056         // Unroll the first two cycles of the loop
2057         // and remember the previous character to
2058         // remove unnecessary getChar() calls
2059         pos_type i = 0;
2060         pos_type end = size();
2061         if (i < end && !(isNewline(i) || isEnvSeparator(i))) {
2062                 ++i;
2063                 if (i < end) {
2064                         char_type previous_char = d->text_[i];
2065                         if (!(isNewline(i) || isEnvSeparator(i))) {
2066                                 ++i;
2067                                 while (i < end && previous_char != ' ') {
2068                                         char_type temp = d->text_[i];
2069                                         if (isNewline(i) || isEnvSeparator(i))
2070                                                 break;
2071                                         ++i;
2072                                         previous_char = temp;
2073                                 }
2074                         }
2075                 }
2076         }
2077
2078         d->begin_of_body_ = i;
2079 }
2080
2081
2082 bool Paragraph::allowParagraphCustomization() const
2083 {
2084         return inInset().allowParagraphCustomization();
2085 }
2086
2087
2088 bool Paragraph::usePlainLayout() const
2089 {
2090         return inInset().usePlainLayout();
2091 }
2092
2093
2094 bool Paragraph::isPassThru() const
2095 {
2096         return inInset().isPassThru() || d->layout_->pass_thru;
2097 }
2098
2099 namespace {
2100
2101 // paragraphs inside floats need different alignment tags to avoid
2102 // unwanted space
2103
2104 bool noTrivlistCentering(InsetCode code)
2105 {
2106         return code == FLOAT_CODE
2107                || code == WRAP_CODE
2108                || code == CELL_CODE;
2109 }
2110
2111
2112 string correction(string const & orig)
2113 {
2114         if (orig == "flushleft")
2115                 return "raggedright";
2116         if (orig == "flushright")
2117                 return "raggedleft";
2118         if (orig == "center")
2119                 return "centering";
2120         return orig;
2121 }
2122
2123
2124 bool corrected_env(otexstream & os, string const & suffix, string const & env,
2125         InsetCode code, bool const lastpar, int & col)
2126 {
2127         string macro = suffix + "{";
2128         if (noTrivlistCentering(code)) {
2129                 if (lastpar) {
2130                         // the last paragraph in non-trivlist-aligned
2131                         // context is special (to avoid unwanted whitespace)
2132                         if (suffix == "\\begin") {
2133                                 macro = "\\" + correction(env) + "{}";
2134                                 os << from_ascii(macro);
2135                                 col += macro.size();
2136                                 return true;
2137                         }
2138                         return false;
2139                 }
2140                 macro += correction(env);
2141         } else
2142                 macro += env;
2143         macro += "}";
2144         if (suffix == "\\par\\end") {
2145                 os << breakln;
2146                 col = 0;
2147         }
2148         os << from_ascii(macro);
2149         col += macro.size();
2150         if (suffix == "\\begin") {
2151                 os << breakln;
2152                 col = 0;
2153         }
2154         return true;
2155 }
2156
2157 } // namespace anon
2158
2159
2160 int Paragraph::Private::startTeXParParams(BufferParams const & bparams,
2161                         otexstream & os, OutputParams const & runparams) const
2162 {
2163         int column = 0;
2164
2165         bool canindent =
2166                 (bparams.paragraph_separation == BufferParams::ParagraphIndentSeparation) ?
2167                         (layout_->toggle_indent != ITOGGLE_NEVER) :
2168                         (layout_->toggle_indent == ITOGGLE_ALWAYS);
2169
2170         if (canindent && params_.noindent() && !layout_->pass_thru) {
2171                 os << "\\noindent ";
2172                 column += 10;
2173         }
2174
2175         LyXAlignment const curAlign = params_.align();
2176
2177         if (curAlign == layout_->align)
2178                 return column;
2179
2180         switch (curAlign) {
2181         case LYX_ALIGN_NONE:
2182         case LYX_ALIGN_BLOCK:
2183         case LYX_ALIGN_LAYOUT:
2184         case LYX_ALIGN_SPECIAL:
2185         case LYX_ALIGN_DECIMAL:
2186                 break;
2187         case LYX_ALIGN_LEFT:
2188         case LYX_ALIGN_RIGHT:
2189         case LYX_ALIGN_CENTER:
2190                 if (runparams.moving_arg) {
2191                         os << "\\protect";
2192                         column += 8;
2193                 }
2194                 break;
2195         }
2196
2197         string const begin_tag = "\\begin";
2198         InsetCode code = ownerCode();
2199         bool const lastpar = runparams.isLastPar;
2200
2201         switch (curAlign) {
2202         case LYX_ALIGN_NONE:
2203         case LYX_ALIGN_BLOCK:
2204         case LYX_ALIGN_LAYOUT:
2205         case LYX_ALIGN_SPECIAL:
2206         case LYX_ALIGN_DECIMAL:
2207                 break;
2208         case LYX_ALIGN_LEFT: {
2209                 if (owner_->getParLanguage(bparams)->babel() != "hebrew")
2210                         corrected_env(os, begin_tag, "flushleft", code, lastpar, column);
2211                 else
2212                         corrected_env(os, begin_tag, "flushright", code, lastpar, column);
2213                 break;
2214         } case LYX_ALIGN_RIGHT: {
2215                 if (owner_->getParLanguage(bparams)->babel() != "hebrew")
2216                         corrected_env(os, begin_tag, "flushright", code, lastpar, column);
2217                 else
2218                         corrected_env(os, begin_tag, "flushleft", code, lastpar, column);
2219                 break;
2220         } case LYX_ALIGN_CENTER: {
2221                 corrected_env(os, begin_tag, "center", code, lastpar, column);
2222                 break;
2223         }
2224         }
2225
2226         return column;
2227 }
2228
2229
2230 bool Paragraph::Private::endTeXParParams(BufferParams const & bparams,
2231                         otexstream & os, OutputParams const & runparams) const
2232 {
2233         LyXAlignment const curAlign = params_.align();
2234
2235         if (curAlign == layout_->align)
2236                 return false;
2237
2238         switch (curAlign) {
2239         case LYX_ALIGN_NONE:
2240         case LYX_ALIGN_BLOCK:
2241         case LYX_ALIGN_LAYOUT:
2242         case LYX_ALIGN_SPECIAL:
2243         case LYX_ALIGN_DECIMAL:
2244                 break;
2245         case LYX_ALIGN_LEFT:
2246         case LYX_ALIGN_RIGHT:
2247         case LYX_ALIGN_CENTER:
2248                 if (runparams.moving_arg)
2249                         os << "\\protect";
2250                 break;
2251         }
2252
2253         bool output = false;
2254         int col = 0;
2255         string const end_tag = "\\par\\end";
2256         InsetCode code = ownerCode();
2257         bool const lastpar = runparams.isLastPar;
2258
2259         switch (curAlign) {
2260         case LYX_ALIGN_NONE:
2261         case LYX_ALIGN_BLOCK:
2262         case LYX_ALIGN_LAYOUT:
2263         case LYX_ALIGN_SPECIAL:
2264         case LYX_ALIGN_DECIMAL:
2265                 break;
2266         case LYX_ALIGN_LEFT: {
2267                 if (owner_->getParLanguage(bparams)->babel() != "hebrew")
2268                         output = corrected_env(os, end_tag, "flushleft", code, lastpar, col);
2269                 else
2270                         output = corrected_env(os, end_tag, "flushright", code, lastpar, col);
2271                 break;
2272         } case LYX_ALIGN_RIGHT: {
2273                 if (owner_->getParLanguage(bparams)->babel() != "hebrew")
2274                         output = corrected_env(os, end_tag, "flushright", code, lastpar, col);
2275                 else
2276                         output = corrected_env(os, end_tag, "flushleft", code, lastpar, col);
2277                 break;
2278         } case LYX_ALIGN_CENTER: {
2279                 corrected_env(os, end_tag, "center", code, lastpar, col);
2280                 break;
2281         }
2282         }
2283
2284         return output || lastpar;
2285 }
2286
2287
2288 // This one spits out the text of the paragraph
2289 void Paragraph::latex(BufferParams const & bparams,
2290         Font const & outerfont,
2291         otexstream & os,
2292         OutputParams const & runparams,
2293         int start_pos, int end_pos, bool force) const
2294 {
2295         LYXERR(Debug::LATEX, "Paragraph::latex...     " << this);
2296
2297         // FIXME This check should not be needed. Perhaps issue an
2298         // error if it triggers.
2299         Layout const & style = inInset().forcePlainLayout() ?
2300                 bparams.documentClass().plainLayout() : *d->layout_;
2301
2302         if (!force && style.inpreamble)
2303                 return;
2304
2305         bool const allowcust = allowParagraphCustomization();
2306
2307         // Current base font for all inherited font changes, without any
2308         // change caused by an individual character, except for the language:
2309         // It is set to the language of the first character.
2310         // As long as we are in the label, this font is the base font of the
2311         // label. Before the first body character it is set to the base font
2312         // of the body.
2313         Font basefont;
2314
2315         // Maybe we have to create a optional argument.
2316         pos_type body_pos = beginOfBody();
2317         unsigned int column = 0;
2318
2319         if (body_pos > 0) {
2320                 // the optional argument is kept in curly brackets in
2321                 // case it contains a ']'
2322                 // This is not strictly needed, but if this is changed it
2323                 // would be a file format change, and tex2lyx would need
2324                 // to be adjusted, since it unconditionally removes the
2325                 // braces when it parses \item.
