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 "WordLangTuple.h"
  47 #include "WordList.h"
  48
  49 #include "frontends/alert.h"
  50
  51 #include "insets/InsetBibitem.h"
  52 #include "insets/InsetLabel.h"
  53 #include "insets/InsetSpecialChar.h"
  54
  55 #include "support/debug.h"
  56 #include "support/docstring_list.h"
  57 #include "support/ExceptionMessage.h"
  58 #include "support/gettext.h"
  59 #include "support/lassert.h"
  60 #include "support/lstrings.h"
  61 #include "support/textutils.h"
  62
  63 #include <sstream>
  64 #include <vector>
  65
  66 using namespace std;
  67 using namespace lyx::support;
  68
  69 namespace lyx {
  70
  71 namespace {
  72 /// Inset identifier (above 0x10ffff, for ucs-4)
  73 char_type const META_INSET = 0x200001;
  74 }
  75
  76
  77 /////////////////////////////////////////////////////////////////////
  78 //
  79 // SpellResultRange
  80 //
  81 /////////////////////////////////////////////////////////////////////
  82
  83 class SpellResultRange {
  84 public:
  85         SpellResultRange(FontSpan range, SpellChecker::Result result)
  86         : range_(range), result_(result)
  87         {}
  88         ///
  89         FontSpan const & range() const { return range_; }
  90         ///
  91         void range(FontSpan const & r) { range_ = r; }
  92         ///
  93         SpellChecker::Result result() const { return result_; }
  94         ///
  95         void result(SpellChecker::Result r) { result_ = r; }
  96         ///
  97         bool inside(pos_type pos) const { return range_.inside(pos); }
  98         ///
  99         bool covered(FontSpan const & r) const
 100         {
 101                 // 1. first of new range inside current range or
 102                 // 2. last of new range inside current range or
 103                 // 3. first of current range inside new range or
 104                 // 4. last of current range inside new range
 105                 return range_.inside(r.first) || range_.inside(r.last) ||
 106                         r.inside(range_.first) || r.inside(range_.last);
 107         }
 108         ///
 109         void shift(pos_type pos, int offset)
 110         {
 111                 if (range_.first > pos) {
 112                         range_.first += offset;
 113                         range_.last += offset;
 114                 } else if (range_.last >= pos) {
 115                         range_.last += offset;
 116                 }
 117         }
 118 private:
 119         FontSpan range_ ;
 120         SpellChecker::Result result_ ;
 121 };
 122
 123
 124 /////////////////////////////////////////////////////////////////////
 125 //
 126 // SpellCheckerState
 127 //
 128 /////////////////////////////////////////////////////////////////////
 129
 130 class SpellCheckerState {
 131 public:
 132         SpellCheckerState() {
 133                 needs_refresh_ = true;
 134                 current_change_number_ = 0;
 135         }
 136
 137         void setRange(FontSpan const fp, SpellChecker::Result state)
 138         {
 139                 Ranges result;
 140                 RangesIterator et = ranges_.end();
 141                 RangesIterator it = ranges_.begin();
 142                 for (; it != et; ++it) {
 143                         if (!it->covered(fp))
 144                                 result.push_back(SpellResultRange(it->range(), it->result()));
 145                         else if (state == SpellChecker::WORD_OK) {
 146                                 // trim or split the current misspelled range
 147                                 // store misspelled ranges only
 148                                 FontSpan range = it->range();
 149                                 if (fp.first > range.first) {
 150                                         // misspelled area in front of WORD_OK
 151                                         range.last = fp.first - 1;
 152                                         result.push_back(SpellResultRange(range, it->result()));
 153                                         range = it->range();
 154                                 }
 155                                 if (fp.last < range.last) {
 156                                         // misspelled area after WORD_OK range
 157                                         range.first = fp.last + 1;
 158                                         result.push_back(SpellResultRange(range, it->result()));
 159                                 }
 160                         }
 161                 }
 162                 ranges_ = result;
 163                 if (state != SpellChecker::WORD_OK)
 164                         ranges_.push_back(SpellResultRange(fp, state));
 165         }
 166
 167         void increasePosAfterPos(pos_type pos)
 168         {
 169                 correctRangesAfterPos(pos, 1);
 170                 needsRefresh(pos);
 171         }
 172
 173         void decreasePosAfterPos(pos_type pos)
 174         {
 175                 correctRangesAfterPos(pos, -1);
 176                 needsRefresh(pos);
 177         }
 178
 179         void refreshLast(pos_type pos)
 180         {
 181                 if (pos < refresh_.last)
 182                         refresh_.last = pos;
 183         }
 184
 185         SpellChecker::Result getState(pos_type pos) const
 186         {
 187                 SpellChecker::Result result = SpellChecker::WORD_OK;
 188                 RangesIterator et = ranges_.end();
 189                 RangesIterator it = ranges_.begin();
 190                 for (; it != et; ++it) {
 191                         if(it->inside(pos)) {
 192                                 return it->result();
 193                         }
 194                 }
 195                 return result;
 196         }
 197
 198         FontSpan const & getRange(pos_type pos) const
 199         {
 200                 /// empty span to indicate mismatch
 201                 static FontSpan empty_;
 202                 RangesIterator et = ranges_.end();
 203                 RangesIterator it = ranges_.begin();
 204                 for (; it != et; ++it) {
 205                         if(it->inside(pos)) {
 206                                 return it->range();
 207                         }
 208                 }
 209                 return empty_;
 210         }
 211
 212         bool needsRefresh() const {
 213                 return needs_refresh_;
 214         }
 215
 216         SpellChecker::ChangeNumber currentChangeNumber() const {
 217                 return current_change_number_;
 218         }
 219
 220         void refreshRange(pos_type & first, pos_type & last) const {
 221                 first = refresh_.first;
 222                 last = refresh_.last;
 223         }
 224
 225         void needsRefresh(pos_type pos) {
 226                 if (needs_refresh_ && pos != -1) {
 227                         if (pos < refresh_.first)
 228                                 refresh_.first = pos;
 229                         if (pos > refresh_.last)
 230                                 refresh_.last = pos;
 231                 } else if (pos != -1) {
 232                         // init request check for neighbour positions too
 233                         refresh_.first = pos > 0 ? pos - 1 : 0;
 234                         // no need for special end of paragraph check
 235                         refresh_.last = pos + 1;
 236                 }
 237                 needs_refresh_ = pos != -1;
 238         }
 239
 240         void needsCompleteRefresh(SpellChecker::ChangeNumber change_number) {
 241                 needs_refresh_ = true;
 242                 refresh_.first = 0;
 243                 refresh_.last = -1;
 244                 current_change_number_ = change_number;
 245         }
 246
 247 private:
 248         typedef vector<SpellResultRange> Ranges;
 249         typedef Ranges::const_iterator RangesIterator;
 250         Ranges ranges_;
 251         /// the area of the paragraph with pending spell check
 252         FontSpan refresh_;
 253         bool needs_refresh_;
 254         /// spell state cache version number
 255         SpellChecker::ChangeNumber current_change_number_;
 256
 257
 258         void correctRangesAfterPos(pos_type pos, int offset)
 259         {
 260                 RangesIterator et = ranges_.end();
 261                 Ranges::iterator it = ranges_.begin();
 262                 for (; it != et; ++it) {
 263                         it->shift(pos, offset);
 264                 }
 265         }
 266
 267 };
 268
 269 /////////////////////////////////////////////////////////////////////
 270 //
 271 // Paragraph::Private
 272 //
 273 /////////////////////////////////////////////////////////////////////
 274
 275 class Paragraph::Private
 276 {
 277 public:
 278         ///
 279         Private(Paragraph * owner, Layout const & layout);
 280         /// "Copy constructor"
 281         Private(Private const &, Paragraph * owner);
 282         /// Copy constructor from \p beg  to \p end
 283         Private(Private const &, Paragraph * owner, pos_type beg, pos_type end);
 284
 285         ///
 286         void insertChar(pos_type pos, char_type c, Change const & change);
 287
 288         /// Output the surrogate pair formed by \p c and \p next to \p os.
 289         /// \return the number of characters written.
 290         int latexSurrogatePair(otexstream & os, char_type c, char_type next,
 291                                OutputParams const &);
 292
 293         /// Output a space in appropriate formatting (or a surrogate pair
 294         /// if the next character is a combining character).
 295         /// \return whether a surrogate pair was output.
 296         bool simpleTeXBlanks(OutputParams const &,
 297                              otexstream &,
 298                              pos_type i,
 299                              unsigned int & column,
 300                              Font const & font,
 301                              Layout const & style);
 302
 303         /// Output consecutive unicode chars, belonging to the same script as
 304         /// specified by the latex macro \p ltx, to \p os starting from \p i.
 305         /// \return the number of characters written.
 306         int writeScriptChars(otexstream & os, docstring const & ltx,
 307                            Change const &, Encoding const &, pos_type & i);
 308
 309         /// This could go to ParagraphParameters if we want to.
 310         int startTeXParParams(BufferParams const &, otexstream &,
 311                               OutputParams const &) const;
 312
 313         /// This could go to ParagraphParameters if we want to.
 314         bool endTeXParParams(BufferParams const &, otexstream &,
 315                              OutputParams const &) const;
 316
 317         ///
 318         void latexInset(BufferParams const &,
 319                                    otexstream &,
 320                                    OutputParams &,
 321                                    Font & running_font,
 322                                    Font & basefont,
 323                                    Font const & outerfont,
 324                                    bool & open_font,
 325                                    Change & running_change,
 326                                    Layout const & style,
 327                                    pos_type & i,
 328                                    unsigned int & column);
 329
 330         ///
 331         void latexSpecialChar(
 332                                    otexstream & os,
 333                                    OutputParams const & runparams,
 334                                    Font const & running_font,
 335                                    Change const & running_change,
 336                                    Layout const & style,
 337                                    pos_type & i,
 338                                    pos_type end_pos,
 339                                    unsigned int & column);
 340
 341         ///
 342         bool latexSpecialT1(
 343                 char_type const c,
 344                 otexstream & os,
 345                 pos_type i,
 346                 unsigned int & column);
 347         ///
 348         bool latexSpecialTypewriter(
 349                 char_type const c,
 350                 otexstream & os,
 351                 pos_type i,
 352                 unsigned int & column);
 353         ///
 354         bool latexSpecialPhrase(
 355                 otexstream & os,
 356                 pos_type & i,
 357                 pos_type end_pos,
 358                 unsigned int & column,
 359                 OutputParams const & runparams);
 360
 361         ///
 362         void validate(LaTeXFeatures & features) const;
 363
 364         /// Checks if the paragraph contains only text and no inset or font change.
 365         bool onlyText(Buffer const & buf, Font const & outerfont,
 366                       pos_type initial) const;
 367
 368         /// match a string against a particular point in the paragraph
 369         bool isTextAt(string const & str, pos_type pos) const;
 370
 371         /// a vector of speller skip positions
 372         typedef vector<FontSpan> SkipPositions;
 373         typedef SkipPositions::const_iterator SkipPositionsIterator;
 374
 375         void appendSkipPosition(SkipPositions & skips, pos_type const pos) const;
 376
 377         Language * getSpellLanguage(pos_type const from) const;
 378
 379         Language * locateSpellRange(pos_type & from, pos_type & to,
 380                                                                 SkipPositions & skips) const;
 381
 382         bool hasSpellerChange() const {
 383                 SpellChecker::ChangeNumber speller_change_number = 0;
 384                 if (theSpellChecker())
 385                         speller_change_number = theSpellChecker()->changeNumber();
 386                 return speller_change_number > speller_state_.currentChangeNumber();
 387         }
 388
 389         bool ignoreWord(docstring const & word) const ;
 390
 391         void setMisspelled(pos_type from, pos_type to, SpellChecker::Result state)
 392         {
 393                 pos_type textsize = owner_->size();
 394                 // check for sane arguments
 395                 if (to <= from || from >= textsize)
 396                         return;
 397                 FontSpan fp = FontSpan(from, to - 1);
 398                 speller_state_.setRange(fp, state);
 399         }
 400
 401         void requestSpellCheck(pos_type pos) {
 402                 if (pos == -1)
 403                         speller_state_.needsCompleteRefresh(speller_state_.currentChangeNumber());
 404                 else
 405                         speller_state_.needsRefresh(pos);
 406         }
 407
 408         void readySpellCheck() {
 409                 speller_state_.needsRefresh(-1);
 410         }
 411
 412         bool needsSpellCheck() const
 413         {
 414                 return speller_state_.needsRefresh();
 415         }
 416
 417         void rangeOfSpellCheck(pos_type & first, pos_type & last) const
 418         {
 419                 speller_state_.refreshRange(first, last);
 420                 if (last == -1) {
 421                         last = owner_->size();
 422                         return;
 423                 }
 424                 pos_type endpos = last;
 425                 owner_->locateWord(first, endpos, WHOLE_WORD);
 426                 if (endpos < last) {
 427                         endpos = last;
 428                         owner_->locateWord(last, endpos, WHOLE_WORD);
 429                 }
 430                 last = endpos;
 431         }
 432
 433         int countSkips(SkipPositionsIterator & it, SkipPositionsIterator const et,
 434                             int & start) const
 435         {
 436                 int numskips = 0;
 437                 while (it != et && it->first < start) {
 438                         int skip = it->last - it->first + 1;
 439                         start += skip;
 440                         numskips += skip;
 441                         ++it;
 442                 }
 443                 return numskips;
 444         }
 445
 446         void markMisspelledWords(pos_type const & first, pos_type const & last,
 447                                                          SpellChecker::Result result,
 448                                                          docstring const & word,
 449                                                          SkipPositions const & skips);
 450
 451         InsetCode ownerCode() const
 452         {
 453                 return inset_owner_ ? inset_owner_->lyxCode() : NO_CODE;
 454         }
 455
 456         /// Which Paragraph owns us?
 457         Paragraph * owner_;
 458
 459         /// In which Inset?
 460         Inset const * inset_owner_;
 461
 462         ///
 463         FontList fontlist_;
 464
 465         ///
 466         int id_;
 467
 468         ///
 469         ParagraphParameters params_;
 470
 471         /// for recording and looking up changes
 472         Changes changes_;
 473
 474         ///
 475         InsetList insetlist_;
 476
 477         /// end of label
 478         pos_type begin_of_body_;
 479
 480         typedef docstring TextContainer;
 481         ///
 482         TextContainer text_;
 483
 484         typedef set<docstring> Words;
 485         typedef map<Language, Words> LangWordsMap;
 486         ///
 487         LangWordsMap words_;
 488         ///
 489         Layout const * layout_;
 490         ///
 491         SpellCheckerState speller_state_;
 492 };
 493
 494
 495 namespace {
 496
 497 struct special_phrase {
 498         string phrase;
 499         docstring macro;
 500         bool builtin;
 501 };
 502
 503 special_phrase const special_phrases[] = {
 504         { "LyX", from_ascii("\\LyX{}"), false },
 505         { "TeX", from_ascii("\\TeX{}"), true },
 506         { "LaTeX2e", from_ascii("\\LaTeXe{}"), true },
 507         { "LaTeX", from_ascii("\\LaTeX{}"), true },
 508 };
 509
 510 size_t const phrases_nr = sizeof(special_phrases)/sizeof(special_phrase);
 511
 512 } // namespace anon
 513
 514
 515 Paragraph::Private::Private(Paragraph * owner, Layout const & layout)
 516         : owner_(owner), inset_owner_(0), id_(-1), begin_of_body_(0), layout_(&layout)
 517 {
 518         text_.reserve(100);
 519 }
 520
 521
 522 // Initialization of the counter for the paragraph id's,
 523 //
 524 // FIXME: There should be a more intelligent way to generate and use the
 525 // paragraph ids per buffer instead a global static counter for all InsetText
 526 // in the running program.
