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                          latex[latex.length() - 1] != '-') {
1272                         // Prevent eating of a following
1273                         // space or command corruption by
1274                         // following characters
1275                         column += latex.length() + 1;
1276                         os << latex << "{}";
1277                 } else {
1278                         column += latex.length() - 1;
1279                         os << latex;
1280                 }
1281                 break;
1282         }
1283 }
1284
1285
1286 bool Paragraph::Private::latexSpecialT1(char_type const c, otexstream & os,
1287         pos_type i, unsigned int & column)
1288 {
1289         switch (c) {
1290         case '>':
1291         case '<':
1292                 os.put(c);
1293                 // In T1 encoding, these characters exist
1294                 // but we should avoid ligatures
1295                 if (i + 1 >= int(text_.size()) || text_[i + 1] != c)
1296                         return true;
1297                 os << "\\textcompwordmark{}";
1298                 column += 19;
1299                 return true;
1300         case '|':
1301                 os.put(c);
1302                 return true;
1303         case '\"':
1304                 // soul.sty breaks with \char`\"
1305                 os << "\\textquotedbl{}";
1306                 column += 14;
1307                 return true;
1308         default:
1309                 return false;
1310         }
1311 }
1312
1313
1314 bool Paragraph::Private::latexSpecialTypewriter(char_type const c, otexstream & os,
1315         pos_type i, unsigned int & column)
1316 {
1317         switch (c) {
1318         case '-':
1319                 // within \ttfamily, "--" is merged to "-" (no endash)
1320                 // so we avoid this rather irritating ligature
1321                 if (i + 1 < int(text_.size()) && text_[i + 1] == '-') {
1322                         os << "-{}";
1323                         column += 2;
1324                 } else
1325                         os << '-';
1326                 return true;
1327
1328         // everything else has to be checked separately
1329         // (depending on the encoding)
1330         default:
1331                 return false;
1332         }
1333 }
1334
1335
1336 /// \param end_pos
1337 ///   If [start_pos, end_pos) does not include entirely the special phrase, then
1338 ///   do not apply the macro transformation.
1339 bool Paragraph::Private::latexSpecialPhrase(otexstream & os, pos_type & i, pos_type end_pos,
1340         unsigned int & column, OutputParams const & runparams)
1341 {
1342         // FIXME: if we have "LaTeX" with a font
1343         // change in the middle (before the 'T', then
1344         // the "TeX" part is still special cased.
1345         // Really we should only operate this on
1346         // "words" for some definition of word
1347
1348         for (size_t pnr = 0; pnr < phrases_nr; ++pnr) {
1349                 if (!isTextAt(special_phrases[pnr].phrase, i)
1350                     || (end_pos != -1 && i + int(special_phrases[pnr].phrase.size()) > end_pos))
1351                         continue;
1352                 if (runparams.moving_arg)
1353                         os << "\\protect";
1354                 os << special_phrases[pnr].macro;
1355                 i += special_phrases[pnr].phrase.length() - 1;
1356                 column += special_phrases[pnr].macro.length() - 1;
1357                 return true;
1358         }
1359         return false;
1360 }
1361
1362
1363 void Paragraph::Private::validate(LaTeXFeatures & features) const
1364 {
1365         if (layout_->inpreamble && inset_owner_) {
1366                 bool const is_command = layout_->latextype == LATEX_COMMAND;
1367                 Buffer const & buf = inset_owner_->buffer();
1368                 BufferParams const & bp = buf.params();
1369                 Font f;
1370                 TexRow texrow;
1371                 // Using a string stream here circumvents the encoding
1372                 // switching machinery of odocstream. Therefore the
1373                 // output is wrong if this paragraph contains content
1374                 // that needs to switch encoding.
1375                 odocstringstream ods;
1376                 otexstream os(ods, texrow);
1377                 if (is_command) {
1378                         os << '\\' << from_ascii(layout_->latexname());
1379                         // we have to provide all the optional arguments here, even though
1380                         // the last one is the only one we care about.
1381                         // Separate handling of optional argument inset.
1382                         if (layout_->optargs != 0 || layout_->reqargs != 0)
1383                                 latexArgInsets(*owner_, os, features.runparams(),
1384                                         layout_->reqargs, layout_->optargs);
1385                         else
1386                                 os << from_ascii(layout_->latexparam());
1387                 }
1388                 docstring::size_type const length = ods.str().length();
1389                 // this will output "{" at the beginning, but not at the end
1390                 owner_->latex(bp, f, os, features.runparams(), 0, -1, true);
1391                 if (ods.str().length() > length) {
1392                         if (is_command)
1393                                 ods << '}';
1394                         string const snippet = to_utf8(ods.str());
1395                         features.addPreambleSnippet(snippet);
1396                 }
1397         }
1398
1399         if (features.runparams().flavor == OutputParams::HTML
1400             && layout_->htmltitle()) {
1401                 features.setHTMLTitle(owner_->asString(AS_STR_INSETS));
1402         }
1403
1404         // check the params.
1405         if (!params_.spacing().isDefault())
1406                 features.require("setspace");
1407
1408         // then the layouts
1409         features.useLayout(layout_->name());
1410
1411         // then the fonts
1412         fontlist_.validate(features);
1413
1414         // then the indentation
1415         if (!params_.leftIndent().zero())
1416                 features.require("ParagraphLeftIndent");
1417
1418         // then the insets
1419         InsetList::const_iterator icit = insetlist_.begin();
1420         InsetList::const_iterator iend = insetlist_.end();
1421         for (; icit != iend; ++icit) {
1422                 if (icit->inset) {
1423                         icit->inset->validate(features);
1424                         if (layout_->needprotect &&
1425                             icit->inset->lyxCode() == FOOT_CODE)
1426                                 features.require("NeedLyXFootnoteCode");
1427                 }
1428         }
1429
1430         // then the contents
1431         for (pos_type i = 0; i < int(text_.size()) ; ++i) {
1432                 for (size_t pnr = 0; pnr < phrases_nr; ++pnr) {
1433                         if (!special_phrases[pnr].builtin
1434                             && isTextAt(special_phrases[pnr].phrase, i)) {
1435                                 features.require(special_phrases[pnr].phrase);
1436                                 break;
1437                         }
1438                 }
1439                 Encodings::validate(text_[i], features);
1440         }
1441 }
1442
1443 /////////////////////////////////////////////////////////////////////
1444 //
1445 // Paragraph
1446 //
1447 /////////////////////////////////////////////////////////////////////
1448
1449 namespace {
1450         Layout const emptyParagraphLayout;
1451 }
1452
1453 Paragraph::Paragraph()
1454         : d(new Paragraph::Private(this, emptyParagraphLayout))
1455 {
1456         itemdepth = 0;
1457         d->params_.clear();
1458 }
1459
1460
1461 Paragraph::Paragraph(Paragraph const & par)
1462         : itemdepth(par.itemdepth),
1463         d(new Paragraph::Private(*par.d, this))
1464 {
1465         registerWords();
1466 }
1467
1468
1469 Paragraph::Paragraph(Paragraph const & par, pos_type beg, pos_type end)
1470         : itemdepth(par.itemdepth),
1471         d(new Paragraph::Private(*par.d, this, beg, end))
1472 {
1473         registerWords();
1474 }
1475
1476
1477 Paragraph & Paragraph::operator=(Paragraph const & par)
1478 {
1479         // needed as we will destroy the private part before copying it
1480         if (&par != this) {
1481                 itemdepth = par.itemdepth;
1482
1483                 deregisterWords();
1484                 delete d;
1485                 d = new Private(*par.d, this);
1486                 registerWords();
1487         }
1488         return *this;
1489 }
1490
1491
1492 Paragraph::~Paragraph()
1493 {
1494         deregisterWords();
1495         delete d;
1496 }
1497
1498
1499 namespace {
1500
1501 // this shall be called just before every "os << ..." action.
1502 void flushString(ostream & os, docstring & s)
1503 {
1504         os << to_utf8(s);
1505         s.erase();
1506 }
1507
1508 }
1509
1510
1511 void Paragraph::write(ostream & os, BufferParams const & bparams,
1512         depth_type & dth) const
1513 {
1514         // The beginning or end of a deeper (i.e. nested) area?
1515         if (dth != d->params_.depth()) {
1516                 if (d->params_.depth() > dth) {
1517                         while (d->params_.depth() > dth) {
1518                                 os << "\n\\begin_deeper";
1519                                 ++dth;
1520                         }
1521                 } else {
1522                         while (d->params_.depth() < dth) {
1523                                 os << "\n\\end_deeper";
1524                                 --dth;
1525                         }
1526                 }
1527         }
1528
1529         // First write the layout
1530         os << "\n\\begin_layout " << to_utf8(d->layout_->name()) << '\n';
1531
1532         d->params_.write(os);
1533
1534         Font font1(inherit_font, bparams.language);
1535
1536         Change running_change = Change(Change::UNCHANGED);
1537
1538         // this string is used as a buffer to avoid repetitive calls
1539         // to to_utf8(), which turn out to be expensive (JMarc)
1540         docstring write_buffer;
1541
1542         int column = 0;
1543         for (pos_type i = 0; i <= size(); ++i) {
1544
1545                 Change const change = lookupChange(i);
1546                 if (change != running_change)
1547                         flushString(os, write_buffer);
1548                 Changes::lyxMarkChange(os, bparams, column, running_change, change);
1549                 running_change = change;
1550
1551                 if (i == size())
1552                         break;
1553
1554                 // Write font changes
1555                 Font font2 = getFontSettings(bparams, i);
1556                 if (font2 != font1) {
1557                         flushString(os, write_buffer);
1558                         font2.lyxWriteChanges(font1, os);
1559                         column = 0;
1560                         font1 = font2;
1561                 }
1562
1563                 char_type const c = d->text_[i];
1564                 switch (c) {
1565                 case META_INSET:
1566                         if (Inset const * inset = getInset(i)) {
1567                                 flushString(os, write_buffer);
1568                                 if (inset->directWrite()) {
1569                                         // international char, let it write
1570                                         // code directly so it's shorter in
1571                                         // the file
1572                                         inset->write(os);
1573                                 } else {
1574                                         if (i)
1575                                                 os << '\n';
1576                                         os << "\\begin_inset ";
1577                                         inset->write(os);
1578                                         os << "\n\\end_inset\n\n";
1579                                         column = 0;
1580                                 }
1581                         }
1582                         break;
1583                 case '\\':
1584                         flushString(os, write_buffer);
1585                         os << "\n\\backslash\n";
1586                         column = 0;
1587                         break;
1588                 case '.':
1589                         flushString(os, write_buffer);
1590                         if (i + 1 < size() && d->text_[i + 1] == ' ') {
1591                                 os << ".\n";
1592                                 column = 0;
1593                         } else
1594                                 os << '.';
1595                         break;
1596                 default:
1597                         if ((column > 70 && c == ' ')
1598                             || column > 79) {
1599                                 flushString(os, write_buffer);
1600                                 os << '\n';
1601                                 column = 0;
1602                         }
1603                         // this check is to amend a bug. LyX sometimes
1604                         // inserts '\0' this could cause problems.