2326                 os << "[{";
2327                 column += 2;
2328                 basefont = getLabelFont(bparams, outerfont);
2329         } else {
2330                 basefont = getLayoutFont(bparams, outerfont);
2331         }
2332
2333         // Which font is currently active?
2334         Font running_font(basefont);
2335         // Do we have an open font change?
2336         bool open_font = false;
2337
2338         Change runningChange = Change(Change::UNCHANGED);
2339
2340         Encoding const * const prev_encoding = runparams.encoding;
2341
2342         os.texrow().start(id(), 0);
2343
2344         // if the paragraph is empty, the loop will not be entered at all
2345         if (empty()) {
2346                 if (style.isCommand()) {
2347                         os << '{';
2348                         ++column;
2349                 }
2350                 if (!style.leftdelim().empty()) {
2351                         os << style.leftdelim();
2352                         column += style.leftdelim().size();
2353                 }
2354                 if (allowcust)
2355                         column += d->startTeXParParams(bparams, os, runparams);
2356         }
2357
2358         for (pos_type i = 0; i < size(); ++i) {
2359                 // First char in paragraph or after label?
2360                 if (i == body_pos) {
2361                         if (body_pos > 0) {
2362                                 if (open_font) {
2363                                         column += running_font.latexWriteEndChanges(
2364                                                 os, bparams, runparams,
2365                                                 basefont, basefont);
2366                                         open_font = false;
2367                                 }
2368                                 basefont = getLayoutFont(bparams, outerfont);
2369                                 running_font = basefont;
2370
2371                                 column += Changes::latexMarkChange(os, bparams,
2372                                                 runningChange, Change(Change::UNCHANGED),
2373                                                 runparams);
2374                                 runningChange = Change(Change::UNCHANGED);
2375
2376                                 os << "}] ";
2377                                 column +=3;
2378                         }
2379                         if (style.isCommand()) {
2380                                 os << '{';
2381                                 ++column;
2382                         }
2383
2384                         if (!style.leftdelim().empty()) {
2385                                 os << style.leftdelim();
2386                                 column += style.leftdelim().size();
2387                         }
2388
2389                         if (allowcust)
2390                                 column += d->startTeXParParams(bparams, os,
2391                                                             runparams);
2392                 }
2393
2394                 runparams.wasDisplayMath = runparams.inDisplayMath;
2395                 runparams.inDisplayMath = false;
2396                 bool deleted_display_math = false;
2397
2398                 // Check whether a display math inset follows
2399                 if (d->text_[i] == META_INSET
2400                     && i >= start_pos && (end_pos == -1 || i < end_pos)) {
2401                         InsetMath const * im = getInset(i)->asInsetMath();
2402                         if (im && im->asHullInset()
2403                             && im->asHullInset()->outerDisplay()) {
2404                                 runparams.inDisplayMath = true;
2405                                 // runparams.inDeletedInset will be set by
2406                                 // latexInset later, but we need this info
2407                                 // before it is called. On the other hand, we
2408                                 // cannot set it here because it is a counter.
2409                                 deleted_display_math = isDeleted(i);
2410                         }
2411                 }
2412
2413                 Change const & change = runparams.inDeletedInset
2414                         ? runparams.changeOfDeletedInset : lookupChange(i);
2415
2416                 if (bparams.output_changes && runningChange != change) {
2417                         if (open_font) {
2418                                 column += running_font.latexWriteEndChanges(
2419                                                 os, bparams, runparams, basefont, basefont);
2420                                 open_font = false;
2421                         }
2422                         basefont = getLayoutFont(bparams, outerfont);
2423                         running_font = basefont;
2424                         column += Changes::latexMarkChange(os, bparams, runningChange,
2425                                                            change, runparams);
2426                         runningChange = change;
2427                 }
2428
2429                 // do not output text which is marked deleted
2430                 // if change tracking output is disabled
2431                 if (!bparams.output_changes && change.deleted()) {
2432                         continue;
2433                 }
2434
2435                 ++column;
2436
2437                 // Fully instantiated font
2438                 Font const current_font = getFont(bparams, i, outerfont);
2439
2440                 Font const last_font = running_font;
2441
2442                 // Do we need to close the previous font?
2443                 if (open_font &&
2444                     (current_font != running_font ||
2445                      current_font.language() != running_font.language()))
2446                 {
2447                         column += running_font.latexWriteEndChanges(
2448                                         os, bparams, runparams, basefont,
2449                                         (i == body_pos-1) ? basefont : current_font);
2450                         running_font = basefont;
2451                         open_font = false;
2452                 }
2453
2454                 string const running_lang = runparams.use_polyglossia ?
2455                         running_font.language()->polyglossia() : running_font.language()->babel();
2456                 // close babel's font environment before opening CJK.
2457                 string const lang_end_command = runparams.use_polyglossia ?
2458                         "\\end{$$lang}" : lyxrc.language_command_end;
2459                 if (!running_lang.empty() &&
2460                     current_font.language()->encoding()->package() == Encoding::CJK) {
2461                                 string end_tag = subst(lang_end_command,
2462                                                         "$$lang",
2463                                                         running_lang);
2464                                 os << from_ascii(end_tag);
2465                                 column += end_tag.length();
2466                                 if (runparams.use_polyglossia)
2467                                         popPolyglossiaLang();
2468                 }
2469
2470                 // Switch file encoding if necessary (and allowed)
2471                 if (!runparams.pass_thru && !style.pass_thru &&
2472                     runparams.encoding->package() != Encoding::none &&
2473                     current_font.language()->encoding()->package() != Encoding::none) {
2474                         pair<bool, int> const enc_switch =
2475                                 switchEncoding(os.os(), bparams, runparams,
2476                                         *(current_font.language()->encoding()));
2477                         if (enc_switch.first) {
2478                                 column += enc_switch.second;
2479                                 runparams.encoding = current_font.language()->encoding();
2480                         }
2481                 }
2482
2483                 char_type const c = d->text_[i];
2484
2485                 // A display math inset inside an ulem command will be output
2486                 // as a box of width \columnwidth, so we have to either disable
2487                 // indentation if the inset starts a paragraph, or start a new
2488                 // line to accommodate such box. This has to be done before
2489                 // writing any font changing commands.
2490                 if (runparams.inDisplayMath && !deleted_display_math
2491                     && runparams.inulemcmd) {
2492                         if (os.afterParbreak())
2493                                 os << "\\noindent";
2494                         else
2495                                 os << "\\\\\n";
2496                 }
2497
2498                 // Do we need to change font?
2499                 if ((current_font != running_font ||
2500                      current_font.language() != running_font.language()) &&
2501                         i != body_pos - 1)
2502                 {
2503                         odocstringstream ods;
2504                         column += current_font.latexWriteStartChanges(ods, bparams,
2505                                                               runparams, basefont,
2506                                                               last_font);
2507                         // Check again for display math in ulem commands as a
2508                         // font change may also occur just before a math inset.
2509                         if (runparams.inDisplayMath && !deleted_display_math
2510                             && runparams.inulemcmd) {
2511                                 if (os.afterParbreak())
2512                                         os << "\\noindent";
2513                                 else
2514                                         os << "\\\\\n";
2515                         }
2516                         running_font = current_font;
2517                         open_font = true;
2518                         docstring fontchange = ods.str();
2519                         // check whether the fontchange ends with a \\textcolor
2520                         // modifier and the text starts with a space (bug 4473)
2521                         docstring const last_modifier = rsplit(fontchange, '\\');
2522                         if (prefixIs(last_modifier, from_ascii("textcolor")) && c == ' ')
2523                                 os << fontchange << from_ascii("{}");
2524                         // check if the fontchange ends with a trailing blank
2525                         // (like "\small " (see bug 3382)
2526                         else if (suffixIs(fontchange, ' ') && c == ' ')
2527                                 os << fontchange.substr(0, fontchange.size() - 1)
2528                                    << from_ascii("{}");
2529                         else
2530                                 os << fontchange;
2531                 }
2532
2533                 // FIXME: think about end_pos implementation...
2534                 if (c == ' ' && i >= start_pos && (end_pos == -1 || i < end_pos)) {
2535                         // FIXME: integrate this case in latexSpecialChar
2536                         // Do not print the separation of the optional argument
2537                         // if style.pass_thru is false. This works because
2538                         // latexSpecialChar ignores spaces if
2539                         // style.pass_thru is false.
2540                         if (i != body_pos - 1) {
2541                                 if (d->simpleTeXBlanks(runparams, os,
2542                                                 i, column, current_font, style)) {
2543                                         // A surrogate pair was output. We
2544                                         // must not call latexSpecialChar
2545                                         // in this iteration, since it would output
2546                                         // the combining character again.
2547                                         ++i;
2548                                         continue;
2549                                 }
2550                         }
2551                 }
2552
2553                 OutputParams rp = runparams;
2554                 rp.free_spacing = style.free_spacing;
2555                 rp.local_font = &current_font;
2556                 rp.intitle = style.intitle;
2557
2558                 // Two major modes:  LaTeX or plain
2559                 // Handle here those cases common to both modes
2560                 // and then split to handle the two modes separately.