 527 static int paragraph_id = -1;
 528
 529 Paragraph::Private::Private(Private const & p, Paragraph * owner)
 530         : owner_(owner), inset_owner_(p.inset_owner_), fontlist_(p.fontlist_),
 531           params_(p.params_), changes_(p.changes_), insetlist_(p.insetlist_),
 532           begin_of_body_(p.begin_of_body_), text_(p.text_), words_(p.words_),
 533           layout_(p.layout_)
 534 {
 535         id_ = ++paragraph_id;
 536         requestSpellCheck(p.text_.size());
 537 }
 538
 539
 540 Paragraph::Private::Private(Private const & p, Paragraph * owner,
 541         pos_type beg, pos_type end)
 542         : owner_(owner), inset_owner_(p.inset_owner_),
 543           params_(p.params_), changes_(p.changes_),
 544           insetlist_(p.insetlist_, beg, end),
 545           begin_of_body_(p.begin_of_body_), words_(p.words_),
 546           layout_(p.layout_)
 547 {
 548         id_ = ++paragraph_id;
 549         if (beg >= pos_type(p.text_.size()))
 550                 return;
 551         text_ = p.text_.substr(beg, end - beg);
 552
 553         FontList::const_iterator fcit = fontlist_.begin();
 554         FontList::const_iterator fend = fontlist_.end();
 555         for (; fcit != fend; ++fcit) {
 556                 if (fcit->pos() < beg)
 557                         continue;
 558                 if (fcit->pos() >= end) {
 559                         // Add last entry in the fontlist_.
 560                         fontlist_.set(text_.size() - 1, fcit->font());
 561                         break;
 562                 }
 563                 // Add a new entry in the fontlist_.
 564                 fontlist_.set(fcit->pos() - beg, fcit->font());
 565         }
 566         requestSpellCheck(p.text_.size());
 567 }
 568
 569
 570 void Paragraph::addChangesToToc(DocIterator const & cdit,
 571         Buffer const & buf) const
 572 {
 573         d->changes_.addToToc(cdit, buf);
 574 }
 575
 576
 577 bool Paragraph::isDeleted(pos_type start, pos_type end) const
 578 {
 579         LASSERT(start >= 0 && start <= size(), /**/);
 580         LASSERT(end > start && end <= size() + 1, /**/);
 581
 582         return d->changes_.isDeleted(start, end);
 583 }
 584
 585
 586 bool Paragraph::isChanged(pos_type start, pos_type end) const
 587 {
 588         LASSERT(start >= 0 && start <= size(), /**/);
 589         LASSERT(end > start && end <= size() + 1, /**/);
 590
 591         return d->changes_.isChanged(start, end);
 592 }
 593
 594
 595 bool Paragraph::isMergedOnEndOfParDeletion(bool trackChanges) const
 596 {
 597         // keep the logic here in sync with the logic of eraseChars()
 598         if (!trackChanges)
 599                 return true;
 600
 601         Change const change = d->changes_.lookup(size());
 602         return change.inserted() && change.currentAuthor();
 603 }
 604
 605
 606 void Paragraph::setChange(Change const & change)
 607 {
 608         // beware of the imaginary end-of-par character!
 609         d->changes_.set(change, 0, size() + 1);
 610
 611         /*
 612          * Propagate the change recursively - but not in case of DELETED!
 613          *
 614          * Imagine that your co-author makes changes in an existing inset. He
 615          * sends your document to you and you come to the conclusion that the
 616          * inset should go completely. If you erase it, LyX must not delete all
 617          * text within the inset. Otherwise, the change tracked insertions of
 618          * your co-author get lost and there is no way to restore them later.
 619          *
 620          * Conclusion: An inset's content should remain untouched if you delete it
 621          */
 622
 623         if (!change.deleted()) {
 624                 for (pos_type pos = 0; pos < size(); ++pos) {
 625                         if (Inset * inset = getInset(pos))
 626                                 inset->setChange(change);
 627                 }
 628         }
 629 }
 630
 631
 632 void Paragraph::setChange(pos_type pos, Change const & change)
 633 {
 634         LASSERT(pos >= 0 && pos <= size(), /**/);
 635         d->changes_.set(change, pos);
 636
 637         // see comment in setChange(Change const &) above
 638         if (!change.deleted() && pos < size())
 639                         if (Inset * inset = getInset(pos))
 640                                 inset->setChange(change);
 641 }
 642
 643
 644 Change const & Paragraph::lookupChange(pos_type pos) const
 645 {
 646         LASSERT(pos >= 0 && pos <= size(), /**/);
 647         return d->changes_.lookup(pos);
 648 }
 649
 650
 651 void Paragraph::acceptChanges(pos_type start, pos_type end)
 652 {
 653         LASSERT(start >= 0 && start <= size(), /**/);
 654         LASSERT(end > start && end <= size() + 1, /**/);
 655
 656         for (pos_type pos = start; pos < end; ++pos) {
 657                 switch (lookupChange(pos).type) {
 658                         case Change::UNCHANGED:
 659                                 // accept changes in nested inset
 660                                 if (Inset * inset = getInset(pos))
 661                                         inset->acceptChanges();
 662                                 break;
 663
 664                         case Change::INSERTED:
 665                                 d->changes_.set(Change(Change::UNCHANGED), pos);
 666                                 // also accept changes in nested inset
 667                                 if (Inset * inset = getInset(pos))
 668                                         inset->acceptChanges();
 669                                 break;
 670
 671                         case Change::DELETED:
 672                                 // Suppress access to non-existent
 673                                 // "end-of-paragraph char"
 674                                 if (pos < size()) {
 675                                         eraseChar(pos, false);
 676                                         --end;
 677                                         --pos;
 678                                 }
 679                                 break;
 680                 }
 681
 682         }
 683 }
 684
 685
 686 void Paragraph::rejectChanges(pos_type start, pos_type end)
 687 {
 688         LASSERT(start >= 0 && start <= size(), /**/);
 689         LASSERT(end > start && end <= size() + 1, /**/);
 690
 691         for (pos_type pos = start; pos < end; ++pos) {
 692                 switch (lookupChange(pos).type) {
 693                         case Change::UNCHANGED:
 694                                 // reject changes in nested inset
 695                                 if (Inset * inset = getInset(pos))
 696                                                 inset->rejectChanges();
 697                                 break;
 698
 699                         case Change::INSERTED:
 700                                 // Suppress access to non-existent
 701                                 // "end-of-paragraph char"
 702                                 if (pos < size()) {
 703                                         eraseChar(pos, false);
 704                                         --end;
 705                                         --pos;
 706                                 }
 707                                 break;
 708
 709                         case Change::DELETED:
 710                                 d->changes_.set(Change(Change::UNCHANGED), pos);
 711
 712                                 // Do NOT reject changes within a deleted inset!
 713                                 // There may be insertions of a co-author inside of it!
 714
 715                                 break;
 716                 }
 717         }
 718 }
 719
 720
 721 void Paragraph::Private::insertChar(pos_type pos, char_type c,
 722                 Change const & change)
 723 {
 724         LASSERT(pos >= 0 && pos <= int(text_.size()), /**/);
 725
 726         // track change
 727         changes_.insert(change, pos);
 728
 729         // This is actually very common when parsing buffers (and
 730         // maybe inserting ascii text)
 731         if (pos == pos_type(text_.size())) {
 732                 // when appending characters, no need to update tables
 733                 text_.push_back(c);
 734                 // but we want spell checking
 735                 requestSpellCheck(pos);
 736                 return;
 737         }
 738
 739         text_.insert(text_.begin() + pos, c);
 740
 741         // Update the font table.
 742         fontlist_.increasePosAfterPos(pos);
 743
 744         // Update the insets
 745         insetlist_.increasePosAfterPos(pos);
 746
 747         // Update list of misspelled positions
 748         speller_state_.increasePosAfterPos(pos);
 749 }
 750
 751
 752 bool Paragraph::insertInset(pos_type pos, Inset * inset,
 753                                    Change const & change)
 754 {
 755         LASSERT(inset, /**/);
 756         LASSERT(pos >= 0 && pos <= size(), /**/);
 757
 758         // Paragraph::insertInset() can be used in cut/copy/paste operation where
 759         // d->inset_owner_ is not set yet.
 760         if (d->inset_owner_ && !d->inset_owner_->insetAllowed(inset->lyxCode()))
 761                 return false;
 762
 763         d->insertChar(pos, META_INSET, change);
 764         LASSERT(d->text_[pos] == META_INSET, /**/);
 765
 766         // Add a new entry in the insetlist_.
 767         d->insetlist_.insert(inset, pos);
 768
 769         // Some insets require run of spell checker
 770         requestSpellCheck(pos);
 771         return true;
 772 }
 773
 774
 775 bool Paragraph::eraseChar(pos_type pos, bool trackChanges)
 776 {
 777         LASSERT(pos >= 0 && pos <= size(), return false);
 778
 779         // keep the logic here in sync with the logic of isMergedOnEndOfParDeletion()
 780
 781         if (trackChanges) {
 782                 Change change = d->changes_.lookup(pos);
 783
 784                 // set the character to DELETED if
 785                 //  a) it was previously unchanged or
 786                 //  b) it was inserted by a co-author
 787
 788                 if (!change.changed() ||
 789                       (change.inserted() && !change.currentAuthor())) {
 790                         setChange(pos, Change(Change::DELETED));
 791                         // request run of spell checker
 792                         requestSpellCheck(pos);
 793                         return false;
 794                 }
 795
 796                 if (change.deleted())
 797                         return false;
 798         }
 799
 800         // Don't physically access the imaginary end-of-paragraph character.
 801         // eraseChar() can only mark it as DELETED. A physical deletion of
 802         // end-of-par must be handled externally.
 803         if (pos == size()) {
 804                 return false;
 805         }
 806
 807         // track change
 808         d->changes_.erase(pos);
 809
 810         // if it is an inset, delete the inset entry
 811         if (d->text_[pos] == META_INSET)
 812                 d->insetlist_.erase(pos);
 813
 814         d->text_.erase(d->text_.begin() + pos);
 815
 816         // Update the fontlist_
 817         d->fontlist_.erase(pos);
 818
 819         // Update the insetlist_
 820         d->insetlist_.decreasePosAfterPos(pos);
 821
 822         // Update list of misspelled positions
 823         d->speller_state_.decreasePosAfterPos(pos);
 824         d->speller_state_.refreshLast(size());
 825
 826         return true;
 827 }
 828
 829
 830 int Paragraph::eraseChars(pos_type start, pos_type end, bool trackChanges)
 831 {
 832         LASSERT(start >= 0 && start <= size(), /**/);
 833         LASSERT(end >= start && end <= size() + 1, /**/);
 834
 835         pos_type i = start;
 836         for (pos_type count = end - start; count; --count) {
 837                 if (!eraseChar(i, trackChanges))
 838                         ++i;
 839         }
 840         return end - i;
 841 }
 842
 843
 844 int Paragraph::Private::latexSurrogatePair(otexstream & os, char_type c,
 845                 char_type next, OutputParams const & runparams)
 846 {
 847         // Writing next here may circumvent a possible font change between
 848         // c and next. Since next is only output if it forms a surrogate pair
 849         // with c we can ignore this:
 850         // A font change inside a surrogate pair does not make sense and is
 851         // hopefully impossible to input.
 852         // FIXME: change tracking
 853         // Is this correct WRT change tracking?
 854         Encoding const & encoding = *(runparams.encoding);
 855         docstring const latex1 = encoding.latexChar(next);
 856         docstring const latex2 = encoding.latexChar(c);
 857         if (docstring(1, next) == latex1) {
 858                 // the encoding supports the combination
 859                 os << latex2 << latex1;
 860                 return latex1.length() + latex2.length();
 861         } else if (runparams.local_font &&
 862                    runparams.local_font->language()->lang() == "polutonikogreek") {
 863                 // polutonikogreek only works without the brackets
 864                 os << latex1 << latex2;
 865                 return latex1.length() + latex2.length();
 866         } else
 867                 os << latex1 << '{' << latex2 << '}';
 868         return latex1.length() + latex2.length() + 2;
 869 }
 870
 871
 872 bool Paragraph::Private::simpleTeXBlanks(OutputParams const & runparams,
 873                                        otexstream & os,
 874                                        pos_type i,
 875                                        unsigned int & column,
 876                                        Font const & font,
 877                                        Layout const & style)
 878 {
 879         if (style.pass_thru || runparams.pass_thru)
 880                 return false;
 881
 882         if (i + 1 < int(text_.size())) {
 883                 char_type next = text_[i + 1];
 884                 if (Encodings::isCombiningChar(next)) {
 885                         // This space has an accent, so we must always output it.
 886                         column += latexSurrogatePair(os, ' ', next, runparams) - 1;
 887                         return true;
 888                 }
 889         }
 890
 891         if (runparams.linelen > 0
 892             && column > runparams.linelen
 893             && i
 894             && text_[i - 1] != ' '
 895             && (i + 1 < int(text_.size()))
 896             // same in FreeSpacing mode
 897             && !owner_->isFreeSpacing()
 898             // In typewriter mode, we want to avoid
 899             // ! . ? : at the end of a line
 900             && !(font.fontInfo().family() == TYPEWRITER_FAMILY
 901                  && (text_[i - 1] == '.'
 902                      || text_[i - 1] == '?'
 903                      || text_[i - 1] == ':'
 904                      || text_[i - 1] == '!'))) {
 905                 os << '\n';
 906                 os.texrow().start(owner_->id(), i + 1);
 907                 column = 0;
 908         } else if (style.free_spacing) {
 909                 os << '~';
 910         } else {
 911                 os << ' ';
 912         }
 913         return false;
 914 }
 915
 916
 917 int Paragraph::Private::writeScriptChars(otexstream & os,
 918                                          docstring const & ltx,
 919                                          Change const & runningChange,
 920                                          Encoding const & encoding,
 921                                          pos_type & i)
 922 {
 923         // FIXME: modifying i here is not very nice...
 924
 925         // We only arrive here when a proper language for character text_[i] has
 926         // not been specified (i.e., it could not be translated in the current
 927         // latex encoding) or its latex translation has been forced, and it
 928         // belongs to a known script.
 929         // Parameter ltx contains the latex translation of text_[i] as specified
 930         // in the unicodesymbols file and is something like "\textXXX{<spec>}".
 931         // The latex macro name "textXXX" specifies the script to which text_[i]
 932         // belongs and we use it in order to check whether characters from the
 933         // same script immediately follow, such that we can collect them in a
 934         // single "\textXXX" macro. So, we have to retain "\textXXX{<spec>"
 935         // for the first char but only "<spec>" for all subsequent chars.
 936         docstring::size_type const brace1 = ltx.find_first_of(from_ascii("{"));
 937         docstring::size_type const brace2 = ltx.find_last_of(from_ascii("}"));
 938         string script = to_ascii(ltx.substr(1, brace1 - 1));
 939         int pos = 0;
 940         int length = brace2;
 941         bool closing_brace = true;
 942         if (script == "textgreek" && encoding.latexName() == "iso-8859-7") {
 943                 // Correct encoding is being used, so we can avoid \textgreek.
 944                 pos = brace1 + 1;
 945                 length -= pos;
 946                 closing_brace = false;
 947         }
 948         os << ltx.substr(pos, length);
 949         int size = text_.size();
 950         while (i + 1 < size) {
 951                 char_type const next = text_[i + 1];
 952                 // Stop here if next character belongs to another script
 953                 // or there is a change in change tracking status.
