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 "Encoding.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 "TextClass.h"
  44 #include "TexRow.h"
  45 #include "Text.h"
  46 #include "VSpace.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 "support/debug.h"
  57 #include "support/docstring_list.h"
  58 #include "support/ExceptionMessage.h"
  59 #include "support/gettext.h"
  60 #include "support/lassert.h"
  61 #include "support/lstrings.h"
  62 #include "support/textutils.h"
  63
  64 #include <sstream>
  65 #include <vector>
  66
  67 using namespace std;
  68 using namespace lyx::support;
  69
  70 namespace lyx {
  71
  72 namespace {
  73 /// Inset identifier (above 0x10ffff, for ucs-4)
  74 char_type const META_INSET = 0x200001;
  75 }
  76
  77
  78 /////////////////////////////////////////////////////////////////////
  79 //
  80 // SpellResultRange
  81 //
  82 /////////////////////////////////////////////////////////////////////
  83
  84 class SpellResultRange {
  85 public:
  86         SpellResultRange(FontSpan range, SpellChecker::Result result)
  87         : range_(range), result_(result)
  88         {}
  89         ///
  90         FontSpan const & range() const { return range_; }
  91         ///
  92         void range(FontSpan const & r) { range_ = r; }
  93         ///
  94         SpellChecker::Result result() const { return result_; }
  95         ///
  96         void result(SpellChecker::Result r) { result_ = r; }
  97         ///
  98         bool inside(pos_type pos) const { return range_.inside(pos); }
  99         ///
 100         bool covered(FontSpan const & r) const
 101         {
 102                 // 1. first of new range inside current range or
 103                 // 2. last of new range inside current range or
 104                 // 3. first of current range inside new range or
 105                 // 4. last of current range inside new range
 106                 return range_.inside(r.first) || range_.inside(r.last) ||
 107                         r.inside(range_.first) || r.inside(range_.last);
 108         }
 109         ///
 110         void shift(pos_type pos, int offset)
 111         {
 112                 if (range_.first > pos) {
 113                         range_.first += offset;
 114                         range_.last += offset;
 115                 } else if (range_.last >= pos) {
 116                         range_.last += offset;
 117                 }
 118         }
 119 private:
 120         FontSpan range_ ;
 121         SpellChecker::Result result_ ;
 122 };
 123
 124
 125 /////////////////////////////////////////////////////////////////////
 126 //
 127 // SpellCheckerState
 128 //
 129 /////////////////////////////////////////////////////////////////////
 130
 131 class SpellCheckerState {
 132 public:
 133         SpellCheckerState() {
 134                 needs_refresh_ = true;
 135                 current_change_number_ = 0;
 136         }
 137
 138         void setRange(FontSpan const fp, SpellChecker::Result state)
 139         {
 140                 Ranges result;
 141                 RangesIterator et = ranges_.end();
 142                 RangesIterator it = ranges_.begin();
 143                 for (; it != et; ++it) {
 144                         if (!it->covered(fp))
 145                                 result.push_back(SpellResultRange(it->range(), it->result()));
 146                         else if (state == SpellChecker::WORD_OK) {
 147                                 // trim or split the current misspelled range
 148                                 // store misspelled ranges only
 149                                 FontSpan range = it->range();
 150                                 if (fp.first > range.first) {
 151                                         // misspelled area in front of WORD_OK
 152                                         range.last = fp.first - 1;
 153                                         result.push_back(SpellResultRange(range, it->result()));
 154                                         range = it->range();
 155                                 }
 156                                 if (fp.last < range.last) {
 157                                         // misspelled area after WORD_OK range
 158                                         range.first = fp.last + 1;
 159                                         result.push_back(SpellResultRange(range, it->result()));
 160                                 }
 161                         }
 162                 }
 163                 ranges_ = result;
 164                 if (state != SpellChecker::WORD_OK)
 165                         ranges_.push_back(SpellResultRange(fp, state));
 166         }
 167
 168         void increasePosAfterPos(pos_type pos)
 169         {
 170                 correctRangesAfterPos(pos, 1);
 171                 needsRefresh(pos);
 172         }
 173
 174         void decreasePosAfterPos(pos_type pos)
 175         {
 176                 correctRangesAfterPos(pos, -1);
 177                 needsRefresh(pos);
 178         }
 179
 180         void refreshLast(pos_type pos)
 181         {
 182                 if (pos < refresh_.last)
 183                         refresh_.last = pos;
 184         }
 185
 186         SpellChecker::Result getState(pos_type pos) const
 187         {
 188                 SpellChecker::Result result = SpellChecker::WORD_OK;
 189                 RangesIterator et = ranges_.end();
 190                 RangesIterator it = ranges_.begin();
 191                 for (; it != et; ++it) {
 192                         if(it->inside(pos)) {
 193                                 return it->result();
 194                         }
 195                 }
 196                 return result;
 197         }
 198
 199         FontSpan const & getRange(pos_type pos) const
 200         {
 201                 /// empty span to indicate mismatch
 202                 static FontSpan empty_;
 203                 RangesIterator et = ranges_.end();
 204                 RangesIterator it = ranges_.begin();
 205                 for (; it != et; ++it) {
 206                         if(it->inside(pos)) {
 207                                 return it->range();
 208                         }
 209                 }
 210                 return empty_;
 211         }
 212
 213         bool needsRefresh() const {
 214                 return needs_refresh_;
 215         }
 216
 217         SpellChecker::ChangeNumber currentChangeNumber() const {
 218                 return current_change_number_;
 219         }
 220
 221         void refreshRange(pos_type & first, pos_type & last) const {
 222                 first = refresh_.first;
 223                 last = refresh_.last;
 224         }
 225
 226         void needsRefresh(pos_type pos) {
 227                 if (needs_refresh_ && pos != -1) {
 228                         if (pos < refresh_.first)
 229                                 refresh_.first = pos;
 230                         if (pos > refresh_.last)
 231                                 refresh_.last = pos;
 232                 } else if (pos != -1) {
 233                         // init request check for neighbour positions too
 234                         refresh_.first = pos > 0 ? pos - 1 : 0;
 235                         // no need for special end of paragraph check
 236                         refresh_.last = pos + 1;
 237                 }
 238                 needs_refresh_ = pos != -1;
 239         }
 240
 241         void needsCompleteRefresh(SpellChecker::ChangeNumber change_number) {
 242                 needs_refresh_ = true;
 243                 refresh_.first = 0;
 244                 refresh_.last = -1;
 245                 current_change_number_ = change_number;
 246         }
 247
 248 private:
 249         typedef vector<SpellResultRange> Ranges;
 250         typedef Ranges::const_iterator RangesIterator;
 251         Ranges ranges_;
 252         /// the area of the paragraph with pending spell check
 253         FontSpan refresh_;
 254         bool needs_refresh_;
 255         /// spell state cache version number
 256         SpellChecker::ChangeNumber current_change_number_;
 257
 258
 259         void correctRangesAfterPos(pos_type pos, int offset)
 260         {
 261                 RangesIterator et = ranges_.end();
 262                 Ranges::iterator it = ranges_.begin();
 263                 for (; it != et; ++it) {
 264                         it->shift(pos, offset);
 265                 }
 266         }
 267
 268 };
 269
 270 /////////////////////////////////////////////////////////////////////
 271 //
 272 // Paragraph::Private
 273 //
 274 /////////////////////////////////////////////////////////////////////
 275
 276 class Paragraph::Private
 277 {
 278 public:
 279         ///
 280         Private(Paragraph * owner, Layout const & layout);
 281         /// "Copy constructor"
 282         Private(Private const &, Paragraph * owner);
 283         /// Copy constructor from \p beg  to \p end
 284         Private(Private const &, Paragraph * owner, pos_type beg, pos_type end);
 285
 286         ///
 287         void insertChar(pos_type pos, char_type c, Change const & change);
 288
 289         /// Output the surrogate pair formed by \p c and \p next to \p os.
 290         /// \return the number of characters written.
 291         int latexSurrogatePair(otexstream & os, char_type c, char_type next,
 292                                OutputParams const &);
 293
 294         /// Output a space in appropriate formatting (or a surrogate pair
 295         /// if the next character is a combining character).
 296         /// \return whether a surrogate pair was output.
 297         bool simpleTeXBlanks(OutputParams const &,
 298                              otexstream &,
 299                              pos_type i,
 300                              unsigned int & column,
 301                              Font const & font,
 302                              Layout const & style);
 303
 304         /// Output consecutive unicode chars, belonging to the same script as
 305         /// specified by the latex macro \p ltx, to \p os starting from \p i.
 306         /// \return the number of characters written.
 307         int writeScriptChars(otexstream & os, docstring const & ltx,
 308                            Change const &, Encoding const &, pos_type & i);
 309
 310         /// This could go to ParagraphParameters if we want to.
 311         int startTeXParParams(BufferParams const &, otexstream &,
 312                               OutputParams const &) const;
 313
 314         /// This could go to ParagraphParameters if we want to.
 315         int endTeXParParams(BufferParams const &, otexstream &,
 316                             OutputParams const &) const;
 317
 318         ///
 319         void latexInset(BufferParams const &,
 320                                    otexstream &,
 321                                    OutputParams &,
 322                                    Font & running_font,
 323                                    Font & basefont,
 324                                    Font const & outerfont,
 325                                    bool & open_font,
 326                                    Change & running_change,
 327                                    Layout const & style,
 328                                    pos_type & i,
 329                                    unsigned int & column);
 330
 331         ///
 332         void latexSpecialChar(
 333                                    otexstream & os,
 334                                    OutputParams const & runparams,
 335                                    Font const & running_font,
 336                                    Change const & running_change,
 337                                    Layout const & style,
 338                                    pos_type & i,
 339                                    pos_type end_pos,
 340                                    unsigned int & column);
 341
 342         ///
 343         bool latexSpecialT1(
 344                 char_type const c,
 345                 otexstream & os,
 346                 pos_type i,
 347                 unsigned int & column);
 348         ///
 349         bool latexSpecialTypewriter(
 350                 char_type const c,
 351                 otexstream & os,
 352                 pos_type i,
 353                 unsigned int & column);
 354         ///
 355         bool latexSpecialPhrase(
 356                 otexstream & os,
 357                 pos_type & i,
 358                 pos_type end_pos,
 359                 unsigned int & column,
 360                 OutputParams const & runparams);
 361
 362         ///
 363         void validate(LaTeXFeatures & features) const;
 364
 365         /// Checks if the paragraph contains only text and no inset or font change.
 366         bool onlyText(Buffer const & buf, Font const & outerfont,
 367                       pos_type initial) const;
 368
 369         /// match a string against a particular point in the paragraph
 370         bool isTextAt(string const & str, pos_type pos) const;
 371
 372         /// a vector of speller skip positions
 373         typedef vector<FontSpan> SkipPositions;
 374         typedef SkipPositions::const_iterator SkipPositionsIterator;
 375
 376         void appendSkipPosition(SkipPositions & skips, pos_type const pos) const;
 377
 378         Language * getSpellLanguage(pos_type const from) const;
 379
 380         Language * locateSpellRange(pos_type & from, pos_type & to,
 381                                                                 SkipPositions & skips) const;
 382
 383         bool hasSpellerChange() const {
 384                 SpellChecker::ChangeNumber speller_change_number = 0;
 385                 if (theSpellChecker())
 386                         speller_change_number = theSpellChecker()->changeNumber();
 387                 return speller_change_number > speller_state_.currentChangeNumber();
 388         }
 389
 390         bool ignoreWord(docstring const & word) const ;
 391
 392         void setMisspelled(pos_type from, pos_type to, SpellChecker::Result state)
 393         {
 394                 pos_type textsize = owner_->size();
 395                 // check for sane arguments
 396                 if (to <= from || from >= textsize)
 397                         return;
 398                 FontSpan fp = FontSpan(from, to - 1);
 399                 speller_state_.setRange(fp, state);
 400         }
 401
 402         void requestSpellCheck(pos_type pos) {
 403                 speller_state_.needsRefresh(pos);
 404         }
 405
 406         void readySpellCheck() {
 407                 speller_state_.needsRefresh(-1);
 408         }
 409
 410         bool needsSpellCheck() const
 411         {
 412                 return speller_state_.needsRefresh();
 413         }
 414
 415         void rangeOfSpellCheck(pos_type & first, pos_type & last) const
 416         {
 417                 speller_state_.refreshRange(first, last);
 418                 if (last == -1) {
 419                         last = owner_->size();
 420                         return;
 421                 }
 422                 pos_type endpos = last;
 423                 owner_->locateWord(first, endpos, WHOLE_WORD);
 424                 if (endpos < last) {
 425                         endpos = last;
 426                         owner_->locateWord(last, endpos, WHOLE_WORD);
 427                 }
 428                 last = endpos;
 429         }
 430
 431         int countSkips(SkipPositionsIterator & it, SkipPositionsIterator const et,
 432                             int & start) const
 433         {
 434                 int numskips = 0;
 435                 while (it != et && it->first < start) {
 436                         int skip = it->last - it->first + 1;
 437                         start += skip;
 438                         numskips += skip;
 439                         ++it;
 440                 }
 441                 return numskips;
 442         }
 443
 444         void markMisspelledWords(pos_type const & first, pos_type const & last,
 445                                                          SpellChecker::Result result,
 446                                                          docstring const & word,
 447                                                          SkipPositions const & skips);
 448
 449         InsetCode ownerCode() const
 450         {
 451                 return inset_owner_ ? inset_owner_->lyxCode() : NO_CODE;
 452         }
 453
 454         /// Which Paragraph owns us?
 455         Paragraph * owner_;
 456
 457         /// In which Inset?
 458         Inset const * inset_owner_;
 459
 460         ///
 461         FontList fontlist_;
 462
 463         ///
 464         int id_;
 465
 466         ///
 467         ParagraphParameters params_;
 468
 469         /// for recording and looking up changes
 470         Changes changes_;
 471
 472         ///
 473         InsetList insetlist_;
 474
 475         /// end of label
 476         pos_type begin_of_body_;
 477
 478         typedef docstring TextContainer;
 479         ///
 480         TextContainer text_;
 481
 482         typedef set<docstring> Words;
 483         typedef map<Language, Words> LangWordsMap;
 484         ///
 485         LangWordsMap words_;
 486         ///
 487         Layout const * layout_;
 488         ///
 489         SpellCheckerState speller_state_;
 490 };
 491
 492
 493 namespace {
 494
 495 struct special_phrase {
 496         string phrase;
 497         docstring macro;
 498         bool builtin;
 499 };
 500
 501 special_phrase const special_phrases[] = {
 502         { "LyX", from_ascii("\\LyX{}"), false },
 503         { "TeX", from_ascii("\\TeX{}"), true },
 504         { "LaTeX2e", from_ascii("\\LaTeXe{}"), true },
 505         { "LaTeX", from_ascii("\\LaTeX{}"), true },
 506 };
 507
 508 size_t const phrases_nr = sizeof(special_phrases)/sizeof(special_phrase);
 509
 510 } // namespace anon
 511
 512
 513 Paragraph::Private::Private(Paragraph * owner, Layout const & layout)
 514         : owner_(owner), inset_owner_(0), id_(-1), begin_of_body_(0), layout_(&layout)
 515 {
 516         text_.reserve(100);
 517 }
 518
 519
 520 // Initialization of the counter for the paragraph id's,
 521 //
 522 // FIXME: There should be a more intelligent way to generate and use the
 523 // paragraph ids per buffer instead a global static counter for all InsetText
 524 // in the running program.