2561                 if (c == META_INSET) {
2562                         if (i >= start_pos && (end_pos == -1 || i < end_pos)) {
2563                                 d->latexInset(bparams, os, rp, running_font,
2564                                                 basefont, outerfont, open_font,
2565                                                 runningChange, style, i, column);
2566                         }
2567                 } else if (i >= start_pos && (end_pos == -1 || i < end_pos)) {
2568                         try {
2569                                 d->latexSpecialChar(os, bparams, rp,
2570                                                     running_font, runningChange,
2571                                                     style, i, end_pos, column);
2572                         } catch (EncodingException & e) {
2573                                 if (runparams.dryrun) {
2574                                         os << "<" << _("LyX Warning: ")
2575                                            << _("uncodable character") << " '";
2576                                         os.put(c);
2577                                         os << "'>";
2578                                 } else {
2579                                         // add location information and throw again.
2580                                         e.par_id = id();
2581                                         e.pos = i;
2582                                         throw(e);
2583                                 }
2584                         }
2585                 }
2586
2587                 // Set the encoding to that returned from latexSpecialChar (see
2588                 // comment for encoding member in OutputParams.h)
2589                 runparams.encoding = rp.encoding;
2590
2591                 // Also carry on the info on a closed ulem command for insets
2592                 // such as Note that do not produce any output, so that no
2593                 // command is ever executed but its opening was recorded.
2594                 runparams.inulemcmd = rp.inulemcmd;
2595         }
2596
2597         // If we have an open font definition, we have to close it
2598         if (open_font) {
2599 #ifdef FIXED_LANGUAGE_END_DETECTION
2600                 if (next_) {
2601                         running_font.latexWriteEndChanges(os, bparams,
2602                                         runparams, basefont,
2603                                         next_->getFont(bparams, 0, outerfont));
2604                 } else {
2605                         running_font.latexWriteEndChanges(os, bparams,
2606                                         runparams, basefont, basefont);
2607                 }
2608 #else
2609 //FIXME: For now we ALWAYS have to close the foreign font settings if they are
2610 //FIXME: there as we start another \selectlanguage with the next paragraph if
2611 //FIXME: we are in need of this. This should be fixed sometime (Jug)
2612                 running_font.latexWriteEndChanges(os, bparams, runparams,
2613                                 basefont, basefont);
2614 #endif
2615         }
2616
2617         column += Changes::latexMarkChange(os, bparams, runningChange,
2618                                            Change(Change::UNCHANGED), runparams);
2619
2620         // Needed if there is an optional argument but no contents.
2621         if (body_pos > 0 && body_pos == size()) {
2622                 os << "}]~";
2623         }
2624
2625         if (!style.rightdelim().empty()) {
2626                 os << style.rightdelim();
2627                 column += style.rightdelim().size();
2628         }
2629
2630         if (allowcust && d->endTeXParParams(bparams, os, runparams)
2631             && runparams.encoding != prev_encoding) {
2632                 runparams.encoding = prev_encoding;
2633                 os << setEncoding(prev_encoding->iconvName());
2634         }
2635
2636         LYXERR(Debug::LATEX, "Paragraph::latex... done " << this);
2637 }
2638
2639
2640 bool Paragraph::emptyTag() const
2641 {
2642         for (pos_type i = 0; i < size(); ++i) {
2643                 if (Inset const * inset = getInset(i)) {
2644                         InsetCode lyx_code = inset->lyxCode();
2645                         // FIXME testing like that is wrong. What is
2646                         // the intent?
2647                         if (lyx_code != TOC_CODE &&
2648                             lyx_code != INCLUDE_CODE &&
2649                             lyx_code != GRAPHICS_CODE &&
2650                             lyx_code != ERT_CODE &&
2651                             lyx_code != LISTINGS_CODE &&
2652                             lyx_code != FLOAT_CODE &&
2653                             lyx_code != TABULAR_CODE) {
2654                                 return false;
2655                         }
2656                 } else {
2657                         char_type c = d->text_[i];
2658                         if (c != ' ' && c != '\t')
2659                                 return false;
2660                 }
2661         }
2662         return true;
2663 }
2664
2665
2666 string Paragraph::getID(Buffer const & buf, OutputParams const & runparams)
2667         const
2668 {
2669         for (pos_type i = 0; i < size(); ++i) {
2670                 if (Inset const * inset = getInset(i)) {
2671                         InsetCode lyx_code = inset->lyxCode();
2672                         if (lyx_code == LABEL_CODE) {
2673                                 InsetLabel const * const il = static_cast<InsetLabel const *>(inset);
2674                                 docstring const & id = il->getParam("name");
2675                                 return "id='" + to_utf8(sgml::cleanID(buf, runparams, id)) + "'";
2676                         }
2677                 }
2678         }
2679         return string();
2680 }
2681
2682
2683 pos_type Paragraph::firstWordDocBook(odocstream & os, OutputParams const & runparams)
2684         const
2685 {
2686         pos_type i;
2687         for (i = 0; i < size(); ++i) {
2688                 if (Inset const * inset = getInset(i)) {
2689                         inset->docbook(os, runparams);
2690                 } else {
2691                         char_type c = d->text_[i];
2692                         if (c == ' ')
2693                                 break;
2694                         os << sgml::escapeChar(c);
2695                 }
2696         }
2697         return i;
2698 }
2699
2700
2701 pos_type Paragraph::firstWordLyXHTML(XHTMLStream & xs, OutputParams const & runparams)
2702         const
2703 {
2704         pos_type i;
2705         for (i = 0; i < size(); ++i) {
2706                 if (Inset const * inset = getInset(i)) {
2707                         inset->xhtml(xs, runparams);
2708                 } else {
2709                         char_type c = d->text_[i];
2710                         if (c == ' ')
2711                                 break;
2712                         xs << c;
2713                 }
2714         }
2715         return i;
2716 }
2717
2718
2719 bool Paragraph::Private::onlyText(Buffer const & buf, Font const & outerfont, pos_type initial) const
2720 {
2721         Font font_old;
2722         pos_type size = text_.size();
2723         for (pos_type i = initial; i < size; ++i) {
2724                 Font font = owner_->getFont(buf.params(), i, outerfont);
2725                 if (text_[i] == META_INSET)
2726                         return false;
2727                 if (i != initial && font != font_old)
2728                         return false;
2729                 font_old = font;
2730         }
2731
2732         return true;
2733 }
2734
2735
2736 void Paragraph::simpleDocBookOnePar(Buffer const & buf,
2737                                     odocstream & os,
2738                                     OutputParams const & runparams,
2739                                     Font const & outerfont,
2740                                     pos_type initial) const
2741 {
2742         bool emph_flag = false;
2743
2744         Layout const & style = *d->layout_;
2745         FontInfo font_old =
2746                 style.labeltype == LABEL_MANUAL ? style.labelfont : style.font;
2747
2748         if (style.pass_thru && !d->onlyText(buf, outerfont, initial))
2749                 os << "]]>";
2750
2751         // parsing main loop
2752         for (pos_type i = initial; i < size(); ++i) {
2753                 Font font = getFont(buf.params(), i, outerfont);
2754
2755                 // handle <emphasis> tag
2756                 if (font_old.emph() != font.fontInfo().emph()) {
2757                         if (font.fontInfo().emph() == FONT_ON) {
2758                                 os << "<emphasis>";
2759                                 emph_flag = true;
2760                         } else if (i != initial) {
2761                                 os << "</emphasis>";
2762                                 emph_flag = false;
2763                         }
2764                 }
2765
2766                 if (Inset const * inset = getInset(i)) {
2767                         inset->docbook(os, runparams);
2768                 } else {
2769                         char_type c = d->text_[i];
2770
2771                         if (style.pass_thru)
2772                                 os.put(c);
2773                         else
2774                                 os << sgml::escapeChar(c);
2775                 }
2776                 font_old = font.fontInfo();
2777         }
2778
2779         if (emph_flag) {
2780                 os << "</emphasis>";
2781         }
2782
2783         if (style.free_spacing)
2784                 os << '\n';
2785         if (style.pass_thru && !d->onlyText(buf, outerfont, initial))
2786                 os << "<![CDATA[";
2787 }
2788
2789
2790 namespace {
2791 void doFontSwitch(vector<html::FontTag> & tagsToOpen,
2792                   vector<html::EndFontTag> & tagsToClose,
2793                   bool & flag, FontState curstate, html::FontTypes type)
2794 {
2795         if (curstate == FONT_ON) {
2796                 tagsToOpen.push_back(html::FontTag(type));
2797                 flag = true;
2798         } else if (flag) {
2799                 tagsToClose.push_back(html::EndFontTag(type));
2800                 flag = false;
2801         }
2802 }
2803 }
2804
2805
2806 docstring Paragraph::simpleLyXHTMLOnePar(Buffer const & buf,
2807                                     XHTMLStream & xs,
2808                                     OutputParams const & runparams,
2809                                     Font const & outerfont,