 954                 if (!Encodings::isKnownScriptChar(next, script) ||
 955                     runningChange != owner_->lookupChange(i + 1))
 956                         break;
 957                 Font prev_font;
 958                 bool found = false;
 959                 FontList::const_iterator cit = fontlist_.begin();
 960                 FontList::const_iterator end = fontlist_.end();
 961                 for (; cit != end; ++cit) {
 962                         if (cit->pos() >= i && !found) {
 963                                 prev_font = cit->font();
 964                                 found = true;
 965                         }
 966                         if (cit->pos() >= i + 1)
 967                                 break;
 968                 }
 969                 // Stop here if there is a font attribute or encoding change.
 970                 if (found && cit != end && prev_font != cit->font())
 971                         break;
 972                 docstring const latex = encoding.latexChar(next);
 973                 docstring::size_type const b1 =
 974                                         latex.find_first_of(from_ascii("{"));
 975                 docstring::size_type const b2 =
 976                                         latex.find_last_of(from_ascii("}"));
 977                 int const len = b2 - b1 - 1;
 978                 os << latex.substr(b1 + 1, len);
 979                 length += len;
 980                 ++i;
 981         }
 982         if (closing_brace) {
 983                 os << '}';
 984                 ++length;
 985         }
 986         return length;
 987 }
 988
 989
 990 bool Paragraph::Private::isTextAt(string const & str, pos_type pos) const
 991 {
 992         pos_type const len = str.length();
 993
 994         // is the paragraph large enough?
 995         if (pos + len > int(text_.size()))
 996                 return false;
 997
 998         // does the wanted text start at point?
 999         for (string::size_type i = 0; i < str.length(); ++i) {
1000                 // Caution: direct comparison of characters works only
1001                 // because str is pure ASCII.
1002                 if (str[i] != text_[pos + i])
1003                         return false;
1004         }
1005
1006         return fontlist_.hasChangeInRange(pos, len);
1007 }
1008
1009
1010 void Paragraph::Private::latexInset(BufferParams const & bparams,
1011                                     otexstream & os,
1012                                     OutputParams & runparams,
1013                                     Font & running_font,
1014                                     Font & basefont,
1015                                     Font const & outerfont,
1016                                     bool & open_font,
1017                                     Change & running_change,
1018                                     Layout const & style,
1019                                     pos_type & i,
1020                                     unsigned int & column)
1021 {
1022         Inset * inset = owner_->getInset(i);
1023         LASSERT(inset, /**/);
1024
1025         if (style.pass_thru) {
1026                 inset->plaintext(os.os(), runparams);
1027                 return;
1028         }
1029
1030         // FIXME: move this to InsetNewline::latex
1031         if (inset->lyxCode() == NEWLINE_CODE) {
1032                 // newlines are handled differently here than
1033                 // the default in simpleTeXSpecialChars().
1034                 if (!style.newline_allowed) {
1035                         os << '\n';
1036                 } else {
1037                         if (open_font) {
1038                                 column += running_font.latexWriteEndChanges(
1039                                         os, bparams, runparams,
1040                                         basefont, basefont);
1041                                 open_font = false;
1042                         }
1043
1044                         if (running_font.fontInfo().family() == TYPEWRITER_FAMILY)
1045                                 os << '~';
1046
1047                         basefont = owner_->getLayoutFont(bparams, outerfont);
1048                         running_font = basefont;
1049
1050                         if (runparams.moving_arg)
1051                                 os << "\\protect ";
1052
1053                 }
1054                 os.texrow().start(owner_->id(), i + 1);
1055                 column = 0;
1056         }
1057
1058         if (owner_->isDeleted(i)) {
1059                 if( ++runparams.inDeletedInset == 1)
1060                         runparams.changeOfDeletedInset = owner_->lookupChange(i);
1061         }
1062
1063         if (inset->canTrackChanges()) {
1064                 column += Changes::latexMarkChange(os, bparams, running_change,
1065                         Change(Change::UNCHANGED), runparams);
1066                 running_change = Change(Change::UNCHANGED);
1067         }
1068
1069         bool close = false;
1070         odocstream::pos_type const len = os.os().tellp();
1071
1072         if (inset->forceLTR()
1073             && running_font.isRightToLeft()
1074             // ERT is an exception, it should be output with no
1075             // decorations at all
1076             && inset->lyxCode() != ERT_CODE) {
1077                 if (running_font.language()->lang() == "farsi")
1078                         os << "\\beginL{}";
1079                 else
1080                         os << "\\L{";
1081                 close = true;
1082         }
1083
1084         // FIXME: Bug: we can have an empty font change here!
1085         // if there has just been a font change, we are going to close it
1086         // right now, which means stupid latex code like \textsf{}. AFAIK,
1087         // this does not harm dvi output. A minor bug, thus (JMarc)
1088
1089         // Some insets cannot be inside a font change command.
1090         // However, even such insets *can* be placed in \L or \R
1091         // or their equivalents (for RTL language switches), so we don't
1092         // close the language in those cases.
1093         // ArabTeX, though, cannot handle this special behavior, it seems.
1094         bool arabtex = basefont.language()->lang() == "arabic_arabtex"
1095                 || running_font.language()->lang() == "arabic_arabtex";
1096         if (open_font && !inset->inheritFont()) {
1097                 bool closeLanguage = arabtex
1098                         || basefont.isRightToLeft() == running_font.isRightToLeft();
1099                 unsigned int count = running_font.latexWriteEndChanges(os,
1100                         bparams, runparams, basefont, basefont, closeLanguage);
1101                 column += count;
1102                 // if any font properties were closed, update the running_font,
1103                 // making sure, however, to leave the language as it was
1104                 if (count > 0) {
1105                         // FIXME: probably a better way to keep track of the old
1106                         // language, than copying the entire font?
1107                         Font const copy_font(running_font);
1108                         basefont = owner_->getLayoutFont(bparams, outerfont);
1109                         running_font = basefont;
1110                         if (!closeLanguage)
1111                                 running_font.setLanguage(copy_font.language());
1112                         // leave font open if language is still open
1113                         open_font = (running_font.language() == basefont.language());
1114                         if (closeLanguage)
1115                                 runparams.local_font = &basefont;
1116                 }
1117         }
1118
1119         int prev_rows = os.texrow().rows();
1120
1121         try {
1122                 runparams.lastid = id_;
1123                 runparams.lastpos = i;
1124                 inset->latex(os, runparams);
1125         } catch (EncodingException & e) {
1126                 // add location information and throw again.
1127                 e.par_id = id_;
1128                 e.pos = i;
1129                 throw(e);
1130         }
1131
1132         if (close) {
1133                 if (running_font.language()->lang() == "farsi")
1134                                 os << "\\endL{}";
1135                         else
1136                                 os << '}';
1137         }
1138
1139         if (os.texrow().rows() > prev_rows) {
1140                 os.texrow().start(owner_->id(), i + 1);
1141                 column = 0;
1142         } else {
1143                 column += (unsigned int)(os.os().tellp() - len);
1144         }
1145
1146         if (owner_->isDeleted(i))
1147                 --runparams.inDeletedInset;
1148 }
1149
1150
1151 void Paragraph::Private::latexSpecialChar(otexstream & os,
1152                                           OutputParams const & runparams,
1153                                           Font const & running_font,
1154                                           Change const & running_change,
1155                                           Layout const & style,
1156                                           pos_type & i,
1157                                           pos_type end_pos,
1158                                           unsigned int & column)
1159 {
1160         char_type const c = text_[i];
1161
1162         if (style.pass_thru || runparams.pass_thru) {
1163                 if (c != '\0') {
1164                         Encoding const * const enc = runparams.encoding;
1165                         if (enc && enc->latexChar(c, true).empty())
1166                                 throw EncodingException(c);
1167                         os.put(c);
1168                 }
1169                 return;
1170         }
1171
1172         // If T1 font encoding is used, use the special
1173         // characters it provides.
1174         // NOTE: some languages reset the font encoding
1175         // internally
1176         if (!running_font.language()->internalFontEncoding()
1177             && lyxrc.fontenc == "T1" && latexSpecialT1(c, os, i, column))
1178                 return;
1179
1180         // \tt font needs special treatment
1181         if (running_font.fontInfo().family() == TYPEWRITER_FAMILY
1182                 && latexSpecialTypewriter(c, os, i, column))
1183                 return;
1184
1185         // Otherwise, we use what LaTeX provides us.
1186         switch (c) {
1187         case '\\':
1188                 os << "\\textbackslash{}";
1189                 column += 15;
1190                 break;
1191         case '<':
1192                 os << "\\textless{}";
1193                 column += 10;
1194                 break;
1195         case '>':
1196                 os << "\\textgreater{}";
1197                 column += 13;
1198                 break;
1199         case '|':
1200                 os << "\\textbar{}";
1201                 column += 9;
1202                 break;
1203         case '-':
1204                 os << '-';
1205                 break;
1206         case '\"':
1207                 os << "\\char`\\\"{}";
1208                 column += 9;
1209                 break;
1210
1211         case '$': case '&':
1212         case '%': case '#': case '{':
1213         case '}': case '_':
1214                 os << '\\';
1215                 os.put(c);
1216                 column += 1;
1217                 break;
1218
1219         case '~':
1220                 os << "\\textasciitilde{}";
1221                 column += 16;
1222                 break;
1223
1224         case '^':
1225                 os << "\\textasciicircum{}";
1226                 column += 17;
1227                 break;
1228
1229         case '*':
1230         case '[':
1231         case ']':
1232                 // avoid being mistaken for optional arguments
1233                 os << '{';
1234                 os.put(c);
1235                 os << '}';
1236                 column += 2;
1237                 break;
1238
1239         case ' ':
1240                 // Blanks are printed before font switching.
1241                 // Sure? I am not! (try nice-latex)
1242                 // I am sure it's correct. LyX might be smarter
1243                 // in the future, but for now, nothing wrong is
1244                 // written. (Asger)
1245                 break;
1246
1247         default:
1248                 // LyX, LaTeX etc.
1249                 if (latexSpecialPhrase(os, i, end_pos, column, runparams))
1250                         return;
1251
1252                 if (c == '\0')
1253                         return;
1254
1255                 Encoding const & encoding = *(runparams.encoding);
1256                 if (i + 1 < int(text_.size())) {
1257                         char_type next = text_[i + 1];
1258                         if (Encodings::isCombiningChar(next)) {
1259                                 column += latexSurrogatePair(os, c, next, runparams) - 1;
1260                                 ++i;
1261                                 break;
1262                         }
1263                 }
1264                 string script;
1265                 docstring const latex = encoding.latexChar(c);
1266                 if (Encodings::isKnownScriptChar(c, script)
1267                     && prefixIs(latex, from_ascii("\\" + script)))
1268                         column += writeScriptChars(os, latex,
1269                                         running_change, encoding, i) - 1;
1270                 else if (latex.length() > 1 && latex[latex.length() - 1] != '}') {
1271                         // Prevent eating of a following
1272                         // space or command corruption by
1273                         // following characters
1274                         column += latex.length() + 1;
1275                         os << latex << "{}";
1276                 } else {
1277                         column += latex.length() - 1;
1278                         os << latex;
1279                 }
1280                 break;
1281         }
1282 }
1283
1284
1285 bool Paragraph::Private::latexSpecialT1(char_type const c, otexstream & os,
1286         pos_type i, unsigned int & column)
1287 {
1288         switch (c) {
1289         case '>':
1290         case '<':
1291                 os.put(c);
1292                 // In T1 encoding, these characters exist
1293                 // but we should avoid ligatures
1294                 if (i + 1 >= int(text_.size()) || text_[i + 1] != c)
1295                         return true;
1296                 os << "\\textcompwordmark{}";
1297                 column += 19;
1298                 return true;
1299         case '|':
1300                 os.put(c);
1301                 return true;
1302         case '\"':
1303                 // soul.sty breaks with \char`\"
1304                 os << "\\textquotedbl{}";
1305                 column += 14;
1306                 return true;
1307         default:
1308                 return false;
1309         }
1310 }
1311
1312
1313 bool Paragraph::Private::latexSpecialTypewriter(char_type const c, otexstream & os,
1314         pos_type i, unsigned int & column)
1315 {
1316         switch (c) {
1317         case '-':
1318                 // within \ttfamily, "--" is merged to "-" (no endash)
1319                 // so we avoid this rather irritating ligature
1320                 if (i + 1 < int(text_.size()) && text_[i + 1] == '-') {
1321                         os << "-{}";
1322                         column += 2;
1323                 } else
1324                         os << '-';
1325                 return true;
1326
1327         // everything else has to be checked separately
1328         // (depending on the encoding)
1329         default:
1330                 return false;
1331         }
1332 }
1333
1334
1335 /// \param end_pos
1336 ///   If [start_pos, end_pos) does not include entirely the special phrase, then
1337 ///   do not apply the macro transformation.
1338 bool Paragraph::Private::latexSpecialPhrase(otexstream & os, pos_type & i, pos_type end_pos,
1339         unsigned int & column, OutputParams const & runparams)
1340 {
1341         // FIXME: if we have "LaTeX" with a font
1342         // change in the middle (before the 'T', then
1343         // the "TeX" part is still special cased.
1344         // Really we should only operate this on
1345         // "words" for some definition of word
1346
1347         for (size_t pnr = 0; pnr < phrases_nr; ++pnr) {
1348                 if (!isTextAt(special_phrases[pnr].phrase, i)
1349                     || (end_pos != -1 && i + int(special_phrases[pnr].phrase.size()) > end_pos))
1350                         continue;
1351                 if (runparams.moving_arg)
1352                         os << "\\protect";
1353                 os << special_phrases[pnr].macro;
1354                 i += special_phrases[pnr].phrase.length() - 1;
1355                 column += special_phrases[pnr].macro.length() - 1;
1356                 return true;
1357         }
1358         return false;
1359 }
1360
1361
1362 void Paragraph::Private::validate(LaTeXFeatures & features) const
1363 {
1364         if (layout_->inpreamble && inset_owner_) {
1365                 bool const is_command = layout_->latextype == LATEX_COMMAND;
1366                 Buffer const & buf = inset_owner_->buffer();
1367                 BufferParams const & bp = buf.params();
1368                 Font f;
1369                 TexRow texrow;
1370                 // Using a string stream here circumvents the encoding
1371                 // switching machinery of odocstream. Therefore the
1372                 // output is wrong if this paragraph contains content
1373                 // that needs to switch encoding.
1374                 odocstringstream ods;
1375                 otexstream os(ods, texrow);
1376                 if (is_command) {
1377                         os << '\\' << from_ascii(layout_->latexname());
1378                         // we have to provide all the optional arguments here, even though
1379                         // the last one is the only one we care about.
1380                         // Separate handling of optional argument inset.
1381                         if (layout_->optargs != 0 || layout_->reqargs != 0)
1382                                 latexArgInsets(*owner_, os, features.runparams(),
1383                                         layout_->reqargs, layout_->optargs);
1384                         else
1385                                 os << from_ascii(layout_->latexparam());
1386                 }
1387                 docstring::size_type const length = ods.str().length();
1388                 // this will output "{" at the beginning, but not at the end
1389                 owner_->latex(bp, f, os, features.runparams(), 0, -1, true);
1390                 if (ods.str().length() > length) {
1391                         if (is_command)
1392                                 ods << '}';
1393                         string const snippet = to_utf8(ods.str());
1394                         features.addPreambleSnippet(snippet);
1395                 }
1396         }
1397
1398         if (features.runparams().flavor == OutputParams::HTML
1399             && layout_->htmltitle()) {
1400                 features.setHTMLTitle(owner_->asString(AS_STR_INSETS));
1401         }
1402
1403         // check the params.