 525 static int paragraph_id = -1;
 526
 527 Paragraph::Private::Private(Private const & p, Paragraph * owner)
 528         : owner_(owner), inset_owner_(p.inset_owner_), fontlist_(p.fontlist_),
 529           params_(p.params_), changes_(p.changes_), insetlist_(p.insetlist_),
 530           begin_of_body_(p.begin_of_body_), text_(p.text_), words_(p.words_),
 531           layout_(p.layout_)
 532 {
 533         id_ = ++paragraph_id;
 534         requestSpellCheck(p.text_.size());
 535 }
 536
 537
 538 Paragraph::Private::Private(Private const & p, Paragraph * owner,
 539         pos_type beg, pos_type end)
 540         : owner_(owner), inset_owner_(p.inset_owner_),
 541           params_(p.params_), changes_(p.changes_),
 542           insetlist_(p.insetlist_, beg, end),
 543           begin_of_body_(p.begin_of_body_), words_(p.words_),
 544           layout_(p.layout_)
 545 {
 546         id_ = ++paragraph_id;
 547         if (beg >= pos_type(p.text_.size()))
 548                 return;
 549         text_ = p.text_.substr(beg, end - beg);
 550
 551         FontList::const_iterator fcit = fontlist_.begin();
 552         FontList::const_iterator fend = fontlist_.end();
 553         for (; fcit != fend; ++fcit) {
 554                 if (fcit->pos() < beg)
 555                         continue;
 556                 if (fcit->pos() >= end) {
 557                         // Add last entry in the fontlist_.
 558                         fontlist_.set(text_.size() - 1, fcit->font());
 559                         break;
 560                 }
 561                 // Add a new entry in the fontlist_.
 562                 fontlist_.set(fcit->pos() - beg, fcit->font());
 563         }
 564         requestSpellCheck(p.text_.size());
 565 }
 566
 567
 568 void Paragraph::addChangesToToc(DocIterator const & cdit,
 569         Buffer const & buf) const
 570 {
 571         d->changes_.addToToc(cdit, buf);
 572 }
 573
 574
 575 bool Paragraph::isDeleted(pos_type start, pos_type end) const
 576 {
 577         LASSERT(start >= 0 && start <= size(), /**/);
 578         LASSERT(end > start && end <= size() + 1, /**/);
 579
 580         return d->changes_.isDeleted(start, end);
 581 }
 582
 583
 584 bool Paragraph::isChanged(pos_type start, pos_type end) const
 585 {
 586         LASSERT(start >= 0 && start <= size(), /**/);
 587         LASSERT(end > start && end <= size() + 1, /**/);
 588
 589         return d->changes_.isChanged(start, end);
 590 }
 591
 592
 593 bool Paragraph::isMergedOnEndOfParDeletion(bool trackChanges) const
 594 {
 595         // keep the logic here in sync with the logic of eraseChars()
 596         if (!trackChanges)
 597                 return true;
 598
 599         Change const change = d->changes_.lookup(size());
 600         return change.inserted() && change.currentAuthor();
 601 }
 602
 603
 604 void Paragraph::setChange(Change const & change)
 605 {
 606         // beware of the imaginary end-of-par character!
 607         d->changes_.set(change, 0, size() + 1);
 608
 609         /*
 610          * Propagate the change recursively - but not in case of DELETED!
 611          *
 612          * Imagine that your co-author makes changes in an existing inset. He
 613          * sends your document to you and you come to the conclusion that the
 614          * inset should go completely. If you erase it, LyX must not delete all
 615          * text within the inset. Otherwise, the change tracked insertions of
 616          * your co-author get lost and there is no way to restore them later.
 617          *
 618          * Conclusion: An inset's content should remain untouched if you delete it
 619          */
 620
 621         if (!change.deleted()) {
 622                 for (pos_type pos = 0; pos < size(); ++pos) {
 623                         if (Inset * inset = getInset(pos))
 624                                 inset->setChange(change);
 625                 }
 626         }
 627 }
 628
 629
 630 void Paragraph::setChange(pos_type pos, Change const & change)
 631 {
 632         LASSERT(pos >= 0 && pos <= size(), /**/);
 633         d->changes_.set(change, pos);
 634
 635         // see comment in setChange(Change const &) above
 636         if (!change.deleted() && pos < size())
 637                         if (Inset * inset = getInset(pos))
 638                                 inset->setChange(change);
 639 }
 640
 641
 642 Change const & Paragraph::lookupChange(pos_type pos) const
 643 {
 644         LASSERT(pos >= 0 && pos <= size(), /**/);
 645         return d->changes_.lookup(pos);
 646 }
 647
 648
 649 void Paragraph::acceptChanges(pos_type start, pos_type end)
 650 {
 651         LASSERT(start >= 0 && start <= size(), /**/);
 652         LASSERT(end > start && end <= size() + 1, /**/);
 653
 654         for (pos_type pos = start; pos < end; ++pos) {
 655                 switch (lookupChange(pos).type) {
 656                         case Change::UNCHANGED:
 657                                 // accept changes in nested inset
 658                                 if (Inset * inset = getInset(pos))
 659                                         inset->acceptChanges();
 660                                 break;
 661
 662                         case Change::INSERTED:
 663                                 d->changes_.set(Change(Change::UNCHANGED), pos);
 664                                 // also accept changes in nested inset
 665                                 if (Inset * inset = getInset(pos))
 666                                         inset->acceptChanges();
 667                                 break;
 668
 669                         case Change::DELETED:
 670                                 // Suppress access to non-existent
 671                                 // "end-of-paragraph char"
 672                                 if (pos < size()) {
 673                                         eraseChar(pos, false);
 674                                         --end;
 675                                         --pos;
 676                                 }
 677                                 break;
 678                 }
 679
 680         }
 681 }
 682
 683
 684 void Paragraph::rejectChanges(pos_type start, pos_type end)
 685 {
 686         LASSERT(start >= 0 && start <= size(), /**/);
 687         LASSERT(end > start && end <= size() + 1, /**/);
 688
 689         for (pos_type pos = start; pos < end; ++pos) {
 690                 switch (lookupChange(pos).type) {
 691                         case Change::UNCHANGED:
 692                                 // reject changes in nested inset
 693                                 if (Inset * inset = getInset(pos))
 694                                                 inset->rejectChanges();
 695                                 break;
 696
 697                         case Change::INSERTED:
 698                                 // Suppress access to non-existent
 699                                 // "end-of-paragraph char"
 700                                 if (pos < size()) {
 701                                         eraseChar(pos, false);
 702                                         --end;
 703                                         --pos;
 704                                 }
 705                                 break;
 706
 707                         case Change::DELETED:
 708                                 d->changes_.set(Change(Change::UNCHANGED), pos);
 709
 710                                 // Do NOT reject changes within a deleted inset!
 711                                 // There may be insertions of a co-author inside of it!
 712
 713                                 break;
 714                 }
 715         }
 716 }
 717
 718
 719 void Paragraph::Private::insertChar(pos_type pos, char_type c,
 720                 Change const & change)
 721 {
 722         LASSERT(pos >= 0 && pos <= int(text_.size()), /**/);
 723
 724         // track change
 725         changes_.insert(change, pos);
 726
 727         // This is actually very common when parsing buffers (and
 728         // maybe inserting ascii text)
 729         if (pos == pos_type(text_.size())) {
 730                 // when appending characters, no need to update tables
 731                 text_.push_back(c);
 732                 // but we want spell checking
 733                 requestSpellCheck(pos);
 734                 return;
 735         }
 736
 737         text_.insert(text_.begin() + pos, c);
 738
 739         // Update the font table.
 740         fontlist_.increasePosAfterPos(pos);
 741
 742         // Update the insets
 743         insetlist_.increasePosAfterPos(pos);
 744
 745         // Update list of misspelled positions
 746         speller_state_.increasePosAfterPos(pos);
 747 }
 748
 749
 750 bool Paragraph::insertInset(pos_type pos, Inset * inset,
 751                                    Change const & change)
 752 {
 753         LASSERT(inset, /**/);
 754         LASSERT(pos >= 0 && pos <= size(), /**/);
 755
 756         // Paragraph::insertInset() can be used in cut/copy/paste operation where
 757         // d->inset_owner_ is not set yet.
 758         if (d->inset_owner_ && !d->inset_owner_->insetAllowed(inset->lyxCode()))
 759                 return false;
 760
 761         d->insertChar(pos, META_INSET, change);
 762         LASSERT(d->text_[pos] == META_INSET, /**/);
 763
 764         // Add a new entry in the insetlist_.
 765         d->insetlist_.insert(inset, pos);
 766
 767         // Some insets require run of spell checker
 768         requestSpellCheck(pos);
 769         return true;
 770 }
 771
 772
 773 bool Paragraph::eraseChar(pos_type pos, bool trackChanges)
 774 {
 775         LASSERT(pos >= 0 && pos <= size(), return false);
 776
 777         // keep the logic here in sync with the logic of isMergedOnEndOfParDeletion()
 778
 779         if (trackChanges) {
 780                 Change change = d->changes_.lookup(pos);
 781
 782                 // set the character to DELETED if
 783                 //  a) it was previously unchanged or
 784                 //  b) it was inserted by a co-author
 785
 786                 if (!change.changed() ||
 787                       (change.inserted() && !change.currentAuthor())) {
 788                         setChange(pos, Change(Change::DELETED));
 789                         // request run of spell checker
 790                         requestSpellCheck(pos);
 791                         return false;
 792                 }
 793
 794                 if (change.deleted())
 795                         return false;
 796         }
 797
 798         // Don't physically access the imaginary end-of-paragraph character.
 799         // eraseChar() can only mark it as DELETED. A physical deletion of
 800         // end-of-par must be handled externally.
 801         if (pos == size()) {
 802                 return false;
 803         }
 804
 805         // track change
 806         d->changes_.erase(pos);
 807
 808         // if it is an inset, delete the inset entry
 809         if (d->text_[pos] == META_INSET)
 810                 d->insetlist_.erase(pos);
 811
 812         d->text_.erase(d->text_.begin() + pos);
 813
 814         // Update the fontlist_
 815         d->fontlist_.erase(pos);
 816
 817         // Update the insetlist_
 818         d->insetlist_.decreasePosAfterPos(pos);
 819
 820         // Update list of misspelled positions
 821         d->speller_state_.decreasePosAfterPos(pos);
 822         d->speller_state_.refreshLast(size());
 823
 824         return true;
 825 }
 826
 827
 828 int Paragraph::eraseChars(pos_type start, pos_type end, bool trackChanges)
 829 {
 830         LASSERT(start >= 0 && start <= size(), /**/);
 831         LASSERT(end >= start && end <= size() + 1, /**/);
 832
 833         pos_type i = start;
 834         for (pos_type count = end - start; count; --count) {
 835                 if (!eraseChar(i, trackChanges))
 836                         ++i;
 837         }
 838         return end - i;
 839 }
 840
 841
 842 int Paragraph::Private::latexSurrogatePair(otexstream & os, char_type c,
 843                 char_type next, OutputParams const & runparams)
 844 {
 845         // Writing next here may circumvent a possible font change between
 846         // c and next. Since next is only output if it forms a surrogate pair
 847         // with c we can ignore this:
 848         // A font change inside a surrogate pair does not make sense and is
 849         // hopefully impossible to input.
 850         // FIXME: change tracking
 851         // Is this correct WRT change tracking?
 852         Encoding const & encoding = *(runparams.encoding);
 853         docstring const latex1 = encoding.latexChar(next);
 854         docstring const latex2 = encoding.latexChar(c);
 855         if (docstring(1, next) == latex1) {
 856                 // the encoding supports the combination
 857                 os << latex2 << latex1;
 858                 return latex1.length() + latex2.length();
 859         } else if (runparams.local_font &&
 860                    runparams.local_font->language()->lang() == "polutonikogreek") {
 861                 // polutonikogreek only works without the brackets
 862                 os << latex1 << latex2;
 863                 return latex1.length() + latex2.length();
 864         } else
 865                 os << latex1 << '{' << latex2 << '}';
 866         return latex1.length() + latex2.length() + 2;
 867 }
 868
 869
 870 bool Paragraph::Private::simpleTeXBlanks(OutputParams const & runparams,
 871                                        otexstream & os,
 872                                        pos_type i,
 873                                        unsigned int & column,
 874                                        Font const & font,
 875                                        Layout const & style)
 876 {
 877         if (style.pass_thru || runparams.pass_thru)
 878                 return false;
 879
 880         if (i + 1 < int(text_.size())) {
 881                 char_type next = text_[i + 1];
 882                 if (Encodings::isCombiningChar(next)) {
 883                         // This space has an accent, so we must always output it.
 884                         column += latexSurrogatePair(os, ' ', next, runparams) - 1;
 885                         return true;
 886                 }
 887         }
 888
 889         if (runparams.linelen > 0
 890             && column > runparams.linelen
 891             && i
 892             && text_[i - 1] != ' '
 893             && (i + 1 < int(text_.size()))
 894             // same in FreeSpacing mode
 895             && !owner_->isFreeSpacing()
 896             // In typewriter mode, we want to avoid
 897             // ! . ? : at the end of a line
 898             && !(font.fontInfo().family() == TYPEWRITER_FAMILY
 899                  && (text_[i - 1] == '.'
 900                      || text_[i - 1] == '?'
 901                      || text_[i - 1] == ':'
 902                      || text_[i - 1] == '!'))) {
 903                 os << '\n';
 904                 os.texrow().start(owner_->id(), i + 1);
 905                 column = 0;
 906         } else if (style.free_spacing) {
 907                 os << '~';
 908         } else {
 909                 os << ' ';
 910         }
 911         return false;
 912 }
 913
 914
 915 int Paragraph::Private::writeScriptChars(otexstream & os,
 916                                          docstring const & ltx,
 917                                          Change const & runningChange,
 918                                          Encoding const & encoding,
 919                                          pos_type & i)
 920 {
 921         // FIXME: modifying i here is not very nice...
 922
 923         // We only arrive here when a proper language for character text_[i] has
 924         // not been specified (i.e., it could not be translated in the current
 925         // latex encoding) or its latex translation has been forced, and it
 926         // belongs to a known script.
 927         // Parameter ltx contains the latex translation of text_[i] as specified
 928         // in the unicodesymbols file and is something like "\textXXX{<spec>}".
 929         // The latex macro name "textXXX" specifies the script to which text_[i]
 930         // belongs and we use it in order to check whether characters from the
 931         // same script immediately follow, such that we can collect them in a
 932         // single "\textXXX" macro. So, we have to retain "\textXXX{<spec>"
 933         // for the first char but only "<spec>" for all subsequent chars.
 934         docstring::size_type const brace1 = ltx.find_first_of(from_ascii("{"));
 935         docstring::size_type const brace2 = ltx.find_last_of(from_ascii("}"));
 936         string script = to_ascii(ltx.substr(1, brace1 - 1));
 937         int pos = 0;
 938         int length = brace2;
 939         bool closing_brace = true;
 940         if (script == "textgreek" && encoding.latexName() == "iso-8859-7") {
 941                 // Correct encoding is being used, so we can avoid \textgreek.
 942                 pos = brace1 + 1;
 943                 length -= pos;
 944                 closing_brace = false;
 945         }
 946         os << ltx.substr(pos, length);
 947         int size = text_.size();
 948         while (i + 1 < size) {
 949                 char_type const next = text_[i + 1];
 950                 // Stop here if next character belongs to another script
 951                 // or there is a change in change tracking status.
 952                 if (!Encodings::isKnownScriptChar(next, script) ||
 953                     runningChange != owner_->lookupChange(i + 1))
 954                         break;
 955                 Font prev_font;
 956                 bool found = false;
 957                 FontList::const_iterator cit = fontlist_.begin();
 958                 FontList::const_iterator end = fontlist_.end();
 959                 for (; cit != end; ++cit) {
 960                         if (cit->pos() >= i && !found) {
 961                                 prev_font = cit->font();
 962                                 found = true;
 963                         }
 964                         if (cit->pos() >= i + 1)
 965                                 break;
 966                 }
 967                 // Stop here if there is a font attribute or encoding change.
