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 latexSpecialT3(
 349                 char_type const c,
 350                 otexstream & os,
 351                 pos_type i,
 352                 unsigned int & column);
 353         ///
 354         bool latexSpecialTypewriter(
 355                 char_type const c,
 356                 otexstream & os,
 357                 pos_type i,
 358                 unsigned int & column);
 359         ///
 360         bool latexSpecialPhrase(
 361                 otexstream & os,
 362                 pos_type & i,
 363                 pos_type end_pos,
 364                 unsigned int & column,
 365                 OutputParams const & runparams);
 366
 367         ///
 368         void validate(LaTeXFeatures & features) const;
 369
 370         /// Checks if the paragraph contains only text and no inset or font change.
 371         bool onlyText(Buffer const & buf, Font const & outerfont,
 372                       pos_type initial) const;
 373
 374         /// match a string against a particular point in the paragraph
 375         bool isTextAt(string const & str, pos_type pos) const;
 376
 377         /// a vector of speller skip positions
 378         typedef vector<FontSpan> SkipPositions;
 379         typedef SkipPositions::const_iterator SkipPositionsIterator;
 380
 381         void appendSkipPosition(SkipPositions & skips, pos_type const pos) const;
 382
 383         Language * getSpellLanguage(pos_type const from) const;
 384
 385         Language * locateSpellRange(pos_type & from, pos_type & to,
 386                                                                 SkipPositions & skips) const;
 387
 388         bool hasSpellerChange() const {
 389                 SpellChecker::ChangeNumber speller_change_number = 0;
 390                 if (theSpellChecker())
 391                         speller_change_number = theSpellChecker()->changeNumber();
 392                 return speller_change_number > speller_state_.currentChangeNumber();
 393         }
 394
 395         bool ignoreWord(docstring const & word) const ;
 396
 397         void setMisspelled(pos_type from, pos_type to, SpellChecker::Result state)
 398         {
 399                 pos_type textsize = owner_->size();
 400                 // check for sane arguments
 401                 if (to <= from || from >= textsize)
 402                         return;
 403                 FontSpan fp = FontSpan(from, to - 1);
 404                 speller_state_.setRange(fp, state);
 405         }
 406
 407         void requestSpellCheck(pos_type pos) {
 408                 if (pos == -1)
 409                         speller_state_.needsCompleteRefresh(speller_state_.currentChangeNumber());
 410                 else
 411                         speller_state_.needsRefresh(pos);
 412         }
 413
 414         void readySpellCheck() {
 415                 speller_state_.needsRefresh(-1);
 416         }
 417
 418         bool needsSpellCheck() const
 419         {
 420                 return speller_state_.needsRefresh();
 421         }
 422
 423         void rangeOfSpellCheck(pos_type & first, pos_type & last) const
 424         {
 425                 speller_state_.refreshRange(first, last);
 426                 if (last == -1) {
 427                         last = owner_->size();
 428                         return;
 429                 }
 430                 pos_type endpos = last;
 431                 owner_->locateWord(first, endpos, WHOLE_WORD);
 432                 if (endpos < last) {
 433                         endpos = last;
 434                         owner_->locateWord(last, endpos, WHOLE_WORD);
 435                 }
 436                 last = endpos;
 437         }
 438
 439         int countSkips(SkipPositionsIterator & it, SkipPositionsIterator const et,
 440                             int & start) const
 441         {
 442                 int numskips = 0;
 443                 while (it != et && it->first < start) {
 444                         int skip = it->last - it->first + 1;
 445                         start += skip;
 446                         numskips += skip;
 447                         ++it;
 448                 }
 449                 return numskips;
 450         }
 451
 452         void markMisspelledWords(pos_type const & first, pos_type const & last,
 453                                                          SpellChecker::Result result,
 454                                                          docstring const & word,
 455                                                          SkipPositions const & skips);
 456
 457         InsetCode ownerCode() const
 458         {
 459                 return inset_owner_ ? inset_owner_->lyxCode() : NO_CODE;
 460         }
 461
 462         /// Which Paragraph owns us?