1404         if (!params_.spacing().isDefault())
1405                 features.require("setspace");
1406
1407         // then the layouts
1408         features.useLayout(layout_->name());
1409
1410         // then the fonts
1411         fontlist_.validate(features);
1412
1413         // then the indentation
1414         if (!params_.leftIndent().zero())
1415                 features.require("ParagraphLeftIndent");
1416
1417         // then the insets
1418         InsetList::const_iterator icit = insetlist_.begin();
1419         InsetList::const_iterator iend = insetlist_.end();
1420         for (; icit != iend; ++icit) {
1421                 if (icit->inset) {
1422                         icit->inset->validate(features);
1423                         if (layout_->needprotect &&
1424                             icit->inset->lyxCode() == FOOT_CODE)
1425                                 features.require("NeedLyXFootnoteCode");
1426                 }
1427         }
1428
1429         // then the contents
1430         for (pos_type i = 0; i < int(text_.size()) ; ++i) {
1431                 for (size_t pnr = 0; pnr < phrases_nr; ++pnr) {
1432                         if (!special_phrases[pnr].builtin
1433                             && isTextAt(special_phrases[pnr].phrase, i)) {
1434                                 features.require(special_phrases[pnr].phrase);
1435                                 break;
1436                         }
1437                 }
1438                 Encodings::validate(text_[i], features);
1439         }
1440 }
1441
1442 /////////////////////////////////////////////////////////////////////
1443 //
1444 // Paragraph
1445 //
1446 /////////////////////////////////////////////////////////////////////
1447
1448 namespace {
1449         Layout const emptyParagraphLayout;
1450 }
1451
1452 Paragraph::Paragraph()
1453         : d(new Paragraph::Private(this, emptyParagraphLayout))
1454 {
1455         itemdepth = 0;
1456         d->params_.clear();
1457 }
1458
1459
1460 Paragraph::Paragraph(Paragraph const & par)
1461         : itemdepth(par.itemdepth),
1462         d(new Paragraph::Private(*par.d, this))
1463 {
1464         registerWords();
1465 }
1466
1467
1468 Paragraph::Paragraph(Paragraph const & par, pos_type beg, pos_type end)
1469         : itemdepth(par.itemdepth),
1470         d(new Paragraph::Private(*par.d, this, beg, end))
1471 {
1472         registerWords();
1473 }
1474
1475
1476 Paragraph & Paragraph::operator=(Paragraph const & par)
1477 {
1478         // needed as we will destroy the private part before copying it
1479         if (&par != this) {
1480                 itemdepth = par.itemdepth;
1481
1482                 deregisterWords();
1483                 delete d;
1484                 d = new Private(*par.d, this);
1485                 registerWords();
1486         }
1487         return *this;
1488 }
1489
1490
1491 Paragraph::~Paragraph()
1492 {
1493         deregisterWords();
1494         delete d;
1495 }
1496
1497
1498 namespace {
1499
1500 // this shall be called just before every "os << ..." action.
1501 void flushString(ostream & os, docstring & s)
1502 {
1503         os << to_utf8(s);
1504         s.erase();
1505 }
1506
1507 }
1508
1509
1510 void Paragraph::write(ostream & os, BufferParams const & bparams,
1511         depth_type & dth) const
1512 {
1513         // The beginning or end of a deeper (i.e. nested) area?
1514         if (dth != d->params_.depth()) {
1515                 if (d->params_.depth() > dth) {
1516                         while (d->params_.depth() > dth) {
1517                                 os << "\n\\begin_deeper";
1518                                 ++dth;
1519                         }
1520                 } else {
1521                         while (d->params_.depth() < dth) {
1522                                 os << "\n\\end_deeper";
1523                                 --dth;
1524                         }
1525                 }
1526         }
1527
1528         // First write the layout
1529         os << "\n\\begin_layout " << to_utf8(d->layout_->name()) << '\n';
1530
1531         d->params_.write(os);
1532
1533         Font font1(inherit_font, bparams.language);
1534
1535         Change running_change = Change(Change::UNCHANGED);
1536
1537         // this string is used as a buffer to avoid repetitive calls
1538         // to to_utf8(), which turn out to be expensive (JMarc)
1539         docstring write_buffer;
1540
1541         int column = 0;
1542         for (pos_type i = 0; i <= size(); ++i) {
1543
1544                 Change const change = lookupChange(i);
1545                 if (change != running_change)
1546                         flushString(os, write_buffer);
1547                 Changes::lyxMarkChange(os, bparams, column, running_change, change);
1548                 running_change = change;
1549
1550                 if (i == size())
1551                         break;
1552
1553                 // Write font changes
1554                 Font font2 = getFontSettings(bparams, i);
1555                 if (font2 != font1) {
1556                         flushString(os, write_buffer);
1557                         font2.lyxWriteChanges(font1, os);
1558                         column = 0;
1559                         font1 = font2;
1560                 }
1561
1562                 char_type const c = d->text_[i];
1563                 switch (c) {
1564                 case META_INSET:
1565                         if (Inset const * inset = getInset(i)) {
1566                                 flushString(os, write_buffer);
1567                                 if (inset->directWrite()) {
1568                                         // international char, let it write
1569                                         // code directly so it's shorter in
1570                                         // the file
1571                                         inset->write(os);
1572                                 } else {
1573                                         if (i)
1574                                                 os << '\n';
1575                                         os << "\\begin_inset ";
1576                                         inset->write(os);
1577                                         os << "\n\\end_inset\n\n";
1578                                         column = 0;
1579                                 }
1580                         }
1581                         break;
1582                 case '\\':
1583                         flushString(os, write_buffer);
1584                         os << "\n\\backslash\n";
1585                         column = 0;
1586                         break;
1587                 case '.':
1588                         flushString(os, write_buffer);
1589                         if (i + 1 < size() && d->text_[i + 1] == ' ') {
1590                                 os << ".\n";
1591                                 column = 0;
1592                         } else
1593                                 os << '.';
1594                         break;
1595                 default:
1596                         if ((column > 70 && c == ' ')
1597                             || column > 79) {
1598                                 flushString(os, write_buffer);
1599                                 os << '\n';
1600                                 column = 0;
1601                         }
1602                         // this check is to amend a bug. LyX sometimes
1603                         // inserts '\0' this could cause problems.
1604                         if (c != '\0')
1605                                 write_buffer.push_back(c);
1606                         else
1607                                 LYXERR0("NUL char in structure.");
1608                         ++column;
1609                         break;
1610                 }
1611         }
1612
1613         flushString(os, write_buffer);
1614         os << "\n\\end_layout\n";
1615 }
1616
1617
1618 void Paragraph::validate(LaTeXFeatures & features) const
1619 {
1620         d->validate(features);
1621 }
1622
1623
1624 void Paragraph::insert(pos_type start, docstring const & str,
1625                        Font const & font, Change const & change)
1626 {
1627         for (size_t i = 0, n = str.size(); i != n ; ++i)
1628                 insertChar(start + i, str[i], font, change);
1629 }
1630
1631
1632 void Paragraph::appendChar(char_type c, Font const & font,
1633                 Change const & change)
1634 {
1635         // track change
1636         d->changes_.insert(change, d->text_.size());
1637         // when appending characters, no need to update tables
1638         d->text_.push_back(c);
1639         setFont(d->text_.size() - 1, font);
1640         d->requestSpellCheck(d->text_.size() - 1);
1641 }
1642
1643
1644 void Paragraph::appendString(docstring const & s, Font const & font,
1645                 Change const & change)
1646 {
1647         pos_type end = s.size();
1648         size_t oldsize = d->text_.size();
1649         size_t newsize = oldsize + end;
1650         size_t capacity = d->text_.capacity();
1651         if (newsize >= capacity)
1652                 d->text_.reserve(max(capacity + 100, newsize));
1653
1654         // when appending characters, no need to update tables
1655         d->text_.append(s);
1656
1657         // FIXME: Optimize this!
1658         for (size_t i = oldsize; i != newsize; ++i) {
1659                 // track change
1660                 d->changes_.insert(change, i);
1661                 d->requestSpellCheck(i);
1662         }
1663         d->fontlist_.set(oldsize, font);
1664         d->fontlist_.set(newsize - 1, font);
1665 }
1666
1667
1668 void Paragraph::insertChar(pos_type pos, char_type c,
1669                            bool trackChanges)
1670 {
1671         d->insertChar(pos, c, Change(trackChanges ?
1672                            Change::INSERTED : Change::UNCHANGED));
1673 }
1674
1675
1676 void Paragraph::insertChar(pos_type pos, char_type c,
1677                            Font const & font, bool trackChanges)
1678 {
1679         d->insertChar(pos, c, Change(trackChanges ?
1680                            Change::INSERTED : Change::UNCHANGED));
1681         setFont(pos, font);
1682 }
1683
1684
1685 void Paragraph::insertChar(pos_type pos, char_type c,
1686                            Font const & font, Change const & change)
1687 {
1688         d->insertChar(pos, c, change);
1689         setFont(pos, font);
1690 }
1691
1692
1693 bool Paragraph::insertInset(pos_type pos, Inset * inset,
1694                             Font const & font, Change const & change)
1695 {
1696         bool const success = insertInset(pos, inset, change);
1697         // Set the font/language of the inset...
1698         setFont(pos, font);
1699         return success;
1700 }
1701
1702
1703 void Paragraph::resetFonts(Font const & font)
1704 {
1705         d->fontlist_.clear();
1706         d->fontlist_.set(0, font);
1707         d->fontlist_.set(d->text_.size() - 1, font);
1708 }
1709
1710 // Gets uninstantiated font setting at position.
1711 Font const & Paragraph::getFontSettings(BufferParams const & bparams,
1712                                          pos_type pos) const
1713 {
1714         if (pos > size()) {
1715                 LYXERR0("pos: " << pos << " size: " << size());
1716                 LASSERT(pos <= size(), /**/);
1717         }
1718
1719         FontList::const_iterator cit = d->fontlist_.fontIterator(pos);
1720         if (cit != d->fontlist_.end())
1721                 return cit->font();
1722
1723         if (pos == size() && !empty())
1724                 return getFontSettings(bparams, pos - 1);
1725
1726         // Optimisation: avoid a full font instantiation if there is no
1727         // language change from previous call.
1728         static Font previous_font;
1729         static Language const * previous_lang = 0;
1730         Language const * lang = getParLanguage(bparams);
1731         if (lang != previous_lang) {
1732                 previous_lang = lang;
1733                 previous_font = Font(inherit_font, lang);
1734         }
1735         return previous_font;
1736 }
1737
1738
1739 FontSpan Paragraph::fontSpan(pos_type pos) const
1740 {
1741         LASSERT(pos <= size(), /**/);
1742         pos_type start = 0;
1743
1744         FontList::const_iterator cit = d->fontlist_.begin();
1745         FontList::const_iterator end = d->fontlist_.end();
1746         for (; cit != end; ++cit) {
1747                 if (cit->pos() >= pos) {
1748                         if (pos >= beginOfBody())
1749                                 return FontSpan(max(start, beginOfBody()),
1750                                                 cit->pos());
1751                         else
1752                                 return FontSpan(start,
1753                                                 min(beginOfBody() - 1,
1754                                                          cit->pos()));
1755                 }
1756                 start = cit->pos() + 1;
1757         }
1758
1759         // This should not happen, but if so, we take no chances.
1760         // LYXERR0("Paragraph::getEndPosOfFontSpan: This should not happen!");
1761         return FontSpan(pos, pos);
1762 }
1763
1764
1765 // Gets uninstantiated font setting at position 0
1766 Font const & Paragraph::getFirstFontSettings(BufferParams const & bparams) const
1767 {
1768         if (!empty() && !d->fontlist_.empty())
1769                 return d->fontlist_.begin()->font();
1770
1771         // Optimisation: avoid a full font instantiation if there is no
1772         // language change from previous call.
1773         static Font previous_font;
1774         static Language const * previous_lang = 0;
1775         if (bparams.language != previous_lang) {
1776                 previous_lang = bparams.language;
1777                 previous_font = Font(inherit_font, bparams.language);
1778         }
1779
1780         return previous_font;
1781 }
1782
1783
1784 // Gets the fully instantiated font at a given position in a paragraph
1785 // This is basically the same function as Text::GetFont() in text2.cpp.