 968                 if (found && cit != end && prev_font != cit->font())
 969                         break;
 970                 docstring const latex = encoding.latexChar(next);
 971                 docstring::size_type const b1 =
 972                                         latex.find_first_of(from_ascii("{"));
 973                 docstring::size_type const b2 =
 974                                         latex.find_last_of(from_ascii("}"));
 975                 int const len = b2 - b1 - 1;
 976                 os << latex.substr(b1 + 1, len);
 977                 length += len;
 978                 ++i;
 979         }
 980         if (closing_brace) {
 981                 os << '}';
 982                 ++length;
 983         }
 984         return length;
 985 }
 986
 987
 988 bool Paragraph::Private::isTextAt(string const & str, pos_type pos) const
 989 {
 990         pos_type const len = str.length();
 991
 992         // is the paragraph large enough?
 993         if (pos + len > int(text_.size()))
 994                 return false;
 995
 996         // does the wanted text start at point?
 997         for (string::size_type i = 0; i < str.length(); ++i) {
 998                 // Caution: direct comparison of characters works only
 999                 // because str is pure ASCII.
1000                 if (str[i] != text_[pos + i])
1001                         return false;
1002         }
1003
1004         return fontlist_.hasChangeInRange(pos, len);
1005 }
1006
1007
1008 void Paragraph::Private::latexInset(BufferParams const & bparams,
1009                                     otexstream & os,
1010                                     OutputParams & runparams,
1011                                     Font & running_font,
1012                                     Font & basefont,
1013                                     Font const & outerfont,
1014                                     bool & open_font,
1015                                     Change & running_change,
1016                                     Layout const & style,
1017                                     pos_type & i,
1018                                     unsigned int & column)
1019 {
1020         Inset * inset = owner_->getInset(i);
1021         LASSERT(inset, /**/);
1022
1023         if (style.pass_thru) {
1024                 inset->plaintext(os.os(), runparams);
1025                 return;
1026         }
1027
1028         // FIXME: move this to InsetNewline::latex
1029         if (inset->lyxCode() == NEWLINE_CODE) {
1030                 // newlines are handled differently here than
1031                 // the default in simpleTeXSpecialChars().
1032                 if (!style.newline_allowed) {
1033                         os << '\n';
1034                 } else {
1035                         if (open_font) {
1036                                 column += running_font.latexWriteEndChanges(
1037                                         os, bparams, runparams,
1038                                         basefont, basefont);
1039                                 open_font = false;
1040                         }
1041
1042                         if (running_font.fontInfo().family() == TYPEWRITER_FAMILY)
1043                                 os << '~';
1044
1045                         basefont = owner_->getLayoutFont(bparams, outerfont);
1046                         running_font = basefont;
1047
1048                         if (runparams.moving_arg)
1049                                 os << "\\protect ";
1050
1051                 }
1052                 os.texrow().start(owner_->id(), i + 1);
1053                 column = 0;
1054         }
1055
1056         if (owner_->isDeleted(i)) {
1057                 if( ++runparams.inDeletedInset == 1)
1058                         runparams.changeOfDeletedInset = owner_->lookupChange(i);
1059         }
1060
1061         if (inset->canTrackChanges()) {
1062                 column += Changes::latexMarkChange(os, bparams, running_change,
1063                         Change(Change::UNCHANGED), runparams);
1064                 running_change = Change(Change::UNCHANGED);
1065         }
1066
1067         bool close = false;
1068         odocstream::pos_type const len = os.os().tellp();
1069
1070         if (inset->forceLTR()
1071             && running_font.isRightToLeft()
1072             // ERT is an exception, it should be output with no
1073             // decorations at all
1074             && inset->lyxCode() != ERT_CODE) {
1075                 if (running_font.language()->lang() == "farsi")
1076                         os << "\\beginL{}";
1077                 else
1078                         os << "\\L{";
1079                 close = true;
1080         }
1081
1082         // FIXME: Bug: we can have an empty font change here!
1083         // if there has just been a font change, we are going to close it
1084         // right now, which means stupid latex code like \textsf{}. AFAIK,
1085         // this does not harm dvi output. A minor bug, thus (JMarc)
1086
1087         // Some insets cannot be inside a font change command.
1088         // However, even such insets *can* be placed in \L or \R
1089         // or their equivalents (for RTL language switches), so we don't
1090         // close the language in those cases.
1091         // ArabTeX, though, cannot handle this special behavior, it seems.
1092         bool arabtex = basefont.language()->lang() == "arabic_arabtex"
1093                 || running_font.language()->lang() == "arabic_arabtex";
1094         if (open_font && !inset->inheritFont()) {
1095                 bool closeLanguage = arabtex
1096                         || basefont.isRightToLeft() == running_font.isRightToLeft();
1097                 unsigned int count = running_font.latexWriteEndChanges(os,
1098                         bparams, runparams, basefont, basefont, closeLanguage);
1099                 column += count;
1100                 // if any font properties were closed, update the running_font,
1101                 // making sure, however, to leave the language as it was
1102                 if (count > 0) {
1103                         // FIXME: probably a better way to keep track of the old
1104                         // language, than copying the entire font?
1105                         Font const copy_font(running_font);
1106                         basefont = owner_->getLayoutFont(bparams, outerfont);
1107                         running_font = basefont;
1108                         if (!closeLanguage)
1109                                 running_font.setLanguage(copy_font.language());
1110                         // leave font open if language is still open
1111                         open_font = (running_font.language() == basefont.language());
1112                         if (closeLanguage)
1113                                 runparams.local_font = &basefont;
1114                 }
1115         }
1116
1117         int prev_rows = os.texrow().rows();
1118
1119         try {
1120                 runparams.lastid = id_;
1121                 runparams.lastpos = i;
1122                 inset->latex(os, runparams);
1123         } catch (EncodingException & e) {
1124                 // add location information and throw again.
1125                 e.par_id = id_;
1126                 e.pos = i;
1127                 throw(e);
1128         }
1129
1130         if (close) {
1131                 if (running_font.language()->lang() == "farsi")
1132                                 os << "\\endL{}";
1133                         else
1134                                 os << '}';
1135         }
1136
1137         if (os.texrow().rows() > prev_rows) {
1138                 os.texrow().start(owner_->id(), i + 1);
1139                 column = 0;
1140         } else {
1141                 column += (unsigned int)(os.os().tellp() - len);
1142         }
1143
1144         if (owner_->isDeleted(i))
1145                 --runparams.inDeletedInset;
1146 }
1147
1148
1149 void Paragraph::Private::latexSpecialChar(otexstream & os,
1150                                           OutputParams const & runparams,
1151                                           Font const & running_font,
1152                                           Change const & running_change,
1153                                           Layout const & style,
1154                                           pos_type & i,
1155                                           pos_type end_pos,
1156                                           unsigned int & column)
1157 {
1158         char_type const c = text_[i];
1159
1160         if (style.pass_thru || runparams.pass_thru) {
1161                 if (c != '\0') {
1162                         Encoding const * const enc = runparams.encoding;
1163                         if (enc && enc->latexChar(c, true).empty())
1164                                 throw EncodingException(c);
1165                         os.put(c);
1166                 }
1167                 return;
1168         }
1169
1170         // If T1 font encoding is used, use the special
1171         // characters it provides.
1172         // NOTE: some languages reset the font encoding
1173         // internally
1174         if (!running_font.language()->internalFontEncoding()
1175             && lyxrc.fontenc == "T1" && latexSpecialT1(c, os, i, column))
1176                 return;
1177
1178         // \tt font needs special treatment
1179         if (running_font.fontInfo().family() == TYPEWRITER_FAMILY
1180                 && latexSpecialTypewriter(c, os, i, column))
1181                 return;
1182
1183         // Otherwise, we use what LaTeX provides us.
1184         switch (c) {
1185         case '\\':
1186                 os << "\\textbackslash{}";
1187                 column += 15;
1188                 break;
1189         case '<':
1190                 os << "\\textless{}";
1191                 column += 10;
1192                 break;
1193         case '>':
1194                 os << "\\textgreater{}";
1195                 column += 13;
1196                 break;
1197         case '|':
1198                 os << "\\textbar{}";
1199                 column += 9;
1200                 break;
1201         case '-':
1202                 os << '-';
1203                 break;
1204         case '\"':
1205                 os << "\\char`\\\"{}";
1206                 column += 9;
1207                 break;
1208
1209         case '$': case '&':
1210         case '%': case '#': case '{':
1211         case '}': case '_':
1212                 os << '\\';
1213                 os.put(c);
1214                 column += 1;
1215                 break;
1216
1217         case '~':
1218                 os << "\\textasciitilde{}";
1219                 column += 16;
1220                 break;
1221
1222         case '^':
1223                 os << "\\textasciicircum{}";
1224                 column += 17;
1225                 break;
1226
1227         case '*':
1228         case '[':
1229         case ']':
1230                 // avoid being mistaken for optional arguments
1231                 os << '{';
1232                 os.put(c);
1233                 os << '}';
1234                 column += 2;
1235                 break;
1236
1237         case ' ':
1238                 // Blanks are printed before font switching.
1239                 // Sure? I am not! (try nice-latex)
1240                 // I am sure it's correct. LyX might be smarter
1241                 // in the future, but for now, nothing wrong is
1242                 // written. (Asger)
1243                 break;
1244
1245         default:
1246                 // LyX, LaTeX etc.
1247                 if (latexSpecialPhrase(os, i, end_pos, column, runparams))
1248                         return;
1249
1250                 if (c == '\0')
1251                         return;
1252
1253                 Encoding const & encoding = *(runparams.encoding);
1254                 if (i + 1 < int(text_.size())) {
1255                         char_type next = text_[i + 1];
1256                         if (Encodings::isCombiningChar(next)) {
1257                                 column += latexSurrogatePair(os, c, next, runparams) - 1;
1258                                 ++i;
1259                                 break;
1260                         }
1261                 }
1262                 string script;
1263                 docstring const latex = encoding.latexChar(c);
1264                 if (Encodings::isKnownScriptChar(c, script)
1265                     && prefixIs(latex, from_ascii("\\" + script)))
1266                         column += writeScriptChars(os, latex,
1267                                         running_change, encoding, i) - 1;
1268                 else if (latex.length() > 1 && latex[latex.length() - 1] != '}') {
1269                         // Prevent eating of a following
1270                         // space or command corruption by
1271                         // following characters
1272                         column += latex.length() + 1;
1273                         os << latex << "{}";
1274                 } else {
1275                         column += latex.length() - 1;
1276                         os << latex;
1277                 }
1278                 break;
1279         }
1280 }
1281
1282
1283 bool Paragraph::Private::latexSpecialT1(char_type const c, otexstream & os,
1284         pos_type i, unsigned int & column)
1285 {
1286         switch (c) {
1287         case '>':
1288         case '<':
1289                 os.put(c);
1290                 // In T1 encoding, these characters exist
1291                 // but we should avoid ligatures
1292                 if (i + 1 >= int(text_.size()) || text_[i + 1] != c)
1293                         return true;
1294                 os << "\\textcompwordmark{}";
1295                 column += 19;
1296                 return true;
1297         case '|':
1298                 os.put(c);
1299                 return true;
1300         case '\"':
1301                 // soul.sty breaks with \char`\"
1302                 os << "\\textquotedbl{}";
1303                 column += 14;
1304                 return true;
1305         default:
1306                 return false;
1307         }
1308 }
1309
1310
1311 bool Paragraph::Private::latexSpecialTypewriter(char_type const c, otexstream & os,
1312         pos_type i, unsigned int & column)
1313 {
1314         switch (c) {
1315         case '-':
1316                 // within \ttfamily, "--" is merged to "-" (no endash)
1317                 // so we avoid this rather irritating ligature
1318                 if (i + 1 < int(text_.size()) && text_[i + 1] == '-') {
1319                         os << "-{}";
1320                         column += 2;
1321                 } else
1322                         os << '-';
1323                 return true;
1324
1325         // everything else has to be checked separately
1326         // (depending on the encoding)
1327         default:
1328                 return false;
1329         }
1330 }
1331
1332
1333 /// \param end_pos
1334 ///   If [start_pos, end_pos) does not include entirely the special phrase, then
1335 ///   do not apply the macro transformation.
1336 bool Paragraph::Private::latexSpecialPhrase(otexstream & os, pos_type & i, pos_type end_pos,
1337         unsigned int & column, OutputParams const & runparams)
1338 {
1339         // FIXME: if we have "LaTeX" with a font
1340         // change in the middle (before the 'T', then
1341         // the "TeX" part is still special cased.
1342         // Really we should only operate this on
1343         // "words" for some definition of word
1344
1345         for (size_t pnr = 0; pnr < phrases_nr; ++pnr) {
1346                 if (!isTextAt(special_phrases[pnr].phrase, i)
1347                     || (end_pos != -1 && i + int(special_phrases[pnr].phrase.size()) > end_pos))
1348                         continue;
1349                 if (runparams.moving_arg)
1350                         os << "\\protect";
1351                 os << special_phrases[pnr].macro;
1352                 i += special_phrases[pnr].phrase.length() - 1;
1353                 column += special_phrases[pnr].macro.length() - 1;
1354                 return true;
1355         }
1356         return false;
1357 }
1358
1359
1360 void Paragraph::Private::validate(LaTeXFeatures & features) const
1361 {
1362         if (layout_->inpreamble && inset_owner_) {
1363                 bool const is_command = layout_->latextype == LATEX_COMMAND;
1364                 Buffer const & buf = inset_owner_->buffer();
1365                 BufferParams const & bp = buf.params();
1366                 Font f;
1367                 TexRow texrow;
1368                 // Using a string stream here circumvents the encoding
1369                 // switching machinery of odocstream. Therefore the
1370                 // output is wrong if this paragraph contains content
1371                 // that needs to switch encoding.
1372                 odocstringstream ods;
1373                 otexstream os(ods, texrow);
1374                 if (is_command) {
1375                         os << '\\' << from_ascii(layout_->latexname());
1376                         // we have to provide all the optional arguments here, even though
1377                         // the last one is the only one we care about.
1378                         // Separate handling of optional argument inset.
1379                         if (layout_->optargs != 0 || layout_->reqargs != 0)
1380                                 latexArgInsets(*owner_, os, features.runparams(),
1381                                         layout_->reqargs, layout_->optargs);
1382                         else
1383                                 os << from_ascii(layout_->latexparam());
1384                 }
1385                 docstring::size_type const length = ods.str().length();
1386                 // this will output "{" at the beginning, but not at the end
1387                 owner_->latex(bp, f, os, features.runparams(), 0, -1, true);
1388                 if (ods.str().length() > length) {
1389                         if (is_command)
1390                                 ods << '}';
1391                         string const snippet = to_utf8(ods.str());
1392                         features.addPreambleSnippet(snippet);
1393                 }
1394         }
1395
1396         if (features.runparams().flavor == OutputParams::HTML
1397             && layout_->htmltitle()) {
1398                 features.setHTMLTitle(owner_->asString(AS_STR_INSETS));
1399         }
1400
1401         // check the params.