 463         Paragraph * owner_;
 464
 465         /// In which Inset?
 466         Inset const * inset_owner_;
 467
 468         ///
 469         FontList fontlist_;
 470
 471         ///
 472         int id_;
 473
 474         ///
 475         ParagraphParameters params_;
 476
 477         /// for recording and looking up changes
 478         Changes changes_;
 479
 480         ///
 481         InsetList insetlist_;
 482
 483         /// end of label
 484         pos_type begin_of_body_;
 485
 486         typedef docstring TextContainer;
 487         ///
 488         TextContainer text_;
 489
 490         typedef set<docstring> Words;
 491         typedef map<Language, Words> LangWordsMap;
 492         ///
 493         LangWordsMap words_;
 494         ///
 495         Layout const * layout_;
 496         ///
 497         SpellCheckerState speller_state_;
 498 };
 499
 500
 501 namespace {
 502
 503 struct special_phrase {
 504         string phrase;
 505         docstring macro;
 506         bool builtin;
 507 };
 508
 509 special_phrase const special_phrases[] = {
 510         { "LyX", from_ascii("\\LyX{}"), false },
 511         { "TeX", from_ascii("\\TeX{}"), true },
 512         { "LaTeX2e", from_ascii("\\LaTeXe{}"), true },
 513         { "LaTeX", from_ascii("\\LaTeX{}"), true },
 514 };
 515
 516 size_t const phrases_nr = sizeof(special_phrases)/sizeof(special_phrase);
 517
 518 } // namespace anon
 519
 520
 521 Paragraph::Private::Private(Paragraph * owner, Layout const & layout)
 522         : owner_(owner), inset_owner_(0), id_(-1), begin_of_body_(0), layout_(&layout)
 523 {
 524         text_.reserve(100);
 525 }
 526
 527
 528 // Initialization of the counter for the paragraph id's,
 529 //
 530 // FIXME: There should be a more intelligent way to generate and use the
 531 // paragraph ids per buffer instead a global static counter for all InsetText
 532 // in the running program.
 533 static int paragraph_id = -1;
 534
 535 Paragraph::Private::Private(Private const & p, Paragraph * owner)
 536         : owner_(owner), inset_owner_(p.inset_owner_), fontlist_(p.fontlist_),
 537           params_(p.params_), changes_(p.changes_), insetlist_(p.insetlist_),
 538           begin_of_body_(p.begin_of_body_), text_(p.text_), words_(p.words_),
 539           layout_(p.layout_)
 540 {
 541         id_ = ++paragraph_id;
 542         requestSpellCheck(p.text_.size());
 543 }
 544
 545
 546 Paragraph::Private::Private(Private const & p, Paragraph * owner,
 547         pos_type beg, pos_type end)
 548         : owner_(owner), inset_owner_(p.inset_owner_),
 549           params_(p.params_), changes_(p.changes_),
 550           insetlist_(p.insetlist_, beg, end),
 551           begin_of_body_(p.begin_of_body_), words_(p.words_),
 552           layout_(p.layout_)
 553 {
 554         id_ = ++paragraph_id;
 555         if (beg >= pos_type(p.text_.size()))
 556                 return;
 557         text_ = p.text_.substr(beg, end - beg);
 558
 559         FontList::const_iterator fcit = fontlist_.begin();
 560         FontList::const_iterator fend = fontlist_.end();
 561         for (; fcit != fend; ++fcit) {
 562                 if (fcit->pos() < beg)
 563                         continue;
 564                 if (fcit->pos() >= end) {
 565                         // Add last entry in the fontlist_.
 566                         fontlist_.set(text_.size() - 1, fcit->font());
 567                         break;
 568                 }
 569                 // Add a new entry in the fontlist_.