1605                         if (c != '\0')
1606                                 write_buffer.push_back(c);
1607                         else
1608                                 LYXERR0("NUL char in structure.");
1609                         ++column;
1610                         break;
1611                 }
1612         }
1613
1614         flushString(os, write_buffer);
1615         os << "\n\\end_layout\n";
1616 }
1617
1618
1619 void Paragraph::validate(LaTeXFeatures & features) const
1620 {
1621         d->validate(features);
1622 }
1623
1624
1625 void Paragraph::insert(pos_type start, docstring const & str,
1626                        Font const & font, Change const & change)
1627 {
1628         for (size_t i = 0, n = str.size(); i != n ; ++i)
1629                 insertChar(start + i, str[i], font, change);
1630 }
1631
1632
1633 void Paragraph::appendChar(char_type c, Font const & font,
1634                 Change const & change)
1635 {
1636         // track change
1637         d->changes_.insert(change, d->text_.size());
1638         // when appending characters, no need to update tables
1639         d->text_.push_back(c);
1640         setFont(d->text_.size() - 1, font);
1641         d->requestSpellCheck(d->text_.size() - 1);
1642 }
1643
1644
1645 void Paragraph::appendString(docstring const & s, Font const & font,
1646                 Change const & change)
1647 {
1648         pos_type end = s.size();
1649         size_t oldsize = d->text_.size();
1650         size_t newsize = oldsize + end;
1651         size_t capacity = d->text_.capacity();
1652         if (newsize >= capacity)
1653                 d->text_.reserve(max(capacity + 100, newsize));
1654
1655         // when appending characters, no need to update tables
1656         d->text_.append(s);
1657
1658         // FIXME: Optimize this!
1659         for (size_t i = oldsize; i != newsize; ++i) {
1660                 // track change
1661                 d->changes_.insert(change, i);
1662                 d->requestSpellCheck(i);
1663         }
1664         d->fontlist_.set(oldsize, font);
1665         d->fontlist_.set(newsize - 1, font);
1666 }
1667
1668
1669 void Paragraph::insertChar(pos_type pos, char_type c,
1670                            bool trackChanges)
1671 {
1672         d->insertChar(pos, c, Change(trackChanges ?
1673                            Change::INSERTED : Change::UNCHANGED));
1674 }
1675
1676
1677 void Paragraph::insertChar(pos_type pos, char_type c,
1678                            Font const & font, bool trackChanges)
1679 {
1680         d->insertChar(pos, c, Change(trackChanges ?
1681                            Change::INSERTED : Change::UNCHANGED));
1682         setFont(pos, font);
1683 }
1684
1685
1686 void Paragraph::insertChar(pos_type pos, char_type c,
1687                            Font const & font, Change const & change)
1688 {
1689         d->insertChar(pos, c, change);
1690         setFont(pos, font);
1691 }
1692
1693
1694 bool Paragraph::insertInset(pos_type pos, Inset * inset,
1695                             Font const & font, Change const & change)
1696 {
1697         bool const success = insertInset(pos, inset, change);
1698         // Set the font/language of the inset...
1699         setFont(pos, font);
1700         return success;
1701 }
1702
1703
1704 void Paragraph::resetFonts(Font const & font)
1705 {
1706         d->fontlist_.clear();
1707         d->fontlist_.set(0, font);
1708         d->fontlist_.set(d->text_.size() - 1, font);
1709 }
1710
1711 // Gets uninstantiated font setting at position.
1712 Font const & Paragraph::getFontSettings(BufferParams const & bparams,
1713                                          pos_type pos) const
1714 {
1715         if (pos > size()) {
1716                 LYXERR0("pos: " << pos << " size: " << size());
1717                 LASSERT(pos <= size(), /**/);
1718         }
1719
1720         FontList::const_iterator cit = d->fontlist_.fontIterator(pos);
1721         if (cit != d->fontlist_.end())
1722                 return cit->font();
1723
1724         if (pos == size() && !empty())
1725                 return getFontSettings(bparams, pos - 1);
1726
1727         // Optimisation: avoid a full font instantiation if there is no
1728         // language change from previous call.
1729         static Font previous_font;
1730         static Language const * previous_lang = 0;
1731         Language const * lang = getParLanguage(bparams);
1732         if (lang != previous_lang) {
1733                 previous_lang = lang;
1734                 previous_font = Font(inherit_font, lang);
1735         }
1736         return previous_font;
1737 }
1738
1739
1740 FontSpan Paragraph::fontSpan(pos_type pos) const
1741 {
1742         LASSERT(pos <= size(), /**/);
1743         pos_type start = 0;
1744
1745         FontList::const_iterator cit = d->fontlist_.begin();
1746         FontList::const_iterator end = d->fontlist_.end();
1747         for (; cit != end; ++cit) {
1748                 if (cit->pos() >= pos) {
1749                         if (pos >= beginOfBody())
1750                                 return FontSpan(max(start, beginOfBody()),
1751                                                 cit->pos());
1752                         else
1753                                 return FontSpan(start,
1754                                                 min(beginOfBody() - 1,
1755                                                          cit->pos()));
1756                 }
1757                 start = cit->pos() + 1;
1758         }
1759
1760         // This should not happen, but if so, we take no chances.
1761         // LYXERR0("Paragraph::getEndPosOfFontSpan: This should not happen!");
1762         return FontSpan(pos, pos);
1763 }
1764
1765
1766 // Gets uninstantiated font setting at position 0
1767 Font const & Paragraph::getFirstFontSettings(BufferParams const & bparams) const
1768 {
1769         if (!empty() && !d->fontlist_.empty())
1770                 return d->fontlist_.begin()->font();
1771
1772         // Optimisation: avoid a full font instantiation if there is no
1773         // language change from previous call.
1774         static Font previous_font;
1775         static Language const * previous_lang = 0;
1776         if (bparams.language != previous_lang) {
1777                 previous_lang = bparams.language;
1778                 previous_font = Font(inherit_font, bparams.language);
1779         }
1780
1781         return previous_font;
1782 }
1783
1784
1785 // Gets the fully instantiated font at a given position in a paragraph
1786 // This is basically the same function as Text::GetFont() in text2.cpp.
1787 // The difference is that this one is used for generating the LaTeX file,
1788 // and thus cosmetic "improvements" are disallowed: This has to deliver
1789 // the true picture of the buffer. (Asger)
1790 Font const Paragraph::getFont(BufferParams const & bparams, pos_type pos,
1791                                  Font const & outerfont) const
1792 {
1793         LASSERT(pos >= 0, /**/);
1794
1795         Font font = getFontSettings(bparams, pos);
1796
1797         pos_type const body_pos = beginOfBody();
1798         FontInfo & fi = font.fontInfo();
1799         if (pos < body_pos)
1800                 fi.realize(d->layout_->labelfont);
1801         else
1802                 fi.realize(d->layout_->font);
1803
1804         fi.realize(outerfont.fontInfo());
1805         fi.realize(bparams.getFont().fontInfo());
1806
1807         return font;
1808 }
1809
1810
1811 Font const Paragraph::getLabelFont
1812         (BufferParams const & bparams, Font const & outerfont) const
1813 {
1814         FontInfo tmpfont = d->layout_->labelfont;
1815         tmpfont.realize(outerfont.fontInfo());
1816         tmpfont.realize(bparams.getFont().fontInfo());
1817         return Font(tmpfont, getParLanguage(bparams));
1818 }
1819
1820
1821 Font const Paragraph::getLayoutFont
1822         (BufferParams const & bparams, Font const & outerfont) const
1823 {
1824         FontInfo tmpfont = d->layout_->font;
1825         tmpfont.realize(outerfont.fontInfo());
1826         tmpfont.realize(bparams.getFont().fontInfo());
1827         return Font(tmpfont, getParLanguage(bparams));
1828 }
1829
1830
1831 /// Returns the height of the highest font in range
1832 FontSize Paragraph::highestFontInRange
1833         (pos_type startpos, pos_type endpos, FontSize def_size) const
1834 {
1835         return d->fontlist_.highestInRange(startpos, endpos, def_size);
1836 }
1837
1838
1839 char_type Paragraph::getUChar(BufferParams const & bparams, pos_type pos) const
1840 {
1841         char_type c = d->text_[pos];
1842         if (!lyxrc.rtl_support)
1843                 return c;
1844
1845         char_type uc = c;
1846         switch (c) {
1847         case '(':
1848                 uc = ')';
1849                 break;
1850         case ')':
1851                 uc = '(';
1852                 break;
1853         case '[':
1854                 uc = ']';
1855                 break;
1856         case ']':
1857                 uc = '[';
1858                 break;
1859         case '{':
1860                 uc = '}';
1861                 break;
1862         case '}':
1863                 uc = '{';
1864                 break;
1865         case '<':
1866                 uc = '>';
1867                 break;
1868         case '>':
1869                 uc = '<';
1870                 break;
1871         }
1872         if (uc != c && getFontSettings(bparams, pos).isRightToLeft())
1873                 return uc;
1874         return c;
1875 }
1876
1877
1878 void Paragraph::setFont(pos_type pos, Font const & font)
1879 {
1880         LASSERT(pos <= size(), /**/);
1881
1882         // First, reduce font against layout/label font
1883         // Update: The setCharFont() routine in text2.cpp already
1884         // reduces font, so we don't need to do that here. (Asger)
1885
1886         d->fontlist_.set(pos, font);
1887 }
1888
1889
1890 void Paragraph::makeSameLayout(Paragraph const & par)
1891 {
1892         d->layout_ = par.d->layout_;
1893         d->params_ = par.d->params_;
1894 }
1895
1896
1897 bool Paragraph::stripLeadingSpaces(bool trackChanges)
1898 {
1899         if (isFreeSpacing())
1900                 return false;
1901
1902         int pos = 0;
1903         int count = 0;
1904
1905         while (pos < size() && (isNewline(pos) || isLineSeparator(pos))) {
1906                 if (eraseChar(pos, trackChanges))
1907                         ++count;
1908                 else
1909                         ++pos;
1910         }
1911
1912         return count > 0 || pos > 0;
1913 }
1914
1915
1916 bool Paragraph::hasSameLayout(Paragraph const & par) const
1917 {
1918         return par.d->layout_ == d->layout_
1919                 && d->params_.sameLayout(par.d->params_);
1920 }
1921
1922
1923 depth_type Paragraph::getDepth() const
1924 {
1925         return d->params_.depth();
1926 }
1927
1928
1929 depth_type Paragraph::getMaxDepthAfter() const
1930 {
1931         if (d->layout_->isEnvironment())
1932                 return d->params_.depth() + 1;
1933         else
1934                 return d->params_.depth();
1935 }
1936
1937
1938 char Paragraph::getAlign() const
1939 {
1940         if (d->params_.align() == LYX_ALIGN_LAYOUT)
1941                 return d->layout_->align;
1942         else
1943                 return d->params_.align();
1944 }
1945
1946
1947 docstring const & Paragraph::labelString() const
1948 {
1949         return d->params_.labelString();
1950 }
1951
1952
1953 // the next two functions are for the manual labels
1954 docstring const Paragraph::getLabelWidthString() const
1955 {
1956         if (d->layout_->margintype == MARGIN_MANUAL
1957             || d->layout_->latextype == LATEX_BIB_ENVIRONMENT)
1958                 return d->params_.labelWidthString();
1959         else
1960                 return _("Senseless with this layout!");