2810                                     bool start_paragraph, bool close_paragraph,
2811                                     pos_type initial) const
2812 {
2813         docstring retval;
2814
2815         // track whether we have opened these tags
2816         bool emph_flag = false;
2817         bool bold_flag = false;
2818         bool noun_flag = false;
2819         bool ubar_flag = false;
2820         bool dbar_flag = false;
2821         bool sout_flag = false;
2822         bool wave_flag = false;
2823         // shape tags
2824         bool shap_flag = false;
2825         // family tags
2826         bool faml_flag = false;
2827         // size tags
2828         bool size_flag = false;
2829
2830         Layout const & style = *d->layout_;
2831
2832         if (start_paragraph)
2833                 xs.startDivision(allowEmpty());
2834
2835         FontInfo font_old =
2836                 style.labeltype == LABEL_MANUAL ? style.labelfont : style.font;
2837
2838         FontShape  curr_fs   = INHERIT_SHAPE;
2839         FontFamily curr_fam  = INHERIT_FAMILY;
2840         FontSize   curr_size = FONT_SIZE_INHERIT;
2841
2842         string const default_family =
2843                 buf.masterBuffer()->params().fonts_default_family;
2844
2845         vector<html::FontTag> tagsToOpen;
2846         vector<html::EndFontTag> tagsToClose;
2847
2848         // parsing main loop
2849         for (pos_type i = initial; i < size(); ++i) {
2850                 // let's not show deleted material in the output
2851                 if (isDeleted(i))
2852                         continue;
2853
2854                 Font const font = getFont(buf.masterBuffer()->params(), i, outerfont);
2855
2856                 // emphasis
2857                 FontState curstate = font.fontInfo().emph();
2858                 if (font_old.emph() != curstate)
2859                         doFontSwitch(tagsToOpen, tagsToClose, emph_flag, curstate, html::FT_EMPH);
2860
2861                 // noun
2862                 curstate = font.fontInfo().noun();
2863                 if (font_old.noun() != curstate)
2864                         doFontSwitch(tagsToOpen, tagsToClose, noun_flag, curstate, html::FT_NOUN);
2865
2866                 // underbar
2867                 curstate = font.fontInfo().underbar();
2868                 if (font_old.underbar() != curstate)
2869                         doFontSwitch(tagsToOpen, tagsToClose, ubar_flag, curstate, html::FT_UBAR);
2870
2871                 // strikeout
2872                 curstate = font.fontInfo().strikeout();
2873                 if (font_old.strikeout() != curstate)
2874                         doFontSwitch(tagsToOpen, tagsToClose, sout_flag, curstate, html::FT_SOUT);
2875
2876                 // double underbar
2877                 curstate = font.fontInfo().uuline();
2878                 if (font_old.uuline() != curstate)
2879                         doFontSwitch(tagsToOpen, tagsToClose, dbar_flag, curstate, html::FT_DBAR);
2880
2881                 // wavy line
2882                 curstate = font.fontInfo().uwave();
2883                 if (font_old.uwave() != curstate)
2884                         doFontSwitch(tagsToOpen, tagsToClose, wave_flag, curstate, html::FT_WAVE);
2885
2886                 // bold
2887                 // a little hackish, but allows us to reuse what we have.
2888                 curstate = (font.fontInfo().series() == BOLD_SERIES ? FONT_ON : FONT_OFF);
2889                 if (font_old.series() != font.fontInfo().series())
2890                         doFontSwitch(tagsToOpen, tagsToClose, bold_flag, curstate, html::FT_BOLD);
2891
2892                 // Font shape
2893                 curr_fs = font.fontInfo().shape();
2894                 FontShape old_fs = font_old.shape();
2895                 if (old_fs != curr_fs) {
2896                         if (shap_flag) {
2897                                 switch (old_fs) {
2898                                 case ITALIC_SHAPE:
2899                                         tagsToClose.push_back(html::EndFontTag(html::FT_ITALIC));
2900                                         break;
2901                                 case SLANTED_SHAPE:
2902                                         tagsToClose.push_back(html::EndFontTag(html::FT_SLANTED));
2903                                         break;
2904                                 case SMALLCAPS_SHAPE:
2905                                         tagsToClose.push_back(html::EndFontTag(html::FT_SMALLCAPS));
2906                                         break;
2907                                 case UP_SHAPE:
2908                                 case INHERIT_SHAPE:
2909                                         break;
2910                                 default:
2911                                         // the other tags are for internal use
2912                                         LATTEST(false);
2913                                         break;
2914                                 }
2915                                 shap_flag = false;
2916                         }
2917                         switch (curr_fs) {
2918                         case ITALIC_SHAPE:
2919                                 tagsToOpen.push_back(html::FontTag(html::FT_ITALIC));
2920                                 shap_flag = true;
2921                                 break;
2922                         case SLANTED_SHAPE:
2923                                 tagsToOpen.push_back(html::FontTag(html::FT_SLANTED));
2924                                 shap_flag = true;
2925                                 break;
2926                         case SMALLCAPS_SHAPE:
2927                                 tagsToOpen.push_back(html::FontTag(html::FT_SMALLCAPS));
2928                                 shap_flag = true;
2929                                 break;
2930                         case UP_SHAPE:
2931                         case INHERIT_SHAPE:
2932                                 break;
2933                         default:
2934                                 // the other tags are for internal use
2935                                 LATTEST(false);
2936                                 break;
2937                         }
2938                 }
2939
2940                 // Font family
2941                 curr_fam = font.fontInfo().family();
2942                 FontFamily old_fam = font_old.family();
2943                 if (old_fam != curr_fam) {
2944                         if (faml_flag) {
2945                                 switch (old_fam) {
2946                                 case ROMAN_FAMILY:
2947                                         tagsToClose.push_back(html::EndFontTag(html::FT_ROMAN));
2948                                         break;
2949                                 case SANS_FAMILY:
2950                                         tagsToClose.push_back(html::EndFontTag(html::FT_SANS));
2951                                         break;
2952                                 case TYPEWRITER_FAMILY:
2953                                         tagsToClose.push_back(html::EndFontTag(html::FT_TYPE));
2954                                         break;
2955                                 case INHERIT_FAMILY:
2956                                         break;
2957                                 default:
2958                                         // the other tags are for internal use
2959                                         LATTEST(false);
2960                                         break;
2961                                 }
2962                                 faml_flag = false;
2963                         }
2964                         switch (curr_fam) {
2965                         case ROMAN_FAMILY:
2966                                 // we will treat a "default" font family as roman, since we have
2967                                 // no other idea what to do.
2968                                 if (default_family != "rmdefault" && default_family != "default") {
2969                                         tagsToOpen.push_back(html::FontTag(html::FT_ROMAN));
2970                                         faml_flag = true;
2971                                 }
2972                                 break;
2973                         case SANS_FAMILY:
2974                                 if (default_family != "sfdefault") {
2975                                         tagsToOpen.push_back(html::FontTag(html::FT_SANS));
2976                                         faml_flag = true;
2977                                 }
2978                                 break;
2979                         case TYPEWRITER_FAMILY:
2980                                 if (default_family != "ttdefault") {
2981                                         tagsToOpen.push_back(html::FontTag(html::FT_TYPE));
2982                                         faml_flag = true;
2983                                 }
2984                                 break;
2985                         case INHERIT_FAMILY:
2986                                 break;
2987                         default:
2988                                 // the other tags are for internal use
2989                                 LATTEST(false);
2990                                 break;
2991                         }
2992                 }
2993
2994                 // Font size
2995                 curr_size = font.fontInfo().size();
2996                 FontSize old_size = font_old.size();
2997                 if (old_size != curr_size) {
2998                         if (size_flag) {
2999                                 switch (old_size) {
3000                                 case FONT_SIZE_TINY:
3001                                         tagsToClose.push_back(html::EndFontTag(html::FT_SIZE_TINY));
3002                                         break;
3003                                 case FONT_SIZE_SCRIPT:
3004                                         tagsToClose.push_back(html::EndFontTag(html::FT_SIZE_SCRIPT));
3005                                         break;
3006                                 case FONT_SIZE_FOOTNOTE:
3007                                         tagsToClose.push_back(html::EndFontTag(html::FT_SIZE_FOOTNOTE));
3008                                         break;
3009                                 case FONT_SIZE_SMALL:
3010                                         tagsToClose.push_back(html::EndFontTag(html::FT_SIZE_SMALL));
3011                                         break;
3012                                 case FONT_SIZE_LARGE:
3013                                         tagsToClose.push_back(html::EndFontTag(html::FT_SIZE_LARGE));