 570                 fontlist_.set(fcit->pos() - beg, fcit->font());
 571         }
 572         requestSpellCheck(p.text_.size());
 573 }
 574
 575
 576 void Paragraph::addChangesToToc(DocIterator const & cdit,
 577         Buffer const & buf) const
 578 {
 579         d->changes_.addToToc(cdit, buf);
 580 }
 581
 582
 583 bool Paragraph::isDeleted(pos_type start, pos_type end) const
 584 {
 585         LASSERT(start >= 0 && start <= size(), /**/);
 586         LASSERT(end > start && end <= size() + 1, /**/);
 587
 588         return d->changes_.isDeleted(start, end);
 589 }
 590
 591
 592 bool Paragraph::isChanged(pos_type start, pos_type end) const
 593 {
 594         LASSERT(start >= 0 && start <= size(), /**/);
 595         LASSERT(end > start && end <= size() + 1, /**/);
 596
 597         return d->changes_.isChanged(start, end);
 598 }
 599
 600
 601 bool Paragraph::isMergedOnEndOfParDeletion(bool trackChanges) const
 602 {
 603         // keep the logic here in sync with the logic of eraseChars()
 604         if (!trackChanges)
 605                 return true;
 606
 607         Change const change = d->changes_.lookup(size());
 608         return change.inserted() && change.currentAuthor();
 609 }
 610
 611
 612 void Paragraph::setChange(Change const & change)
 613 {
 614         // beware of the imaginary end-of-par character!
 615         d->changes_.set(change, 0, size() + 1);
 616
 617         /*
 618          * Propagate the change recursively - but not in case of DELETED!
 619          *
 620          * Imagine that your co-author makes changes in an existing inset. He
 621          * sends your document to you and you come to the conclusion that the
 622          * inset should go completely. If you erase it, LyX must not delete all
 623          * text within the inset. Otherwise, the change tracked insertions of
 624          * your co-author get lost and there is no way to restore them later.
 625          *
 626          * Conclusion: An inset's content should remain untouched if you delete it
 627          */
 628
 629         if (!change.deleted()) {
 630                 for (pos_type pos = 0; pos < size(); ++pos) {
 631                         if (Inset * inset = getInset(pos))
 632                                 inset->setChange(change);
 633                 }
 634         }
 635 }
 636
 637
 638 void Paragraph::setChange(pos_type pos, Change const & change)
 639 {
 640         LASSERT(pos >= 0 && pos <= size(), /**/);
 641         d->changes_.set(change, pos);
 642
 643         // see comment in setChange(Change const &) above
 644         if (!change.deleted() && pos < size())
 645                         if (Inset * inset = getInset(pos))
 646                                 inset->setChange(change);
 647 }
 648
 649
 650 Change const & Paragraph::lookupChange(pos_type pos) const
 651 {
 652         LASSERT(pos >= 0 && pos <= size(), /**/);
 653         return d->changes_.lookup(pos);
 654 }
 655
 656
 657 void Paragraph::acceptChanges(pos_type start, pos_type end)
 658 {
 659         LASSERT(start >= 0 && start <= size(), /**/);
 660         LASSERT(end > start && end <= size() + 1, /**/);
 661
 662         for (pos_type pos = start; pos < end; ++pos) {
 663                 switch (lookupChange(pos).type) {
 664                         case Change::UNCHANGED:
 665                                 // accept changes in nested inset
 666                                 if (Inset * inset = getInset(pos))
 667                                         inset->acceptChanges();
 668                                 break;
 669
 670                         case Change::INSERTED:
 671                                 d->changes_.set(Change(Change::UNCHANGED), pos);
 672                                 // also accept changes in nested inset
 673                                 if (Inset * inset = getInset(pos))
 674                                         inset->acceptChanges();
 675                                 break;
 676
 677                         case Change::DELETED:
 678                                 // Suppress access to non-existent
 679                                 // "end-of-paragraph char"
 680                                 if (pos < size()) {
 681                                         eraseChar(pos, false);
 682                                         --end;
 683                                         --pos;
 684                                 }
 685                                 break;
 686                 }
 687
 688         }
 689 }
 690
 691
 692 void Paragraph::rejectChanges(pos_type start, pos_type end)
 693 {
 694         LASSERT(start >= 0 && start <= size(), /**/);
 695         LASSERT(end > start && end <= size() + 1, /**/);
 696
 697         for (pos_type pos = start; pos < end; ++pos) {
 698                 switch (lookupChange(pos).type) {
 699                         case Change::UNCHANGED:
 700                                 // reject changes in nested inset
 701                                 if (Inset * inset = getInset(pos))
 702                                                 inset->rejectChanges();
 703                                 break;
 704
 705                         case Change::INSERTED:
 706                                 // Suppress access to non-existent
 707                                 // "end-of-paragraph char"
 708                                 if (pos < size()) {
 709                                         eraseChar(pos, false);
 710                                         --end;
 711                                         --pos;
 712                                 }
 713                                 break;
 714
 715                         case Change::DELETED:
 716                                 d->changes_.set(Change(Change::UNCHANGED), pos);
 717
 718                                 // Do NOT reject changes within a deleted inset!