1961 }
1962
1963
1964 void Paragraph::setLabelWidthString(docstring const & s)
1965 {
1966         d->params_.labelWidthString(s);
1967 }
1968
1969
1970 docstring Paragraph::expandLabel(Layout const & layout,
1971                 BufferParams const & bparams) const
1972 {
1973         return expandParagraphLabel(layout, bparams, true);
1974 }
1975
1976
1977 docstring Paragraph::expandDocBookLabel(Layout const & layout,
1978                 BufferParams const & bparams) const
1979 {
1980         return expandParagraphLabel(layout, bparams, false);
1981 }
1982
1983
1984 docstring Paragraph::expandParagraphLabel(Layout const & layout,
1985                 BufferParams const & bparams, bool process_appendix) const
1986 {
1987         DocumentClass const & tclass = bparams.documentClass();
1988         string const & lang = getParLanguage(bparams)->code();
1989         bool const in_appendix = process_appendix && d->params_.appendix();
1990         docstring fmt = translateIfPossible(layout.labelstring(in_appendix), lang);
1991
1992         if (fmt.empty() && layout.labeltype == LABEL_COUNTER
1993             && !layout.counter.empty())
1994                 return tclass.counters().theCounter(layout.counter, lang);
1995
1996         // handle 'inherited level parts' in 'fmt',
1997         // i.e. the stuff between '@' in   '@Section@.\arabic{subsection}'
1998         size_t const i = fmt.find('@', 0);
1999         if (i != docstring::npos) {
2000                 size_t const j = fmt.find('@', i + 1);
2001                 if (j != docstring::npos) {
2002                         docstring parent(fmt, i + 1, j - i - 1);
2003                         docstring label = from_ascii("??");
2004                         if (tclass.hasLayout(parent))
2005                                 docstring label = expandParagraphLabel(tclass[parent], bparams,
2006                                                       process_appendix);
2007                         fmt = docstring(fmt, 0, i) + label
2008                                 + docstring(fmt, j + 1, docstring::npos);
2009                 }
2010         }
2011
2012         return tclass.counters().counterLabel(fmt, lang);
2013 }
2014
2015
2016 void Paragraph::applyLayout(Layout const & new_layout)
2017 {
2018         d->layout_ = &new_layout;
2019         LyXAlignment const oldAlign = d->params_.align();
2020
2021         if (!(oldAlign & d->layout_->alignpossible)) {
2022                 frontend::Alert::warning(_("Alignment not permitted"),
2023                         _("The new layout does not permit the alignment previously used.\nSetting to default."));
2024                 d->params_.align(LYX_ALIGN_LAYOUT);
2025         }
2026 }
2027
2028
2029 pos_type Paragraph::beginOfBody() const
2030 {
2031         return d->begin_of_body_;
2032 }
2033
2034
2035 void Paragraph::setBeginOfBody()
2036 {
2037         if (d->layout_->labeltype != LABEL_MANUAL) {
2038                 d->begin_of_body_ = 0;
2039                 return;
2040         }
2041
2042         // Unroll the first two cycles of the loop
2043         // and remember the previous character to
2044         // remove unnecessary getChar() calls
2045         pos_type i = 0;
2046         pos_type end = size();
2047         if (i < end && !isNewline(i)) {
2048                 ++i;
2049                 char_type previous_char = 0;
2050                 char_type temp = 0;
2051                 if (i < end) {
2052                         previous_char = d->text_[i];
2053                         if (!isNewline(i)) {
2054                                 ++i;
2055                                 while (i < end && previous_char != ' ') {
2056                                         temp = d->text_[i];
2057                                         if (isNewline(i))
2058                                                 break;
2059                                         ++i;
2060                                         previous_char = temp;
2061                                 }
2062                         }
2063                 }
2064         }
2065
2066         d->begin_of_body_ = i;
2067 }
2068
2069
2070 bool Paragraph::allowParagraphCustomization() const
2071 {
2072         return inInset().allowParagraphCustomization();
2073 }
2074
2075
2076 bool Paragraph::usePlainLayout() const
2077 {
2078         return inInset().usePlainLayout();
2079 }
2080
2081
2082 bool Paragraph::isPassThru() const
2083 {
2084         return inInset().getLayout().isPassThru() || d->layout_->pass_thru;
2085 }
2086
2087 namespace {
2088
2089 // paragraphs inside floats need different alignment tags to avoid
2090 // unwanted space
2091
2092 bool noTrivlistCentering(InsetCode code)
2093 {
2094         return code == FLOAT_CODE
2095                || code == WRAP_CODE
2096                || code == CELL_CODE;
2097 }
2098
2099
2100 string correction(string const & orig)
2101 {
2102         if (orig == "flushleft")
2103                 return "raggedright";
2104         if (orig == "flushright")
2105                 return "raggedleft";
2106         if (orig == "center")
2107                 return "centering";
2108         return orig;
2109 }
2110
2111
2112 string const corrected_env(string const & suffix, string const & env,
2113         InsetCode code, bool const lastpar)
2114 {
2115         string output = suffix + "{";
2116         if (noTrivlistCentering(code)) {
2117                 if (lastpar) {
2118                         // the last paragraph in non-trivlist-aligned
2119                         // context is special (to avoid unwanted whitespace)
2120                         if (suffix == "\\begin")
2121                                 return "\\" + correction(env) + "{}";
2122                         return string();
2123                 }
2124                 output += correction(env);
2125         } else
2126                 output += env;
2127         output += "}";
2128         if (suffix == "\\begin")
2129                 output += "\n";
2130         return output;
2131 }
2132
2133
2134 void adjust_column(string const & str, int & column)
2135 {
2136         if (!contains(str, "\n"))
2137                 column += str.size();
2138         else {
2139                 string tmp;
2140                 column = rsplit(str, tmp, '\n').size();
2141         }
2142 }
2143
2144 } // namespace anon
2145
2146
2147 int Paragraph::Private::startTeXParParams(BufferParams const & bparams,
2148                         otexstream & os, OutputParams const & runparams) const
2149 {
2150         int column = 0;
2151
2152         if (params_.noindent() && !layout_->pass_thru) {
2153                 os << "\\noindent ";
2154                 column += 10;
2155         }
2156
2157         LyXAlignment const curAlign = params_.align();
2158
2159         if (curAlign == layout_->align)
2160                 return column;
2161
2162         switch (curAlign) {
2163         case LYX_ALIGN_NONE:
2164         case LYX_ALIGN_BLOCK:
2165         case LYX_ALIGN_LAYOUT:
2166         case LYX_ALIGN_SPECIAL:
2167         case LYX_ALIGN_DECIMAL:
2168                 break;
2169         case LYX_ALIGN_LEFT:
2170         case LYX_ALIGN_RIGHT:
2171         case LYX_ALIGN_CENTER:
2172                 if (runparams.moving_arg) {
2173                         os << "\\protect";
2174                         column += 8;
2175                 }
2176                 break;
2177         }
2178
2179         string const begin_tag = "\\begin";
2180         InsetCode code = ownerCode();
2181         bool const lastpar = runparams.isLastPar;
2182
2183         switch (curAlign) {
2184         case LYX_ALIGN_NONE:
2185         case LYX_ALIGN_BLOCK:
2186         case LYX_ALIGN_LAYOUT:
2187         case LYX_ALIGN_SPECIAL:
2188         case LYX_ALIGN_DECIMAL:
2189                 break;
2190         case LYX_ALIGN_LEFT: {
2191                 string output;
2192                 if (owner_->getParLanguage(bparams)->babel() != "hebrew")
2193                         output = corrected_env(begin_tag, "flushleft", code, lastpar);
2194                 else
2195                         output = corrected_env(begin_tag, "flushright", code, lastpar);
2196                 os << from_ascii(output);
2197                 adjust_column(output, column);
2198                 break;
2199         } case LYX_ALIGN_RIGHT: {
2200                 string output;
2201                 if (owner_->getParLanguage(bparams)->babel() != "hebrew")
2202                         output = corrected_env(begin_tag, "flushright", code, lastpar);
2203                 else
2204                         output = corrected_env(begin_tag, "flushleft", code, lastpar);
2205                 os << from_ascii(output);
2206                 adjust_column(output, column);
2207                 break;
2208         } case LYX_ALIGN_CENTER: {
2209                 string output;
2210                 output = corrected_env(begin_tag, "center", code, lastpar);
2211                 os << from_ascii(output);
2212                 adjust_column(output, column);
2213                 break;
2214         }
2215         }
2216
2217         return column;
2218 }
2219
2220
2221 bool Paragraph::Private::endTeXParParams(BufferParams const & bparams,
2222                         otexstream & os, OutputParams const & runparams) const
2223 {
2224         LyXAlignment const curAlign = params_.align();
2225
2226         if (curAlign == layout_->align)
2227                 return false;
2228
2229         switch (curAlign) {
2230         case LYX_ALIGN_NONE:
2231         case LYX_ALIGN_BLOCK:
2232         case LYX_ALIGN_LAYOUT:
2233         case LYX_ALIGN_SPECIAL:
2234         case LYX_ALIGN_DECIMAL:
2235                 break;
2236         case LYX_ALIGN_LEFT:
2237         case LYX_ALIGN_RIGHT:
2238         case LYX_ALIGN_CENTER:
2239                 if (runparams.moving_arg)
2240                         os << "\\protect";
2241                 break;
2242         }
2243
2244         string output;
2245         string const end_tag = "\n\\par\\end";
2246         InsetCode code = ownerCode();
2247         bool const lastpar = runparams.isLastPar;
2248
2249         switch (curAlign) {
2250         case LYX_ALIGN_NONE:
2251         case LYX_ALIGN_BLOCK:
2252         case LYX_ALIGN_LAYOUT:
2253         case LYX_ALIGN_SPECIAL:
2254         case LYX_ALIGN_DECIMAL:
2255                 break;
2256         case LYX_ALIGN_LEFT: {
2257                 if (owner_->getParLanguage(bparams)->babel() != "hebrew")
2258                         output = corrected_env(end_tag, "flushleft", code, lastpar);
2259                 else
2260                         output = corrected_env(end_tag, "flushright", code, lastpar);
2261                 os << from_ascii(output);
2262                 break;
2263         } case LYX_ALIGN_RIGHT: {
2264                 if (owner_->getParLanguage(bparams)->babel() != "hebrew")
2265                         output = corrected_env(end_tag, "flushright", code, lastpar);
2266                 else
2267                         output = corrected_env(end_tag, "flushleft", code, lastpar);
2268                 os << from_ascii(output);
2269                 break;
2270         } case LYX_ALIGN_CENTER: {
2271                 output = corrected_env(end_tag, "center", code, lastpar);
2272                 os << from_ascii(output);
2273                 break;
2274         }
2275         }
2276
2277         return !output.empty() || lastpar;
2278 }
2279
2280
2281 // This one spits out the text of the paragraph
2282 void Paragraph::latex(BufferParams const & bparams,
2283         Font const & outerfont,
2284         otexstream & os,
2285         OutputParams const & runparams,
2286         int start_pos, int end_pos, bool force) const
2287 {
2288         LYXERR(Debug::LATEX, "Paragraph::latex...     " << this);
2289
2290         // FIXME This check should not be needed. Perhaps issue an
2291         // error if it triggers.