1402         if (!params_.spacing().isDefault())
1403                 features.require("setspace");
1404
1405         // then the layouts
1406         features.useLayout(layout_->name());
1407
1408         // then the fonts
1409         fontlist_.validate(features);
1410
1411         // then the indentation
1412         if (!params_.leftIndent().zero())
1413                 features.require("ParagraphLeftIndent");
1414
1415         // then the insets
1416         InsetList::const_iterator icit = insetlist_.begin();
1417         InsetList::const_iterator iend = insetlist_.end();
1418         for (; icit != iend; ++icit) {
1419                 if (icit->inset) {
1420                         icit->inset->validate(features);
1421                         if (layout_->needprotect &&
1422                             icit->inset->lyxCode() == FOOT_CODE)
1423                                 features.require("NeedLyXFootnoteCode");
1424                 }
1425         }
1426
1427         // then the contents
1428         for (pos_type i = 0; i < int(text_.size()) ; ++i) {
1429                 for (size_t pnr = 0; pnr < phrases_nr; ++pnr) {
1430                         if (!special_phrases[pnr].builtin
1431                             && isTextAt(special_phrases[pnr].phrase, i)) {
1432                                 features.require(special_phrases[pnr].phrase);
1433                                 break;
1434                         }
1435                 }
1436                 Encodings::validate(text_[i], features);
1437         }
1438 }
1439
1440 /////////////////////////////////////////////////////////////////////
1441 //
1442 // Paragraph
1443 //
1444 /////////////////////////////////////////////////////////////////////
1445
1446 namespace {
1447         Layout const emptyParagraphLayout;
1448 }
1449
1450 Paragraph::Paragraph()
1451         : d(new Paragraph::Private(this, emptyParagraphLayout))
1452 {
1453         itemdepth = 0;
1454         d->params_.clear();
1455 }
1456
1457
1458 Paragraph::Paragraph(Paragraph const & par)
1459         : itemdepth(par.itemdepth),
1460         d(new Paragraph::Private(*par.d, this))
1461 {
1462         registerWords();
1463 }
1464
1465
1466 Paragraph::Paragraph(Paragraph const & par, pos_type beg, pos_type end)
1467         : itemdepth(par.itemdepth),
1468         d(new Paragraph::Private(*par.d, this, beg, end))
1469 {
1470         registerWords();
1471 }
1472
1473
1474 Paragraph & Paragraph::operator=(Paragraph const & par)
1475 {
1476         // needed as we will destroy the private part before copying it
1477         if (&par != this) {
1478                 itemdepth = par.itemdepth;
1479
1480                 deregisterWords();
1481                 delete d;
1482                 d = new Private(*par.d, this);
1483                 registerWords();
1484         }
1485         return *this;
1486 }
1487
1488
1489 Paragraph::~Paragraph()
1490 {
1491         deregisterWords();
1492         delete d;
1493 }
1494
1495
1496 namespace {
1497
1498 // this shall be called just before every "os << ..." action.
1499 void flushString(ostream & os, docstring & s)
1500 {
1501         os << to_utf8(s);
1502         s.erase();
1503 }
1504
1505 }
1506
1507
1508 void Paragraph::write(ostream & os, BufferParams const & bparams,
1509         depth_type & dth) const
1510 {
1511         // The beginning or end of a deeper (i.e. nested) area?
1512         if (dth != d->params_.depth()) {
1513                 if (d->params_.depth() > dth) {
1514                         while (d->params_.depth() > dth) {
1515                                 os << "\n\\begin_deeper";
1516                                 ++dth;
1517                         }
1518                 } else {
1519                         while (d->params_.depth() < dth) {
1520                                 os << "\n\\end_deeper";
1521                                 --dth;
1522                         }
1523                 }
1524         }
1525
1526         // First write the layout
1527         os << "\n\\begin_layout " << to_utf8(d->layout_->name()) << '\n';
1528
1529         d->params_.write(os);
1530
1531         Font font1(inherit_font, bparams.language);
1532
1533         Change running_change = Change(Change::UNCHANGED);
1534
1535         // this string is used as a buffer to avoid repetitive calls
1536         // to to_utf8(), which turn out to be expensive (JMarc)
1537         docstring write_buffer;
1538
1539         int column = 0;
1540         for (pos_type i = 0; i <= size(); ++i) {
1541
1542                 Change const change = lookupChange(i);
1543                 if (change != running_change)
1544                         flushString(os, write_buffer);
1545                 Changes::lyxMarkChange(os, bparams, column, running_change, change);
1546                 running_change = change;
1547
1548                 if (i == size())
1549                         break;
1550
1551                 // Write font changes
1552                 Font font2 = getFontSettings(bparams, i);
1553                 if (font2 != font1) {
1554                         flushString(os, write_buffer);
1555                         font2.lyxWriteChanges(font1, os);
1556                         column = 0;
1557                         font1 = font2;
1558                 }
1559
1560                 char_type const c = d->text_[i];
1561                 switch (c) {
1562                 case META_INSET:
1563                         if (Inset const * inset = getInset(i)) {
1564                                 flushString(os, write_buffer);
1565                                 if (inset->directWrite()) {
1566                                         // international char, let it write
1567                                         // code directly so it's shorter in
1568                                         // the file
1569                                         inset->write(os);
1570                                 } else {
1571                                         if (i)
1572                                                 os << '\n';
1573                                         os << "\\begin_inset ";
1574                                         inset->write(os);
1575                                         os << "\n\\end_inset\n\n";
1576                                         column = 0;
1577                                 }
1578                         }
1579                         break;
1580                 case '\\':
1581                         flushString(os, write_buffer);
1582                         os << "\n\\backslash\n";
1583                         column = 0;
1584                         break;
1585                 case '.':
1586                         flushString(os, write_buffer);
1587                         if (i + 1 < size() && d->text_[i + 1] == ' ') {
1588                                 os << ".\n";
1589                                 column = 0;
1590                         } else
1591                                 os << '.';
1592                         break;
1593                 default:
1594                         if ((column > 70 && c == ' ')
1595                             || column > 79) {
1596                                 flushString(os, write_buffer);
1597                                 os << '\n';
1598                                 column = 0;
1599                         }
1600                         // this check is to amend a bug. LyX sometimes
1601                         // inserts '\0' this could cause problems.
1602                         if (c != '\0')
1603                                 write_buffer.push_back(c);
1604                         else
1605                                 LYXERR0("NUL char in structure.");
1606                         ++column;
1607                         break;
1608                 }
1609         }
1610
1611         flushString(os, write_buffer);
1612         os << "\n\\end_layout\n";
1613 }
1614
1615
1616 void Paragraph::validate(LaTeXFeatures & features) const
1617 {
1618         d->validate(features);
1619 }
1620
1621
1622 void Paragraph::insert(pos_type start, docstring const & str,
1623                        Font const & font, Change const & change)
1624 {
1625         for (size_t i = 0, n = str.size(); i != n ; ++i)
1626                 insertChar(start + i, str[i], font, change);
1627 }
1628
1629
1630 void Paragraph::appendChar(char_type c, Font const & font,
1631                 Change const & change)
1632 {
1633         // track change
1634         d->changes_.insert(change, d->text_.size());
1635         // when appending characters, no need to update tables
1636         d->text_.push_back(c);
1637         setFont(d->text_.size() - 1, font);
1638 }
1639
1640
1641 void Paragraph::appendString(docstring const & s, Font const & font,
1642                 Change const & change)
1643 {
1644         pos_type end = s.size();
1645         size_t oldsize = d->text_.size();
1646         size_t newsize = oldsize + end;
1647         size_t capacity = d->text_.capacity();
1648         if (newsize >= capacity)
1649                 d->text_.reserve(max(capacity + 100, newsize));
1650
1651         // when appending characters, no need to update tables
1652         d->text_.append(s);
1653
1654         // FIXME: Optimize this!
1655         for (size_t i = oldsize; i != newsize; ++i) {
1656                 // track change
1657                 d->changes_.insert(change, i);
1658         }
1659         d->fontlist_.set(oldsize, font);
1660         d->fontlist_.set(newsize - 1, font);
1661 }
1662
1663
1664 void Paragraph::insertChar(pos_type pos, char_type c,
1665                            bool trackChanges)
1666 {
1667         d->insertChar(pos, c, Change(trackChanges ?
1668                            Change::INSERTED : Change::UNCHANGED));
1669 }
1670
1671
1672 void Paragraph::insertChar(pos_type pos, char_type c,
1673                            Font const & font, bool trackChanges)
1674 {
1675         d->insertChar(pos, c, Change(trackChanges ?
1676                            Change::INSERTED : Change::UNCHANGED));
1677         setFont(pos, font);
1678 }
1679
1680
1681 void Paragraph::insertChar(pos_type pos, char_type c,
1682                            Font const & font, Change const & change)
1683 {
1684         d->insertChar(pos, c, change);
1685         setFont(pos, font);
1686 }
1687
1688
1689 bool Paragraph::insertInset(pos_type pos, Inset * inset,
1690                             Font const & font, Change const & change)
1691 {
1692         bool const success = insertInset(pos, inset, change);
1693         // Set the font/language of the inset...
1694         setFont(pos, font);
1695         return success;
1696 }
1697
1698
1699 void Paragraph::resetFonts(Font const & font)
1700 {
1701         d->fontlist_.clear();
1702         d->fontlist_.set(0, font);
1703         d->fontlist_.set(d->text_.size() - 1, font);
1704 }
1705
1706 // Gets uninstantiated font setting at position.
1707 Font const & Paragraph::getFontSettings(BufferParams const & bparams,
1708                                          pos_type pos) const
1709 {
1710         if (pos > size()) {
1711                 LYXERR0("pos: " << pos << " size: " << size());
1712                 LASSERT(pos <= size(), /**/);
1713         }
1714
1715         FontList::const_iterator cit = d->fontlist_.fontIterator(pos);
1716         if (cit != d->fontlist_.end())
1717                 return cit->font();
1718
1719         if (pos == size() && !empty())
1720                 return getFontSettings(bparams, pos - 1);
1721
1722         // Optimisation: avoid a full font instantiation if there is no
1723         // language change from previous call.
1724         static Font previous_font;
1725         static Language const * previous_lang = 0;
1726         Language const * lang = getParLanguage(bparams);
1727         if (lang != previous_lang) {
1728                 previous_lang = lang;
1729                 previous_font = Font(inherit_font, lang);
1730         }
1731         return previous_font;
1732 }
1733
1734
1735 FontSpan Paragraph::fontSpan(pos_type pos) const
1736 {
1737         LASSERT(pos <= size(), /**/);
1738         pos_type start = 0;
1739
1740         FontList::const_iterator cit = d->fontlist_.begin();
1741         FontList::const_iterator end = d->fontlist_.end();
1742         for (; cit != end; ++cit) {
1743                 if (cit->pos() >= pos) {
1744                         if (pos >= beginOfBody())
1745                                 return FontSpan(max(start, beginOfBody()),
1746                                                 cit->pos());
1747                         else
1748                                 return FontSpan(start,
1749                                                 min(beginOfBody() - 1,
1750                                                          cit->pos()));
1751                 }
1752                 start = cit->pos() + 1;
1753         }
1754
1755         // This should not happen, but if so, we take no chances.
1756         // LYXERR0("Paragraph::getEndPosOfFontSpan: This should not happen!");
1757         return FontSpan(pos, pos);
1758 }
1759
1760
1761 // Gets uninstantiated font setting at position 0
1762 Font const & Paragraph::getFirstFontSettings(BufferParams const & bparams) const
1763 {
1764         if (!empty() && !d->fontlist_.empty())
1765                 return d->fontlist_.begin()->font();
1766
1767         // Optimisation: avoid a full font instantiation if there is no
1768         // language change from previous call.
1769         static Font previous_font;
1770         static Language const * previous_lang = 0;
1771         if (bparams.language != previous_lang) {
1772                 previous_lang = bparams.language;
1773                 previous_font = Font(inherit_font, bparams.language);
1774         }
1775
1776         return previous_font;
1777 }
1778
1779
1780 // Gets the fully instantiated font at a given position in a paragraph
1781 // This is basically the same function as Text::GetFont() in text2.cpp.
1782 // The difference is that this one is used for generating the LaTeX file,
1783 // and thus cosmetic "improvements" are disallowed: This has to deliver
1784 // the true picture of the buffer. (Asger)
1785 Font const Paragraph::getFont(BufferParams const & bparams, pos_type pos,
1786                                  Font const & outerfont) const
1787 {
1788         LASSERT(pos >= 0, /**/);
1789
1790         Font font = getFontSettings(bparams, pos);
1791
1792         pos_type const body_pos = beginOfBody();
1793         FontInfo & fi = font.fontInfo();
1794         if (pos < body_pos)
1795                 fi.realize(d->layout_->labelfont);
1796         else
1797                 fi.realize(d->layout_->font);
1798
1799         fi.realize(outerfont.fontInfo());
1800         fi.realize(bparams.getFont().fontInfo());
1801
1802         return font;
1803 }
1804
1805
1806 Font const Paragraph::getLabelFont
1807         (BufferParams const & bparams, Font const & outerfont) const
1808 {
1809         FontInfo tmpfont = d->layout_->labelfont;
1810         tmpfont.realize(outerfont.fontInfo());
1811         tmpfont.realize(bparams.getFont().fontInfo());
1812         return Font(tmpfont, getParLanguage(bparams));
1813 }
1814
1815
1816 Font const Paragraph::getLayoutFont
1817         (BufferParams const & bparams, Font const & outerfont) const
1818 {
1819         FontInfo tmpfont = d->layout_->font;
1820         tmpfont.realize(outerfont.fontInfo());
1821         tmpfont.realize(bparams.getFont().fontInfo());
1822         return Font(tmpfont, getParLanguage(bparams));
1823 }
1824
1825
1826 /// Returns the height of the highest font in range
1827 FontSize Paragraph::highestFontInRange
1828         (pos_type startpos, pos_type endpos, FontSize def_size) const
1829 {
1830         return d->fontlist_.highestInRange(startpos, endpos, def_size);
1831 }
1832
1833
1834 char_type Paragraph::getUChar(BufferParams const & bparams, pos_type pos) const
1835 {
1836         char_type c = d->text_[pos];
1837         if (!lyxrc.rtl_support)
1838                 return c;
1839
1840         char_type uc = c;
1841         switch (c) {
1842         case '(':
1843                 uc = ')';
1844                 break;
1845         case ')':
1846                 uc = '(';
1847                 break;
1848         case '[':
1849                 uc = ']';
1850                 break;
1851         case ']':
1852                 uc = '[';
1853                 break;
1854         case '{':
1855                 uc = '}';
1856                 break;
1857         case '}':
1858                 uc = '{';
1859                 break;
1860         case '<':
1861                 uc = '>';
1862                 break;
1863         case '>':
1864                 uc = '<';
1865                 break;
1866         }
1867         if (uc != c && getFontSettings(bparams, pos).isRightToLeft())
1868                 return uc;
1869         return c;
1870 }
1871
1872
1873 void Paragraph::setFont(pos_type pos, Font const & font)
1874 {
1875         LASSERT(pos <= size(), /**/);
1876
1877         // First, reduce font against layout/label font
1878         // Update: The setCharFont() routine in text2.cpp already
1879         // reduces font, so we don't need to do that here. (Asger)
1880
1881         d->fontlist_.set(pos, font);
1882 }
1883
1884
1885 void Paragraph::makeSameLayout(Paragraph const & par)
1886 {
1887         d->layout_ = par.d->layout_;
1888         d->params_ = par.d->params_;
1889 }
1890
1891
1892 bool Paragraph::stripLeadingSpaces(bool trackChanges)
1893 {
1894         if (isFreeSpacing())
1895                 return false;
1896
1897         int pos = 0;
1898         int count = 0;
1899
1900         while (pos < size() && (isNewline(pos) || isLineSeparator(pos))) {
1901                 if (eraseChar(pos, trackChanges))
1902                         ++count;
1903                 else
1904                         ++pos;
1905         }
1906
1907         return count > 0 || pos > 0;
1908 }
1909
1910
1911 bool Paragraph::hasSameLayout(Paragraph const & par) const
1912 {
1913         return par.d->layout_ == d->layout_
1914                 && d->params_.sameLayout(par.d->params_);
1915 }
1916
1917
1918 depth_type Paragraph::getDepth() const
1919 {
1920         return d->params_.depth();
1921 }
1922
1923
1924 depth_type Paragraph::getMaxDepthAfter() const
1925 {
1926         if (d->layout_->isEnvironment())
1927                 return d->params_.depth() + 1;
1928         else
1929                 return d->params_.depth();
1930 }
1931
1932
1933 char Paragraph::getAlign() const
1934 {