1786 // The difference is that this one is used for generating the LaTeX file,
1787 // and thus cosmetic "improvements" are disallowed: This has to deliver
1788 // the true picture of the buffer. (Asger)
1789 Font const Paragraph::getFont(BufferParams const & bparams, pos_type pos,
1790                                  Font const & outerfont) const
1791 {
1792         LASSERT(pos >= 0, /**/);
1793
1794         Font font = getFontSettings(bparams, pos);
1795
1796         pos_type const body_pos = beginOfBody();
1797         FontInfo & fi = font.fontInfo();
1798         if (pos < body_pos)
1799                 fi.realize(d->layout_->labelfont);
1800         else
1801                 fi.realize(d->layout_->font);
1802
1803         fi.realize(outerfont.fontInfo());
1804         fi.realize(bparams.getFont().fontInfo());
1805
1806         return font;
1807 }
1808
1809
1810 Font const Paragraph::getLabelFont
1811         (BufferParams const & bparams, Font const & outerfont) const
1812 {
1813         FontInfo tmpfont = d->layout_->labelfont;
1814         tmpfont.realize(outerfont.fontInfo());
1815         tmpfont.realize(bparams.getFont().fontInfo());
1816         return Font(tmpfont, getParLanguage(bparams));
1817 }
1818
1819
1820 Font const Paragraph::getLayoutFont
1821         (BufferParams const & bparams, Font const & outerfont) const
1822 {
1823         FontInfo tmpfont = d->layout_->font;
1824         tmpfont.realize(outerfont.fontInfo());
1825         tmpfont.realize(bparams.getFont().fontInfo());
1826         return Font(tmpfont, getParLanguage(bparams));
1827 }
1828
1829
1830 /// Returns the height of the highest font in range
1831 FontSize Paragraph::highestFontInRange
1832         (pos_type startpos, pos_type endpos, FontSize def_size) const
1833 {
1834         return d->fontlist_.highestInRange(startpos, endpos, def_size);
1835 }
1836
1837
1838 char_type Paragraph::getUChar(BufferParams const & bparams, pos_type pos) const
1839 {
1840         char_type c = d->text_[pos];
1841         if (!lyxrc.rtl_support)
1842                 return c;
1843
1844         char_type uc = c;
1845         switch (c) {
1846         case '(':
1847                 uc = ')';
1848                 break;
1849         case ')':
1850                 uc = '(';
1851                 break;
1852         case '[':
1853                 uc = ']';
1854                 break;
1855         case ']':
1856                 uc = '[';
1857                 break;
1858         case '{':
1859                 uc = '}';
1860                 break;
1861         case '}':
1862                 uc = '{';
1863                 break;
1864         case '<':
1865                 uc = '>';
1866                 break;
1867         case '>':
1868                 uc = '<';
1869                 break;
1870         }
1871         if (uc != c && getFontSettings(bparams, pos).isRightToLeft())
1872                 return uc;
1873         return c;
1874 }
1875
1876
1877 void Paragraph::setFont(pos_type pos, Font const & font)
1878 {
1879         LASSERT(pos <= size(), /**/);
1880
1881         // First, reduce font against layout/label font
1882         // Update: The setCharFont() routine in text2.cpp already
1883         // reduces font, so we don't need to do that here. (Asger)
1884
1885         d->fontlist_.set(pos, font);
1886 }
1887
1888
1889 void Paragraph::makeSameLayout(Paragraph const & par)
1890 {
1891         d->layout_ = par.d->layout_;
1892         d->params_ = par.d->params_;
1893 }
1894
1895
1896 bool Paragraph::stripLeadingSpaces(bool trackChanges)
1897 {
1898         if (isFreeSpacing())
1899                 return false;
1900
1901         int pos = 0;
1902         int count = 0;
1903
1904         while (pos < size() && (isNewline(pos) || isLineSeparator(pos))) {
1905                 if (eraseChar(pos, trackChanges))
1906                         ++count;
1907                 else
1908                         ++pos;
1909         }
1910
1911         return count > 0 || pos > 0;
1912 }
1913
1914
1915 bool Paragraph::hasSameLayout(Paragraph const & par) const
1916 {
1917         return par.d->layout_ == d->layout_
1918                 && d->params_.sameLayout(par.d->params_);
1919 }
1920
1921
1922 depth_type Paragraph::getDepth() const
1923 {
1924         return d->params_.depth();
1925 }
1926
1927
1928 depth_type Paragraph::getMaxDepthAfter() const
1929 {
1930         if (d->layout_->isEnvironment())
1931                 return d->params_.depth() + 1;
1932         else
1933                 return d->params_.depth();
1934 }
1935
1936
1937 char Paragraph::getAlign() const
1938 {
1939         if (d->params_.align() == LYX_ALIGN_LAYOUT)
1940                 return d->layout_->align;
1941         else
1942                 return d->params_.align();
1943 }
1944
1945
1946 docstring const & Paragraph::labelString() const
1947 {
1948         return d->params_.labelString();
1949 }
1950
1951
1952 // the next two functions are for the manual labels
1953 docstring const Paragraph::getLabelWidthString() const
1954 {
1955         if (d->layout_->margintype == MARGIN_MANUAL
1956             || d->layout_->latextype == LATEX_BIB_ENVIRONMENT)
1957                 return d->params_.labelWidthString();
1958         else
1959                 return _("Senseless with this layout!");
1960 }
1961
1962
1963 void Paragraph::setLabelWidthString(docstring const & s)
1964 {
1965         d->params_.labelWidthString(s);
1966 }
1967
1968
1969 docstring Paragraph::expandLabel(Layout const & layout,
1970                 BufferParams const & bparams) const
1971 {
1972         return expandParagraphLabel(layout, bparams, true);
1973 }
1974
1975
1976 docstring Paragraph::expandDocBookLabel(Layout const & layout,
1977                 BufferParams const & bparams) const
1978 {
1979         return expandParagraphLabel(layout, bparams, false);
1980 }
1981
1982
1983 docstring Paragraph::expandParagraphLabel(Layout const & layout,
1984                 BufferParams const & bparams, bool process_appendix) const
1985 {
1986         DocumentClass const & tclass = bparams.documentClass();
1987         string const & lang = getParLanguage(bparams)->code();
1988         bool const in_appendix = process_appendix && d->params_.appendix();
1989         docstring fmt = translateIfPossible(layout.labelstring(in_appendix), lang);
1990
1991         if (fmt.empty() && layout.labeltype == LABEL_COUNTER
1992             && !layout.counter.empty())
1993                 return tclass.counters().theCounter(layout.counter, lang);
1994
1995         // handle 'inherited level parts' in 'fmt',
1996         // i.e. the stuff between '@' in   '@Section@.\arabic{subsection}'
1997         size_t const i = fmt.find('@', 0);
1998         if (i != docstring::npos) {
1999                 size_t const j = fmt.find('@', i + 1);
2000                 if (j != docstring::npos) {
2001                         docstring parent(fmt, i + 1, j - i - 1);
2002                         docstring label = from_ascii("??");
2003                         if (tclass.hasLayout(parent))
2004                                 docstring label = expandParagraphLabel(tclass[parent], bparams,
2005                                                       process_appendix);
2006                         fmt = docstring(fmt, 0, i) + label
2007                                 + docstring(fmt, j + 1, docstring::npos);
2008                 }
2009         }
2010
2011         return tclass.counters().counterLabel(fmt, lang);
2012 }
2013
2014
2015 void Paragraph::applyLayout(Layout const & new_layout)
2016 {
2017         d->layout_ = &new_layout;
2018         LyXAlignment const oldAlign = d->params_.align();
2019
2020         if (!(oldAlign & d->layout_->alignpossible)) {
2021                 frontend::Alert::warning(_("Alignment not permitted"),
2022                         _("The new layout does not permit the alignment previously used.\nSetting to default."));
2023                 d->params_.align(LYX_ALIGN_LAYOUT);
2024         }
2025 }
2026
2027
2028 pos_type Paragraph::beginOfBody() const
2029 {
2030         return d->begin_of_body_;
2031 }
2032
2033
2034 void Paragraph::setBeginOfBody()
2035 {
2036         if (d->layout_->labeltype != LABEL_MANUAL) {
2037                 d->begin_of_body_ = 0;
2038                 return;
2039         }
2040
2041         // Unroll the first two cycles of the loop
2042         // and remember the previous character to
2043         // remove unnecessary getChar() calls
2044         pos_type i = 0;
2045         pos_type end = size();
2046         if (i < end && !isNewline(i)) {
2047                 ++i;
2048                 char_type previous_char = 0;
2049                 char_type temp = 0;
2050                 if (i < end) {
2051                         previous_char = d->text_[i];
2052                         if (!isNewline(i)) {
2053                                 ++i;
2054                                 while (i < end && previous_char != ' ') {
2055                                         temp = d->text_[i];
2056                                         if (isNewline(i))
2057                                                 break;
2058                                         ++i;
2059                                         previous_char = temp;
2060                                 }
2061                         }
2062                 }
2063         }
2064
2065         d->begin_of_body_ = i;
2066 }
2067
2068
2069 bool Paragraph::allowParagraphCustomization() const
2070 {
2071         return inInset().allowParagraphCustomization();
2072 }
2073
2074
2075 bool Paragraph::usePlainLayout() const
2076 {
2077         return inInset().usePlainLayout();
2078 }
2079
2080
2081 bool Paragraph::isPassThru() const
2082 {
2083         return inInset().getLayout().isPassThru() || d->layout_->pass_thru;
2084 }
2085
2086 namespace {
2087
2088 // paragraphs inside floats need different alignment tags to avoid
2089 // unwanted space
2090
2091 bool noTrivlistCentering(InsetCode code)
2092 {
2093         return code == FLOAT_CODE
2094                || code == WRAP_CODE
2095                || code == CELL_CODE;
2096 }
2097
2098
2099 string correction(string const & orig)
2100 {
2101         if (orig == "flushleft")
2102                 return "raggedright";
2103         if (orig == "flushright")
2104                 return "raggedleft";
2105         if (orig == "center")
2106                 return "centering";
2107         return orig;
2108 }
2109
2110
2111 string const corrected_env(string const & suffix, string const & env,
2112         InsetCode code, bool const lastpar)
2113 {
2114         string output = suffix + "{";
2115         if (noTrivlistCentering(code)) {
2116                 if (lastpar) {
2117                         // the last paragraph in non-trivlist-aligned
2118                         // context is special (to avoid unwanted whitespace)
2119                         if (suffix == "\\begin")
2120                                 return "\\" + correction(env) + "{}";
2121                         return string();
2122                 }
2123                 output += correction(env);
2124         } else
2125                 output += env;
2126         output += "}";
2127         if (suffix == "\\begin")
2128                 output += "\n";
2129         return output;
2130 }
2131
2132
2133 void adjust_column(string const & str, int & column)
2134 {
2135         if (!contains(str, "\n"))
2136                 column += str.size();
2137         else {
2138                 string tmp;
2139                 column = rsplit(str, tmp, '\n').size();
2140         }
2141 }
2142
2143 } // namespace anon
2144
2145
2146 int Paragraph::Private::startTeXParParams(BufferParams const & bparams,
2147                         otexstream & os, OutputParams const & runparams) const
2148 {
2149         int column = 0;
2150
2151         if (params_.noindent() && !layout_->pass_thru) {
2152                 os << "\\noindent ";
2153                 column += 10;
2154         }
2155
2156         LyXAlignment const curAlign = params_.align();
2157
2158         if (curAlign == layout_->align)
2159                 return column;
2160
2161         switch (curAlign) {
2162         case LYX_ALIGN_NONE:
2163         case LYX_ALIGN_BLOCK:
2164         case LYX_ALIGN_LAYOUT:
2165         case LYX_ALIGN_SPECIAL:
2166         case LYX_ALIGN_DECIMAL:
2167                 break;
2168         case LYX_ALIGN_LEFT:
2169         case LYX_ALIGN_RIGHT:
2170         case LYX_ALIGN_CENTER:
2171                 if (runparams.moving_arg) {
2172                         os << "\\protect";
2173                         column += 8;
2174                 }
2175                 break;
2176         }
2177
2178         string const begin_tag = "\\begin";
2179         InsetCode code = ownerCode();
2180         bool const lastpar = runparams.isLastPar;
2181
2182         switch (curAlign) {
2183         case LYX_ALIGN_NONE:
2184         case LYX_ALIGN_BLOCK:
2185         case LYX_ALIGN_LAYOUT:
2186         case LYX_ALIGN_SPECIAL:
2187         case LYX_ALIGN_DECIMAL:
2188                 break;
2189         case LYX_ALIGN_LEFT: {
2190                 string output;
2191                 if (owner_->getParLanguage(bparams)->babel() != "hebrew")
2192                         output = corrected_env(begin_tag, "flushleft", code, lastpar);
2193                 else
2194                         output = corrected_env(begin_tag, "flushright", code, lastpar);
2195                 os << from_ascii(output);
2196                 adjust_column(output, column);
2197                 break;
2198         } case LYX_ALIGN_RIGHT: {
2199                 string output;
2200                 if (owner_->getParLanguage(bparams)->babel() != "hebrew")
2201                         output = corrected_env(begin_tag, "flushright", code, lastpar);
2202                 else
2203                         output = corrected_env(begin_tag, "flushleft", code, lastpar);
2204                 os << from_ascii(output);
2205                 adjust_column(output, column);
2206                 break;
2207         } case LYX_ALIGN_CENTER: {
2208                 string output;
2209                 output = corrected_env(begin_tag, "center", code, lastpar);
2210                 os << from_ascii(output);
2211                 adjust_column(output, column);
2212                 break;
2213         }
2214         }
2215
2216         return column;
2217 }
2218
2219
2220 bool Paragraph::Private::endTeXParParams(BufferParams const & bparams,
2221                         otexstream & os, OutputParams const & runparams) const
2222 {
2223         LyXAlignment const curAlign = params_.align();
2224
2225         if (curAlign == layout_->align)
2226                 return false;
2227
2228         switch (curAlign) {
2229         case LYX_ALIGN_NONE:
2230         case LYX_ALIGN_BLOCK:
2231         case LYX_ALIGN_LAYOUT:
2232         case LYX_ALIGN_SPECIAL:
2233         case LYX_ALIGN_DECIMAL:
2234                 break;
2235         case LYX_ALIGN_LEFT:
2236         case LYX_ALIGN_RIGHT:
2237         case LYX_ALIGN_CENTER:
2238                 if (runparams.moving_arg)
2239                         os << "\\protect";
2240                 break;
2241         }
2242
2243         string output;
2244         string const end_tag = "\n\\par\\end";
2245         InsetCode code = ownerCode();
2246         bool const lastpar = runparams.isLastPar;
2247
2248         switch (curAlign) {
2249         case LYX_ALIGN_NONE:
2250         case LYX_ALIGN_BLOCK:
2251         case LYX_ALIGN_LAYOUT:
2252         case LYX_ALIGN_SPECIAL:
2253         case LYX_ALIGN_DECIMAL:
2254                 break;
2255         case LYX_ALIGN_LEFT: {
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                 break;
2262         } case LYX_ALIGN_RIGHT: {
2263                 if (owner_->getParLanguage(bparams)->babel() != "hebrew")
2264                         output = corrected_env(end_tag, "flushright", code, lastpar);
2265                 else
2266                         output = corrected_env(end_tag, "flushleft", code, lastpar);
2267                 os << from_ascii(output);
2268                 break;
2269         } case LYX_ALIGN_CENTER: {
2270                 output = corrected_env(end_tag, "center", code, lastpar);
2271                 os << from_ascii(output);
2272                 break;
2273         }
2274         }
2275
2276         return !output.empty() || lastpar;
2277 }
2278
2279
2280 // This one spits out the text of the paragraph
2281 void Paragraph::latex(BufferParams const & bparams,
2282         Font const & outerfont,
2283         otexstream & os,
2284         OutputParams const & runparams,
2285         int start_pos, int end_pos, bool force) const
2286 {
2287         LYXERR(Debug::LATEX, "Paragraph::latex...     " << this);
2288
2289         // FIXME This check should not be needed. Perhaps issue an
2290         // error if it triggers.
2291         Layout const & style = inInset().forcePlainLayout() ?
2292                 bparams.documentClass().plainLayout() : *d->layout_;
2293
2294         if (!force && style.inpreamble)
2295                 return;
2296
2297         bool const allowcust = allowParagraphCustomization();
2298
2299         // Current base font for all inherited font changes, without any
2300         // change caused by an individual character, except for the language:
2301         // It is set to the language of the first character.
2302         // As long as we are in the label, this font is the base font of the
2303         // label. Before the first body character it is set to the base font
2304         // of the body.
2305         Font basefont;
2306
2307         // Maybe we have to create a optional argument.
2308         pos_type body_pos = beginOfBody();
2309         unsigned int column = 0;
2310
2311         if (body_pos > 0) {
2312                 // the optional argument is kept in curly brackets in
2313                 // case it contains a ']'
2314                 // This is not strictly needed, but if this is changed it
2315                 // would be a file format change, and tex2lyx would need
2316                 // to be adjusted, since it unconditionally removes the
2317                 // braces when it parses \item.
2318                 os << "[{";
2319                 column += 2;
2320                 basefont = getLabelFont(bparams, outerfont);
2321         } else {
2322                 basefont = getLayoutFont(bparams, outerfont);
2323         }
2324
2325         // Which font is currently active?
2326         Font running_font(basefont);
2327         // Do we have an open font change?
2328         bool open_font = false;
2329
2330         Change runningChange = Change(Change::UNCHANGED);
2331
2332         Encoding const * const prev_encoding = runparams.encoding;
2333
2334         os.texrow().start(id(), 0);
2335
2336         // if the paragraph is empty, the loop will not be entered at all
2337         if (empty()) {
2338                 if (style.isCommand()) {
2339                         os << '{';
2340                         ++column;
2341                 }
2342                 if (allowcust)
2343                         column += d->startTeXParParams(bparams, os, runparams);
2344         }
2345
2346         for (pos_type i = 0; i < size(); ++i) {
2347                 // First char in paragraph or after label?
2348                 if (i == body_pos) {
2349                         if (body_pos > 0) {
2350                                 if (open_font) {
2351                                         column += running_font.latexWriteEndChanges(
2352                                                 os, bparams, runparams,
2353                                                 basefont, basefont);
2354                                         open_font = false;
2355                                 }
2356                                 basefont = getLayoutFont(bparams, outerfont);
2357                                 running_font = basefont;
2358
2359                                 column += Changes::latexMarkChange(os, bparams,
2360                                                 runningChange, Change(Change::UNCHANGED),
2361                                                 runparams);
2362                                 runningChange = Change(Change::UNCHANGED);
2363
2364                                 os << "}] ";
2365                                 column +=3;
2366                         }
2367                         if (style.isCommand()) {
2368                                 os << '{';
2369                                 ++column;
2370                         }
2371
2372                         if (allowcust)
2373                                 column += d->startTeXParParams(bparams, os,
2374                                                             runparams);
2375                 }
2376
2377                 Change const & change = runparams.inDeletedInset
2378                         ? runparams.changeOfDeletedInset : lookupChange(i);
2379
2380                 if (bparams.outputChanges && runningChange != change) {
2381                         if (open_font) {
2382                                 column += running_font.latexWriteEndChanges(
2383                                                 os, bparams, runparams, basefont, basefont);
2384                                 open_font = false;
2385                         }
2386                         basefont = getLayoutFont(bparams, outerfont);
2387                         running_font = basefont;
2388
2389                         column += Changes::latexMarkChange(os, bparams, runningChange,
2390                                                            change, runparams);
2391                         runningChange = change;
2392                 }
2393
2394                 // do not output text which is marked deleted
2395                 // if change tracking output is disabled
2396                 if (!bparams.outputChanges && change.deleted()) {
2397                         continue;
2398                 }
2399
2400                 ++column;
2401
2402                 // Fully instantiated font
2403                 Font const font = getFont(bparams, i, outerfont);
2404
2405                 Font const last_font = running_font;
2406
2407                 // Do we need to close the previous font?