2292         Layout const & style = inInset().forcePlainLayout() ?
2293                 bparams.documentClass().plainLayout() : *d->layout_;
2294
2295         if (!force && style.inpreamble)
2296                 return;
2297
2298         bool const allowcust = allowParagraphCustomization();
2299
2300         // Current base font for all inherited font changes, without any
2301         // change caused by an individual character, except for the language:
2302         // It is set to the language of the first character.
2303         // As long as we are in the label, this font is the base font of the
2304         // label. Before the first body character it is set to the base font
2305         // of the body.
2306         Font basefont;
2307
2308         // Maybe we have to create a optional argument.
2309         pos_type body_pos = beginOfBody();
2310         unsigned int column = 0;
2311
2312         if (body_pos > 0) {
2313                 // the optional argument is kept in curly brackets in
2314                 // case it contains a ']'
2315                 // This is not strictly needed, but if this is changed it
2316                 // would be a file format change, and tex2lyx would need
2317                 // to be adjusted, since it unconditionally removes the
2318                 // braces when it parses \item.
2319                 os << "[{";
2320                 column += 2;
2321                 basefont = getLabelFont(bparams, outerfont);
2322         } else {
2323                 basefont = getLayoutFont(bparams, outerfont);
2324         }
2325
2326         // Which font is currently active?
2327         Font running_font(basefont);
2328         // Do we have an open font change?
2329         bool open_font = false;
2330
2331         Change runningChange = Change(Change::UNCHANGED);
2332
2333         Encoding const * const prev_encoding = runparams.encoding;
2334
2335         os.texrow().start(id(), 0);
2336
2337         // if the paragraph is empty, the loop will not be entered at all
2338         if (empty()) {
2339                 if (style.isCommand()) {
2340                         os << '{';
2341                         ++column;
2342                 }
2343                 if (allowcust)
2344                         column += d->startTeXParParams(bparams, os, runparams);
2345         }
2346
2347         for (pos_type i = 0; i < size(); ++i) {
2348                 // First char in paragraph or after label?
2349                 if (i == body_pos) {
2350                         if (body_pos > 0) {
2351                                 if (open_font) {
2352                                         column += running_font.latexWriteEndChanges(
2353                                                 os, bparams, runparams,
2354                                                 basefont, basefont);
2355                                         open_font = false;
2356                                 }
2357                                 basefont = getLayoutFont(bparams, outerfont);
2358                                 running_font = basefont;
2359
2360                                 column += Changes::latexMarkChange(os, bparams,
2361                                                 runningChange, Change(Change::UNCHANGED),
2362                                                 runparams);
2363                                 runningChange = Change(Change::UNCHANGED);
2364
2365                                 os << "}] ";
2366                                 column +=3;
2367                         }
2368                         if (style.isCommand()) {
2369                                 os << '{';
2370                                 ++column;
2371                         }
2372
2373                         if (allowcust)
2374                                 column += d->startTeXParParams(bparams, os,
2375                                                             runparams);
2376                 }
2377
2378                 Change const & change = runparams.inDeletedInset
2379                         ? runparams.changeOfDeletedInset : lookupChange(i);
2380
2381                 if (bparams.outputChanges && runningChange != change) {
2382                         if (open_font) {
2383                                 column += running_font.latexWriteEndChanges(
2384                                                 os, bparams, runparams, basefont, basefont);
2385                                 open_font = false;
2386                         }
2387                         basefont = getLayoutFont(bparams, outerfont);
2388                         running_font = basefont;
2389
2390                         column += Changes::latexMarkChange(os, bparams, runningChange,
2391                                                            change, runparams);
2392                         runningChange = change;
2393                 }
2394
2395                 // do not output text which is marked deleted
2396                 // if change tracking output is disabled
2397                 if (!bparams.outputChanges && change.deleted()) {
2398                         continue;
2399                 }
2400
2401                 ++column;
2402
2403                 // Fully instantiated font
2404                 Font const font = getFont(bparams, i, outerfont);
2405
2406                 Font const last_font = running_font;
2407
2408                 // Do we need to close the previous font?
2409                 if (open_font &&
2410                     (font != running_font ||
2411                      font.language() != running_font.language()))
2412                 {
2413                         column += running_font.latexWriteEndChanges(
2414                                         os, bparams, runparams, basefont,
2415                                         (i == body_pos-1) ? basefont : font);
2416                         running_font = basefont;
2417                         open_font = false;
2418                 }
2419
2420                 string const running_lang = runparams.use_polyglossia ?
2421                         running_font.language()->polyglossia() : running_font.language()->babel();
2422                 // close babel's font environment before opening CJK.
2423                 string const lang_end_command = runparams.use_polyglossia ?
2424                         "\\end{$$lang}" : lyxrc.language_command_end;
2425                 if (!running_lang.empty() &&
2426                     font.language()->encoding()->package() == Encoding::CJK) {
2427                                 string end_tag = subst(lang_end_command,
2428                                                         "$$lang",
2429                                                         running_lang);
2430                                 os << from_ascii(end_tag);
2431                                 column += end_tag.length();
2432                 }
2433
2434                 // Switch file encoding if necessary (and allowed)
2435                 if (!runparams.pass_thru && !style.pass_thru &&
2436                     runparams.encoding->package() != Encoding::none &&
2437                     font.language()->encoding()->package() != Encoding::none) {
2438                         pair<bool, int> const enc_switch =
2439                                 switchEncoding(os.os(), bparams, runparams,
2440                                         *(font.language()->encoding()));
2441                         if (enc_switch.first) {
2442                                 column += enc_switch.second;
2443                                 runparams.encoding = font.language()->encoding();
2444                         }
2445                 }
2446
2447                 char_type const c = d->text_[i];
2448
2449                 // Do we need to change font?
2450                 if ((font != running_font ||
2451                      font.language() != running_font.language()) &&
2452                         i != body_pos - 1)
2453                 {
2454                         odocstringstream ods;
2455                         column += font.latexWriteStartChanges(ods, bparams,
2456                                                               runparams, basefont,
2457                                                               last_font);
2458                         running_font = font;
2459                         open_font = true;
2460                         docstring fontchange = ods.str();
2461                         // check whether the fontchange ends with a \\textcolor
2462                         // modifier and the text starts with a space (bug 4473)
2463                         docstring const last_modifier = rsplit(fontchange, '\\');
2464                         if (prefixIs(last_modifier, from_ascii("textcolor")) && c == ' ')
2465                                 os << fontchange << from_ascii("{}");
2466                         // check if the fontchange ends with a trailing blank
2467                         // (like "\small " (see bug 3382)
2468                         else if (suffixIs(fontchange, ' ') && c == ' ')
2469                                 os << fontchange.substr(0, fontchange.size() - 1)
2470                                    << from_ascii("{}");
2471                         else
2472                                 os << fontchange;
2473                 }
2474
2475                 // FIXME: think about end_pos implementation...
2476                 if (c == ' ' && i >= start_pos && (end_pos == -1 || i < end_pos)) {
2477                         // FIXME: integrate this case in latexSpecialChar
2478                         // Do not print the separation of the optional argument
2479                         // if style.pass_thru is false. This works because
2480                         // latexSpecialChar ignores spaces if
2481                         // style.pass_thru is false.
2482                         if (i != body_pos - 1) {
2483                                 if (d->simpleTeXBlanks(runparams, os,
2484                                                 i, column, font, style)) {
2485                                         // A surrogate pair was output. We
2486                                         // must not call latexSpecialChar
2487                                         // in this iteration, since it would output
2488                                         // the combining character again.
2489                                         ++i;
2490                                         continue;
2491                                 }
2492                         }
2493                 }
2494
2495                 OutputParams rp = runparams;
2496                 rp.free_spacing = style.free_spacing;
2497                 rp.local_font = &font;
2498                 rp.intitle = style.intitle;
2499
2500                 // Two major modes:  LaTeX or plain
2501                 // Handle here those cases common to both modes
2502                 // and then split to handle the two modes separately.
2503                 if (c == META_INSET) {
2504                         if (i >= start_pos && (end_pos == -1 || i < end_pos)) {
2505                                 d->latexInset(bparams, os, rp, running_font,
2506                                                 basefont, outerfont, open_font,
2507                                                 runningChange, style, i, column);
2508                         }
2509                 } else {
2510                         if (i >= start_pos && (end_pos == -1 || i < end_pos)) {
2511                                 try {
2512                                         d->latexSpecialChar(os, rp, running_font, runningChange,
2513                                                             style, i, end_pos, column);
2514                                 } catch (EncodingException & e) {
2515                                 if (runparams.dryrun) {
2516                                         os << "<" << _("LyX Warning: ")
2517                                            << _("uncodable character") << " '";
2518                                         os.put(c);
2519                                         os << "'>";
2520                                 } else {
2521                                         // add location information and throw again.
2522                                         e.par_id = id();
2523                                         e.pos = i;
2524                                         throw(e);
2525                                 }
2526                         }
2527                 }
2528                 }
2529
2530                 // Set the encoding to that returned from latexSpecialChar (see
2531                 // comment for encoding member in OutputParams.h)
2532                 runparams.encoding = rp.encoding;
2533         }
2534
2535         // If we have an open font definition, we have to close it
2536         if (open_font) {
2537 #ifdef FIXED_LANGUAGE_END_DETECTION
2538                 if (next_) {
2539                         running_font.latexWriteEndChanges(os, bparams,
2540                                         runparams, basefont,
2541                                         next_->getFont(bparams, 0, outerfont));
2542                 } else {
2543                         running_font.latexWriteEndChanges(os, bparams,
2544                                         runparams, basefont, basefont);
2545                 }
2546 #else
2547 //FIXME: For now we ALWAYS have to close the foreign font settings if they are
2548 //FIXME: there as we start another \selectlanguage with the next paragraph if
2549 //FIXME: we are in need of this. This should be fixed sometime (Jug)
2550                 running_font.latexWriteEndChanges(os, bparams, runparams,
2551                                 basefont, basefont);
2552 #endif
2553         }
2554
2555         column += Changes::latexMarkChange(os, bparams, runningChange,
2556                                            Change(Change::UNCHANGED), runparams);
2557
2558         // Needed if there is an optional argument but no contents.