1935         if (d->params_.align() == LYX_ALIGN_LAYOUT)
1936                 return d->layout_->align;
1937         else
1938                 return d->params_.align();
1939 }
1940
1941
1942 docstring const & Paragraph::labelString() const
1943 {
1944         return d->params_.labelString();
1945 }
1946
1947
1948 // the next two functions are for the manual labels
1949 docstring const Paragraph::getLabelWidthString() const
1950 {
1951         if (d->layout_->margintype == MARGIN_MANUAL
1952             || d->layout_->latextype == LATEX_BIB_ENVIRONMENT)
1953                 return d->params_.labelWidthString();
1954         else
1955                 return _("Senseless with this layout!");
1956 }
1957
1958
1959 void Paragraph::setLabelWidthString(docstring const & s)
1960 {
1961         d->params_.labelWidthString(s);
1962 }
1963
1964
1965 docstring Paragraph::expandLabel(Layout const & layout,
1966                 BufferParams const & bparams) const
1967 {
1968         return expandParagraphLabel(layout, bparams, true);
1969 }
1970
1971
1972 docstring Paragraph::expandDocBookLabel(Layout const & layout,
1973                 BufferParams const & bparams) const
1974 {
1975         return expandParagraphLabel(layout, bparams, false);
1976 }
1977
1978
1979 docstring Paragraph::expandParagraphLabel(Layout const & layout,
1980                 BufferParams const & bparams, bool process_appendix) const
1981 {
1982         DocumentClass const & tclass = bparams.documentClass();
1983         string const & lang = getParLanguage(bparams)->code();
1984         bool const in_appendix = process_appendix && d->params_.appendix();
1985         docstring fmt = translateIfPossible(layout.labelstring(in_appendix), lang);
1986
1987         if (fmt.empty() && layout.labeltype == LABEL_COUNTER
1988             && !layout.counter.empty())
1989                 return tclass.counters().theCounter(layout.counter, lang);
1990
1991         // handle 'inherited level parts' in 'fmt',
1992         // i.e. the stuff between '@' in   '@Section@.\arabic{subsection}'
1993         size_t const i = fmt.find('@', 0);
1994         if (i != docstring::npos) {
1995                 size_t const j = fmt.find('@', i + 1);
1996                 if (j != docstring::npos) {
1997                         docstring parent(fmt, i + 1, j - i - 1);
1998                         docstring label = from_ascii("??");
1999                         if (tclass.hasLayout(parent))
2000                                 docstring label = expandParagraphLabel(tclass[parent], bparams,
2001                                                       process_appendix);
2002                         fmt = docstring(fmt, 0, i) + label
2003                                 + docstring(fmt, j + 1, docstring::npos);
2004                 }
2005         }
2006
2007         return tclass.counters().counterLabel(fmt, lang);
2008 }
2009
2010
2011 void Paragraph::applyLayout(Layout const & new_layout)
2012 {
2013         d->layout_ = &new_layout;
2014         LyXAlignment const oldAlign = d->params_.align();
2015
2016         if (!(oldAlign & d->layout_->alignpossible)) {
2017                 frontend::Alert::warning(_("Alignment not permitted"),
2018                         _("The new layout does not permit the alignment previously used.\nSetting to default."));
2019                 d->params_.align(LYX_ALIGN_LAYOUT);
2020         }
2021 }
2022
2023
2024 pos_type Paragraph::beginOfBody() const
2025 {
2026         return d->begin_of_body_;
2027 }
2028
2029
2030 void Paragraph::setBeginOfBody()
2031 {
2032         if (d->layout_->labeltype != LABEL_MANUAL) {
2033                 d->begin_of_body_ = 0;
2034                 return;
2035         }
2036
2037         // Unroll the first two cycles of the loop
2038         // and remember the previous character to
2039         // remove unnecessary getChar() calls
2040         pos_type i = 0;
2041         pos_type end = size();
2042         if (i < end && !isNewline(i)) {
2043                 ++i;
2044                 char_type previous_char = 0;
2045                 char_type temp = 0;
2046                 if (i < end) {
2047                         previous_char = d->text_[i];
2048                         if (!isNewline(i)) {
2049                                 ++i;
2050                                 while (i < end && previous_char != ' ') {
2051                                         temp = d->text_[i];
2052                                         if (isNewline(i))
2053                                                 break;
2054                                         ++i;
2055                                         previous_char = temp;
2056                                 }
2057                         }
2058                 }
2059         }
2060
2061         d->begin_of_body_ = i;
2062 }
2063
2064
2065 bool Paragraph::allowParagraphCustomization() const
2066 {
2067         return inInset().allowParagraphCustomization();
2068 }
2069
2070
2071 bool Paragraph::usePlainLayout() const
2072 {
2073         return inInset().usePlainLayout();
2074 }
2075
2076
2077 bool Paragraph::isPassThru() const
2078 {
2079         return inInset().getLayout().isPassThru() || d->layout_->pass_thru;
2080 }
2081
2082 namespace {
2083
2084 // paragraphs inside floats need different alignment tags to avoid
2085 // unwanted space
2086
2087 bool noTrivlistCentering(InsetCode code)
2088 {
2089         return code == FLOAT_CODE
2090                || code == WRAP_CODE
2091                || code == CELL_CODE;
2092 }
2093
2094
2095 string correction(string const & orig)
2096 {
2097         if (orig == "flushleft")
2098                 return "raggedright";
2099         if (orig == "flushright")
2100                 return "raggedleft";
2101         if (orig == "center")
2102                 return "centering";
2103         return orig;
2104 }
2105
2106
2107 string const corrected_env(string const & suffix, string const & env,
2108         InsetCode code, bool const lastpar)
2109 {
2110         string output = suffix + "{";
2111         if (noTrivlistCentering(code)) {
2112                 if (lastpar) {
2113                         // the last paragraph in non-trivlist-aligned
2114                         // context is special (to avoid unwanted whitespace)
2115                         if (suffix == "\\begin")
2116                                 return "\\" + correction(env) + "{}";
2117                         return string();
2118                 }
2119                 output += correction(env);
2120         } else
2121                 output += env;
2122         output += "}";
2123         if (suffix == "\\begin")
2124                 output += "\n";
2125         return output;
2126 }
2127
2128
2129 void adjust_column(string const & str, int & column)
2130 {
2131         if (!contains(str, "\n"))
2132                 column += str.size();
2133         else {
2134                 string tmp;
2135                 column = rsplit(str, tmp, '\n').size();
2136         }
2137 }
2138
2139 } // namespace anon
2140
2141
2142 int Paragraph::Private::startTeXParParams(BufferParams const & bparams,
2143                         otexstream & os, OutputParams const & runparams) const
2144 {
2145         int column = 0;
2146
2147         if (params_.noindent() && !layout_->pass_thru) {
2148                 os << "\\noindent ";
2149                 column += 10;
2150         }
2151
2152         LyXAlignment const curAlign = params_.align();
2153
2154         if (curAlign == layout_->align)
2155                 return column;
2156
2157         switch (curAlign) {
2158         case LYX_ALIGN_NONE:
2159         case LYX_ALIGN_BLOCK:
2160         case LYX_ALIGN_LAYOUT:
2161         case LYX_ALIGN_SPECIAL:
2162         case LYX_ALIGN_DECIMAL:
2163                 break;
2164         case LYX_ALIGN_LEFT:
2165         case LYX_ALIGN_RIGHT:
2166         case LYX_ALIGN_CENTER:
2167                 if (runparams.moving_arg) {
2168                         os << "\\protect";
2169                         column += 8;
2170                 }
2171                 break;
2172         }
2173
2174         string const begin_tag = "\\begin";
2175         InsetCode code = ownerCode();
2176         bool const lastpar = runparams.isLastPar;
2177
2178         switch (curAlign) {
2179         case LYX_ALIGN_NONE:
2180         case LYX_ALIGN_BLOCK:
2181         case LYX_ALIGN_LAYOUT:
2182         case LYX_ALIGN_SPECIAL:
2183         case LYX_ALIGN_DECIMAL:
2184                 break;
2185         case LYX_ALIGN_LEFT: {
2186                 string output;
2187                 if (owner_->getParLanguage(bparams)->babel() != "hebrew")
2188                         output = corrected_env(begin_tag, "flushleft", code, lastpar);
2189                 else
2190                         output = corrected_env(begin_tag, "flushright", code, lastpar);
2191                 os << from_ascii(output);
2192                 adjust_column(output, column);
2193                 break;
2194         } case LYX_ALIGN_RIGHT: {
2195                 string output;
2196                 if (owner_->getParLanguage(bparams)->babel() != "hebrew")
2197                         output = corrected_env(begin_tag, "flushright", code, lastpar);
2198                 else
2199                         output = corrected_env(begin_tag, "flushleft", code, lastpar);
2200                 os << from_ascii(output);
2201                 adjust_column(output, column);
2202                 break;
2203         } case LYX_ALIGN_CENTER: {
2204                 string output;
2205                 output = corrected_env(begin_tag, "center", code, lastpar);
2206                 os << from_ascii(output);
2207                 adjust_column(output, column);
2208                 break;
2209         }
2210         }
2211
2212         return column;
2213 }
2214
2215
2216 int Paragraph::Private::endTeXParParams(BufferParams const & bparams,
2217                         otexstream & os, OutputParams const & runparams) const
2218 {
2219         int column = 0;
2220
2221         LyXAlignment const curAlign = params_.align();
2222
2223         if (curAlign == layout_->align)
2224                 return column;
2225
2226         switch (curAlign) {
2227         case LYX_ALIGN_NONE:
2228         case LYX_ALIGN_BLOCK:
2229         case LYX_ALIGN_LAYOUT:
2230         case LYX_ALIGN_SPECIAL:
2231         case LYX_ALIGN_DECIMAL:
2232                 break;
2233         case LYX_ALIGN_LEFT:
2234         case LYX_ALIGN_RIGHT:
2235         case LYX_ALIGN_CENTER:
2236                 if (runparams.moving_arg) {
2237                         os << "\\protect";
2238                         column = 8;
2239                 }
2240                 break;
2241         }
2242
2243         string const end_tag = "\n\\par\\end";
2244         InsetCode code = ownerCode();
2245         bool const lastpar = runparams.isLastPar;
2246
2247         switch (curAlign) {
2248         case LYX_ALIGN_NONE:
2249         case LYX_ALIGN_BLOCK:
2250         case LYX_ALIGN_LAYOUT:
2251         case LYX_ALIGN_SPECIAL:
2252         case LYX_ALIGN_DECIMAL:
2253                 break;
2254         case LYX_ALIGN_LEFT: {
2255                 string output;
2256                 if (owner_->getParLanguage(bparams)->babel() != "hebrew")
2257                         output = corrected_env(end_tag, "flushleft", code, lastpar);
2258                 else
2259                         output = corrected_env(end_tag, "flushright", code, lastpar);
2260                 os << from_ascii(output);
2261                 adjust_column(output, column);
2262                 break;
2263         } case LYX_ALIGN_RIGHT: {
2264                 string output;
2265                 if (owner_->getParLanguage(bparams)->babel() != "hebrew")
2266                         output = corrected_env(end_tag, "flushright", code, lastpar);
2267                 else
2268                         output = corrected_env(end_tag, "flushleft", code, lastpar);
2269                 os << from_ascii(output);
2270                 adjust_column(output, column);
2271                 break;
2272         } case LYX_ALIGN_CENTER: {
2273                 string output;
2274                 output = corrected_env(end_tag, "center", code, lastpar);
2275                 os << from_ascii(output);
2276                 adjust_column(output, column);
2277                 break;
2278         }
2279         }
2280
2281         return column;
2282 }
2283
2284
2285 // This one spits out the text of the paragraph
2286 void Paragraph::latex(BufferParams const & bparams,
2287         Font const & outerfont,
2288         otexstream & os,
2289         OutputParams const & runparams,
2290         int start_pos, int end_pos, bool force) const
2291 {
2292         LYXERR(Debug::LATEX, "Paragraph::latex...     " << this);
2293
2294         // FIXME This check should not be needed. Perhaps issue an
2295         // error if it triggers.
2296         Layout const & style = inInset().forcePlainLayout() ?
2297                 bparams.documentClass().plainLayout() : *d->layout_;
2298
2299         if (!force && style.inpreamble)
2300                 return;
2301
2302         bool const allowcust = allowParagraphCustomization();
2303
2304         // Current base font for all inherited font changes, without any
2305         // change caused by an individual character, except for the language:
2306         // It is set to the language of the first character.
2307         // As long as we are in the label, this font is the base font of the
2308         // label. Before the first body character it is set to the base font
2309         // of the body.
2310         Font basefont;
2311
2312         // Maybe we have to create a optional argument.
2313         pos_type body_pos = beginOfBody();
2314         unsigned int column = 0;
2315
2316         if (body_pos > 0) {
2317                 // the optional argument is kept in curly brackets in
2318                 // case it contains a ']'
2319                 os << "[{";
2320                 column += 2;
2321                 basefont = getLabelFont(bparams, outerfont);
2322         } else {
2323                 basefont = getLayoutFont(bparams, outerfont);
2324         }
2325
2326         // Which font is currently active?
2327         Font running_font(basefont);
2328         // Do we have an open font change?
2329         bool open_font = false;
2330
2331         Change runningChange = Change(Change::UNCHANGED);
2332
2333         Encoding const * const prev_encoding = runparams.encoding;
2334
2335         os.texrow().start(id(), 0);
2336
2337         // if the paragraph is empty, the loop will not be entered at all
2338         if (empty()) {
2339                 if (style.isCommand()) {
2340                         os << '{';
2341                         ++column;
2342                 }
2343                 if (allowcust)
2344                         column += d->startTeXParParams(bparams, os, runparams);
2345         }
2346
2347         for (pos_type i = 0; i < size(); ++i) {
2348                 // First char in paragraph or after label?
2349                 if (i == body_pos) {
2350                         if (body_pos > 0) {
2351                                 if (open_font) {
2352                                         column += running_font.latexWriteEndChanges(
2353                                                 os, bparams, runparams,
2354                                                 basefont, basefont);
2355                                         open_font = false;
2356                                 }
2357                                 basefont = getLayoutFont(bparams, outerfont);
2358                                 running_font = basefont;
2359
2360                                 column += Changes::latexMarkChange(os, bparams,
2361                                                 runningChange, Change(Change::UNCHANGED),
2362                                                 runparams);
2363                                 runningChange = Change(Change::UNCHANGED);
2364
2365                                 os << "}] ";
2366                                 column +=3;
2367                         }
2368                         if (style.isCommand()) {
2369                                 os << '{';
2370                                 ++column;
2371                         }
2372
2373                         if (allowcust)
2374                                 column += d->startTeXParParams(bparams, os,
2375                                                             runparams);
2376                 }
2377
2378                 Change const & change = runparams.inDeletedInset
2379                         ? runparams.changeOfDeletedInset : lookupChange(i);
2380
2381                 if (bparams.outputChanges && runningChange != change) {
2382                         if (open_font) {
2383                                 column += running_font.latexWriteEndChanges(
2384                                                 os, bparams, runparams, basefont, basefont);
2385                                 open_font = false;
2386                         }
2387                         basefont = getLayoutFont(bparams, outerfont);
2388                         running_font = basefont;
2389
2390                         column += Changes::latexMarkChange(os, bparams, runningChange,
2391                                                            change, runparams);
2392                         runningChange = change;
2393                 }
2394
2395                 // do not output text which is marked deleted
2396                 // if change tracking output is disabled
2397                 if (!bparams.outputChanges && change.deleted()) {
2398                         continue;
2399                 }
2400
2401                 ++column;
2402
2403                 // Fully instantiated font
2404                 Font const font = getFont(bparams, i, outerfont);
2405
2406                 Font const last_font = running_font;
2407
2408                 // Do we need to close the previous font?
2409                 if (open_font &&
2410                     (font != running_font ||
2411                      font.language() != running_font.language()))
2412                 {
2413                         column += running_font.latexWriteEndChanges(
2414                                         os, bparams, runparams, basefont,
2415                                         (i == body_pos-1) ? basefont : font);
2416                         running_font = basefont;
2417                         open_font = false;
2418                 }
2419
2420                 string const running_lang = runparams.use_polyglossia ?