 719                                 // There may be insertions of a co-author inside of it!
 720
 721                                 break;
 722                 }
 723         }
 724 }
 725
 726
 727 void Paragraph::Private::insertChar(pos_type pos, char_type c,
 728                 Change const & change)
 729 {
 730         LASSERT(pos >= 0 && pos <= int(text_.size()), /**/);
 731
 732         // track change
 733         changes_.insert(change, pos);
 734
 735         // This is actually very common when parsing buffers (and
 736         // maybe inserting ascii text)
 737         if (pos == pos_type(text_.size())) {
 738                 // when appending characters, no need to update tables
 739                 text_.push_back(c);
 740                 // but we want spell checking
 741                 requestSpellCheck(pos);
 742                 return;
 743         }
 744
 745         text_.insert(text_.begin() + pos, c);
 746
 747         // Update the font table.
 748         fontlist_.increasePosAfterPos(pos);
 749
 750         // Update the insets
 751         insetlist_.increasePosAfterPos(pos);
 752
 753         // Update list of misspelled positions
 754         speller_state_.increasePosAfterPos(pos);
 755 }
 756
 757
 758 bool Paragraph::insertInset(pos_type pos, Inset * inset,
 759                                    Change const & change)
 760 {
 761         LASSERT(inset, /**/);
 762         LASSERT(pos >= 0 && pos <= size(), /**/);
 763
 764         // Paragraph::insertInset() can be used in cut/copy/paste operation where
 765         // d->inset_owner_ is not set yet.
 766         if (d->inset_owner_ && !d->inset_owner_->insetAllowed(inset->lyxCode()))
 767                 return false;
 768
 769         d->insertChar(pos, META_INSET, change);
 770         LASSERT(d->text_[pos] == META_INSET, /**/);
 771
 772         // Add a new entry in the insetlist_.
 773         d->insetlist_.insert(inset, pos);
 774
 775         // Some insets require run of spell checker
 776         requestSpellCheck(pos);
 777         return true;
 778 }
 779
 780
 781 bool Paragraph::eraseChar(pos_type pos, bool trackChanges)
 782 {
 783         LASSERT(pos >= 0 && pos <= size(), return false);
 784
 785         // keep the logic here in sync with the logic of isMergedOnEndOfParDeletion()
 786
 787         if (trackChanges) {
 788                 Change change = d->changes_.lookup(pos);
 789
 790                 // set the character to DELETED if
 791                 //  a) it was previously unchanged or
 792                 //  b) it was inserted by a co-author
 793
 794                 if (!change.changed() ||
 795                       (change.inserted() && !change.currentAuthor())) {
 796                         setChange(pos, Change(Change::DELETED));
 797                         // request run of spell checker
 798                         requestSpellCheck(pos);
 799                         return false;
 800                 }
 801
 802                 if (change.deleted())
 803                         return false;
 804         }
 805
 806         // Don't physically access the imaginary end-of-paragraph character.
 807         // eraseChar() can only mark it as DELETED. A physical deletion of
 808         // end-of-par must be handled externally.