3014                                         break;
3015                                 case FONT_SIZE_LARGER:
3016                                         tagsToClose.push_back(html::EndFontTag(html::FT_SIZE_LARGER));
3017                                         break;
3018                                 case FONT_SIZE_LARGEST:
3019                                         tagsToClose.push_back(html::EndFontTag(html::FT_SIZE_LARGEST));
3020                                         break;
3021                                 case FONT_SIZE_HUGE:
3022                                         tagsToClose.push_back(html::EndFontTag(html::FT_SIZE_HUGE));
3023                                         break;
3024                                 case FONT_SIZE_HUGER:
3025                                         tagsToClose.push_back(html::EndFontTag(html::FT_SIZE_HUGER));
3026                                         break;
3027                                 case FONT_SIZE_INCREASE:
3028                                         tagsToClose.push_back(html::EndFontTag(html::FT_SIZE_INCREASE));
3029                                         break;
3030                                 case FONT_SIZE_DECREASE:
3031                                         tagsToClose.push_back(html::EndFontTag(html::FT_SIZE_DECREASE));
3032                                         break;
3033                                 case FONT_SIZE_INHERIT:
3034                                 case FONT_SIZE_NORMAL:
3035                                         break;
3036                                 default:
3037                                         // the other tags are for internal use
3038                                         LATTEST(false);
3039                                         break;
3040                                 }
3041                                 size_flag = false;
3042                         }
3043                         switch (curr_size) {
3044                         case FONT_SIZE_TINY:
3045                                 tagsToOpen.push_back(html::FontTag(html::FT_SIZE_TINY));
3046                                 size_flag = true;
3047                                 break;
3048                         case FONT_SIZE_SCRIPT:
3049                                 tagsToOpen.push_back(html::FontTag(html::FT_SIZE_SCRIPT));
3050                                 size_flag = true;
3051                                 break;
3052                         case FONT_SIZE_FOOTNOTE:
3053                                 tagsToOpen.push_back(html::FontTag(html::FT_SIZE_FOOTNOTE));
3054                                 size_flag = true;
3055                                 break;
3056                         case FONT_SIZE_SMALL:
3057                                 tagsToOpen.push_back(html::FontTag(html::FT_SIZE_SMALL));
3058                                 size_flag = true;
3059                                 break;
3060                         case FONT_SIZE_LARGE:
3061                                 tagsToOpen.push_back(html::FontTag(html::FT_SIZE_LARGE));
3062                                 size_flag = true;
3063                                 break;
3064                         case FONT_SIZE_LARGER:
3065                                 tagsToOpen.push_back(html::FontTag(html::FT_SIZE_LARGER));
3066                                 size_flag = true;
3067                                 break;
3068                         case FONT_SIZE_LARGEST:
3069                                 tagsToOpen.push_back(html::FontTag(html::FT_SIZE_LARGEST));
3070                                 size_flag = true;
3071                                 break;
3072                         case FONT_SIZE_HUGE:
3073                                 tagsToOpen.push_back(html::FontTag(html::FT_SIZE_HUGE));
3074                                 size_flag = true;
3075                                 break;
3076                         case FONT_SIZE_HUGER:
3077                                 tagsToOpen.push_back(html::FontTag(html::FT_SIZE_HUGER));
3078                                 size_flag = true;
3079                                 break;
3080                         case FONT_SIZE_INCREASE:
3081                                 tagsToOpen.push_back(html::FontTag(html::FT_SIZE_INCREASE));
3082                                 size_flag = true;
3083                                 break;
3084                         case FONT_SIZE_DECREASE:
3085                                 tagsToOpen.push_back(html::FontTag(html::FT_SIZE_DECREASE));
3086                                 size_flag = true;
3087                                 break;
3088                         case FONT_SIZE_NORMAL:
3089                         case FONT_SIZE_INHERIT:
3090                                 break;
3091                         default:
3092                                 // the other tags are for internal use
3093                                 LATTEST(false);
3094                                 break;
3095                         }
3096                 }
3097
3098                 // FIXME XHTML
3099                 // Other such tags? What about the other text ranges?
3100
3101                 vector<html::EndFontTag>::const_iterator cit = tagsToClose.begin();
3102                 vector<html::EndFontTag>::const_iterator cen = tagsToClose.end();
3103                 for (; cit != cen; ++cit)
3104                         xs << *cit;
3105
3106                 vector<html::FontTag>::const_iterator sit = tagsToOpen.begin();
3107                 vector<html::FontTag>::const_iterator sen = tagsToOpen.end();
3108                 for (; sit != sen; ++sit)
3109                         xs << *sit;
3110
3111                 tagsToClose.clear();
3112                 tagsToOpen.clear();
3113
3114                 Inset const * inset = getInset(i);
3115                 if (inset) {
3116                         if (!runparams.for_toc || inset->isInToc()) {
3117                                 OutputParams np = runparams;
3118                                 np.local_font = &font;
3119                                 // If the paragraph has size 1, then we are in the "special
3120                                 // case" where we do not output the containing paragraph info
3121                                 if (!inset->getLayout().htmlisblock() && size() != 1)
3122                                         np.html_in_par = true;
3123                                 retval += inset->xhtml(xs, np);
3124                         }
3125                 } else {
3126                         char_type c = getUChar(buf.masterBuffer()->params(), i);
3127                         xs << c;
3128                 }
3129                 font_old = font.fontInfo();
3130         }
3131
3132         // FIXME XHTML
3133         // I'm worried about what happens if a branch, say, is itself
3134         // wrapped in some font stuff. I think that will not work.
3135         xs.closeFontTags();
3136         if (close_paragraph)
3137                 xs.endDivision();
3138
3139         return retval;
3140 }
3141
3142
3143 bool Paragraph::isHfill(pos_type pos) const
3144 {
3145         Inset const * inset = getInset(pos);
3146         return inset && inset->isHfill();
3147 }
3148
3149
3150 bool Paragraph::isNewline(pos_type pos) const
3151 {
3152         // U+2028 LINE SEPARATOR
3153         // U+2029 PARAGRAPH SEPARATOR
3154         char_type const c = d->text_[pos];
3155         if (c == 0x2028 || c == 0x2029)
3156                 return true;
3157         Inset const * inset = getInset(pos);
3158         return inset && inset->lyxCode() == NEWLINE_CODE;
3159 }
3160
3161
3162 bool Paragraph::isEnvSeparator(pos_type pos) const
3163 {
3164         Inset const * inset = getInset(pos);
3165         return inset && inset->lyxCode() == SEPARATOR_CODE;
3166 }
3167
3168
3169 bool Paragraph::isLineSeparator(pos_type pos) const
3170 {
3171         char_type const c = d->text_[pos];
3172         if (isLineSeparatorChar(c))
3173                 return true;
3174         Inset const * inset = getInset(pos);
3175         return inset && inset->isLineSeparator();
3176 }
3177
3178
3179 bool Paragraph::isWordSeparator(pos_type pos) const
3180 {
3181         if (pos == size())
3182                 return true;
3183         if (Inset const * inset = getInset(pos))
3184                 return !inset->isLetter();
3185         // if we have a hard hyphen (no en- or emdash) or apostrophe
3186         // we pass this to the spell checker
3187         // FIXME: this method is subject to change, visit
3188         // https://bugzilla.mozilla.org/show_bug.cgi?id=355178
3189         // to get an impression how complex this is.
3190         if (isHardHyphenOrApostrophe(pos))
3191                 return false;
3192         char_type const c = d->text_[pos];
3193         // We want to pass the escape chars to the spellchecker
3194         docstring const escape_chars = from_utf8(lyxrc.spellchecker_esc_chars);
3195         return !isLetterChar(c) && !isDigitASCII(c) && !contains(escape_chars, c);
3196 }
3197
3198
3199 bool Paragraph::isHardHyphenOrApostrophe(pos_type pos) const
3200 {
3201         pos_type const psize = size();
3202         if (pos >= psize)
3203                 return false;
3204         char_type const c = d->text_[pos];
3205         if (c != '-' && c != '\'')
3206                 return false;
3207         int nextpos = pos + 1;
3208         int prevpos = pos > 0 ? pos - 1 : 0;
3209         if ((nextpos == psize || isSpace(nextpos))
3210                 && (pos == 0 || isSpace(prevpos)))
3211                 return false;
3212         return true;
3213 }
3214
3215
3216 FontSpan const & Paragraph::getSpellRange(pos_type pos) const
3217 {
3218         return d->speller_state_.getRange(pos);
3219 }
3220
3221
3222 bool Paragraph::isChar(pos_type pos) const
3223 {
3224         if (Inset const * inset = getInset(pos))
3225                 return inset->isChar();
3226         char_type const c = d->text_[pos];
3227         return !isLetterChar(c) && !isDigitASCII(c) && !lyx::isSpace(c);
3228 }
3229
3230
3231 bool Paragraph::isSpace(pos_type pos) const
3232 {
3233         if (Inset const * inset = getInset(pos))
3234                 return inset->isSpace();
3235         char_type const c = d->text_[pos];
3236         return lyx::isSpace(c);
3237 }
3238
3239
3240 Language const *
3241 Paragraph::getParLanguage(BufferParams const & bparams) const
3242 {
3243         if (!empty())
3244                 return getFirstFontSettings(bparams).language();
3245         // FIXME: we should check the prev par as well (Lgb)
3246         return bparams.language;