2421                         running_font.language()->polyglossia() : running_font.language()->babel();
2422                 // close babel's font environment before opening CJK.
2423                 string const lang_end_command = runparams.use_polyglossia ?
2424                         "\\end{$$lang}" : lyxrc.language_command_end;
2425                 if (!running_lang.empty() &&
2426                     font.language()->encoding()->package() == Encoding::CJK) {
2427                                 string end_tag = subst(lang_end_command,
2428                                                         "$$lang",
2429                                                         running_lang);
2430                                 os << from_ascii(end_tag);
2431                                 column += end_tag.length();
2432                 }
2433
2434                 // Switch file encoding if necessary (and allowed)
2435                 if (!runparams.pass_thru && !style.pass_thru &&
2436                     runparams.encoding->package() != Encoding::none &&
2437                     font.language()->encoding()->package() != Encoding::none) {
2438                         pair<bool, int> const enc_switch =
2439                                 switchEncoding(os.os(), bparams, runparams,
2440                                         *(font.language()->encoding()));
2441                         if (enc_switch.first) {
2442                                 column += enc_switch.second;
2443                                 runparams.encoding = font.language()->encoding();
2444                         }
2445                 }
2446
2447                 char_type const c = d->text_[i];
2448
2449                 // Do we need to change font?
2450                 if ((font != running_font ||
2451                      font.language() != running_font.language()) &&
2452                         i != body_pos - 1)
2453                 {
2454                         odocstringstream ods;
2455                         column += font.latexWriteStartChanges(ods, bparams,
2456                                                               runparams, basefont,
2457                                                               last_font);
2458                         running_font = font;
2459                         open_font = true;
2460                         docstring fontchange = ods.str();
2461                         // check whether the fontchange ends with a \\textcolor
2462                         // modifier and the text starts with a space (bug 4473)
2463                         docstring const last_modifier = rsplit(fontchange, '\\');
2464                         if (prefixIs(last_modifier, from_ascii("textcolor")) && c == ' ')
2465                                 os << fontchange << from_ascii("{}");
2466                         // check if the fontchange ends with a trailing blank
2467                         // (like "\small " (see bug 3382)
2468                         else if (suffixIs(fontchange, ' ') && c == ' ')
2469                                 os << fontchange.substr(0, fontchange.size() - 1)
2470                                    << from_ascii("{}");
2471                         else
2472                                 os << fontchange;
2473                 }
2474
2475                 // FIXME: think about end_pos implementation...
2476                 if (c == ' ' && i >= start_pos && (end_pos == -1 || i < end_pos)) {
2477                         // FIXME: integrate this case in latexSpecialChar
2478                         // Do not print the separation of the optional argument
2479                         // if style.pass_thru is false. This works because
2480                         // latexSpecialChar ignores spaces if
2481                         // style.pass_thru is false.
2482                         if (i != body_pos - 1) {
2483                                 if (d->simpleTeXBlanks(runparams, os,
2484                                                 i, column, font, style)) {
2485                                         // A surrogate pair was output. We
2486                                         // must not call latexSpecialChar
2487                                         // in this iteration, since it would output
2488                                         // the combining character again.
2489                                         ++i;
2490                                         continue;
2491                                 }
2492                         }
2493                 }
2494
2495                 OutputParams rp = runparams;
2496                 rp.free_spacing = style.free_spacing;
2497                 rp.local_font = &font;
2498                 rp.intitle = style.intitle;
2499
2500                 // Two major modes:  LaTeX or plain
2501                 // Handle here those cases common to both modes
2502                 // and then split to handle the two modes separately.
2503                 if (c == META_INSET) {
2504                         if (i >= start_pos && (end_pos == -1 || i < end_pos)) {
2505                                 d->latexInset(bparams, os, rp, running_font,
2506                                                 basefont, outerfont, open_font,
2507                                                 runningChange, style, i, column);
2508                         }
2509                 } else {
2510                         if (i >= start_pos && (end_pos == -1 || i < end_pos)) {
2511                                 try {
2512                                         d->latexSpecialChar(os, rp, running_font, runningChange,
2513                                                             style, i, end_pos, column);
2514                                 } catch (EncodingException & e) {
2515                                 if (runparams.dryrun) {
2516                                         os << "<" << _("LyX Warning: ")
2517                                            << _("uncodable character") << " '";
2518                                         os.put(c);
2519                                         os << "'>";
2520                                 } else {
2521                                         // add location information and throw again.
2522                                         e.par_id = id();
2523                                         e.pos = i;
2524                                         throw(e);
2525                                 }
2526                         }
2527                 }
2528                 }
2529
2530                 // Set the encoding to that returned from latexSpecialChar (see
2531                 // comment for encoding member in OutputParams.h)
2532                 runparams.encoding = rp.encoding;
2533         }
2534
2535         // If we have an open font definition, we have to close it
2536         if (open_font) {
2537 #ifdef FIXED_LANGUAGE_END_DETECTION
2538                 if (next_) {
2539                         running_font.latexWriteEndChanges(os, bparams,
2540                                         runparams, basefont,
2541                                         next_->getFont(bparams, 0, outerfont));
2542                 } else {
2543                         running_font.latexWriteEndChanges(os, bparams,
2544                                         runparams, basefont, basefont);
2545                 }
2546 #else
2547 //FIXME: For now we ALWAYS have to close the foreign font settings if they are
2548 //FIXME: there as we start another \selectlanguage with the next paragraph if
2549 //FIXME: we are in need of this. This should be fixed sometime (Jug)
2550                 running_font.latexWriteEndChanges(os, bparams, runparams,
2551                                 basefont, basefont);
2552 #endif
2553         }
2554
2555         column += Changes::latexMarkChange(os, bparams, runningChange,
2556                                            Change(Change::UNCHANGED), runparams);
2557
2558         // Needed if there is an optional argument but no contents.
2559         if (body_pos > 0 && body_pos == size()) {
2560                 os << "}]~";
2561         }
2562
2563         if (allowcust && d->endTeXParParams(bparams, os, runparams)
2564             && runparams.encoding != prev_encoding) {
2565                 runparams.encoding = prev_encoding;
2566                 if (!runparams.isFullUnicode())
2567                         os << setEncoding(prev_encoding->iconvName());
2568         }
2569
2570         LYXERR(Debug::LATEX, "Paragraph::latex... done " << this);
2571 }
2572
2573
2574 bool Paragraph::emptyTag() const
2575 {
2576         for (pos_type i = 0; i < size(); ++i) {
2577                 if (Inset const * inset = getInset(i)) {
2578                         InsetCode lyx_code = inset->lyxCode();
2579                         // FIXME testing like that is wrong. What is
2580                         // the intent?
2581                         if (lyx_code != TOC_CODE &&
2582                             lyx_code != INCLUDE_CODE &&
2583                             lyx_code != GRAPHICS_CODE &&
2584                             lyx_code != ERT_CODE &&
2585                             lyx_code != LISTINGS_CODE &&
2586                             lyx_code != FLOAT_CODE &&
2587                             lyx_code != TABULAR_CODE) {
2588                                 return false;
2589                         }
2590                 } else {
2591                         char_type c = d->text_[i];
2592                         if (c != ' ' && c != '\t')
2593                                 return false;
2594                 }
2595         }
2596         return true;
2597 }
2598
2599
2600 string Paragraph::getID(Buffer const & buf, OutputParams const & runparams)
2601         const
2602 {
2603         for (pos_type i = 0; i < size(); ++i) {
2604                 if (Inset const * inset = getInset(i)) {
2605                         InsetCode lyx_code = inset->lyxCode();
2606                         if (lyx_code == LABEL_CODE) {
2607                                 InsetLabel const * const il = static_cast<InsetLabel const *>(inset);
2608                                 docstring const & id = il->getParam("name");
2609                                 return "id='" + to_utf8(sgml::cleanID(buf, runparams, id)) + "'";
2610                         }
2611                 }
2612         }
2613         return string();
2614 }
2615
2616
2617 pos_type Paragraph::firstWordDocBook(odocstream & os, OutputParams const & runparams)
2618         const
2619 {
2620         pos_type i;
2621         for (i = 0; i < size(); ++i) {
2622                 if (Inset const * inset = getInset(i)) {
2623                         inset->docbook(os, runparams);
2624                 } else {
2625                         char_type c = d->text_[i];
2626                         if (c == ' ')
2627                                 break;
2628                         os << sgml::escapeChar(c);
2629                 }
2630         }
2631         return i;
2632 }
2633
2634
2635 pos_type Paragraph::firstWordLyXHTML(XHTMLStream & xs, OutputParams const & runparams)
2636         const
2637 {
2638         pos_type i;
2639         for (i = 0; i < size(); ++i) {
2640                 if (Inset const * inset = getInset(i)) {
2641                         inset->xhtml(xs, runparams);
2642                 } else {
2643                         char_type c = d->text_[i];
2644                         if (c == ' ')
2645                                 break;
2646                         xs << c;
2647                 }
2648         }
2649         return i;
2650 }
2651
2652
2653 bool Paragraph::Private::onlyText(Buffer const & buf, Font const & outerfont, pos_type initial) const
2654 {
2655         Font font_old;
2656         pos_type size = text_.size();
2657         for (pos_type i = initial; i < size; ++i) {
2658                 Font font = owner_->getFont(buf.params(), i, outerfont);
2659                 if (text_[i] == META_INSET)
2660                         return false;
2661                 if (i != initial && font != font_old)
2662                         return false;
2663                 font_old = font;
2664         }
2665
2666         return true;
2667 }
2668
2669
2670 void Paragraph::simpleDocBookOnePar(Buffer const & buf,
2671                                     odocstream & os,
2672                                     OutputParams const & runparams,
2673                                     Font const & outerfont,
2674                                     pos_type initial) const
2675 {
2676         bool emph_flag = false;
2677
2678         Layout const & style = *d->layout_;
2679         FontInfo font_old =
2680                 style.labeltype == LABEL_MANUAL ? style.labelfont : style.font;
2681
2682         if (style.pass_thru && !d->onlyText(buf, outerfont, initial))
2683                 os << "]]>";
2684
2685         // parsing main loop
2686         for (pos_type i = initial; i < size(); ++i) {
2687                 Font font = getFont(buf.params(), i, outerfont);
2688
2689                 // handle <emphasis> tag
2690                 if (font_old.emph() != font.fontInfo().emph()) {
2691                         if (font.fontInfo().emph() == FONT_ON) {
2692                                 os << "<emphasis>";
2693                                 emph_flag = true;
2694                         } else if (i != initial) {
2695                                 os << "</emphasis>";
2696                                 emph_flag = false;
2697                         }
2698                 }
2699
2700                 if (Inset const * inset = getInset(i)) {
2701                         inset->docbook(os, runparams);
2702                 } else {
2703                         char_type c = d->text_[i];
2704
2705                         if (style.pass_thru)
2706                                 os.put(c);
2707                         else
2708                                 os << sgml::escapeChar(c);
2709                 }
2710                 font_old = font.fontInfo();
2711         }
2712
2713         if (emph_flag) {
2714                 os << "</emphasis>";
2715         }
2716
2717         if (style.free_spacing)
2718                 os << '\n';
2719         if (style.pass_thru && !d->onlyText(buf, outerfont, initial))
2720                 os << "<![CDATA[";
2721 }
2722
2723
2724 docstring Paragraph::simpleLyXHTMLOnePar(Buffer const & buf,
2725                                     XHTMLStream & xs,
2726                                     OutputParams const & runparams,
2727                                     Font const & outerfont,
2728                                     pos_type initial) const
2729 {
2730         docstring retval;
2731
2732         bool emph_flag = false;
2733         bool bold_flag = false;
2734         string closing_tag;
2735
2736         Layout const & style = *d->layout_;
2737
2738         if (!runparams.for_toc && runparams.html_make_pars) {
2739                 // generate a magic label for this paragraph
2740                 string const attr = "id='" + magicLabel() + "'";
2741                 xs << html::CompTag("a", attr);
2742         }
2743
2744         FontInfo font_old =
2745                 style.labeltype == LABEL_MANUAL ? style.labelfont : style.font;
2746
2747         // parsing main loop
2748         for (pos_type i = initial; i < size(); ++i) {
2749                 // let's not show deleted material in the output
2750                 if (isDeleted(i))
2751                         continue;
2752
2753                 Font font = getFont(buf.params(), i, outerfont);
2754
2755                 // emphasis
2756                 if (font_old.emph() != font.fontInfo().emph()) {
2757                         if (font.fontInfo().emph() == FONT_ON) {
2758                                 xs << html::StartTag("em");
2759                                 emph_flag = true;
2760                         } else if (emph_flag && i != initial) {
2761                                 xs << html::EndTag("em");
2762                                 emph_flag = false;
2763                         }
2764                 }
2765                 // bold
2766                 if (font_old.series() != font.fontInfo().series()) {
2767                         if (font.fontInfo().series() == BOLD_SERIES) {
2768                                 xs << html::StartTag("strong");
2769                                 bold_flag = true;
2770                         } else if (bold_flag && i != initial) {
2771                                 xs << html::EndTag("strong");
2772                                 bold_flag = false;
2773                         }
2774                 }
2775                 // FIXME XHTML
2776                 // Other such tags? What about the other text ranges?
2777
2778                 Inset const * inset = getInset(i);
2779                 if (inset) {
2780                         if (!runparams.for_toc || inset->isInToc()) {
2781                                 OutputParams np = runparams;
2782                                 if (!inset->getLayout().htmlisblock())
2783                                         np.html_in_par = true;
2784                                 retval += inset->xhtml(xs, np);
2785                         }
2786                 } else {
2787                         char_type c = d->text_[i];
2788
2789                         if (style.pass_thru)
2790                                 xs << c;
2791                         else if (c == '-') {
2792                                 docstring str;
2793                                 int j = i + 1;
2794                                 if (j < size() && d->text_[j] == '-') {
2795                                         j += 1;
2796                                         if (j < size() && d->text_[j] == '-') {
2797                                                 str += from_ascii("&mdash;");
2798                                                 i += 2;
2799                                         } else {
2800                                                 str += from_ascii("&ndash;");
2801                                                 i += 1;
2802                                         }
2803                                 }
2804                                 else
2805                                         str += c;
2806                                 // We don't want to escape the entities. Note that
2807                                 // it is safe to do this, since str can otherwise
2808                                 // only be "-". E.g., it can't be "<".