 809         if (pos == size()) {
 810                 return false;
 811         }
 812
 813         // track change
 814         d->changes_.erase(pos);
 815
 816         // if it is an inset, delete the inset entry
 817         if (d->text_[pos] == META_INSET)
 818                 d->insetlist_.erase(pos);
 819
 820         d->text_.erase(d->text_.begin() + pos);
 821
 822         // Update the fontlist_
 823         d->fontlist_.erase(pos);
 824
 825         // Update the insetlist_
 826         d->insetlist_.decreasePosAfterPos(pos);
 827
 828         // Update list of misspelled positions
 829         d->speller_state_.decreasePosAfterPos(pos);
 830         d->speller_state_.refreshLast(size());
 831
 832         return true;
 833 }
 834
 835
 836 int Paragraph::eraseChars(pos_type start, pos_type end, bool trackChanges)
 837 {
 838         LASSERT(start >= 0 && start <= size(), /**/);
 839         LASSERT(end >= start && end <= size() + 1, /**/);
 840
 841         pos_type i = start;
 842         for (pos_type count = end - start; count; --count) {
 843                 if (!eraseChar(i, trackChanges))
 844                         ++i;
 845         }
 846         return end - i;
 847 }
 848
 849
 850 int Paragraph::Private::latexSurrogatePair(otexstream & os, char_type c,
 851                 char_type next, OutputParams const & runparams)
 852 {
 853         // Writing next here may circumvent a possible font change between
 854         // c and next. Since next is only output if it forms a surrogate pair
 855         // with c we can ignore this:
 856         // A font change inside a surrogate pair does not make sense and is
 857         // hopefully impossible to input.
 858         // FIXME: change tracking
 859         // Is this correct WRT change tracking?
 860         Encoding const & encoding = *(runparams.encoding);
 861         docstring const latex1 = encoding.latexChar(next).first;
 862         docstring const latex2 = encoding.latexChar(c).first;
 863         if (docstring(1, next) == latex1) {
 864                 // the encoding supports the combination
 865                 os << latex2 << latex1;
 866                 return latex1.length() + latex2.length();
 867         } else if (runparams.local_font &&
 868                    runparams.local_font->language()->lang() == "polutonikogreek") {
 869                 // polutonikogreek only works without the brackets
 870                 os << latex1 << latex2;
 871                 return latex1.length() + latex2.length();
 872         } else
 873                 os << latex1 << '{' << latex2 << '}';
 874         return latex1.length() + latex2.length() + 2;
 875 }
 876
 877
 878 bool Paragraph::Private::simpleTeXBlanks(OutputParams const & runparams,
 879                                        otexstream & os,
 880                                        pos_type i,
 881                                        unsigned int & column,
 882                                        Font const & font,
 883                                        Layout const & style)
 884 {
 885         if (style.pass_thru || runparams.pass_thru)
 886                 return false;
 887
 888         if (i + 1 < int(text_.size())) {
 889                 char_type next = text_[i + 1];
 890                 if (Encodings::isCombiningChar(next)) {
 891                         // This space has an accent, so we must always output it.
 892                         column += latexSurrogatePair(os, ' ', next, runparams) - 1;
 893                         return true;
 894                 }
 895         }
 896
 897         if (runparams.linelen > 0
 898             && column > runparams.linelen
 899             && i
 900             && text_[i - 1] != ' '
 901             && (i + 1 < int(text_.size()))
 902             // same in FreeSpacing mode
 903             && !owner_->isFreeSpacing()
 904             // In typewriter mode, we want to avoid
 905             // ! . ? : at the end of a line
 906             && !(font.fontInfo().family() == TYPEWRITER_FAMILY
 907                  && (text_[i - 1] == '.'
 908                      || text_[i - 1] == '?'
 909                      || text_[i - 1] == ':'
 910                      || text_[i - 1] == '!'))) {
 911                 os << '\n';
 912                 os.texrow().start(owner_->id(), i + 1);
 913                 column = 0;
 914         } else if (style.free_spacing) {
 915                 os << '~';
 916         } else {
 917                 os << ' ';
 918         }
 919         return false;
 920 }
 921
 922
 923 int Paragraph::Private::writeScriptChars(otexstream & os,
 924                                          docstring const & ltx,
 925                                          Change const & runningChange,
 926                                          Encoding const & encoding,
 927                                          pos_type & i)
 928 {
 929         // FIXME: modifying i here is not very nice...