3247 }
3248
3249
3250 bool Paragraph::isRTL(BufferParams const & bparams) const
3251 {
3252         return getParLanguage(bparams)->rightToLeft()
3253                 && !inInset().getLayout().forceLTR();
3254 }
3255
3256
3257 void Paragraph::changeLanguage(BufferParams const & bparams,
3258                                Language const * from, Language const * to)
3259 {
3260         // change language including dummy font change at the end
3261         for (pos_type i = 0; i <= size(); ++i) {
3262                 Font font = getFontSettings(bparams, i);
3263                 if (font.language() == from) {
3264                         font.setLanguage(to);
3265                         setFont(i, font);
3266                         d->requestSpellCheck(i);
3267                 }
3268         }
3269 }
3270
3271
3272 bool Paragraph::isMultiLingual(BufferParams const & bparams) const
3273 {
3274         Language const * doc_language = bparams.language;
3275         FontList::const_iterator cit = d->fontlist_.begin();
3276         FontList::const_iterator end = d->fontlist_.end();
3277
3278         for (; cit != end; ++cit)
3279                 if (cit->font().language() != ignore_language &&
3280                     cit->font().language() != latex_language &&
3281                     cit->font().language() != doc_language)
3282                         return true;
3283         return false;
3284 }
3285
3286
3287 void Paragraph::getLanguages(std::set<Language const *> & languages) const
3288 {
3289         FontList::const_iterator cit = d->fontlist_.begin();
3290         FontList::const_iterator end = d->fontlist_.end();
3291
3292         for (; cit != end; ++cit) {
3293                 Language const * lang = cit->font().language();
3294                 if (lang != ignore_language &&
3295                     lang != latex_language)
3296                         languages.insert(lang);
3297         }
3298 }
3299
3300
3301 docstring Paragraph::asString(int options) const
3302 {
3303         return asString(0, size(), options);
3304 }
3305
3306
3307 docstring Paragraph::asString(pos_type beg, pos_type end, int options, const OutputParams *runparams) const
3308 {
3309         odocstringstream os;
3310
3311         if (beg == 0
3312             && options & AS_STR_LABEL
3313             && !d->params_.labelString().empty())
3314                 os << d->params_.labelString() << ' ';
3315
3316         for (pos_type i = beg; i < end; ++i) {
3317                 if ((options & AS_STR_SKIPDELETE) && isDeleted(i))
3318                         continue;
3319                 char_type const c = d->text_[i];
3320                 if (isPrintable(c) || c == '\t'
3321                     || (c == '\n' && (options & AS_STR_NEWLINES)))
3322                         os.put(c);
3323                 else if (c == META_INSET && (options & AS_STR_INSETS)) {
3324                         if (c == META_INSET && (options & AS_STR_PLAINTEXT)) {
3325                                 LASSERT(runparams != 0, return docstring());
3326                                 getInset(i)->plaintext(os, *runparams);
3327                         } else {
3328                                 getInset(i)->toString(os);
3329                         }
3330                 }
3331         }
3332
3333         return os.str();
3334 }
3335
3336
3337 void Paragraph::forOutliner(docstring & os, size_t const maxlen,
3338                             bool const shorten, bool const label) const
3339 {
3340         size_t tmplen = shorten ? maxlen + 1 : maxlen;
3341         if (label && !labelString().empty())
3342                 os += labelString() + ' ';
3343         for (pos_type i = 0; i < size() && os.length() < tmplen; ++i) {
3344                 if (isDeleted(i))
3345                         continue;
3346                 char_type const c = d->text_[i];
3347                 if (isPrintable(c))
3348                         os += c;
3349                 else if (c == META_INSET)
3350                         getInset(i)->forOutliner(os, tmplen, false);
3351         }
3352         if (shorten)
3353                 Text::shortenForOutliner(os, maxlen);
3354 }
3355
3356
3357 void Paragraph::setInsetOwner(Inset const * inset)
3358 {
3359         d->inset_owner_ = inset;
3360 }
3361
3362
3363 int Paragraph::id() const
3364 {
3365         return d->id_;
3366 }
3367
3368
3369 void Paragraph::setId(int id)
3370 {
3371         d->id_ = id;
3372 }
3373
3374
3375 Layout const & Paragraph::layout() const
3376 {
3377         return *d->layout_;
3378 }
3379
3380
3381 void Paragraph::setLayout(Layout const & layout)
3382 {
3383         d->layout_ = &layout;
3384 }
3385
3386
3387 void Paragraph::setDefaultLayout(DocumentClass const & tc)
3388 {
3389         setLayout(tc.defaultLayout());
3390 }
3391
3392
3393 void Paragraph::setPlainLayout(DocumentClass const & tc)
3394 {
3395         setLayout(tc.plainLayout());
3396 }
3397
3398
3399 void Paragraph::setPlainOrDefaultLayout(DocumentClass const & tclass)
3400 {
3401         if (usePlainLayout())
3402                 setPlainLayout(tclass);
3403         else
3404                 setDefaultLayout(tclass);
3405 }
3406
3407
3408 Inset const & Paragraph::inInset() const
3409 {
3410         LBUFERR(d->inset_owner_);
3411         return *d->inset_owner_;
3412 }
3413
3414
3415 ParagraphParameters & Paragraph::params()
3416 {
3417         return d->params_;
3418 }
3419
3420
3421 ParagraphParameters const & Paragraph::params() const
3422 {
3423         return d->params_;
3424 }
3425
3426
3427 bool Paragraph::isFreeSpacing() const
3428 {
3429         if (d->layout_->free_spacing)
3430                 return true;
3431         return d->inset_owner_ && d->inset_owner_->isFreeSpacing();
3432 }
3433
3434
3435 bool Paragraph::allowEmpty() const
3436 {
3437         if (d->layout_->keepempty)
3438                 return true;
3439         return d->inset_owner_ && d->inset_owner_->allowEmpty();
3440 }
3441
3442
3443 bool Paragraph::brokenBiblio() const
3444 {
3445         // there is a problem if there is no bibitem at position 0 or
3446         // if there is another bibitem in the paragraph.
3447         return d->layout_->labeltype == LABEL_BIBLIO
3448                 && (d->insetlist_.find(BIBITEM_CODE) != 0
3449                     || d->insetlist_.find(BIBITEM_CODE, 1) > 0);
3450 }
3451
3452
3453 int Paragraph::fixBiblio(Buffer const & buffer)
3454 {
3455         // FIXME: What about the case where paragraph is not BIBLIO
3456         // but there is an InsetBibitem?
3457         // FIXME: when there was already an inset at 0, the return value is 1,
3458         // which does not tell whether another inset has been remove; the
3459         // cursor cannot be correctly updated.
3460
3461         if (d->layout_->labeltype != LABEL_BIBLIO)
3462                 return 0;
3463
3464         bool const track_changes = buffer.params().track_changes;
3465         int bibitem_pos = d->insetlist_.find(BIBITEM_CODE);
3466         bool const hasbibitem0 = bibitem_pos == 0;
3467
3468         if (hasbibitem0) {
3469                 bibitem_pos = d->insetlist_.find(BIBITEM_CODE, 1);
3470                 // There was an InsetBibitem at pos 0, and no other one => OK
3471                 if (bibitem_pos == -1)
3472                         return 0;
3473                 // there is a bibitem at the 0 position, but since
3474                 // there is a second one, we copy the second on the
3475                 // first. We're assuming there are at most two of
3476                 // these, which there should be.
3477                 // FIXME: why does it make sense to do that rather
3478                 // than keep the first? (JMarc)
3479                 Inset * inset = releaseInset(bibitem_pos);
3480                 d->insetlist_.begin()->inset = inset;
3481                 return -bibitem_pos;
3482         }
3483
3484         // We need to create an inset at the beginning
3485         Inset * inset = 0;
3486         if (bibitem_pos > 0) {
3487                 // there was one somewhere in the paragraph, let's move it
3488                 inset = d->insetlist_.release(bibitem_pos);
3489                 eraseChar(bibitem_pos, track_changes);
3490         } else
3491                 // make a fresh one
3492                 inset = new InsetBibitem(const_cast<Buffer *>(&buffer),
3493                                          InsetCommandParams(BIBITEM_CODE));
3494
3495         Font font(inherit_font, buffer.params().language);
3496         insertInset(0, inset, font, Change(track_changes ? Change::INSERTED
3497                                                    : Change::UNCHANGED));
3498
3499         return 1;
3500 }
3501
3502
3503 void Paragraph::checkAuthors(AuthorList const & authorList)
3504 {
3505         d->changes_.checkAuthors(authorList);
3506 }
3507
3508
3509 bool Paragraph::isChanged(pos_type pos) const
3510 {
3511         return lookupChange(pos).changed();
3512 }
3513
3514
3515 bool Paragraph::isInserted(pos_type pos) const
3516 {
3517         return lookupChange(pos).inserted();
3518 }
3519
3520
3521 bool Paragraph::isDeleted(pos_type pos) const
3522 {
3523         return lookupChange(pos).deleted();
3524 }
3525
3526
3527 InsetList const & Paragraph::insetList() const
3528 {
3529         return d->insetlist_;
3530 }
3531
3532
3533 void Paragraph::setBuffer(Buffer & b)
3534 {
3535         d->insetlist_.setBuffer(b);
3536 }
3537
3538
3539 void Paragraph::resetBuffer()
3540 {
3541         d->insetlist_.resetBuffer();
3542 }
3543
3544
3545 Inset * Paragraph::releaseInset(pos_type pos)
3546 {
3547         Inset * inset = d->insetlist_.release(pos);
3548         /// does not honour change tracking!
3549         eraseChar(pos, false);
3550         return inset;
3551 }
3552
3553
3554 Inset * Paragraph::getInset(pos_type pos)
3555 {
3556         return (pos < pos_type(d->text_.size()) && d->text_[pos] == META_INSET)
3557                  ? d->insetlist_.get(pos) : 0;
3558 }
3559
3560
3561 Inset const * Paragraph::getInset(pos_type pos) const
3562 {
3563         return (pos < pos_type(d->text_.size()) && d->text_[pos] == META_INSET)
3564                  ? d->insetlist_.get(pos) : 0;
3565 }
3566
3567
3568 void Paragraph::changeCase(BufferParams const & bparams, pos_type pos,
3569                 pos_type & right, TextCase action)
3570 {
3571         // process sequences of modified characters; in change
3572         // tracking mode, this approach results in much better
3573         // usability than changing case on a char-by-char basis
3574         // We also need to track the current font, since font
3575         // changes within sequences can occur.