2809                                 xs << XHTMLStream::ESCAPE_NONE << str;
2810                         } else
2811                                 xs << c;
2812                 }
2813                 font_old = font.fontInfo();
2814         }
2815
2816         xs.closeFontTags();
2817         return retval;
2818 }
2819
2820
2821 bool Paragraph::isHfill(pos_type pos) const
2822 {
2823         Inset const * inset = getInset(pos);
2824         return inset && (inset->lyxCode() == SPACE_CODE &&
2825                          inset->isStretchableSpace());
2826 }
2827
2828
2829 bool Paragraph::isNewline(pos_type pos) const
2830 {
2831         Inset const * inset = getInset(pos);
2832         return inset && inset->lyxCode() == NEWLINE_CODE;
2833 }
2834
2835
2836 bool Paragraph::isLineSeparator(pos_type pos) const
2837 {
2838         char_type const c = d->text_[pos];
2839         if (isLineSeparatorChar(c))
2840                 return true;
2841         Inset const * inset = getInset(pos);
2842         return inset && inset->isLineSeparator();
2843 }
2844
2845
2846 bool Paragraph::isWordSeparator(pos_type pos) const
2847 {
2848         if (Inset const * inset = getInset(pos))
2849                 return !inset->isLetter();
2850         if (pos == size())
2851                 return true;
2852         char_type const c = d->text_[pos];
2853         // if we have a hard hyphen (no en- or emdash),
2854         // we pass this to the spell checker
2855         if (c == '-') {
2856                 int j = pos + 1;
2857                 if ((j == size() || d->text_[j] != '-')
2858                     && (pos == 0 || d->text_[pos - 1] != '-'))
2859                         return false;
2860         }
2861         // We want to pass the ' and escape chars to the spellchecker
2862         static docstring const quote = from_utf8(lyxrc.spellchecker_esc_chars + '\'');
2863         return (!isLetterChar(c) && !isDigitASCII(c) && !contains(quote, c));
2864 }
2865
2866
2867 bool Paragraph::isSameSpellRange(pos_type pos1, pos_type pos2) const
2868 {
2869         return pos1 == pos2
2870                 || d->speller_state_.getRange(pos1) == d->speller_state_.getRange(pos2);
2871 }
2872
2873
2874 bool Paragraph::isChar(pos_type pos) const
2875 {
2876         if (Inset const * inset = getInset(pos))
2877                 return inset->isChar();
2878         char_type const c = d->text_[pos];
2879         return !isLetterChar(c) && !isDigitASCII(c) && !lyx::isSpace(c);
2880 }
2881
2882
2883 bool Paragraph::isSpace(pos_type pos) const
2884 {
2885         if (Inset const * inset = getInset(pos))
2886                 return inset->isSpace();
2887         char_type const c = d->text_[pos];
2888         return lyx::isSpace(c);
2889 }
2890
2891
2892 Language const *
2893 Paragraph::getParLanguage(BufferParams const & bparams) const
2894 {
2895         if (!empty())
2896                 return getFirstFontSettings(bparams).language();
2897         // FIXME: we should check the prev par as well (Lgb)
2898         return bparams.language;
2899 }
2900
2901
2902 bool Paragraph::isRTL(BufferParams const & bparams) const
2903 {
2904         return lyxrc.rtl_support
2905                 && getParLanguage(bparams)->rightToLeft()
2906                 && !inInset().getLayout().forceLTR();
2907 }
2908
2909
2910 void Paragraph::changeLanguage(BufferParams const & bparams,
2911                                Language const * from, Language const * to)
2912 {
2913         // change language including dummy font change at the end
2914         for (pos_type i = 0; i <= size(); ++i) {
2915                 Font font = getFontSettings(bparams, i);
2916                 if (font.language() == from) {
2917                         font.setLanguage(to);
2918                         setFont(i, font);
2919                 }
2920         }
2921         d->requestSpellCheck(size());
2922 }
2923
2924
2925 bool Paragraph::isMultiLingual(BufferParams const & bparams) const
2926 {
2927         Language const * doc_language = bparams.language;
2928         FontList::const_iterator cit = d->fontlist_.begin();
2929         FontList::const_iterator end = d->fontlist_.end();
2930
2931         for (; cit != end; ++cit)
2932                 if (cit->font().language() != ignore_language &&
2933                     cit->font().language() != latex_language &&
2934                     cit->font().language() != doc_language)
2935                         return true;
2936         return false;
2937 }
2938
2939
2940 void Paragraph::getLanguages(std::set<Language const *> & languages) const
2941 {
2942         FontList::const_iterator cit = d->fontlist_.begin();
2943         FontList::const_iterator end = d->fontlist_.end();
2944
2945         for (; cit != end; ++cit) {
2946                 Language const * lang = cit->font().language();
2947                 if (lang != ignore_language &&
2948                     lang != latex_language)
2949                         languages.insert(lang);
2950         }
2951 }
2952
2953
2954 docstring Paragraph::asString(int options) const
2955 {
2956         return asString(0, size(), options);
2957 }
2958
2959
2960 docstring Paragraph::asString(pos_type beg, pos_type end, int options) const
2961 {
2962         odocstringstream os;
2963
2964         if (beg == 0
2965             && options & AS_STR_LABEL
2966             && !d->params_.labelString().empty())
2967                 os << d->params_.labelString() << ' ';
2968
2969         for (pos_type i = beg; i < end; ++i) {
2970                 if ((options & AS_STR_SKIPDELETE) && isDeleted(i))
2971                         continue;
2972                 char_type const c = d->text_[i];
2973                 if (isPrintable(c) || c == '\t'
2974                     || (c == '\n' && (options & AS_STR_NEWLINES)))
2975                         os.put(c);
2976                 else if (c == META_INSET && (options & AS_STR_INSETS)) {
2977                         getInset(i)->toString(os);
2978                         if (getInset(i)->asInsetMath())
2979                                 os << " ";
2980                 }
2981         }
2982
2983         return os.str();
2984 }
2985
2986
2987 void Paragraph::forToc(docstring & os, size_t maxlen) const
2988 {
2989         if (!d->params_.labelString().empty())
2990                 os += d->params_.labelString() + ' ';
2991         for (pos_type i = 0; i < size() && os.length() < maxlen; ++i) {
2992                 if (isDeleted(i))
2993                         continue;
2994                 char_type const c = d->text_[i];
2995                 if (isPrintable(c))
2996                         os += c;
2997                 else if (c == '\t' || c == '\n')
2998                         os += ' ';
2999                 else if (c == META_INSET)
3000                         getInset(i)->forToc(os, maxlen);
3001         }
3002 }
3003
3004
3005 docstring Paragraph::stringify(pos_type beg, pos_type end, int options, OutputParams & runparams) const
3006 {
3007         odocstringstream os;
3008
3009         if (beg == 0
3010                 && options & AS_STR_LABEL
3011                 && !d->params_.labelString().empty())
3012                 os << d->params_.labelString() << ' ';
3013
3014         for (pos_type i = beg; i < end; ++i) {
3015                 char_type const c = d->text_[i];
3016                 if (isPrintable(c) || c == '\t'
3017                     || (c == '\n' && (options & AS_STR_NEWLINES)))
3018                         os.put(c);
3019                 else if (c == META_INSET && (options & AS_STR_INSETS)) {
3020                         getInset(i)->plaintext(os, runparams);
3021                 }
3022         }
3023
3024         return os.str();
3025 }
3026
3027
3028 void Paragraph::setInsetOwner(Inset const * inset)
3029 {
3030         d->inset_owner_ = inset;
3031 }
3032
3033
3034 int Paragraph::id() const
3035 {
3036         return d->id_;
3037 }
3038
3039
3040 void Paragraph::setId(int id)
3041 {
3042         d->id_ = id;
3043 }
3044
3045
3046 Layout const & Paragraph::layout() const
3047 {
3048         return *d->layout_;
3049 }
3050
3051
3052 void Paragraph::setLayout(Layout const & layout)
3053 {
3054         d->layout_ = &layout;
3055 }
3056
3057
3058 void Paragraph::setDefaultLayout(DocumentClass const & tc)
3059 {
3060         setLayout(tc.defaultLayout());
3061 }
3062
3063
3064 void Paragraph::setPlainLayout(DocumentClass const & tc)
3065 {
3066         setLayout(tc.plainLayout());
3067 }
3068
3069
3070 void Paragraph::setPlainOrDefaultLayout(DocumentClass const & tclass)
3071 {
3072         if (usePlainLayout())
3073                 setPlainLayout(tclass);
3074         else
3075                 setDefaultLayout(tclass);
3076 }
3077
3078
3079 Inset const & Paragraph::inInset() const
3080 {
3081         LASSERT(d->inset_owner_, throw ExceptionMessage(BufferException,
3082                 _("Memory problem"), _("Paragraph not properly initialized")));
3083         return *d->inset_owner_;
3084 }
3085
3086
3087 ParagraphParameters & Paragraph::params()
3088 {
3089         return d->params_;
3090 }
3091
3092
3093 ParagraphParameters const & Paragraph::params() const
3094 {
3095         return d->params_;
3096 }
3097
3098
3099 bool Paragraph::isFreeSpacing() const
3100 {
3101         if (d->layout_->free_spacing)
3102                 return true;
3103         return d->inset_owner_ && d->inset_owner_->isFreeSpacing();
3104 }
3105
3106
3107 bool Paragraph::allowEmpty() const
3108 {
3109         if (d->layout_->keepempty)
3110                 return true;
3111         return d->inset_owner_ && d->inset_owner_->allowEmpty();
3112 }
3113
3114
3115 char_type Paragraph::transformChar(char_type c, pos_type pos) const
3116 {
3117         if (!Encodings::isArabicChar(c))
3118                 return c;
3119
3120         char_type prev_char = ' ';
3121         char_type next_char = ' ';
3122
3123         for (pos_type i = pos - 1; i >= 0; --i) {
3124                 char_type const par_char = d->text_[i];
3125                 if (!Encodings::isArabicComposeChar(par_char)) {
3126                         prev_char = par_char;
3127                         break;
3128                 }
3129         }
3130
3131         for (pos_type i = pos + 1, end = size(); i < end; ++i) {
3132                 char_type const par_char = d->text_[i];
3133                 if (!Encodings::isArabicComposeChar(par_char)) {
3134                         next_char = par_char;
3135                         break;
3136                 }
3137         }
3138
3139         if (Encodings::isArabicChar(next_char)) {
3140                 if (Encodings::isArabicChar(prev_char) &&
3141                         !Encodings::isArabicSpecialChar(prev_char))
3142                         return Encodings::transformChar(c, Encodings::FORM_MEDIAL);
3143                 else
3144                         return Encodings::transformChar(c, Encodings::FORM_INITIAL);
3145         } else {
3146                 if (Encodings::isArabicChar(prev_char) &&
3147                         !Encodings::isArabicSpecialChar(prev_char))
3148                         return Encodings::transformChar(c, Encodings::FORM_FINAL);
3149                 else
3150                         return Encodings::transformChar(c, Encodings::FORM_ISOLATED);
3151         }
3152 }
3153
3154
3155 int Paragraph::checkBiblio(Buffer const & buffer)
3156 {
3157         // FIXME From JS:
3158         // This is getting more and more a mess. ...We really should clean
3159         // up this bibitem issue for 1.6.
3160
3161         // Add bibitem insets if necessary
3162         if (d->layout_->labeltype != LABEL_BIBLIO)
3163                 return 0;
3164
3165         bool hasbibitem = !d->insetlist_.empty()
3166                 // Insist on it being in pos 0
3167                 && d->text_[0] == META_INSET
3168                 && d->insetlist_.begin()->inset->lyxCode() == BIBITEM_CODE;
3169
3170         bool track_changes = buffer.params().trackChanges;
3171
3172         docstring oldkey;
3173         docstring oldlabel;
3174
3175         // remove a bibitem in pos != 0
3176         // restore it later in pos 0 if necessary
3177         // (e.g. if a user inserts contents _before_ the item)
3178         // we're assuming there's only one of these, which there
3179         // should be.
3180         int erasedInsetPosition = -1;
3181         InsetList::iterator it = d->insetlist_.begin();
3182         InsetList::iterator end = d->insetlist_.end();
3183         for (; it != end; ++it)
3184                 if (it->inset->lyxCode() == BIBITEM_CODE
3185                       && it->pos > 0) {
3186                         InsetCommand * olditem = it->inset->asInsetCommand();
3187                         oldkey = olditem->getParam("key");
3188                         oldlabel = olditem->getParam("label");
3189                         erasedInsetPosition = it->pos;
3190                         eraseChar(erasedInsetPosition, track_changes);
3191                         break;
3192         }
3193
3194         // There was an InsetBibitem at the beginning, and we didn't
3195         // have to erase one.
3196         if (hasbibitem && erasedInsetPosition < 0)
3197                         return 0;
3198
3199         // There was an InsetBibitem at the beginning and we did have to
3200         // erase one. So we give its properties to the beginning inset.
3201         if (hasbibitem) {
3202                 InsetCommand * inset = d->insetlist_.begin()->inset->asInsetCommand();
3203                 if (!oldkey.empty())
3204                         inset->setParam("key", oldkey);
3205                 inset->setParam("label", oldlabel);
3206                 return -erasedInsetPosition;
3207         }
3208
3209         // There was no inset at the beginning, so we need to create one with
3210         // the key and label of the one we erased.
3211         InsetBibitem * inset =
3212                 new InsetBibitem(const_cast<Buffer *>(&buffer), InsetCommandParams(BIBITEM_CODE));
3213         // restore values of previously deleted item in this par.
3214         if (!oldkey.empty())
3215                 inset->setParam("key", oldkey);
3216         inset->setParam("label", oldlabel);
3217         insertInset(0, inset,
3218                     Change(track_changes ? Change::INSERTED : Change::UNCHANGED));
3219
3220         return 1;
3221 }
3222
3223
3224 void Paragraph::checkAuthors(AuthorList const & authorList)
3225 {
3226         d->changes_.checkAuthors(authorList);
3227 }
3228
3229
3230 bool Paragraph::isChanged(pos_type pos) const
3231 {
3232         return lookupChange(pos).changed();
3233 }
3234
3235
3236 bool Paragraph::isInserted(pos_type pos) const
3237 {
3238         return lookupChange(pos).inserted();
3239 }
3240
3241
3242 bool Paragraph::isDeleted(pos_type pos) const
3243 {
3244         return lookupChange(pos).deleted();
3245 }
3246
3247
3248 InsetList const & Paragraph::insetList() const
3249 {
3250         return d->insetlist_;
3251 }
3252
3253
3254 void Paragraph::setBuffer(Buffer & b)
3255 {
3256         d->insetlist_.setBuffer(b);
3257 }
3258
3259
3260 Inset * Paragraph::releaseInset(pos_type pos)
3261 {
3262         Inset * inset = d->insetlist_.release(pos);
3263         /// does not honour change tracking!
3264         eraseChar(pos, false);
3265         return inset;
3266 }
3267
3268
3269 Inset * Paragraph::getInset(pos_type pos)
3270 {
3271         return (pos < pos_type(d->text_.size()) && d->text_[pos] == META_INSET)
3272                  ? d->insetlist_.get(pos) : 0;
3273 }
3274
3275
3276 Inset const * Paragraph::getInset(pos_type pos) const
3277 {
3278         return (pos < pos_type(d->text_.size()) && d->text_[pos] == META_INSET)
3279                  ? d->insetlist_.get(pos) : 0;
3280 }
3281
3282
3283 void Paragraph::changeCase(BufferParams const & bparams, pos_type pos,
3284                 pos_type & right, TextCase action)
3285 {
3286         // process sequences of modified characters; in change
3287         // tracking mode, this approach results in much better
3288         // usability than changing case on a char-by-char basis
3289         docstring changes;
3290
3291         bool const trackChanges = bparams.trackChanges;
3292
3293         bool capitalize = true;
3294
3295         for (; pos < right; ++pos) {
3296                 char_type oldChar = d->text_[pos];
3297                 char_type newChar = oldChar;
3298
3299                 // ignore insets and don't play with deleted text!