2408                 if (open_font &&
2409                     (font != running_font ||
2410                      font.language() != running_font.language()))
2411                 {
2412                         column += running_font.latexWriteEndChanges(
2413                                         os, bparams, runparams, basefont,
2414                                         (i == body_pos-1) ? basefont : font);
2415                         running_font = basefont;
2416                         open_font = false;
2417                 }
2418
2419                 string const running_lang = runparams.use_polyglossia ?
2420                         running_font.language()->polyglossia() : running_font.language()->babel();
2421                 // close babel's font environment before opening CJK.
2422                 string const lang_end_command = runparams.use_polyglossia ?
2423                         "\\end{$$lang}" : lyxrc.language_command_end;
2424                 if (!running_lang.empty() &&
2425                     font.language()->encoding()->package() == Encoding::CJK) {
2426                                 string end_tag = subst(lang_end_command,
2427                                                         "$$lang",
2428                                                         running_lang);
2429                                 os << from_ascii(end_tag);
2430                                 column += end_tag.length();
2431                 }
2432
2433                 // Switch file encoding if necessary (and allowed)
2434                 if (!runparams.pass_thru && !style.pass_thru &&
2435                     runparams.encoding->package() != Encoding::none &&
2436                     font.language()->encoding()->package() != Encoding::none) {
2437                         pair<bool, int> const enc_switch =
2438                                 switchEncoding(os.os(), bparams, runparams,
2439                                         *(font.language()->encoding()));
2440                         if (enc_switch.first) {
2441                                 column += enc_switch.second;
2442                                 runparams.encoding = font.language()->encoding();
2443                         }
2444                 }
2445
2446                 char_type const c = d->text_[i];
2447
2448                 // Do we need to change font?
2449                 if ((font != running_font ||
2450                      font.language() != running_font.language()) &&
2451                         i != body_pos - 1)
2452                 {
2453                         odocstringstream ods;
2454                         column += font.latexWriteStartChanges(ods, bparams,
2455                                                               runparams, basefont,
2456                                                               last_font);
2457                         running_font = font;
2458                         open_font = true;
2459                         docstring fontchange = ods.str();
2460                         // check whether the fontchange ends with a \\textcolor
2461                         // modifier and the text starts with a space (bug 4473)
2462                         docstring const last_modifier = rsplit(fontchange, '\\');
2463                         if (prefixIs(last_modifier, from_ascii("textcolor")) && c == ' ')
2464                                 os << fontchange << from_ascii("{}");
2465                         // check if the fontchange ends with a trailing blank
2466                         // (like "\small " (see bug 3382)
2467                         else if (suffixIs(fontchange, ' ') && c == ' ')
2468                                 os << fontchange.substr(0, fontchange.size() - 1)
2469                                    << from_ascii("{}");
2470                         else
2471                                 os << fontchange;
2472                 }
2473
2474                 // FIXME: think about end_pos implementation...
2475                 if (c == ' ' && i >= start_pos && (end_pos == -1 || i < end_pos)) {
2476                         // FIXME: integrate this case in latexSpecialChar
2477                         // Do not print the separation of the optional argument
2478                         // if style.pass_thru is false. This works because
2479                         // latexSpecialChar ignores spaces if
2480                         // style.pass_thru is false.
2481                         if (i != body_pos - 1) {
2482                                 if (d->simpleTeXBlanks(runparams, os,
2483                                                 i, column, font, style)) {
2484                                         // A surrogate pair was output. We
2485                                         // must not call latexSpecialChar
2486                                         // in this iteration, since it would output
2487                                         // the combining character again.
2488                                         ++i;
2489                                         continue;
2490                                 }
2491                         }
2492                 }
2493
2494                 OutputParams rp = runparams;
2495                 rp.free_spacing = style.free_spacing;
2496                 rp.local_font = &font;
2497                 rp.intitle = style.intitle;
2498
2499                 // Two major modes:  LaTeX or plain
2500                 // Handle here those cases common to both modes
2501                 // and then split to handle the two modes separately.
2502                 if (c == META_INSET) {
2503                         if (i >= start_pos && (end_pos == -1 || i < end_pos)) {
2504                                 d->latexInset(bparams, os, rp, running_font,
2505                                                 basefont, outerfont, open_font,
2506                                                 runningChange, style, i, column);
2507                         }
2508                 } else {
2509                         if (i >= start_pos && (end_pos == -1 || i < end_pos)) {
2510                                 try {
2511                                         d->latexSpecialChar(os, rp, running_font, runningChange,
2512                                                             style, i, end_pos, column);
2513                                 } catch (EncodingException & e) {
2514                                 if (runparams.dryrun) {
2515                                         os << "<" << _("LyX Warning: ")
2516                                            << _("uncodable character") << " '";
2517                                         os.put(c);
2518                                         os << "'>";
2519                                 } else {
2520                                         // add location information and throw again.
2521                                         e.par_id = id();
2522                                         e.pos = i;
2523                                         throw(e);
2524                                 }
2525                         }
2526                 }
2527                 }
2528
2529                 // Set the encoding to that returned from latexSpecialChar (see
2530                 // comment for encoding member in OutputParams.h)
2531                 runparams.encoding = rp.encoding;
2532         }
2533
2534         // If we have an open font definition, we have to close it
2535         if (open_font) {
2536 #ifdef FIXED_LANGUAGE_END_DETECTION
2537                 if (next_) {
2538                         running_font.latexWriteEndChanges(os, bparams,
2539                                         runparams, basefont,
2540                                         next_->getFont(bparams, 0, outerfont));
2541                 } else {
2542                         running_font.latexWriteEndChanges(os, bparams,
2543                                         runparams, basefont, basefont);
2544                 }
2545 #else
2546 //FIXME: For now we ALWAYS have to close the foreign font settings if they are
2547 //FIXME: there as we start another \selectlanguage with the next paragraph if
2548 //FIXME: we are in need of this. This should be fixed sometime (Jug)
2549                 running_font.latexWriteEndChanges(os, bparams, runparams,
2550                                 basefont, basefont);
2551 #endif
2552         }
2553
2554         column += Changes::latexMarkChange(os, bparams, runningChange,
2555                                            Change(Change::UNCHANGED), runparams);
2556
2557         // Needed if there is an optional argument but no contents.
2558         if (body_pos > 0 && body_pos == size()) {
2559                 os << "}]~";
2560         }
2561
2562         if (allowcust && d->endTeXParParams(bparams, os, runparams)
2563             && runparams.encoding != prev_encoding) {
2564                 runparams.encoding = prev_encoding;
2565                 if (!runparams.isFullUnicode())
2566                         os << setEncoding(prev_encoding->iconvName());
2567         }
2568
2569         LYXERR(Debug::LATEX, "Paragraph::latex... done " << this);
2570 }
2571
2572
2573 bool Paragraph::emptyTag() const
2574 {
2575         for (pos_type i = 0; i < size(); ++i) {
2576                 if (Inset const * inset = getInset(i)) {
2577                         InsetCode lyx_code = inset->lyxCode();
2578                         // FIXME testing like that is wrong. What is
2579                         // the intent?
2580                         if (lyx_code != TOC_CODE &&
2581                             lyx_code != INCLUDE_CODE &&
2582                             lyx_code != GRAPHICS_CODE &&
2583                             lyx_code != ERT_CODE &&
2584                             lyx_code != LISTINGS_CODE &&
2585                             lyx_code != FLOAT_CODE &&
2586                             lyx_code != TABULAR_CODE) {
2587                                 return false;
2588                         }
2589                 } else {
2590                         char_type c = d->text_[i];
2591                         if (c != ' ' && c != '\t')
2592                                 return false;
2593                 }
2594         }
2595         return true;
2596 }
2597
2598
2599 string Paragraph::getID(Buffer const & buf, OutputParams const & runparams)
2600         const
2601 {
2602         for (pos_type i = 0; i < size(); ++i) {
2603                 if (Inset const * inset = getInset(i)) {
2604                         InsetCode lyx_code = inset->lyxCode();
2605                         if (lyx_code == LABEL_CODE) {
2606                                 InsetLabel const * const il = static_cast<InsetLabel const *>(inset);
2607                                 docstring const & id = il->getParam("name");
2608                                 return "id='" + to_utf8(sgml::cleanID(buf, runparams, id)) + "'";
2609                         }
2610                 }
2611         }
2612         return string();
2613 }
2614
2615
2616 pos_type Paragraph::firstWordDocBook(odocstream & os, OutputParams const & runparams)
2617         const
2618 {
2619         pos_type i;
2620         for (i = 0; i < size(); ++i) {
2621                 if (Inset const * inset = getInset(i)) {
2622                         inset->docbook(os, runparams);
2623                 } else {
2624                         char_type c = d->text_[i];
2625                         if (c == ' ')
2626                                 break;
2627                         os << sgml::escapeChar(c);
2628                 }
2629         }
2630         return i;
2631 }
2632
2633
2634 pos_type Paragraph::firstWordLyXHTML(XHTMLStream & xs, OutputParams const & runparams)
2635         const
2636 {
2637         pos_type i;
2638         for (i = 0; i < size(); ++i) {
2639                 if (Inset const * inset = getInset(i)) {
2640                         inset->xhtml(xs, runparams);
2641                 } else {
2642                         char_type c = d->text_[i];
2643                         if (c == ' ')
2644                                 break;
2645                         xs << c;
2646                 }
2647         }
2648         return i;
2649 }
2650
2651
2652 bool Paragraph::Private::onlyText(Buffer const & buf, Font const & outerfont, pos_type initial) const
2653 {
2654         Font font_old;
2655         pos_type size = text_.size();
2656         for (pos_type i = initial; i < size; ++i) {
2657                 Font font = owner_->getFont(buf.params(), i, outerfont);
2658                 if (text_[i] == META_INSET)
2659                         return false;
2660                 if (i != initial && font != font_old)
2661                         return false;
2662                 font_old = font;
2663         }
2664
2665         return true;
2666 }
2667
2668
2669 void Paragraph::simpleDocBookOnePar(Buffer const & buf,
2670                                     odocstream & os,
2671                                     OutputParams const & runparams,
2672                                     Font const & outerfont,
2673                                     pos_type initial) const
2674 {
2675         bool emph_flag = false;
2676
2677         Layout const & style = *d->layout_;
2678         FontInfo font_old =
2679                 style.labeltype == LABEL_MANUAL ? style.labelfont : style.font;
2680
2681         if (style.pass_thru && !d->onlyText(buf, outerfont, initial))
2682                 os << "]]>";
2683
2684         // parsing main loop
2685         for (pos_type i = initial; i < size(); ++i) {
2686                 Font font = getFont(buf.params(), i, outerfont);
2687
2688                 // handle <emphasis> tag
2689                 if (font_old.emph() != font.fontInfo().emph()) {
2690                         if (font.fontInfo().emph() == FONT_ON) {
2691                                 os << "<emphasis>";
2692                                 emph_flag = true;
2693                         } else if (i != initial) {
2694                                 os << "</emphasis>";
2695                                 emph_flag = false;
2696                         }
2697                 }
2698
2699                 if (Inset const * inset = getInset(i)) {
2700                         inset->docbook(os, runparams);
2701                 } else {
2702                         char_type c = d->text_[i];
2703
2704                         if (style.pass_thru)
2705                                 os.put(c);
2706                         else
2707                                 os << sgml::escapeChar(c);
2708                 }
2709                 font_old = font.fontInfo();
2710         }
2711
2712         if (emph_flag) {
2713                 os << "</emphasis>";
2714         }
2715
2716         if (style.free_spacing)
2717                 os << '\n';
2718         if (style.pass_thru && !d->onlyText(buf, outerfont, initial))
2719                 os << "<![CDATA[";
2720 }
2721
2722
2723 docstring Paragraph::simpleLyXHTMLOnePar(Buffer const & buf,
2724                                     XHTMLStream & xs,
2725                                     OutputParams const & runparams,
2726                                     Font const & outerfont,
2727                                     pos_type initial) const
2728 {
2729         docstring retval;
2730
2731         bool emph_flag = false;
2732         bool bold_flag = false;
2733
2734         Layout const & style = *d->layout_;
2735
2736         xs.startParagraph(allowEmpty());
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         xs.endParagraph();
2818         return retval;
2819 }
2820
2821
2822 bool Paragraph::isHfill(pos_type pos) const
2823 {
2824         Inset const * inset = getInset(pos);
2825         return inset && (inset->lyxCode() == SPACE_CODE &&
2826                          inset->isStretchableSpace());
2827 }
2828
2829
2830 bool Paragraph::isNewline(pos_type pos) const
2831 {
2832         Inset const * inset = getInset(pos);
2833         return inset && inset->lyxCode() == NEWLINE_CODE;
2834 }
2835
2836
2837 bool Paragraph::isLineSeparator(pos_type pos) const
2838 {
2839         char_type const c = d->text_[pos];
2840         if (isLineSeparatorChar(c))
2841                 return true;
2842         Inset const * inset = getInset(pos);
2843         return inset && inset->isLineSeparator();
2844 }
2845
2846
2847 bool Paragraph::isWordSeparator(pos_type pos) const
2848 {
2849         if (pos == size())
2850                 return true;
2851         if (Inset const * inset = getInset(pos))
2852                 return !inset->isLetter();
2853         // if we have a hard hyphen (no en- or emdash) or apostrophe
2854         // we pass this to the spell checker
2855         // FIXME: this method is subject to change, visit
2856         // https://bugzilla.mozilla.org/show_bug.cgi?id=355178
2857         // to get an impression how complex this is.