3576         vector<pair<char_type, Font> > changes;
3577
3578         bool const trackChanges = bparams.track_changes;
3579
3580         bool capitalize = true;
3581
3582         for (; pos < right; ++pos) {
3583                 char_type oldChar = d->text_[pos];
3584                 char_type newChar = oldChar;
3585
3586                 // ignore insets and don't play with deleted text!
3587                 if (oldChar != META_INSET && !isDeleted(pos)) {
3588                         switch (action) {
3589                                 case text_lowercase:
3590                                         newChar = lowercase(oldChar);
3591                                         break;
3592                                 case text_capitalization:
3593                                         if (capitalize) {
3594                                                 newChar = uppercase(oldChar);
3595                                                 capitalize = false;
3596                                         }
3597                                         break;
3598                                 case text_uppercase:
3599                                         newChar = uppercase(oldChar);
3600                                         break;
3601                         }
3602                 }
3603
3604                 if (isWordSeparator(pos) || isDeleted(pos)) {
3605                         // permit capitalization again
3606                         capitalize = true;
3607                 }
3608
3609                 if (oldChar != newChar) {
3610                         changes.push_back(make_pair(newChar, getFontSettings(bparams, pos)));
3611                         if (pos != right - 1)
3612                                 continue;
3613                         // step behind the changing area
3614                         pos++;
3615                 }
3616
3617                 int erasePos = pos - changes.size();
3618                 for (size_t i = 0; i < changes.size(); i++) {
3619                         insertChar(pos, changes[i].first,
3620                                    changes[i].second,
3621                                    trackChanges);
3622                         if (!eraseChar(erasePos, trackChanges)) {
3623                                 ++erasePos;
3624                                 ++pos; // advance
3625                                 ++right; // expand selection
3626                         }
3627                 }
3628                 changes.clear();
3629         }
3630 }
3631
3632
3633 int Paragraph::find(docstring const & str, bool cs, bool mw,
3634                 pos_type start_pos, bool del) const
3635 {
3636         pos_type pos = start_pos;
3637         int const strsize = str.length();
3638         int i = 0;
3639         pos_type const parsize = d->text_.size();
3640         for (i = 0; i < strsize && pos < parsize; ++i, ++pos) {
3641                 // Ignore "invisible" letters such as ligature breaks
3642                 // and hyphenation chars while searching
3643                 while (pos < parsize - 1 && isInset(pos)) {
3644                         odocstringstream os;
3645                         getInset(pos)->toString(os);
3646                         if (!getInset(pos)->isLetter() || !os.str().empty())
3647                                 break;
3648                         pos++;
3649                 }
3650                 if (cs && str[i] != d->text_[pos])
3651                         break;
3652                 if (!cs && uppercase(str[i]) != uppercase(d->text_[pos]))
3653                         break;
3654                 if (!del && isDeleted(pos))
3655                         break;
3656         }
3657
3658         if (i != strsize)
3659                 return 0;
3660
3661         // if necessary, check whether string matches word
3662         if (mw) {
3663                 if (start_pos > 0 && !isWordSeparator(start_pos - 1))
3664                         return 0;
3665                 if (pos < parsize
3666                         && !isWordSeparator(pos))
3667                         return 0;
3668         }
3669
3670         return pos - start_pos;
3671 }
3672
3673
3674 char_type Paragraph::getChar(pos_type pos) const
3675 {
3676         return d->text_[pos];
3677 }
3678
3679
3680 pos_type Paragraph::size() const
3681 {
3682         return d->text_.size();
3683 }
3684
3685
3686 bool Paragraph::empty() const
3687 {
3688         return d->text_.empty();
3689 }
3690
3691
3692 bool Paragraph::isInset(pos_type pos) const
3693 {
3694         return d->text_[pos] == META_INSET;
3695 }
3696
3697
3698 bool Paragraph::isSeparator(pos_type pos) const
3699 {
3700         //FIXME: Are we sure this can be the only separator?
3701         return d->text_[pos] == ' ';
3702 }
3703
3704
3705 void Paragraph::deregisterWords()
3706 {
3707         Private::LangWordsMap::const_iterator itl = d->words_.begin();
3708         Private::LangWordsMap::const_iterator ite = d->words_.end();
3709         for (; itl != ite; ++itl) {
3710                 WordList & wl = theWordList(itl->first);
3711                 Private::Words::const_iterator it = (itl->second).begin();
3712                 Private::Words::const_iterator et = (itl->second).end();
3713                 for (; it != et; ++it)
3714                         wl.remove(*it);
3715         }
3716         d->words_.clear();
3717 }
3718
3719
3720 void Paragraph::locateWord(pos_type & from, pos_type & to,
3721         word_location const loc) const
3722 {
3723         switch (loc) {
3724         case WHOLE_WORD_STRICT:
3725                 if (from == 0 || from == size()
3726                     || isWordSeparator(from)
3727                     || isWordSeparator(from - 1)) {
3728                         to = from;
3729                         return;
3730                 }
3731                 // no break here, we go to the next
3732
3733         case WHOLE_WORD:
3734                 // If we are already at the beginning of a word, do nothing
3735                 if (!from || isWordSeparator(from - 1))
3736                         break;
3737                 // no break here, we go to the next
3738
3739         case PREVIOUS_WORD:
3740                 // always move the cursor to the beginning of previous word
3741                 while (from && !isWordSeparator(from - 1))
3742                         --from;
3743                 break;
3744         case NEXT_WORD:
3745                 LYXERR0("Paragraph::locateWord: NEXT_WORD not implemented yet");
3746                 break;
3747         case PARTIAL_WORD:
3748                 // no need to move the 'from' cursor
3749                 break;
3750         }
3751         to = from;
3752         while (to < size() && !isWordSeparator(to))
3753                 ++to;
3754 }
3755
3756
3757 void Paragraph::collectWords()
3758 {
3759         for (pos_type pos = 0; pos < size(); ++pos) {
3760                 if (isWordSeparator(pos))
3761                         continue;
3762                 pos_type from = pos;
3763                 locateWord(from, pos, WHOLE_WORD);
3764                 // Work around MSVC warning: The statement
3765                 // if (pos < from + lyxrc.completion_minlength)
3766                 // triggers a signed vs. unsigned warning.
3767                 // I don't know why this happens, it could be a MSVC bug, or
3768                 // related to LLP64 (windows) vs. LP64 (unix) programming
3769                 // model, or the C++ standard might be ambigous in the section
3770                 // defining the "usual arithmetic conversions". However, using
3771                 // a temporary variable is safe and works on all compilers.