2559         if (body_pos > 0 && body_pos == size()) {
2560                 os << "}]~";
2561         }
2562
2563         if (allowcust && d->endTeXParParams(bparams, os, runparams)
2564             && runparams.encoding != prev_encoding) {
2565                 runparams.encoding = prev_encoding;
2566                 if (!runparams.isFullUnicode())
2567                         os << setEncoding(prev_encoding->iconvName());
2568         }
2569
2570         LYXERR(Debug::LATEX, "Paragraph::latex... done " << this);
2571 }
2572
2573
2574 bool Paragraph::emptyTag() const
2575 {
2576         for (pos_type i = 0; i < size(); ++i) {
2577                 if (Inset const * inset = getInset(i)) {
2578                         InsetCode lyx_code = inset->lyxCode();
2579                         // FIXME testing like that is wrong. What is
2580                         // the intent?
2581                         if (lyx_code != TOC_CODE &&
2582                             lyx_code != INCLUDE_CODE &&
2583                             lyx_code != GRAPHICS_CODE &&
2584                             lyx_code != ERT_CODE &&
2585                             lyx_code != LISTINGS_CODE &&
2586                             lyx_code != FLOAT_CODE &&
2587                             lyx_code != TABULAR_CODE) {
2588                                 return false;
2589                         }
2590                 } else {
2591                         char_type c = d->text_[i];
2592                         if (c != ' ' && c != '\t')
2593                                 return false;
2594                 }
2595         }
2596         return true;
2597 }
2598
2599
2600 string Paragraph::getID(Buffer const & buf, OutputParams const & runparams)
2601         const
2602 {
2603         for (pos_type i = 0; i < size(); ++i) {
2604                 if (Inset const * inset = getInset(i)) {
2605                         InsetCode lyx_code = inset->lyxCode();
2606                         if (lyx_code == LABEL_CODE) {
2607                                 InsetLabel const * const il = static_cast<InsetLabel const *>(inset);
2608                                 docstring const & id = il->getParam("name");
2609                                 return "id='" + to_utf8(sgml::cleanID(buf, runparams, id)) + "'";
2610                         }
2611                 }
2612         }
2613         return string();
2614 }
2615
2616
2617 pos_type Paragraph::firstWordDocBook(odocstream & os, OutputParams const & runparams)
2618         const
2619 {
2620         pos_type i;
2621         for (i = 0; i < size(); ++i) {
2622                 if (Inset const * inset = getInset(i)) {
2623                         inset->docbook(os, runparams);
2624                 } else {
2625                         char_type c = d->text_[i];
2626                         if (c == ' ')
2627                                 break;
2628                         os << sgml::escapeChar(c);
2629                 }
2630         }
2631         return i;
2632 }
2633
2634
2635 pos_type Paragraph::firstWordLyXHTML(XHTMLStream & xs, OutputParams const & runparams)
2636         const
2637 {
2638         pos_type i;
2639         for (i = 0; i < size(); ++i) {
2640                 if (Inset const * inset = getInset(i)) {
2641                         inset->xhtml(xs, runparams);
2642                 } else {
2643                         char_type c = d->text_[i];
2644                         if (c == ' ')
2645                                 break;
2646                         xs << c;
2647                 }
2648         }
2649         return i;
2650 }
2651
2652
2653 bool Paragraph::Private::onlyText(Buffer const & buf, Font const & outerfont, pos_type initial) const
2654 {
2655         Font font_old;
2656         pos_type size = text_.size();
2657         for (pos_type i = initial; i < size; ++i) {
2658                 Font font = owner_->getFont(buf.params(), i, outerfont);
2659                 if (text_[i] == META_INSET)
2660                         return false;
2661                 if (i != initial && font != font_old)
2662                         return false;
2663                 font_old = font;
2664         }
2665
2666         return true;
2667 }
2668
2669
2670 void Paragraph::simpleDocBookOnePar(Buffer const & buf,
2671                                     odocstream & os,
2672                                     OutputParams const & runparams,
2673                                     Font const & outerfont,
2674                                     pos_type initial) const
2675 {
2676         bool emph_flag = false;
2677
2678         Layout const & style = *d->layout_;
2679         FontInfo font_old =
2680                 style.labeltype == LABEL_MANUAL ? style.labelfont : style.font;
2681
2682         if (style.pass_thru && !d->onlyText(buf, outerfont, initial))
2683                 os << "]]>";
2684
2685         // parsing main loop
2686         for (pos_type i = initial; i < size(); ++i) {
2687                 Font font = getFont(buf.params(), i, outerfont);
2688
2689                 // handle <emphasis> tag
2690                 if (font_old.emph() != font.fontInfo().emph()) {
2691                         if (font.fontInfo().emph() == FONT_ON) {
2692                                 os << "<emphasis>";
2693                                 emph_flag = true;
2694                         } else if (i != initial) {
2695                                 os << "</emphasis>";
2696                                 emph_flag = false;
2697                         }
2698                 }
2699
2700                 if (Inset const * inset = getInset(i)) {
2701                         inset->docbook(os, runparams);
2702                 } else {
2703                         char_type c = d->text_[i];
2704
2705                         if (style.pass_thru)
2706                                 os.put(c);
2707                         else
2708                                 os << sgml::escapeChar(c);
2709                 }
2710                 font_old = font.fontInfo();
2711         }
2712
2713         if (emph_flag) {
2714                 os << "</emphasis>";
2715         }
2716
2717         if (style.free_spacing)
2718                 os << '\n';
2719         if (style.pass_thru && !d->onlyText(buf, outerfont, initial))
2720                 os << "<![CDATA[";
2721 }
2722
2723
2724 docstring Paragraph::simpleLyXHTMLOnePar(Buffer const & buf,
2725                                     XHTMLStream & xs,
2726                                     OutputParams const & runparams,
2727                                     Font const & outerfont,
2728                                     pos_type initial) const
2729 {
2730         docstring retval;
2731
2732         bool emph_flag = false;
2733         bool bold_flag = false;
2734
2735         Layout const & style = *d->layout_;
2736
2737         xs.startParagraph(allowEmpty());
2738
2739         if (!runparams.for_toc && runparams.html_make_pars) {
2740                 // generate a magic label for this paragraph
2741                 string const attr = "id='" + magicLabel() + "'";
2742                 xs << html::CompTag("a", attr);
2743         }
2744
2745         FontInfo font_old =
2746                 style.labeltype == LABEL_MANUAL ? style.labelfont : style.font;
2747
2748         // parsing main loop
2749         for (pos_type i = initial; i < size(); ++i) {
2750                 // let's not show deleted material in the output
2751                 if (isDeleted(i))
2752                         continue;
2753
2754                 Font font = getFont(buf.params(), i, outerfont);
2755
2756                 // emphasis
2757                 if (font_old.emph() != font.fontInfo().emph()) {
2758                         if (font.fontInfo().emph() == FONT_ON) {
2759                                 xs << html::StartTag("em");
2760                                 emph_flag = true;
2761                         } else if (emph_flag && i != initial) {
2762                                 xs << html::EndTag("em");
2763                                 emph_flag = false;
2764                         }
2765                 }
2766                 // bold
2767                 if (font_old.series() != font.fontInfo().series()) {
2768                         if (font.fontInfo().series() == BOLD_SERIES) {
2769                                 xs << html::StartTag("strong");
2770                                 bold_flag = true;
2771                         } else if (bold_flag && i != initial) {
2772                                 xs << html::EndTag("strong");
2773                                 bold_flag = false;
2774                         }
2775                 }
2776                 // FIXME XHTML
2777                 // Other such tags? What about the other text ranges?
2778
2779                 Inset const * inset = getInset(i);
2780                 if (inset) {
2781                         if (!runparams.for_toc || inset->isInToc()) {
2782                                 OutputParams np = runparams;
2783                                 if (!inset->getLayout().htmlisblock())
2784                                         np.html_in_par = true;
2785                                 retval += inset->xhtml(xs, np);
2786                         }
2787                 } else {
2788                         char_type c = d->text_[i];
2789
2790                         if (style.pass_thru)
2791                                 xs << c;
2792                         else if (c == '-') {
2793                                 docstring str;
2794                                 int j = i + 1;
2795                                 if (j < size() && d->text_[j] == '-') {
2796                                         j += 1;
2797                                         if (j < size() && d->text_[j] == '-') {
2798                                                 str += from_ascii("&mdash;");
2799                                                 i += 2;
2800                                         } else {
2801                                                 str += from_ascii("&ndash;");
2802                                                 i += 1;
2803                                         }
2804                                 }
2805                                 else
2806                                         str += c;
2807                                 // We don't want to escape the entities. Note that
2808                                 // it is safe to do this, since str can otherwise
2809                                 // only be "-". E.g., it can't be "<".
2810                                 xs << XHTMLStream::ESCAPE_NONE << str;
2811                         } else
2812                                 xs << c;
2813                 }
2814                 font_old = font.fontInfo();
2815         }
2816
2817         xs.closeFontTags();
2818         xs.endParagraph();
2819         return retval;
2820 }
2821
2822
2823 bool Paragraph::isHfill(pos_type pos) const
2824 {
2825         Inset const * inset = getInset(pos);
2826         return inset && (inset->lyxCode() == SPACE_CODE &&
2827                          inset->isStretchableSpace());
2828 }
2829
2830
2831 bool Paragraph::isNewline(pos_type pos) const
2832 {
2833         Inset const * inset = getInset(pos);
2834         return inset && inset->lyxCode() == NEWLINE_CODE;
2835 }
2836
2837
2838 bool Paragraph::isLineSeparator(pos_type pos) const
2839 {
2840         char_type const c = d->text_[pos];
2841         if (isLineSeparatorChar(c))
2842                 return true;
2843         Inset const * inset = getInset(pos);
2844         return inset && inset->isLineSeparator();
2845 }
2846
2847
2848 bool Paragraph::isWordSeparator(pos_type pos) const
2849 {
2850         if (pos == size())
2851                 return true;
2852         if (Inset const * inset = getInset(pos))
2853                 return !inset->isLetter();
2854         // if we have a hard hyphen (no en- or emdash) or apostrophe
2855         // we pass this to the spell checker
2856         // FIXME: this method is subject to change, visit
2857         // https://bugzilla.mozilla.org/show_bug.cgi?id=355178
2858         // to get an impression how complex this is.