2858         if (isHardHyphenOrApostrophe(pos))
2859                 return false;
2860         char_type const c = d->text_[pos];
2861         // We want to pass the escape chars to the spellchecker
2862         docstring const escape_chars = from_utf8(lyxrc.spellchecker_esc_chars);
2863         return !isLetterChar(c) && !isDigitASCII(c) && !contains(escape_chars, c);
2864 }
2865
2866
2867 bool Paragraph::isHardHyphenOrApostrophe(pos_type pos) const
2868 {
2869         pos_type const psize = size();
2870         if (pos >= psize)
2871                 return false;
2872         char_type const c = d->text_[pos];
2873         if (c != '-' && c != '\'')
2874                 return false;
2875         int nextpos = pos + 1;
2876         int prevpos = pos > 0 ? pos - 1 : 0;
2877         if ((nextpos == psize || isSpace(nextpos))
2878                 && (pos == 0 || isSpace(prevpos)))
2879                 return false;
2880         return c == '\''
2881                 || ((nextpos == psize || d->text_[nextpos] != '-')
2882                 && (pos == 0 || d->text_[prevpos] != '-'));
2883 }
2884
2885
2886 bool Paragraph::isSameSpellRange(pos_type pos1, pos_type pos2) const
2887 {
2888         return pos1 == pos2
2889                 || d->speller_state_.getRange(pos1) == d->speller_state_.getRange(pos2);
2890 }
2891
2892
2893 bool Paragraph::isChar(pos_type pos) const
2894 {
2895         if (Inset const * inset = getInset(pos))
2896                 return inset->isChar();
2897         char_type const c = d->text_[pos];
2898         return !isLetterChar(c) && !isDigitASCII(c) && !lyx::isSpace(c);
2899 }
2900
2901
2902 bool Paragraph::isSpace(pos_type pos) const
2903 {
2904         if (Inset const * inset = getInset(pos))
2905                 return inset->isSpace();
2906         char_type const c = d->text_[pos];
2907         return lyx::isSpace(c);
2908 }
2909
2910
2911 Language const *
2912 Paragraph::getParLanguage(BufferParams const & bparams) const
2913 {
2914         if (!empty())
2915                 return getFirstFontSettings(bparams).language();
2916         // FIXME: we should check the prev par as well (Lgb)
2917         return bparams.language;
2918 }
2919
2920
2921 bool Paragraph::isRTL(BufferParams const & bparams) const
2922 {
2923         return lyxrc.rtl_support
2924                 && getParLanguage(bparams)->rightToLeft()
2925                 && !inInset().getLayout().forceLTR();
2926 }
2927
2928
2929 void Paragraph::changeLanguage(BufferParams const & bparams,
2930                                Language const * from, Language const * to)
2931 {
2932         // change language including dummy font change at the end
2933         for (pos_type i = 0; i <= size(); ++i) {
2934                 Font font = getFontSettings(bparams, i);
2935                 if (font.language() == from) {
2936                         font.setLanguage(to);
2937                         setFont(i, font);
2938                         d->requestSpellCheck(i);
2939                 }
2940         }
2941 }
2942
2943
2944 bool Paragraph::isMultiLingual(BufferParams const & bparams) const
2945 {
2946         Language const * doc_language = bparams.language;
2947         FontList::const_iterator cit = d->fontlist_.begin();
2948         FontList::const_iterator end = d->fontlist_.end();
2949
2950         for (; cit != end; ++cit)
2951                 if (cit->font().language() != ignore_language &&
2952                     cit->font().language() != latex_language &&
2953                     cit->font().language() != doc_language)
2954                         return true;
2955         return false;
2956 }
2957
2958
2959 void Paragraph::getLanguages(std::set<Language const *> & languages) const
2960 {
2961         FontList::const_iterator cit = d->fontlist_.begin();
2962         FontList::const_iterator end = d->fontlist_.end();
2963
2964         for (; cit != end; ++cit) {
2965                 Language const * lang = cit->font().language();
2966                 if (lang != ignore_language &&
2967                     lang != latex_language)
2968                         languages.insert(lang);
2969         }
2970 }
2971
2972
2973 docstring Paragraph::asString(int options) const
2974 {
2975         return asString(0, size(), options);
2976 }
2977
2978
2979 docstring Paragraph::asString(pos_type beg, pos_type end, int options) const
2980 {
2981         odocstringstream os;
2982
2983         if (beg == 0
2984             && options & AS_STR_LABEL
2985             && !d->params_.labelString().empty())
2986                 os << d->params_.labelString() << ' ';
2987
2988         for (pos_type i = beg; i < end; ++i) {
2989                 if ((options & AS_STR_SKIPDELETE) && isDeleted(i))
2990                         continue;
2991                 char_type const c = d->text_[i];
2992                 if (isPrintable(c) || c == '\t'
2993                     || (c == '\n' && (options & AS_STR_NEWLINES)))
2994                         os.put(c);
2995                 else if (c == META_INSET && (options & AS_STR_INSETS)) {
2996                         getInset(i)->toString(os);
2997                         if (getInset(i)->asInsetMath())
2998                                 os << " ";
2999                 }
3000         }
3001
3002         return os.str();
3003 }
3004
3005
3006 void Paragraph::forToc(docstring & os, size_t maxlen) const
3007 {
3008         if (!d->params_.labelString().empty())
3009                 os += d->params_.labelString() + ' ';
3010         for (pos_type i = 0; i < size() && os.length() < maxlen; ++i) {
3011                 if (isDeleted(i))
3012                         continue;
3013                 char_type const c = d->text_[i];
3014                 if (isPrintable(c))
3015                         os += c;
3016                 else if (c == '\t' || c == '\n')
3017                         os += ' ';
3018                 else if (c == META_INSET)
3019                         getInset(i)->forToc(os, maxlen);
3020         }
3021 }
3022
3023
3024 docstring Paragraph::stringify(pos_type beg, pos_type end, int options, OutputParams & runparams) const
3025 {
3026         odocstringstream os;
3027
3028         if (beg == 0
3029                 && options & AS_STR_LABEL
3030                 && !d->params_.labelString().empty())
3031                 os << d->params_.labelString() << ' ';
3032
3033         for (pos_type i = beg; i < end; ++i) {
3034                 char_type const c = d->text_[i];
3035                 if (isPrintable(c) || c == '\t'
3036                     || (c == '\n' && (options & AS_STR_NEWLINES)))
3037                         os.put(c);
3038                 else if (c == META_INSET && (options & AS_STR_INSETS)) {
3039                         getInset(i)->plaintext(os, runparams);
3040                 }
3041         }
3042
3043         return os.str();
3044 }
3045
3046
3047 void Paragraph::setInsetOwner(Inset const * inset)
3048 {
3049         d->inset_owner_ = inset;
3050 }
3051
3052
3053 int Paragraph::id() const
3054 {
3055         return d->id_;
3056 }
3057
3058
3059 void Paragraph::setId(int id)
3060 {
3061         d->id_ = id;
3062 }
3063
3064
3065 Layout const & Paragraph::layout() const
3066 {
3067         return *d->layout_;
3068 }
3069
3070
3071 void Paragraph::setLayout(Layout const & layout)
3072 {
3073         d->layout_ = &layout;
3074 }
3075
3076
3077 void Paragraph::setDefaultLayout(DocumentClass const & tc)
3078 {
3079         setLayout(tc.defaultLayout());
3080 }
3081
3082
3083 void Paragraph::setPlainLayout(DocumentClass const & tc)
3084 {
3085         setLayout(tc.plainLayout());
3086 }
3087
3088
3089 void Paragraph::setPlainOrDefaultLayout(DocumentClass const & tclass)
3090 {
3091         if (usePlainLayout())
3092                 setPlainLayout(tclass);
3093         else
3094                 setDefaultLayout(tclass);
3095 }
3096
3097
3098 Inset const & Paragraph::inInset() const
3099 {
3100         LASSERT(d->inset_owner_, throw ExceptionMessage(BufferException,
3101                 _("Memory problem"), _("Paragraph not properly initialized")));
3102         return *d->inset_owner_;
3103 }
3104
3105
3106 ParagraphParameters & Paragraph::params()
3107 {
3108         return d->params_;
3109 }
3110
3111
3112 ParagraphParameters const & Paragraph::params() const
3113 {
3114         return d->params_;
3115 }
3116
3117
3118 bool Paragraph::isFreeSpacing() const
3119 {
3120         if (d->layout_->free_spacing)
3121                 return true;
3122         return d->inset_owner_ && d->inset_owner_->isFreeSpacing();
3123 }
3124
3125
3126 bool Paragraph::allowEmpty() const
3127 {
3128         if (d->layout_->keepempty)
3129                 return true;
3130         return d->inset_owner_ && d->inset_owner_->allowEmpty();
3131 }
3132
3133
3134 char_type Paragraph::transformChar(char_type c, pos_type pos) const
3135 {
3136         if (!Encodings::isArabicChar(c))
3137                 return c;
3138
3139         char_type prev_char = ' ';
3140         char_type next_char = ' ';
3141
3142         for (pos_type i = pos - 1; i >= 0; --i) {
3143                 char_type const par_char = d->text_[i];
3144                 if (!Encodings::isArabicComposeChar(par_char)) {
3145                         prev_char = par_char;
3146                         break;
3147                 }
3148         }
3149
3150         for (pos_type i = pos + 1, end = size(); i < end; ++i) {
3151                 char_type const par_char = d->text_[i];
3152                 if (!Encodings::isArabicComposeChar(par_char)) {
3153                         next_char = par_char;
3154                         break;
3155                 }
3156         }
3157
3158         if (Encodings::isArabicChar(next_char)) {
3159                 if (Encodings::isArabicChar(prev_char) &&
3160                         !Encodings::isArabicSpecialChar(prev_char))
3161                         return Encodings::transformChar(c, Encodings::FORM_MEDIAL);
3162                 else
3163                         return Encodings::transformChar(c, Encodings::FORM_INITIAL);
3164         } else {
3165                 if (Encodings::isArabicChar(prev_char) &&
3166                         !Encodings::isArabicSpecialChar(prev_char))
3167                         return Encodings::transformChar(c, Encodings::FORM_FINAL);
3168                 else
3169                         return Encodings::transformChar(c, Encodings::FORM_ISOLATED);
3170         }
3171 }
3172
3173
3174 int Paragraph::checkBiblio(Buffer const & buffer)
3175 {
3176         // FIXME From JS:
3177         // This is getting more and more a mess. ...We really should clean
3178         // up this bibitem issue for 1.6.
3179
3180         // Add bibitem insets if necessary
3181         if (d->layout_->labeltype != LABEL_BIBLIO)
3182                 return 0;
3183
3184         bool hasbibitem = !d->insetlist_.empty()
3185                 // Insist on it being in pos 0
3186                 && d->text_[0] == META_INSET
3187                 && d->insetlist_.begin()->inset->lyxCode() == BIBITEM_CODE;
3188
3189         bool track_changes = buffer.params().trackChanges;
3190
3191         docstring oldkey;
3192         docstring oldlabel;
3193
3194         // remove a bibitem in pos != 0
3195         // restore it later in pos 0 if necessary
3196         // (e.g. if a user inserts contents _before_ the item)
3197         // we're assuming there's only one of these, which there
3198         // should be.
3199         int erasedInsetPosition = -1;
3200         InsetList::iterator it = d->insetlist_.begin();
3201         InsetList::iterator end = d->insetlist_.end();
3202         for (; it != end; ++it)
3203                 if (it->inset->lyxCode() == BIBITEM_CODE
3204                       && it->pos > 0) {
3205                         InsetCommand * olditem = it->inset->asInsetCommand();
3206                         oldkey = olditem->getParam("key");
3207                         oldlabel = olditem->getParam("label");
3208                         erasedInsetPosition = it->pos;
3209                         eraseChar(erasedInsetPosition, track_changes);
3210                         break;
3211         }
3212
3213         // There was an InsetBibitem at the beginning, and we didn't
3214         // have to erase one.
3215         if (hasbibitem && erasedInsetPosition < 0)
3216                         return 0;
3217
3218         // There was an InsetBibitem at the beginning and we did have to
3219         // erase one. So we give its properties to the beginning inset.
3220         if (hasbibitem) {
3221                 InsetCommand * inset = d->insetlist_.begin()->inset->asInsetCommand();
3222                 if (!oldkey.empty())
3223                         inset->setParam("key", oldkey);
3224                 inset->setParam("label", oldlabel);
3225                 return -erasedInsetPosition;
3226         }
3227
3228         // There was no inset at the beginning, so we need to create one with
3229         // the key and label of the one we erased.
3230         InsetBibitem * inset =
3231                 new InsetBibitem(const_cast<Buffer *>(&buffer), InsetCommandParams(BIBITEM_CODE));
3232         // restore values of previously deleted item in this par.
3233         if (!oldkey.empty())
3234                 inset->setParam("key", oldkey);
3235         inset->setParam("label", oldlabel);
3236         insertInset(0, inset,
3237                     Change(track_changes ? Change::INSERTED : Change::UNCHANGED));
3238
3239         return 1;
3240 }
3241
3242
3243 void Paragraph::checkAuthors(AuthorList const & authorList)
3244 {
3245         d->changes_.checkAuthors(authorList);
3246 }
3247
3248
3249 bool Paragraph::isChanged(pos_type pos) const
3250 {
3251         return lookupChange(pos).changed();
3252 }
3253
3254
3255 bool Paragraph::isInserted(pos_type pos) const
3256 {
3257         return lookupChange(pos).inserted();
3258 }
3259
3260
3261 bool Paragraph::isDeleted(pos_type pos) const
3262 {
3263         return lookupChange(pos).deleted();
3264 }
3265
3266
3267 InsetList const & Paragraph::insetList() const
3268 {
3269         return d->insetlist_;
3270 }
3271
3272
3273 void Paragraph::setBuffer(Buffer & b)
3274 {
3275         d->insetlist_.setBuffer(b);
3276 }
3277
3278
3279 Inset * Paragraph::releaseInset(pos_type pos)
3280 {
3281         Inset * inset = d->insetlist_.release(pos);
3282         /// does not honour change tracking!
3283         eraseChar(pos, false);
3284         return inset;
3285 }
3286
3287
3288 Inset * Paragraph::getInset(pos_type pos)
3289 {
3290         return (pos < pos_type(d->text_.size()) && d->text_[pos] == META_INSET)
3291                  ? d->insetlist_.get(pos) : 0;
3292 }
3293
3294
3295 Inset const * Paragraph::getInset(pos_type pos) const
3296 {
3297         return (pos < pos_type(d->text_.size()) && d->text_[pos] == META_INSET)
3298                  ? d->insetlist_.get(pos) : 0;
3299 }
3300
3301
3302 void Paragraph::changeCase(BufferParams const & bparams, pos_type pos,
3303                 pos_type & right, TextCase action)
3304 {
3305         // process sequences of modified characters; in change
3306         // tracking mode, this approach results in much better
3307         // usability than changing case on a char-by-char basis
3308         docstring changes;
3309
3310         bool const trackChanges = bparams.trackChanges;
3311
3312         bool capitalize = true;
3313
3314         for (; pos < right; ++pos) {
3315                 char_type oldChar = d->text_[pos];
3316                 char_type newChar = oldChar;
3317
3318                 // ignore insets and don't play with deleted text!