3772                 pos_type const endpos = from + lyxrc.completion_minlength;
3773                 if (pos < endpos)
3774                         continue;
3775                 FontList::const_iterator cit = d->fontlist_.fontIterator(from);
3776                 if (cit == d->fontlist_.end())
3777                         return;
3778                 Language const * lang = cit->font().language();
3779                 docstring const word = asString(from, pos, AS_STR_NONE);
3780                 d->words_[lang->lang()].insert(word);
3781         }
3782 }
3783
3784
3785 void Paragraph::registerWords()
3786 {
3787         Private::LangWordsMap::const_iterator itl = d->words_.begin();
3788         Private::LangWordsMap::const_iterator ite = d->words_.end();
3789         for (; itl != ite; ++itl) {
3790                 WordList & wl = theWordList(itl->first);
3791                 Private::Words::const_iterator it = (itl->second).begin();
3792                 Private::Words::const_iterator et = (itl->second).end();
3793                 for (; it != et; ++it)
3794                         wl.insert(*it);
3795         }
3796 }
3797
3798
3799 void Paragraph::updateWords()
3800 {
3801         deregisterWords();
3802         collectWords();
3803         registerWords();
3804 }
3805
3806
3807 void Paragraph::Private::appendSkipPosition(SkipPositions & skips, pos_type const pos) const
3808 {
3809         SkipPositionsIterator begin = skips.begin();
3810         SkipPositions::iterator end = skips.end();
3811         if (pos > 0 && begin < end) {
3812                 --end;
3813                 if (end->last == pos - 1) {
3814                         end->last = pos;
3815                         return;
3816                 }
3817         }
3818         skips.insert(end, FontSpan(pos, pos));
3819 }
3820
3821
3822 Language * Paragraph::Private::locateSpellRange(
3823         pos_type & from, pos_type & to,
3824         SkipPositions & skips) const
3825 {
3826         // skip leading white space
3827         while (from < to && owner_->isWordSeparator(from))
3828                 ++from;
3829         // don't check empty range
3830         if (from >= to)
3831                 return 0;
3832         // get current language
3833         Language * lang = getSpellLanguage(from);
3834         pos_type last = from;
3835         bool samelang = true;
3836         bool sameinset = true;
3837         while (last < to && samelang && sameinset) {
3838                 // hop to end of word
3839                 while (last < to && !owner_->isWordSeparator(last)) {
3840                         if (owner_->getInset(last)) {
3841                                 appendSkipPosition(skips, last);
3842                         } else if (owner_->isDeleted(last)) {
3843                                 appendSkipPosition(skips, last);
3844                         }
3845                         ++last;
3846                 }
3847                 // hop to next word while checking for insets
3848                 while (sameinset && last < to && owner_->isWordSeparator(last)) {
3849                         if (Inset const * inset = owner_->getInset(last))
3850                                 sameinset = inset->isChar() && inset->isLetter();
3851                         if (sameinset && owner_->isDeleted(last)) {
3852                                 appendSkipPosition(skips, last);
3853                         }
3854                         if (sameinset)
3855                                 last++;
3856                 }
3857                 if (sameinset && last < to) {
3858                         // now check for language change
3859                         samelang = lang == getSpellLanguage(last);
3860                 }
3861         }
3862         // if language change detected backstep is needed
3863         if (!samelang)
3864                 --last;
3865         to = last;
3866         return lang;
3867 }
3868
3869
3870 Language * Paragraph::Private::getSpellLanguage(pos_type const from) const
3871 {
3872         Language * lang =
3873                 const_cast<Language *>(owner_->getFontSettings(
3874                         inset_owner_->buffer().params(), from).language());
3875         if (lang == inset_owner_->buffer().params().language
3876                 && !lyxrc.spellchecker_alt_lang.empty()) {
3877                 string lang_code;
3878                 string const lang_variety =
3879                         split(lyxrc.spellchecker_alt_lang, lang_code, '-');
3880                 lang->setCode(lang_code);
3881                 lang->setVariety(lang_variety);
3882         }
3883         return lang;
3884 }
3885
3886
3887 void Paragraph::requestSpellCheck(pos_type pos)
3888 {
3889         d->requestSpellCheck(pos);
3890 }
3891
3892
3893 bool Paragraph::needsSpellCheck() const
3894 {
3895         SpellChecker::ChangeNumber speller_change_number = 0;
3896         if (theSpellChecker())
3897                 speller_change_number = theSpellChecker()->changeNumber();
3898         if (speller_change_number > d->speller_state_.currentChangeNumber()) {
3899                 d->speller_state_.needsCompleteRefresh(speller_change_number);
3900         }
3901         return d->needsSpellCheck();
3902 }
3903
3904
3905 bool Paragraph::Private::ignoreWord(docstring const & word) const
3906 {
3907         // Ignore words with digits
3908         // FIXME: make this customizable
3909         // (note that some checkers ignore words with digits by default)
3910         docstring::const_iterator cit = word.begin();
3911         docstring::const_iterator const end = word.end();
3912         for (; cit != end; ++cit) {
3913                 if (isNumber((*cit)))
3914                         return true;
3915         }
3916         return false;
3917 }
3918
3919
3920 SpellChecker::Result Paragraph::spellCheck(pos_type & from, pos_type & to,
3921         WordLangTuple & wl, docstring_list & suggestions,
3922         bool do_suggestion, bool check_learned) const
3923 {
3924         SpellChecker::Result result = SpellChecker::WORD_OK;
3925         SpellChecker * speller = theSpellChecker();
3926         if (!speller)
3927                 return result;
3928
3929         if (!d->layout_->spellcheck || !inInset().allowSpellCheck())
3930                 return result;
3931
3932         locateWord(from, to, WHOLE_WORD);
3933         if (from == to || from >= size())
3934                 return result;
3935
3936         docstring word = asString(from, to, AS_STR_INSETS | AS_STR_SKIPDELETE);
3937         Language * lang = d->getSpellLanguage(from);
3938
3939         wl = WordLangTuple(word, lang);
3940
3941         if (word.empty())
3942                 return result;
3943
3944         if (needsSpellCheck() || check_learned) {
3945                 pos_type end = to;
3946                 if (!d->ignoreWord(word)) {
3947                         bool const trailing_dot = to < size() && d->text_[to] == '.';
3948                         result = speller->check(wl);
3949                         if (SpellChecker::misspelled(result) && trailing_dot) {
3950                                 wl = WordLangTuple(word.append(from_ascii(".")), lang);
3951                                 result = speller->check(wl);
3952                                 if (!SpellChecker::misspelled(result)) {
3953                                         LYXERR(Debug::GUI, "misspelled word is correct with dot: \"" <<
3954                                            word << "\" [" <<
3955                                            from << ".." << to << "]");
3956                                 } else {
3957                                         // spell check with dot appended failed too
3958                                         // restore original word/lang value
3959                                         word = asString(from, to, AS_STR_INSETS | AS_STR_SKIPDELETE);
3960                                         wl = WordLangTuple(word, lang);
3961                                 }
3962                         }
3963                 }
3964                 if (!SpellChecker::misspelled(result)) {
3965                         // area up to the begin of the next word is not misspelled
3966                         while (end < size() && isWordSeparator(end))
3967                                 ++end;
3968                 }
3969                 d->setMisspelled(from, end, result);
3970         } else {
3971                 result = d->speller_state_.getState(from);
3972         }
3973
3974         if (do_suggestion)
3975                 suggestions.clear();
3976
3977         if (SpellChecker::misspelled(result)) {
3978                 LYXERR(Debug::GUI, "misspelled word: \"" <<
3979                            word << "\" [" <<
3980                            from << ".." << to << "]");
3981                 if (do_suggestion)
3982                         speller->suggest(wl, suggestions);
3983         }
3984         return result;
3985 }
3986
3987
3988 void Paragraph::Private::markMisspelledWords(
3989         pos_type const & first, pos_type const & last,
3990         SpellChecker::Result result,
3991         docstring const & word,
3992         SkipPositions const & skips)
3993 {
3994         if (!SpellChecker::misspelled(result)) {
3995                 setMisspelled(first, last, SpellChecker::WORD_OK);
3996                 return;
3997         }
3998         int snext = first;
3999         SpellChecker * speller = theSpellChecker();
4000         // locate and enumerate the error positions
4001         int nerrors = speller->numMisspelledWords();
4002         int numskipped = 0;
4003         SkipPositionsIterator it = skips.begin();
4004         SkipPositionsIterator et = skips.end();
4005         for (int index = 0; index < nerrors; ++index) {
4006                 int wstart;
4007                 int wlen = 0;
4008                 speller->misspelledWord(index, wstart, wlen);
4009                 /// should not happen if speller supports range checks
4010                 if (!wlen) continue;
4011                 docstring const misspelled = word.substr(wstart, wlen);
4012                 wstart += first + numskipped;
4013                 if (snext < wstart) {
4014                         /// mark the range of correct spelling
4015                         numskipped += countSkips(it, et, wstart);
4016                         setMisspelled(snext,
4017                                 wstart - 1, SpellChecker::WORD_OK);
4018                 }
4019                 snext = wstart + wlen;
4020                 numskipped += countSkips(it, et, snext);
4021                 /// mark the range of misspelling
4022                 setMisspelled(wstart, snext, result);
4023                 LYXERR(Debug::GUI, "misspelled word: \"" <<
4024                            misspelled << "\" [" <<
4025                            wstart << ".." << (snext-1) << "]");
4026                 ++snext;
4027         }
4028         if (snext <= last) {
4029                 /// mark the range of correct spelling at end
4030                 setMisspelled(snext, last, SpellChecker::WORD_OK);
4031         }
4032 }
4033
4034
4035 void Paragraph::spellCheck() const
4036 {
4037         SpellChecker * speller = theSpellChecker();
4038         if (!speller || empty() ||!needsSpellCheck())
4039                 return;
4040         pos_type start;
4041         pos_type endpos;
4042         d->rangeOfSpellCheck(start, endpos);
4043         if (speller->canCheckParagraph()) {
4044                 // loop until we leave the range
4045                 for (pos_type first = start; first < endpos; ) {
4046                         pos_type last = endpos;
4047                         Private::SkipPositions skips;
4048                         Language * lang = d->locateSpellRange(first, last, skips);
4049                         if (first >= endpos)
4050                                 break;
4051                         // start the spell checker on the unit of meaning
4052                         docstring word = asString(first, last, AS_STR_INSETS + AS_STR_SKIPDELETE);
4053                         WordLangTuple wl = WordLangTuple(word, lang);
4054                         SpellChecker::Result result = word.size() ?
4055                                 speller->check(wl) : SpellChecker::WORD_OK;
4056                         d->markMisspelledWords(first, last, result, word, skips);
4057                         first = ++last;
4058                 }
4059         } else {
4060                 static docstring_list suggestions;
4061                 pos_type to = endpos;
4062                 while (start < endpos) {
4063                         WordLangTuple wl;
4064                         spellCheck(start, to, wl, suggestions, false);
4065                         start = to + 1;
4066                 }
4067         }
4068         d->readySpellCheck();
4069 }
4070
4071
4072 bool Paragraph::isMisspelled(pos_type pos, bool check_boundary) const
4073 {
4074         bool result = SpellChecker::misspelled(d->speller_state_.getState(pos));
4075         if (result || pos <= 0 || pos > size())
4076                 return result;
4077         if (check_boundary && (pos == size() || isWordSeparator(pos)))
4078                 result = SpellChecker::misspelled(d->speller_state_.getState(pos - 1));
4079         return result;
4080 }
4081
4082
4083 string Paragraph::magicLabel() const
4084 {
4085         stringstream ss;
4086         ss << "magicparlabel-" << id();
4087         return ss.str();
4088 }
4089
4090
4091 } // namespace lyx