2859         if (isHardHyphenOrApostrophe(pos))
2860                 return false;
2861         char_type const c = d->text_[pos];
2862         // We want to pass the escape chars to the spellchecker
2863         docstring const escape_chars = from_utf8(lyxrc.spellchecker_esc_chars);
2864         return !isLetterChar(c) && !isDigitASCII(c) && !contains(escape_chars, c);
2865 }
2866
2867
2868 bool Paragraph::isHardHyphenOrApostrophe(pos_type pos) const
2869 {
2870         pos_type const psize = size();
2871         if (pos >= psize)
2872                 return false;
2873         char_type const c = d->text_[pos];
2874         if (c != '-' && c != '\'')
2875                 return false;
2876         int nextpos = pos + 1;
2877         int prevpos = pos > 0 ? pos - 1 : 0;
2878         if ((nextpos == psize || isSpace(nextpos))
2879                 && (pos == 0 || isSpace(prevpos)))
2880                 return false;
2881         return c == '\''
2882                 || ((nextpos == psize || d->text_[nextpos] != '-')
2883                 && (pos == 0 || d->text_[prevpos] != '-'));
2884 }
2885
2886
2887 bool Paragraph::isSameSpellRange(pos_type pos1, pos_type pos2) const
2888 {
2889         return pos1 == pos2
2890                 || d->speller_state_.getRange(pos1) == d->speller_state_.getRange(pos2);
2891 }
2892
2893
2894 bool Paragraph::isChar(pos_type pos) const
2895 {
2896         if (Inset const * inset = getInset(pos))
2897                 return inset->isChar();
2898         char_type const c = d->text_[pos];
2899         return !isLetterChar(c) && !isDigitASCII(c) && !lyx::isSpace(c);
2900 }
2901
2902
2903 bool Paragraph::isSpace(pos_type pos) const
2904 {
2905         if (Inset const * inset = getInset(pos))
2906                 return inset->isSpace();
2907         char_type const c = d->text_[pos];
2908         return lyx::isSpace(c);
2909 }
2910
2911
2912 Language const *
2913 Paragraph::getParLanguage(BufferParams const & bparams) const
2914 {
2915         if (!empty())
2916                 return getFirstFontSettings(bparams).language();
2917         // FIXME: we should check the prev par as well (Lgb)
2918         return bparams.language;
2919 }
2920
2921
2922 bool Paragraph::isRTL(BufferParams const & bparams) const
2923 {
2924         return lyxrc.rtl_support
2925                 && getParLanguage(bparams)->rightToLeft()
2926                 && !inInset().getLayout().forceLTR();
2927 }
2928
2929
2930 void Paragraph::changeLanguage(BufferParams const & bparams,
2931                                Language const * from, Language const * to)
2932 {
2933         // change language including dummy font change at the end
2934         for (pos_type i = 0; i <= size(); ++i) {
2935                 Font font = getFontSettings(bparams, i);
2936                 if (font.language() == from) {
2937                         font.setLanguage(to);
2938                         setFont(i, font);
2939                         d->requestSpellCheck(i);
2940                 }
2941         }
2942 }
2943
2944
2945 bool Paragraph::isMultiLingual(BufferParams const & bparams) const
2946 {
2947         Language const * doc_language = bparams.language;
2948         FontList::const_iterator cit = d->fontlist_.begin();
2949         FontList::const_iterator end = d->fontlist_.end();
2950
2951         for (; cit != end; ++cit)
2952                 if (cit->font().language() != ignore_language &&
2953                     cit->font().language() != latex_language &&
2954                     cit->font().language() != doc_language)
2955                         return true;
2956         return false;
2957 }
2958
2959
2960 void Paragraph::getLanguages(std::set<Language const *> & languages) const
2961 {
2962         FontList::const_iterator cit = d->fontlist_.begin();
2963         FontList::const_iterator end = d->fontlist_.end();
2964
2965         for (; cit != end; ++cit) {
2966                 Language const * lang = cit->font().language();
2967                 if (lang != ignore_language &&
2968                     lang != latex_language)
2969                         languages.insert(lang);
2970         }
2971 }
2972
2973
2974 docstring Paragraph::asString(int options) const
2975 {
2976         return asString(0, size(), options);
2977 }
2978
2979
2980 docstring Paragraph::asString(pos_type beg, pos_type end, int options) const
2981 {
2982         odocstringstream os;
2983
2984         if (beg == 0
2985             && options & AS_STR_LABEL
2986             && !d->params_.labelString().empty())
2987                 os << d->params_.labelString() << ' ';
2988
2989         for (pos_type i = beg; i < end; ++i) {
2990                 if ((options & AS_STR_SKIPDELETE) && isDeleted(i))
2991                         continue;
2992                 char_type const c = d->text_[i];
2993                 if (isPrintable(c) || c == '\t'
2994                     || (c == '\n' && (options & AS_STR_NEWLINES)))
2995                         os.put(c);
2996                 else if (c == META_INSET && (options & AS_STR_INSETS)) {
2997                         getInset(i)->toString(os);
2998                         if (getInset(i)->asInsetMath())
2999                                 os << " ";
3000                 }
3001         }
3002
3003         return os.str();
3004 }
3005
3006
3007 void Paragraph::forToc(docstring & os, size_t maxlen) const
3008 {
3009         if (!d->params_.labelString().empty())
3010                 os += d->params_.labelString() + ' ';
3011         for (pos_type i = 0; i < size() && os.length() < maxlen; ++i) {
3012                 if (isDeleted(i))
3013                         continue;
3014                 char_type const c = d->text_[i];
3015                 if (isPrintable(c))
3016                         os += c;
3017                 else if (c == '\t' || c == '\n')
3018                         os += ' ';
3019                 else if (c == META_INSET)
3020                         getInset(i)->forToc(os, maxlen);
3021         }
3022 }
3023
3024
3025 docstring Paragraph::stringify(pos_type beg, pos_type end, int options, OutputParams & runparams) const
3026 {
3027         odocstringstream os;
3028
3029         if (beg == 0
3030                 && options & AS_STR_LABEL
3031                 && !d->params_.labelString().empty())
3032                 os << d->params_.labelString() << ' ';
3033
3034         for (pos_type i = beg; i < end; ++i) {
3035                 char_type const c = d->text_[i];
3036                 if (isPrintable(c) || c == '\t'
3037                     || (c == '\n' && (options & AS_STR_NEWLINES)))
3038                         os.put(c);
3039                 else if (c == META_INSET && (options & AS_STR_INSETS)) {
3040                         getInset(i)->plaintext(os, runparams);
3041                 }
3042         }
3043
3044         return os.str();
3045 }
3046
3047
3048 void Paragraph::setInsetOwner(Inset const * inset)
3049 {
3050         d->inset_owner_ = inset;
3051 }
3052
3053
3054 int Paragraph::id() const
3055 {
3056         return d->id_;
3057 }
3058
3059
3060 void Paragraph::setId(int id)
3061 {
3062         d->id_ = id;
3063 }
3064
3065
3066 Layout const & Paragraph::layout() const
3067 {
3068         return *d->layout_;
3069 }
3070
3071
3072 void Paragraph::setLayout(Layout const & layout)
3073 {
3074         d->layout_ = &layout;
3075 }
3076
3077
3078 void Paragraph::setDefaultLayout(DocumentClass const & tc)
3079 {
3080         setLayout(tc.defaultLayout());
3081 }
3082
3083
3084 void Paragraph::setPlainLayout(DocumentClass const & tc)
3085 {
3086         setLayout(tc.plainLayout());
3087 }
3088
3089
3090 void Paragraph::setPlainOrDefaultLayout(DocumentClass const & tclass)
3091 {
3092         if (usePlainLayout())
3093                 setPlainLayout(tclass);
3094         else
3095                 setDefaultLayout(tclass);
3096 }
3097
3098
3099 Inset const & Paragraph::inInset() const
3100 {
3101         LASSERT(d->inset_owner_, throw ExceptionMessage(BufferException,
3102                 _("Memory problem"), _("Paragraph not properly initialized")));
3103         return *d->inset_owner_;
3104 }
3105
3106
3107 ParagraphParameters & Paragraph::params()
3108 {
3109         return d->params_;
3110 }
3111
3112
3113 ParagraphParameters const & Paragraph::params() const
3114 {
3115         return d->params_;
3116 }
3117
3118
3119 bool Paragraph::isFreeSpacing() const
3120 {
3121         if (d->layout_->free_spacing)
3122                 return true;
3123         return d->inset_owner_ && d->inset_owner_->isFreeSpacing();
3124 }
3125
3126
3127 bool Paragraph::allowEmpty() const
3128 {
3129         if (d->layout_->keepempty)
3130                 return true;
3131         return d->inset_owner_ && d->inset_owner_->allowEmpty();
3132 }
3133
3134
3135 char_type Paragraph::transformChar(char_type c, pos_type pos) const
3136 {
3137         if (!Encodings::isArabicChar(c))
3138                 return c;
3139
3140         char_type prev_char = ' ';
3141         char_type next_char = ' ';
3142
3143         for (pos_type i = pos - 1; i >= 0; --i) {
3144                 char_type const par_char = d->text_[i];
3145                 if (!Encodings::isArabicComposeChar(par_char)) {
3146                         prev_char = par_char;
3147                         break;
3148                 }
3149         }
3150
3151         for (pos_type i = pos + 1, end = size(); i < end; ++i) {
3152                 char_type const par_char = d->text_[i];
3153                 if (!Encodings::isArabicComposeChar(par_char)) {
3154                         next_char = par_char;
3155                         break;
3156                 }
3157         }
3158
3159         if (Encodings::isArabicChar(next_char)) {
3160                 if (Encodings::isArabicChar(prev_char) &&
3161                         !Encodings::isArabicSpecialChar(prev_char))
3162                         return Encodings::transformChar(c, Encodings::FORM_MEDIAL);
3163                 else
3164                         return Encodings::transformChar(c, Encodings::FORM_INITIAL);
3165         } else {
3166                 if (Encodings::isArabicChar(prev_char) &&
3167                         !Encodings::isArabicSpecialChar(prev_char))
3168                         return Encodings::transformChar(c, Encodings::FORM_FINAL);
3169                 else
3170                         return Encodings::transformChar(c, Encodings::FORM_ISOLATED);
3171         }
3172 }
3173
3174
3175 int Paragraph::checkBiblio(Buffer const & buffer)
3176 {
3177         // FIXME From JS:
3178         // This is getting more and more a mess. ...We really should clean
3179         // up this bibitem issue for 1.6.
3180
3181         // Add bibitem insets if necessary
3182         if (d->layout_->labeltype != LABEL_BIBLIO)
3183                 return 0;
3184
3185         bool hasbibitem = !d->insetlist_.empty()
3186                 // Insist on it being in pos 0
3187                 && d->text_[0] == META_INSET
3188                 && d->insetlist_.begin()->inset->lyxCode() == BIBITEM_CODE;
3189
3190         bool track_changes = buffer.params().trackChanges;
3191
3192         docstring oldkey;
3193         docstring oldlabel;
3194
3195         // remove a bibitem in pos != 0
3196         // restore it later in pos 0 if necessary
3197         // (e.g. if a user inserts contents _before_ the item)
3198         // we're assuming there's only one of these, which there
3199         // should be.