3319                 if (oldChar != META_INSET && !isDeleted(pos)) {
3320                         switch (action) {
3321                                 case text_lowercase:
3322                                         newChar = lowercase(oldChar);
3323                                         break;
3324                                 case text_capitalization:
3325                                         if (capitalize) {
3326                                                 newChar = uppercase(oldChar);
3327                                                 capitalize = false;
3328                                         }
3329                                         break;
3330                                 case text_uppercase:
3331                                         newChar = uppercase(oldChar);
3332                                         break;
3333                         }
3334                 }
3335
3336                 if (isWordSeparator(pos) || isDeleted(pos)) {
3337                         // permit capitalization again
3338                         capitalize = true;
3339                 }
3340
3341                 if (oldChar != newChar) {
3342                         changes += newChar;
3343                         if (pos != right - 1)
3344                                 continue;
3345                         // step behind the changing area
3346                         pos++;
3347                 }
3348
3349                 int erasePos = pos - changes.size();
3350                 for (size_t i = 0; i < changes.size(); i++) {
3351                         insertChar(pos, changes[i],
3352                                    getFontSettings(bparams,
3353                                                    erasePos),
3354                                    trackChanges);
3355                         if (!eraseChar(erasePos, trackChanges)) {
3356                                 ++erasePos;
3357                                 ++pos; // advance
3358                                 ++right; // expand selection
3359                         }
3360                 }
3361                 changes.clear();
3362         }
3363 }
3364
3365
3366 int Paragraph::find(docstring const & str, bool cs, bool mw,
3367                 pos_type start_pos, bool del) const
3368 {
3369         pos_type pos = start_pos;
3370         int const strsize = str.length();
3371         int i = 0;
3372         pos_type const parsize = d->text_.size();
3373         for (i = 0; i < strsize && pos < parsize; ++i, ++pos) {
3374                 // Ignore "invisible" letters such as ligature breaks
3375                 // and hyphenation chars while searching
3376                 while (pos < parsize - 1 && isInset(pos)) {
3377                         odocstringstream os;
3378                         getInset(pos)->toString(os);
3379                         if (!getInset(pos)->isLetter() || !os.str().empty())
3380                                 break;
3381                         pos++;
3382                 }
3383                 if (cs && str[i] != d->text_[pos])
3384                         break;
3385                 if (!cs && uppercase(str[i]) != uppercase(d->text_[pos]))
3386                         break;
3387                 if (!del && isDeleted(pos))
3388                         break;
3389         }
3390
3391         if (i != strsize)
3392                 return 0;
3393
3394         // if necessary, check whether string matches word
3395         if (mw) {
3396                 if (start_pos > 0 && !isWordSeparator(start_pos - 1))
3397                         return 0;
3398                 if (pos < parsize
3399                         && !isWordSeparator(pos))
3400                         return 0;
3401         }
3402
3403         return pos - start_pos;
3404 }
3405
3406
3407 char_type Paragraph::getChar(pos_type pos) const
3408 {
3409         return d->text_[pos];
3410 }
3411
3412
3413 pos_type Paragraph::size() const
3414 {
3415         return d->text_.size();
3416 }
3417
3418
3419 bool Paragraph::empty() const
3420 {
3421         return d->text_.empty();
3422 }
3423
3424
3425 bool Paragraph::isInset(pos_type pos) const
3426 {
3427         return d->text_[pos] == META_INSET;
3428 }
3429
3430
3431 bool Paragraph::isSeparator(pos_type pos) const
3432 {
3433         //FIXME: Are we sure this can be the only separator?
3434         return d->text_[pos] == ' ';
3435 }
3436
3437
3438 void Paragraph::deregisterWords()
3439 {
3440         Private::LangWordsMap::const_iterator itl = d->words_.begin();
3441         Private::LangWordsMap::const_iterator ite = d->words_.end();
3442         for (; itl != ite; ++itl) {
3443                 WordList * wl = theWordList(itl->first);
3444                 Private::Words::const_iterator it = (itl->second).begin();
3445                 Private::Words::const_iterator et = (itl->second).end();
3446                 for (; it != et; ++it)
3447                         wl->remove(*it);
3448         }
3449         d->words_.clear();
3450 }
3451
3452
3453 void Paragraph::locateWord(pos_type & from, pos_type & to,
3454         word_location const loc) const
3455 {
3456         switch (loc) {
3457         case WHOLE_WORD_STRICT:
3458                 if (from == 0 || from == size()
3459                     || isWordSeparator(from)
3460                     || isWordSeparator(from - 1)) {
3461                         to = from;
3462                         return;
3463                 }
3464                 // no break here, we go to the next
3465
3466         case WHOLE_WORD:
3467                 // If we are already at the beginning of a word, do nothing
3468                 if (!from || isWordSeparator(from - 1))
3469                         break;
3470                 // no break here, we go to the next
3471
3472         case PREVIOUS_WORD:
3473                 // always move the cursor to the beginning of previous word
3474                 while (from && !isWordSeparator(from - 1))
3475                         --from;
3476                 break;
3477         case NEXT_WORD:
3478                 LYXERR0("Paragraph::locateWord: NEXT_WORD not implemented yet");
3479                 break;
3480         case PARTIAL_WORD:
3481                 // no need to move the 'from' cursor
3482                 break;
3483         }
3484         to = from;
3485         while (to < size() && !isWordSeparator(to))
3486                 ++to;
3487 }
3488
3489
3490 void Paragraph::collectWords()
3491 {
3492         // This is the value that needs to be exposed in the preferences
3493         // to resolve bug #6760.
3494         static int minlength = 6;
3495         pos_type n = size();
3496         for (pos_type pos = 0; pos < n; ++pos) {
3497                 if (isWordSeparator(pos))
3498                         continue;
3499                 pos_type from = pos;
3500                 locateWord(from, pos, WHOLE_WORD);
3501                 if (pos - from >= minlength) {
3502                         docstring word = asString(from, pos, AS_STR_NONE);
3503                         FontList::const_iterator cit = d->fontlist_.fontIterator(pos);
3504                         if (cit == d->fontlist_.end())
3505                                 return;
3506                         Language const * lang = cit->font().language();
3507                         d->words_[*lang].insert(word);
3508                 }
3509         }
3510 }
3511
3512
3513 void Paragraph::registerWords()
3514 {
3515         Private::LangWordsMap::const_iterator itl = d->words_.begin();
3516         Private::LangWordsMap::const_iterator ite = d->words_.end();
3517         for (; itl != ite; ++itl) {
3518                 WordList * wl = theWordList(itl->first);
3519                 Private::Words::const_iterator it = (itl->second).begin();
3520                 Private::Words::const_iterator et = (itl->second).end();
3521                 for (; it != et; ++it)
3522                         wl->insert(*it);
3523         }
3524 }
3525
3526
3527 void Paragraph::updateWords()
3528 {
3529         deregisterWords();
3530         collectWords();
3531         registerWords();
3532 }
3533
3534
3535 void Paragraph::Private::appendSkipPosition(SkipPositions & skips, pos_type const pos) const
3536 {
3537         SkipPositionsIterator begin = skips.begin();
3538         SkipPositions::iterator end = skips.end();
3539         if (pos > 0 && begin < end) {
3540                 --end;
3541                 if (end->last == pos - 1) {
3542                         end->last = pos;
3543                         return;
3544                 }
3545         }
3546         skips.insert(end, FontSpan(pos, pos));
3547 }
3548
3549
3550 Language * Paragraph::Private::locateSpellRange(
3551         pos_type & from, pos_type & to,
3552         SkipPositions & skips) const
3553 {
3554         // skip leading white space
3555         while (from < to && owner_->isWordSeparator(from))
3556                 ++from;
3557         // don't check empty range
3558         if (from >= to)
3559                 return 0;
3560         // get current language
3561         Language * lang = getSpellLanguage(from);
3562         pos_type last = from;
3563         bool samelang = true;
3564         bool sameinset = true;
3565         while (last < to && samelang && sameinset) {
3566                 // hop to end of word
3567                 while (last < to && !owner_->isWordSeparator(last)) {
3568                         if (owner_->getInset(last)) {
3569                                 appendSkipPosition(skips, last);
3570                         } else if (owner_->isDeleted(last)) {
3571                                 appendSkipPosition(skips, last);
3572                         }
3573                         ++last;
3574                 }
3575                 // hop to next word while checking for insets
3576                 while (sameinset && last < to && owner_->isWordSeparator(last)) {
3577                         if (Inset const * inset = owner_->getInset(last))
3578                                 sameinset = inset->isChar() && inset->isLetter();
3579                         if (sameinset && owner_->isDeleted(last)) {
3580                                 appendSkipPosition(skips, last);
3581                         }
3582                         if (sameinset)
3583                                 last++;
3584                 }
3585                 if (sameinset && last < to) {
3586                         // now check for language change
3587                         samelang = lang == getSpellLanguage(last);
3588                 }
3589         }
3590         // if language change detected backstep is needed
3591         if (!samelang)
3592                 --last;
3593         to = last;
3594         return lang;
3595 }
3596
3597
3598 Language * Paragraph::Private::getSpellLanguage(pos_type const from) const
3599 {
3600         Language * lang =
3601                 const_cast<Language *>(owner_->getFontSettings(
3602                         inset_owner_->buffer().params(), from).language());
3603         if (lang == inset_owner_->buffer().params().language
3604                 && !lyxrc.spellchecker_alt_lang.empty()) {
3605                 string lang_code;
3606                 string const lang_variety =
3607                         split(lyxrc.spellchecker_alt_lang, lang_code, '-');
3608                 lang->setCode(lang_code);
3609                 lang->setVariety(lang_variety);
3610         }
3611         return lang;
3612 }
3613
3614
3615 void Paragraph::requestSpellCheck(pos_type pos)
3616 {
3617         d->requestSpellCheck(pos);
3618 }
3619
3620
3621 bool Paragraph::needsSpellCheck() const
3622 {
3623         SpellChecker::ChangeNumber speller_change_number = 0;
3624         if (theSpellChecker())
3625                 speller_change_number = theSpellChecker()->changeNumber();
3626         if (speller_change_number > d->speller_state_.currentChangeNumber()) {
3627                 d->speller_state_.needsCompleteRefresh(speller_change_number);
3628         }
3629         return d->needsSpellCheck();
3630 }
3631
3632
3633 bool Paragraph::Private::ignoreWord(docstring const & word) const
3634 {
3635         // Ignore words with digits
3636         // FIXME: make this customizable
3637         // (note that some checkers ignore words with digits by default)
3638         docstring::const_iterator cit = word.begin();
3639         docstring::const_iterator const end = word.end();
3640         for (; cit != end; ++cit) {
3641                 if (isNumber((*cit)))
3642                         return true;
3643         }
3644         return false;
3645 }
3646
3647
3648 SpellChecker::Result Paragraph::spellCheck(pos_type & from, pos_type & to,
3649         WordLangTuple & wl, docstring_list & suggestions,
3650         bool do_suggestion, bool check_learned) const
3651 {
3652         SpellChecker::Result result = SpellChecker::WORD_OK;
3653         SpellChecker * speller = theSpellChecker();
3654         if (!speller)
3655                 return result;
3656
3657         if (!d->layout_->spellcheck || !inInset().allowSpellCheck())
3658                 return result;
3659
3660         locateWord(from, to, WHOLE_WORD);
3661         if (from == to || from >= size())
3662                 return result;
3663
3664         docstring word = asString(from, to, AS_STR_INSETS | AS_STR_SKIPDELETE);
3665         Language * lang = d->getSpellLanguage(from);
3666
3667         wl = WordLangTuple(word, lang);
3668
3669         if (!word.size())
3670                 return result;
3671
3672         if (needsSpellCheck() || check_learned) {
3673                 pos_type end = to;
3674                 if (!d->ignoreWord(word)) {
3675                         bool const trailing_dot = to < size() && d->text_[to] == '.';
3676                         result = speller->check(wl);
3677                         if (SpellChecker::misspelled(result) && trailing_dot) {
3678                                 wl = WordLangTuple(word.append(from_ascii(".")), lang);
3679                                 result = speller->check(wl);
3680                                 if (!SpellChecker::misspelled(result)) {
3681                                         LYXERR(Debug::GUI, "misspelled word is correct with dot: \"" <<
3682                                            word << "\" [" <<
3683                                            from << ".." << to << "]");
3684                                 } else {
3685                                         // spell check with dot appended failed too
3686                                         // restore original word/lang value
3687                                         word = asString(from, to, AS_STR_INSETS | AS_STR_SKIPDELETE);
3688                                         wl = WordLangTuple(word, lang);
3689                                 }
3690                         }
3691                 }
3692                 if (!SpellChecker::misspelled(result)) {
3693                         // area up to the begin of the next word is not misspelled
3694                         while (end < size() && isWordSeparator(end))
3695                                 ++end;
3696                 }
3697                 d->setMisspelled(from, end, result);
3698         } else {
3699                 result = d->speller_state_.getState(from);
3700         }
3701
3702         if (do_suggestion)
3703                 suggestions.clear();
3704
3705         if (SpellChecker::misspelled(result)) {
3706                 LYXERR(Debug::GUI, "misspelled word: \"" <<
3707                            word << "\" [" <<
3708                            from << ".." << to << "]");
3709                 if (do_suggestion)
3710                         speller->suggest(wl, suggestions);
3711         }
3712         return result;
3713 }
3714
3715
3716 void Paragraph::Private::markMisspelledWords(
3717         pos_type const & first, pos_type const & last,
3718         SpellChecker::Result result,
3719         docstring const & word,
3720         SkipPositions const & skips)
3721 {
3722         if (!SpellChecker::misspelled(result)) {
3723                 setMisspelled(first, last, SpellChecker::WORD_OK);
3724                 return;
3725         }
3726         int snext = first;
3727         SpellChecker * speller = theSpellChecker();
3728         // locate and enumerate the error positions
3729         int nerrors = speller->numMisspelledWords();
3730         int numskipped = 0;
3731         SkipPositionsIterator it = skips.begin();
3732         SkipPositionsIterator et = skips.end();
3733         for (int index = 0; index < nerrors; ++index) {
3734                 int wstart;
3735                 int wlen = 0;
3736                 speller->misspelledWord(index, wstart, wlen);
3737                 /// should not happen if speller supports range checks
3738                 if (!wlen) continue;
3739                 docstring const misspelled = word.substr(wstart, wlen);
3740                 wstart += first + numskipped;
3741                 if (snext < wstart) {
3742                         /// mark the range of correct spelling
3743                         numskipped += countSkips(it, et, wstart);
3744                         setMisspelled(snext,
3745                                 wstart - 1, SpellChecker::WORD_OK);
3746                 }
3747                 snext = wstart + wlen;
3748                 numskipped += countSkips(it, et, snext);
3749                 /// mark the range of misspelling
3750                 setMisspelled(wstart, snext, result);
3751                 LYXERR(Debug::GUI, "misspelled word: \"" <<
3752                            misspelled << "\" [" <<
3753                            wstart << ".." << (snext-1) << "]");
3754                 ++snext;
3755         }
3756         if (snext <= last) {
3757                 /// mark the range of correct spelling at end
3758                 setMisspelled(snext, last, SpellChecker::WORD_OK);
3759         }
3760 }
3761
3762
3763 void Paragraph::spellCheck() const
3764 {
3765         SpellChecker * speller = theSpellChecker();
3766         if (!speller || !size() ||!needsSpellCheck())
3767                 return;
3768         pos_type start;
3769         pos_type endpos;
3770         d->rangeOfSpellCheck(start, endpos);
3771         if (speller->canCheckParagraph()) {
3772                 // loop until we leave the range
3773                 for (pos_type first = start; first < endpos; ) {
3774                         pos_type last = endpos;
3775                         Private::SkipPositions skips;
3776                         Language * lang = d->locateSpellRange(first, last, skips);
3777                         if (first >= endpos)
3778                                 break;
3779                         // start the spell checker on the unit of meaning
3780                         docstring word = asString(first, last, AS_STR_INSETS + AS_STR_SKIPDELETE);
3781                         WordLangTuple wl = WordLangTuple(word, lang);
3782                         SpellChecker::Result result = word.size() ?
3783                                 speller->check(wl) : SpellChecker::WORD_OK;
3784                         d->markMisspelledWords(first, last, result, word, skips);
3785                         first = ++last;
3786                 }
3787         } else {
3788                 static docstring_list suggestions;
3789                 pos_type to = endpos;
3790                 while (start < endpos) {
3791                         WordLangTuple wl;
3792                         spellCheck(start, to, wl, suggestions, false);
3793                         start = to + 1;
3794                 }
3795         }
3796         d->readySpellCheck();
3797 }
3798
3799
3800 bool Paragraph::isMisspelled(pos_type pos, bool check_boundary) const
3801 {
3802         bool result = SpellChecker::misspelled(d->speller_state_.getState(pos));
3803         if (result || pos <= 0 || pos > size())
3804                 return result;
3805         if (check_boundary && (pos == size() || isWordSeparator(pos)))
3806                 result = SpellChecker::misspelled(d->speller_state_.getState(pos - 1));
3807         return result;
3808 }
3809
3810
3811 string Paragraph::magicLabel() const
3812 {
3813         stringstream ss;
3814         ss << "magicparlabel-" << id();
3815         return ss.str();
3816 }
3817
3818
3819 } // namespace lyx