3300                 if (oldChar != META_INSET && !isDeleted(pos)) {
3301                         switch (action) {
3302                                 case text_lowercase:
3303                                         newChar = lowercase(oldChar);
3304                                         break;
3305                                 case text_capitalization:
3306                                         if (capitalize) {
3307                                                 newChar = uppercase(oldChar);
3308                                                 capitalize = false;
3309                                         }
3310                                         break;
3311                                 case text_uppercase:
3312                                         newChar = uppercase(oldChar);
3313                                         break;
3314                         }
3315                 }
3316
3317                 if (isWordSeparator(pos) || isDeleted(pos)) {
3318                         // permit capitalization again
3319                         capitalize = true;
3320                 }
3321
3322                 if (oldChar != newChar) {
3323                         changes += newChar;
3324                         if (pos != right - 1)
3325                                 continue;
3326                         // step behind the changing area
3327                         pos++;
3328                 }
3329
3330                 int erasePos = pos - changes.size();
3331                 for (size_t i = 0; i < changes.size(); i++) {
3332                         insertChar(pos, changes[i],
3333                                    getFontSettings(bparams,
3334                                                    erasePos),
3335                                    trackChanges);
3336                         if (!eraseChar(erasePos, trackChanges)) {
3337                                 ++erasePos;
3338                                 ++pos; // advance
3339                                 ++right; // expand selection
3340                         }
3341                 }
3342                 changes.clear();
3343         }
3344 }
3345
3346
3347 int Paragraph::find(docstring const & str, bool cs, bool mw,
3348                 pos_type start_pos, bool del) const
3349 {
3350         pos_type pos = start_pos;
3351         int const strsize = str.length();
3352         int i = 0;
3353         pos_type const parsize = d->text_.size();
3354         for (i = 0; i < strsize && pos < parsize; ++i, ++pos) {
3355                 // Ignore "invisible" letters such as ligature breaks
3356                 // and hyphenation chars while searching
3357                 while (pos < parsize - 1 && isInset(pos)) {
3358                         odocstringstream os;
3359                         getInset(pos)->toString(os);
3360                         if (!getInset(pos)->isLetter() || !os.str().empty())
3361                                 break;
3362                         pos++;
3363                 }
3364                 if (cs && str[i] != d->text_[pos])
3365                         break;
3366                 if (!cs && uppercase(str[i]) != uppercase(d->text_[pos]))
3367                         break;
3368                 if (!del && isDeleted(pos))
3369                         break;
3370         }
3371
3372         if (i != strsize)
3373                 return 0;
3374
3375         // if necessary, check whether string matches word
3376         if (mw) {
3377                 if (start_pos > 0 && !isWordSeparator(start_pos - 1))
3378                         return 0;
3379                 if (pos < parsize
3380                         && !isWordSeparator(pos))
3381                         return 0;
3382         }
3383
3384         return pos - start_pos;
3385 }
3386
3387
3388 char_type Paragraph::getChar(pos_type pos) const
3389 {
3390         return d->text_[pos];
3391 }
3392
3393
3394 pos_type Paragraph::size() const
3395 {
3396         return d->text_.size();
3397 }
3398
3399
3400 bool Paragraph::empty() const
3401 {
3402         return d->text_.empty();
3403 }
3404
3405
3406 bool Paragraph::isInset(pos_type pos) const
3407 {
3408         return d->text_[pos] == META_INSET;
3409 }
3410
3411
3412 bool Paragraph::isSeparator(pos_type pos) const
3413 {
3414         //FIXME: Are we sure this can be the only separator?
3415         return d->text_[pos] == ' ';
3416 }
3417
3418
3419 void Paragraph::deregisterWords()
3420 {
3421         Private::LangWordsMap::const_iterator itl = d->words_.begin();
3422         Private::LangWordsMap::const_iterator ite = d->words_.end();
3423         for (; itl != ite; ++itl) {
3424                 WordList * wl = theWordList(itl->first);
3425                 Private::Words::const_iterator it = (itl->second).begin();
3426                 Private::Words::const_iterator et = (itl->second).end();
3427                 for (; it != et; ++it)
3428                         wl->remove(*it);
3429         }
3430         d->words_.clear();
3431 }
3432
3433
3434 void Paragraph::locateWord(pos_type & from, pos_type & to,
3435         word_location const loc) const
3436 {
3437         switch (loc) {
3438         case WHOLE_WORD_STRICT:
3439                 if (from == 0 || from == size()
3440                     || isWordSeparator(from)
3441                     || isWordSeparator(from - 1)) {
3442                         to = from;
3443                         return;
3444                 }
3445                 // no break here, we go to the next
3446
3447         case WHOLE_WORD:
3448                 // If we are already at the beginning of a word, do nothing
3449                 if (!from || isWordSeparator(from - 1))
3450                         break;
3451                 // no break here, we go to the next
3452
3453         case PREVIOUS_WORD:
3454                 // always move the cursor to the beginning of previous word
3455                 while (from && !isWordSeparator(from - 1))
3456                         --from;
3457                 break;
3458         case NEXT_WORD:
3459                 LYXERR0("Paragraph::locateWord: NEXT_WORD not implemented yet");
3460                 break;
3461         case PARTIAL_WORD:
3462                 // no need to move the 'from' cursor
3463                 break;
3464         }
3465         to = from;
3466         while (to < size() && !isWordSeparator(to))
3467                 ++to;
3468 }
3469
3470
3471 void Paragraph::collectWords()
3472 {
3473         // This is the value that needs to be exposed in the preferences
3474         // to resolve bug #6760.
3475         static int minlength = 6;
3476         pos_type n = size();
3477         for (pos_type pos = 0; pos < n; ++pos) {
3478                 if (isWordSeparator(pos))
3479                         continue;
3480                 pos_type from = pos;
3481                 locateWord(from, pos, WHOLE_WORD);
3482                 if (pos - from >= minlength) {
3483                         docstring word = asString(from, pos, AS_STR_NONE);
3484                         FontList::const_iterator cit = d->fontlist_.fontIterator(pos);
3485                         if (cit == d->fontlist_.end())
3486                                 return;
3487                         Language const * lang = cit->font().language();
3488                         d->words_[*lang].insert(word);
3489                 }
3490         }
3491 }
3492
3493
3494 void Paragraph::registerWords()
3495 {
3496         Private::LangWordsMap::const_iterator itl = d->words_.begin();
3497         Private::LangWordsMap::const_iterator ite = d->words_.end();
3498         for (; itl != ite; ++itl) {
3499                 WordList * wl = theWordList(itl->first);
3500                 Private::Words::const_iterator it = (itl->second).begin();
3501                 Private::Words::const_iterator et = (itl->second).end();
3502                 for (; it != et; ++it)
3503                         wl->insert(*it);
3504         }
3505 }
3506
3507
3508 void Paragraph::updateWords()
3509 {
3510         deregisterWords();
3511         collectWords();
3512         registerWords();
3513 }
3514
3515
3516 void Paragraph::Private::appendSkipPosition(SkipPositions & skips, pos_type const pos) const
3517 {
3518         SkipPositionsIterator begin = skips.begin();
3519         SkipPositions::iterator end = skips.end();
3520         if (pos > 0 && begin < end) {
3521                 --end;
3522                 if (end->last == pos - 1) {
3523                         end->last = pos;
3524                         return;
3525                 }
3526         }
3527         skips.insert(end, FontSpan(pos, pos));
3528 }
3529
3530
3531 Language * Paragraph::Private::locateSpellRange(
3532         pos_type & from, pos_type & to,
3533         SkipPositions & skips) const
3534 {
3535         // skip leading white space
3536         while (from < to && owner_->isWordSeparator(from))
3537                 ++from;
3538         // don't check empty range
3539         if (from >= to)
3540                 return 0;
3541         // get current language
3542         Language * lang = getSpellLanguage(from);
3543         pos_type last = from;
3544         bool samelang = true;
3545         bool sameinset = true;
3546         while (last < to && samelang && sameinset) {
3547                 // hop to end of word
3548                 while (last < to && !owner_->isWordSeparator(last)) {
3549                         if (owner_->getInset(last)) {
3550                                 appendSkipPosition(skips, last);
3551                         } else if (owner_->isDeleted(last)) {
3552                                 appendSkipPosition(skips, last);
3553                         }
3554                         ++last;
3555                 }
3556                 // hop to next word while checking for insets
3557                 while (sameinset && last < to && owner_->isWordSeparator(last)) {
3558                         if (Inset const * inset = owner_->getInset(last))
3559                                 sameinset = inset->isChar() && inset->isLetter();
3560                         if (sameinset && owner_->isDeleted(last)) {
3561                                 appendSkipPosition(skips, last);
3562                         }
3563                         if (sameinset)
3564                                 last++;
3565                 }
3566                 if (sameinset && last < to) {
3567                         // now check for language change
3568                         samelang = lang == getSpellLanguage(last);
3569                 }
3570         }
3571         // if language change detected backstep is needed
3572         if (!samelang)
3573                 --last;
3574         to = last;
3575         return lang;
3576 }
3577
3578
3579 Language * Paragraph::Private::getSpellLanguage(pos_type const from) const
3580 {
3581         Language * lang =
3582                 const_cast<Language *>(owner_->getFontSettings(
3583                         inset_owner_->buffer().params(), from).language());
3584         if (lang == inset_owner_->buffer().params().language
3585                 && !lyxrc.spellchecker_alt_lang.empty()) {
3586                 string lang_code;
3587                 string const lang_variety =
3588                         split(lyxrc.spellchecker_alt_lang, lang_code, '-');
3589                 lang->setCode(lang_code);
3590                 lang->setVariety(lang_variety);
3591         }
3592         return lang;
3593 }
3594
3595
3596 void Paragraph::requestSpellCheck(pos_type pos)
3597 {
3598         d->requestSpellCheck(pos == -1 ? size() : pos);
3599 }
3600
3601
3602 bool Paragraph::needsSpellCheck() const
3603 {
3604         SpellChecker::ChangeNumber speller_change_number = 0;
3605         if (theSpellChecker())
3606                 speller_change_number = theSpellChecker()->changeNumber();
3607         if (speller_change_number > d->speller_state_.currentChangeNumber()) {
3608                 d->speller_state_.needsCompleteRefresh(speller_change_number);
3609         }
3610         return d->needsSpellCheck();
3611 }
3612
3613
3614 bool Paragraph::Private::ignoreWord(docstring const & word) const
3615 {
3616         // Ignore words with digits
3617         // FIXME: make this customizable
3618         // (note that some checkers ignore words with digits by default)
3619         docstring::const_iterator cit = word.begin();
3620         docstring::const_iterator const end = word.end();
3621         for (; cit != end; ++cit) {
3622                 if (isNumber((*cit)))
3623                         return true;
3624         }
3625         return false;
3626 }
3627
3628
3629 SpellChecker::Result Paragraph::spellCheck(pos_type & from, pos_type & to,
3630         WordLangTuple & wl, docstring_list & suggestions,
3631         bool do_suggestion, bool check_learned) const
3632 {
3633         SpellChecker::Result result = SpellChecker::WORD_OK;
3634         SpellChecker * speller = theSpellChecker();
3635         if (!speller)
3636                 return result;
3637
3638         if (!d->layout_->spellcheck || !inInset().allowSpellCheck())
3639                 return result;
3640
3641         locateWord(from, to, WHOLE_WORD);
3642         if (from == to || from >= size())
3643                 return result;
3644
3645         docstring word = asString(from, to, AS_STR_INSETS | AS_STR_SKIPDELETE);
3646         Language * lang = d->getSpellLanguage(from);
3647
3648         wl = WordLangTuple(word, lang);
3649
3650         if (!word.size())
3651                 return result;
3652
3653         if (needsSpellCheck() || check_learned) {
3654                 pos_type end = to;
3655                 if (!d->ignoreWord(word)) {
3656                         bool const trailing_dot = to < size() && d->text_[to] == '.';
3657                         result = speller->check(wl);
3658                         if (SpellChecker::misspelled(result) && trailing_dot) {
3659                                 wl = WordLangTuple(word.append(from_ascii(".")), lang);
3660                                 result = speller->check(wl);
3661                                 if (!SpellChecker::misspelled(result)) {
3662                                         LYXERR(Debug::GUI, "misspelled word is correct with dot: \"" <<
3663                                            word << "\" [" <<
3664                                            from << ".." << to << "]");
3665                                 } else {
3666                                         // spell check with dot appended failed too
3667                                         // restore original word/lang value
3668                                         word = asString(from, to, AS_STR_INSETS | AS_STR_SKIPDELETE);
3669                                         wl = WordLangTuple(word, lang);
3670                                 }
3671                         }
3672                 }
3673                 if (!SpellChecker::misspelled(result)) {
3674                         // area up to the begin of the next word is not misspelled
3675                         while (end < size() && isWordSeparator(end))
3676                                 ++end;
3677                 }
3678                 d->setMisspelled(from, end, result);
3679         } else {
3680                 result = d->speller_state_.getState(from);
3681         }
3682
3683         if (do_suggestion)
3684                 suggestions.clear();
3685
3686         if (SpellChecker::misspelled(result)) {
3687                 LYXERR(Debug::GUI, "misspelled word: \"" <<
3688                            word << "\" [" <<
3689                            from << ".." << to << "]");
3690                 if (do_suggestion)
3691                         speller->suggest(wl, suggestions);
3692         }
3693         return result;
3694 }
3695
3696
3697 void Paragraph::Private::markMisspelledWords(
3698         pos_type const & first, pos_type const & last,
3699         SpellChecker::Result result,
3700         docstring const & word,
3701         SkipPositions const & skips)
3702 {
3703         if (!SpellChecker::misspelled(result)) {
3704                 setMisspelled(first, last, SpellChecker::WORD_OK);
3705                 return;
3706         }
3707         int snext = first;
3708         SpellChecker * speller = theSpellChecker();
3709         // locate and enumerate the error positions
3710         int nerrors = speller->numMisspelledWords();
3711         int numskipped = 0;
3712         SkipPositionsIterator it = skips.begin();
3713         SkipPositionsIterator et = skips.end();
3714         for (int index = 0; index < nerrors; ++index) {
3715                 int wstart;
3716                 int wlen = 0;
3717                 speller->misspelledWord(index, wstart, wlen);
3718                 /// should not happen if speller supports range checks
3719                 if (!wlen) continue;
3720                 docstring const misspelled = word.substr(wstart, wlen);
3721                 wstart += first + numskipped;
3722                 if (snext < wstart) {
3723                         /// mark the range of correct spelling
3724                         numskipped += countSkips(it, et, wstart);
3725                         setMisspelled(snext,
3726                                 wstart - 1, SpellChecker::WORD_OK);
3727                 }
3728                 snext = wstart + wlen;
3729                 numskipped += countSkips(it, et, snext);
3730                 /// mark the range of misspelling
3731                 setMisspelled(wstart, snext, result);
3732                 LYXERR(Debug::GUI, "misspelled word: \"" <<
3733                            misspelled << "\" [" <<
3734                            wstart << ".." << (snext-1) << "]");
3735                 ++snext;
3736         }
3737         if (snext <= last) {
3738                 /// mark the range of correct spelling at end
3739                 setMisspelled(snext, last, SpellChecker::WORD_OK);
3740         }
3741 }
3742
3743
3744 void Paragraph::spellCheck() const
3745 {
3746         SpellChecker * speller = theSpellChecker();
3747         if (!speller || !size() ||!needsSpellCheck())
3748                 return;
3749         pos_type start;
3750         pos_type endpos;
3751         d->rangeOfSpellCheck(start, endpos);
3752         if (speller->canCheckParagraph()) {
3753                 // loop until we leave the range
3754                 for (pos_type first = start; first < endpos; ) {
3755                         pos_type last = endpos;
3756                         Private::SkipPositions skips;
3757                         Language * lang = d->locateSpellRange(first, last, skips);
3758                         if (first >= endpos)
3759                                 break;
3760                         // start the spell checker on the unit of meaning
3761                         docstring word = asString(first, last, AS_STR_INSETS + AS_STR_SKIPDELETE);
3762                         WordLangTuple wl = WordLangTuple(word, lang);
3763                         SpellChecker::Result result = word.size() ?
3764                                 speller->check(wl) : SpellChecker::WORD_OK;
3765                         d->markMisspelledWords(first, last, result, word, skips);
3766                         first = ++last;
3767                 }
3768         } else {
3769                 static docstring_list suggestions;
3770                 pos_type to = endpos;
3771                 while (start < endpos) {
3772                         WordLangTuple wl;
3773                         spellCheck(start, to, wl, suggestions, false);
3774                         start = to + 1;
3775                 }
3776         }
3777         d->readySpellCheck();
3778 }
3779
3780
3781 bool Paragraph::isMisspelled(pos_type pos, bool check_boundary) const
3782 {
3783         bool result = SpellChecker::misspelled(d->speller_state_.getState(pos));
3784         if (result || pos <= 0 || pos >= size())
3785                 return result;
3786         if (check_boundary && isWordSeparator(pos))
3787                 result = SpellChecker::misspelled(d->speller_state_.getState(pos - 1));
3788         return result;
3789 }
3790
3791
3792 string Paragraph::magicLabel() const
3793 {
3794         stringstream ss;
3795         ss << "magicparlabel-" << id();
3796         return ss.str();
3797 }
3798
3799
3800 } // namespace lyx