3200         int erasedInsetPosition = -1;
3201         InsetList::iterator it = d->insetlist_.begin();
3202         InsetList::iterator end = d->insetlist_.end();
3203         for (; it != end; ++it)
3204                 if (it->inset->lyxCode() == BIBITEM_CODE
3205                       && it->pos > 0) {
3206                         InsetCommand * olditem = it->inset->asInsetCommand();
3207                         oldkey = olditem->getParam("key");
3208                         oldlabel = olditem->getParam("label");
3209                         erasedInsetPosition = it->pos;
3210                         eraseChar(erasedInsetPosition, track_changes);
3211                         break;
3212         }
3213
3214         // There was an InsetBibitem at the beginning, and we didn't
3215         // have to erase one.
3216         if (hasbibitem && erasedInsetPosition < 0)
3217                         return 0;
3218
3219         // There was an InsetBibitem at the beginning and we did have to
3220         // erase one. So we give its properties to the beginning inset.
3221         if (hasbibitem) {
3222                 InsetCommand * inset = d->insetlist_.begin()->inset->asInsetCommand();
3223                 if (!oldkey.empty())
3224                         inset->setParam("key", oldkey);
3225                 inset->setParam("label", oldlabel);
3226                 return -erasedInsetPosition;
3227         }
3228
3229         // There was no inset at the beginning, so we need to create one with
3230         // the key and label of the one we erased.
3231         InsetBibitem * inset =
3232                 new InsetBibitem(const_cast<Buffer *>(&buffer), InsetCommandParams(BIBITEM_CODE));
3233         // restore values of previously deleted item in this par.
3234         if (!oldkey.empty())
3235                 inset->setParam("key", oldkey);
3236         inset->setParam("label", oldlabel);
3237         insertInset(0, inset,
3238                     Change(track_changes ? Change::INSERTED : Change::UNCHANGED));
3239
3240         return 1;
3241 }
3242
3243
3244 void Paragraph::checkAuthors(AuthorList const & authorList)
3245 {
3246         d->changes_.checkAuthors(authorList);
3247 }
3248
3249
3250 bool Paragraph::isChanged(pos_type pos) const
3251 {
3252         return lookupChange(pos).changed();
3253 }
3254
3255
3256 bool Paragraph::isInserted(pos_type pos) const
3257 {
3258         return lookupChange(pos).inserted();
3259 }
3260
3261
3262 bool Paragraph::isDeleted(pos_type pos) const
3263 {
3264         return lookupChange(pos).deleted();
3265 }
3266
3267
3268 InsetList const & Paragraph::insetList() const
3269 {
3270         return d->insetlist_;
3271 }
3272
3273
3274 void Paragraph::setBuffer(Buffer & b)
3275 {
3276         d->insetlist_.setBuffer(b);
3277 }
3278
3279
3280 Inset * Paragraph::releaseInset(pos_type pos)
3281 {
3282         Inset * inset = d->insetlist_.release(pos);
3283         /// does not honour change tracking!
3284         eraseChar(pos, false);
3285         return inset;
3286 }
3287
3288
3289 Inset * Paragraph::getInset(pos_type pos)
3290 {
3291         return (pos < pos_type(d->text_.size()) && d->text_[pos] == META_INSET)
3292                  ? d->insetlist_.get(pos) : 0;
3293 }
3294
3295
3296 Inset const * Paragraph::getInset(pos_type pos) const
3297 {
3298         return (pos < pos_type(d->text_.size()) && d->text_[pos] == META_INSET)
3299                  ? d->insetlist_.get(pos) : 0;
3300 }
3301
3302
3303 void Paragraph::changeCase(BufferParams const & bparams, pos_type pos,
3304                 pos_type & right, TextCase action)
3305 {
3306         // process sequences of modified characters; in change
3307         // tracking mode, this approach results in much better
3308         // usability than changing case on a char-by-char basis
3309         docstring changes;
3310
3311         bool const trackChanges = bparams.trackChanges;
3312
3313         bool capitalize = true;
3314
3315         for (; pos < right; ++pos) {
3316                 char_type oldChar = d->text_[pos];
3317                 char_type newChar = oldChar;
3318
3319                 // ignore insets and don't play with deleted text!
3320                 if (oldChar != META_INSET && !isDeleted(pos)) {
3321                         switch (action) {
3322                                 case text_lowercase:
3323                                         newChar = lowercase(oldChar);
3324                                         break;
3325                                 case text_capitalization:
3326                                         if (capitalize) {
3327                                                 newChar = uppercase(oldChar);
3328                                                 capitalize = false;
3329                                         }
3330                                         break;
3331                                 case text_uppercase:
3332                                         newChar = uppercase(oldChar);
3333                                         break;
3334                         }
3335                 }
3336
3337                 if (isWordSeparator(pos) || isDeleted(pos)) {
3338                         // permit capitalization again
3339                         capitalize = true;
3340                 }
3341
3342                 if (oldChar != newChar) {
3343                         changes += newChar;
3344                         if (pos != right - 1)
3345                                 continue;
3346                         // step behind the changing area
3347                         pos++;
3348                 }
3349
3350                 int erasePos = pos - changes.size();
3351                 for (size_t i = 0; i < changes.size(); i++) {
3352                         insertChar(pos, changes[i],
3353                                    getFontSettings(bparams,
3354                                                    erasePos),
3355                                    trackChanges);
3356                         if (!eraseChar(erasePos, trackChanges)) {
3357                                 ++erasePos;
3358                                 ++pos; // advance
3359                                 ++right; // expand selection
3360                         }
3361                 }
3362                 changes.clear();
3363         }
3364 }
3365
3366
3367 int Paragraph::find(docstring const & str, bool cs, bool mw,
3368                 pos_type start_pos, bool del) const
3369 {
3370         pos_type pos = start_pos;
3371         int const strsize = str.length();
3372         int i = 0;
3373         pos_type const parsize = d->text_.size();
3374         for (i = 0; i < strsize && pos < parsize; ++i, ++pos) {
3375                 // Ignore "invisible" letters such as ligature breaks
3376                 // and hyphenation chars while searching
3377                 while (pos < parsize - 1 && isInset(pos)) {
3378                         odocstringstream os;
3379                         getInset(pos)->toString(os);
3380                         if (!getInset(pos)->isLetter() || !os.str().empty())
3381                                 break;
3382                         pos++;
3383                 }
3384                 if (cs && str[i] != d->text_[pos])
3385                         break;
3386                 if (!cs && uppercase(str[i]) != uppercase(d->text_[pos]))
3387                         break;
3388                 if (!del && isDeleted(pos))
3389                         break;
3390         }
3391
3392         if (i != strsize)
3393                 return 0;
3394
3395         // if necessary, check whether string matches word
3396         if (mw) {
3397                 if (start_pos > 0 && !isWordSeparator(start_pos - 1))
3398                         return 0;
3399                 if (pos < parsize
3400                         && !isWordSeparator(pos))
3401                         return 0;
3402         }
3403
3404         return pos - start_pos;
3405 }
3406
3407
3408 char_type Paragraph::getChar(pos_type pos) const
3409 {
3410         return d->text_[pos];
3411 }
3412
3413
3414 pos_type Paragraph::size() const
3415 {
3416         return d->text_.size();
3417 }
3418
3419
3420 bool Paragraph::empty() const
3421 {
3422         return d->text_.empty();
3423 }
3424
3425
3426 bool Paragraph::isInset(pos_type pos) const
3427 {
3428         return d->text_[pos] == META_INSET;
3429 }
3430
3431
3432 bool Paragraph::isSeparator(pos_type pos) const
3433 {
3434         //FIXME: Are we sure this can be the only separator?
3435         return d->text_[pos] == ' ';
3436 }
3437
3438
3439 void Paragraph::deregisterWords()
3440 {
3441         Private::LangWordsMap::const_iterator itl = d->words_.begin();
3442         Private::LangWordsMap::const_iterator ite = d->words_.end();
3443         for (; itl != ite; ++itl) {
3444                 WordList * wl = theWordList(itl->first);
3445                 Private::Words::const_iterator it = (itl->second).begin();
3446                 Private::Words::const_iterator et = (itl->second).end();
3447                 for (; it != et; ++it)
3448                         wl->remove(*it);
3449         }
3450         d->words_.clear();
3451 }
3452
3453
3454 void Paragraph::locateWord(pos_type & from, pos_type & to,
3455         word_location const loc) const
3456 {
3457         switch (loc) {
3458         case WHOLE_WORD_STRICT:
3459                 if (from == 0 || from == size()
3460                     || isWordSeparator(from)
3461                     || isWordSeparator(from - 1)) {
3462                         to = from;
3463                         return;
3464                 }
3465                 // no break here, we go to the next
3466
3467         case WHOLE_WORD:
3468                 // If we are already at the beginning of a word, do nothing
3469                 if (!from || isWordSeparator(from - 1))
3470                         break;
3471                 // no break here, we go to the next
3472
3473         case PREVIOUS_WORD:
3474                 // always move the cursor to the beginning of previous word
3475                 while (from && !isWordSeparator(from - 1))
3476                         --from;
3477                 break;
3478         case NEXT_WORD:
3479                 LYXERR0("Paragraph::locateWord: NEXT_WORD not implemented yet");
3480                 break;
3481         case PARTIAL_WORD:
3482                 // no need to move the 'from' cursor
3483                 break;
3484         }
3485         to = from;
3486         while (to < size() && !isWordSeparator(to))
3487                 ++to;
3488 }
3489
3490
3491 void Paragraph::collectWords()
3492 {
3493         pos_type n = size();
3494         for (pos_type pos = 0; pos < n; ++pos) {
3495                 if (isWordSeparator(pos))
3496                         continue;
3497                 pos_type from = pos;
3498                 locateWord(from, pos, WHOLE_WORD);
3499                 if ((pos - from) >= (int)lyxrc.completion_minlength) {
3500                         docstring word = asString(from, pos, AS_STR_NONE);
3501                         FontList::const_iterator cit = d->fontlist_.fontIterator(pos);
3502                         if (cit == d->fontlist_.end())
3503                                 return;
3504                         Language const * lang = cit->font().language();
3505                         d->words_[*lang].insert(word);
3506                 }
3507         }
3508 }
3509
3510
3511 void Paragraph::registerWords()
3512 {
3513         Private::LangWordsMap::const_iterator itl = d->words_.begin();
3514         Private::LangWordsMap::const_iterator ite = d->words_.end();
3515         for (; itl != ite; ++itl) {
3516                 WordList * wl = theWordList(itl->first);
3517                 Private::Words::const_iterator it = (itl->second).begin();
3518                 Private::Words::const_iterator et = (itl->second).end();
3519                 for (; it != et; ++it)
3520                         wl->insert(*it);
3521         }
3522 }
3523
3524
3525 void Paragraph::updateWords()
3526 {
3527         deregisterWords();
3528         collectWords();
3529         registerWords();
3530 }
3531
3532
3533 void Paragraph::Private::appendSkipPosition(SkipPositions & skips, pos_type const pos) const
3534 {
3535         SkipPositionsIterator begin = skips.begin();
3536         SkipPositions::iterator end = skips.end();
3537         if (pos > 0 && begin < end) {
3538                 --end;
3539                 if (end->last == pos - 1) {
3540                         end->last = pos;
3541                         return;
3542                 }
3543         }
3544         skips.insert(end, FontSpan(pos, pos));
3545 }
3546
3547
3548 Language * Paragraph::Private::locateSpellRange(
3549         pos_type & from, pos_type & to,
3550         SkipPositions & skips) const
3551 {
3552         // skip leading white space
3553         while (from < to && owner_->isWordSeparator(from))
3554                 ++from;
3555         // don't check empty range
3556         if (from >= to)
3557                 return 0;
3558         // get current language
3559         Language * lang = getSpellLanguage(from);
3560         pos_type last = from;
3561         bool samelang = true;
3562         bool sameinset = true;
3563         while (last < to && samelang && sameinset) {
3564                 // hop to end of word
3565                 while (last < to && !owner_->isWordSeparator(last)) {
3566                         if (owner_->getInset(last)) {
3567                                 appendSkipPosition(skips, last);
3568                         } else if (owner_->isDeleted(last)) {
3569                                 appendSkipPosition(skips, last);
3570                         }
3571                         ++last;
3572                 }
3573                 // hop to next word while checking for insets
3574                 while (sameinset && last < to && owner_->isWordSeparator(last)) {
3575                         if (Inset const * inset = owner_->getInset(last))
3576                                 sameinset = inset->isChar() && inset->isLetter();
3577                         if (sameinset && owner_->isDeleted(last)) {
3578                                 appendSkipPosition(skips, last);
3579                         }
3580                         if (sameinset)
3581                                 last++;
3582                 }
3583                 if (sameinset && last < to) {
3584                         // now check for language change
3585                         samelang = lang == getSpellLanguage(last);
3586                 }
3587         }
3588         // if language change detected backstep is needed
3589         if (!samelang)
3590                 --last;
3591         to = last;
3592         return lang;
3593 }
3594
3595
3596 Language * Paragraph::Private::getSpellLanguage(pos_type const from) const
3597 {
3598         Language * lang =
3599                 const_cast<Language *>(owner_->getFontSettings(
3600                         inset_owner_->buffer().params(), from).language());
3601         if (lang == inset_owner_->buffer().params().language
3602                 && !lyxrc.spellchecker_alt_lang.empty()) {
3603                 string lang_code;
3604                 string const lang_variety =
3605                         split(lyxrc.spellchecker_alt_lang, lang_code, '-');
3606                 lang->setCode(lang_code);
3607                 lang->setVariety(lang_variety);
3608         }
3609         return lang;
3610 }
3611
3612
3613 void Paragraph::requestSpellCheck(pos_type pos)
3614 {
3615         d->requestSpellCheck(pos);
3616 }
3617
3618
3619 bool Paragraph::needsSpellCheck() const
3620 {
3621         SpellChecker::ChangeNumber speller_change_number = 0;
3622         if (theSpellChecker())
3623                 speller_change_number = theSpellChecker()->changeNumber();
3624         if (speller_change_number > d->speller_state_.currentChangeNumber()) {
3625                 d->speller_state_.needsCompleteRefresh(speller_change_number);
3626         }
3627         return d->needsSpellCheck();
3628 }
3629
3630
3631 bool Paragraph::Private::ignoreWord(docstring const & word) const
3632 {
3633         // Ignore words with digits
3634         // FIXME: make this customizable
3635         // (note that some checkers ignore words with digits by default)
3636         docstring::const_iterator cit = word.begin();
3637         docstring::const_iterator const end = word.end();
3638         for (; cit != end; ++cit) {
3639                 if (isNumber((*cit)))
3640                         return true;
3641         }
3642         return false;
3643 }
3644
3645
3646 SpellChecker::Result Paragraph::spellCheck(pos_type & from, pos_type & to,
3647         WordLangTuple & wl, docstring_list & suggestions,
3648         bool do_suggestion, bool check_learned) const
3649 {
3650         SpellChecker::Result result = SpellChecker::WORD_OK;
3651         SpellChecker * speller = theSpellChecker();
3652         if (!speller)
3653                 return result;
3654
3655         if (!d->layout_->spellcheck || !inInset().allowSpellCheck())
3656                 return result;
3657
3658         locateWord(from, to, WHOLE_WORD);
3659         if (from == to || from >= size())
3660                 return result;
3661
3662         docstring word = asString(from, to, AS_STR_INSETS | AS_STR_SKIPDELETE);
3663         Language * lang = d->getSpellLanguage(from);
3664
3665         wl = WordLangTuple(word, lang);
3666
3667         if (!word.size())
3668                 return result;
3669
3670         if (needsSpellCheck() || check_learned) {
3671                 pos_type end = to;
3672                 if (!d->ignoreWord(word)) {
3673                         bool const trailing_dot = to < size() && d->text_[to] == '.';
3674                         result = speller->check(wl);
3675                         if (SpellChecker::misspelled(result) && trailing_dot) {
3676                                 wl = WordLangTuple(word.append(from_ascii(".")), lang);
3677                                 result = speller->check(wl);
3678                                 if (!SpellChecker::misspelled(result)) {
3679                                         LYXERR(Debug::GUI, "misspelled word is correct with dot: \"" <<
3680                                            word << "\" [" <<
3681                                            from << ".." << to << "]");
3682                                 } else {
3683                                         // spell check with dot appended failed too
3684                                         // restore original word/lang value
3685                                         word = asString(from, to, AS_STR_INSETS | AS_STR_SKIPDELETE);
3686                                         wl = WordLangTuple(word, lang);
3687                                 }
3688                         }
3689                 }
3690                 if (!SpellChecker::misspelled(result)) {
3691                         // area up to the begin of the next word is not misspelled
3692                         while (end < size() && isWordSeparator(end))
3693                                 ++end;
3694                 }
3695                 d->setMisspelled(from, end, result);
3696         } else {
3697                 result = d->speller_state_.getState(from);
3698         }
3699
3700         if (do_suggestion)
3701                 suggestions.clear();
3702
3703         if (SpellChecker::misspelled(result)) {
3704                 LYXERR(Debug::GUI, "misspelled word: \"" <<
3705                            word << "\" [" <<
3706                            from << ".." << to << "]");
3707                 if (do_suggestion)
3708                         speller->suggest(wl, suggestions);
3709         }
3710         return result;
3711 }
3712
3713
3714 void Paragraph::Private::markMisspelledWords(
3715         pos_type const & first, pos_type const & last,
3716         SpellChecker::Result result,
3717         docstring const & word,
3718         SkipPositions const & skips)
3719 {
3720         if (!SpellChecker::misspelled(result)) {
3721                 setMisspelled(first, last, SpellChecker::WORD_OK);
3722                 return;
3723         }
3724         int snext = first;
3725         SpellChecker * speller = theSpellChecker();
3726         // locate and enumerate the error positions
3727         int nerrors = speller->numMisspelledWords();
3728         int numskipped = 0;
3729         SkipPositionsIterator it = skips.begin();
3730         SkipPositionsIterator et = skips.end();
3731         for (int index = 0; index < nerrors; ++index) {
3732                 int wstart;
3733                 int wlen = 0;
3734                 speller->misspelledWord(index, wstart, wlen);
3735                 /// should not happen if speller supports range checks
3736                 if (!wlen) continue;
3737                 docstring const misspelled = word.substr(wstart, wlen);
3738                 wstart += first + numskipped;
3739                 if (snext < wstart) {
3740                         /// mark the range of correct spelling
3741                         numskipped += countSkips(it, et, wstart);
3742                         setMisspelled(snext,
3743                                 wstart - 1, SpellChecker::WORD_OK);
3744                 }
3745                 snext = wstart + wlen;
3746                 numskipped += countSkips(it, et, snext);
3747                 /// mark the range of misspelling
3748                 setMisspelled(wstart, snext, result);
3749                 LYXERR(Debug::GUI, "misspelled word: \"" <<
3750                            misspelled << "\" [" <<
3751                            wstart << ".." << (snext-1) << "]");
3752                 ++snext;
3753         }
3754         if (snext <= last) {
3755                 /// mark the range of correct spelling at end
3756                 setMisspelled(snext, last, SpellChecker::WORD_OK);
3757         }
3758 }
3759
3760
3761 void Paragraph::spellCheck() const
3762 {
3763         SpellChecker * speller = theSpellChecker();
3764         if (!speller || !size() ||!needsSpellCheck())
3765                 return;
3766         pos_type start;
3767         pos_type endpos;
3768         d->rangeOfSpellCheck(start, endpos);
3769         if (speller->canCheckParagraph()) {
3770                 // loop until we leave the range
3771                 for (pos_type first = start; first < endpos; ) {
3772                         pos_type last = endpos;
3773                         Private::SkipPositions skips;
3774                         Language * lang = d->locateSpellRange(first, last, skips);
3775                         if (first >= endpos)
3776                                 break;
3777                         // start the spell checker on the unit of meaning
3778                         docstring word = asString(first, last, AS_STR_INSETS + AS_STR_SKIPDELETE);
3779                         WordLangTuple wl = WordLangTuple(word, lang);
3780                         SpellChecker::Result result = word.size() ?
3781                                 speller->check(wl) : SpellChecker::WORD_OK;
3782                         d->markMisspelledWords(first, last, result, word, skips);
3783                         first = ++last;
3784                 }
3785         } else {
3786                 static docstring_list suggestions;
3787                 pos_type to = endpos;
3788                 while (start < endpos) {
3789                         WordLangTuple wl;
3790                         spellCheck(start, to, wl, suggestions, false);
3791                         start = to + 1;
3792                 }
3793         }
3794         d->readySpellCheck();
3795 }
3796
3797
3798 bool Paragraph::isMisspelled(pos_type pos, bool check_boundary) const
3799 {
3800         bool result = SpellChecker::misspelled(d->speller_state_.getState(pos));
3801         if (result || pos <= 0 || pos > size())
3802                 return result;
3803         if (check_boundary && (pos == size() || isWordSeparator(pos)))
3804                 result = SpellChecker::misspelled(d->speller_state_.getState(pos - 1));
3805         return result;
3806 }
3807
3808
3809 string Paragraph::magicLabel() const
3810 {
3811         stringstream ss;
3812         ss << "magicparlabel-" << id();
3813         return ss.str();
3814 }
3815
3816
3817 } // namespace lyx