 930
 931         // We only arrive here when a proper language for character text_[i] has
 932         // not been specified (i.e., it could not be translated in the current
 933         // latex encoding) or its latex translation has been forced, and it
 934         // belongs to a known script.
 935         // Parameter ltx contains the latex translation of text_[i] as specified
 936         // in the unicodesymbols file and is something like "\textXXX{<spec>}".
 937         // The latex macro name "textXXX" specifies the script to which text_[i]
 938         // belongs and we use it in order to check whether characters from the
 939         // same script immediately follow, such that we can collect them in a
 940         // single "\textXXX" macro. So, we have to retain "\textXXX{<spec>"
 941         // for the first char but only "<spec>" for all subsequent chars.
 942         docstring::size_type const brace1 = ltx.find_first_of(from_ascii("{"));
 943         docstring::size_type const brace2 = ltx.find_last_of(from_ascii("}"));
 944         string script = to_ascii(ltx.substr(1, brace1 - 1));
 945         int pos = 0;
 946         int length = brace2;
 947         bool closing_brace = true;
 948         if (script == "textgreek" && encoding.latexName() == "iso-8859-7") {
 949                 // Correct encoding is being used, so we can avoid \textgreek.
 950                 pos = brace1 + 1;
 951                 length -= pos;
 952                 closing_brace = false;
 953         }
 954         os << ltx.substr(pos, length);
 955         int size = text_.size();
 956         while (i + 1 < size) {
 957                 char_type const next = text_[i + 1];
 958                 // Stop here if next character belongs to another script
 959                 // or there is a change in change tracking status.
 960                 if (!Encodings::isKnownScriptChar(next, script) ||
 961                     runningChange != owner_->lookupChange(i + 1))
 962                         break;
 963                 Font prev_font;
 964                 bool found = false;
 965                 FontList::const_iterator cit = fontlist_.begin();
 966                 FontList::const_iterator end = fontlist_.end();
 967                 for (; cit != end; ++cit) {
 968                         if (cit->pos() >= i && !found) {
 969                                 prev_font = cit->font();
 970                                 found = true;
 971                         }
 972                         if (cit->pos() >= i + 1)
 973                                 break;
 974                 }
 975                 // Stop here if there is a font attribute or encoding change.
 976                 if (found && cit != end && prev_font != cit->font())
 977                         break;
 978                 docstring const latex = encoding.latexChar(next).first;
 979                 docstring::size_type const b1 =
 980                                         latex.find_first_of(from_ascii("{"));
 981                 docstring::size_type const b2 =
 982                                         latex.find_last_of(from_ascii("}"));
 983                 int const len = b2 - b1 - 1;
 984                 os << latex.substr(b1 + 1, len);
 985                 length += len;
 986                 ++i;
 987         }
 988         if (closing_brace) {
 989                 os << '}';
 990                 ++length;
 991         }
 992         return length;
 993 }
 994
 995
 996 bool Paragraph::Private::isTextAt(string const & str, pos_type pos) const
 997 {
 998         pos_type const len = str.length();
 999
1000         // is the paragraph large enough?
1001         if (pos + len > int(text_.size()))
1002                 return false;
1003
1004         // does the wanted text start at point?
1005         for (string::size_type i = 0; i < str.length(); ++i) {
1006                 // Caution: direct comparison of characters works only
1007                 // because str is pure ASCII.
1008                 if (str[i] != text_[pos + i])
1009                         return false;
1010         }
1011
1012         return fontlist_.hasChangeInRange(pos, len);
1013 }
1014
1015
1016 void Paragraph::Private::latexInset(BufferParams const & bparams,
1017                                     otexstream & os,
1018                                     OutputParams & runparams,
1019                                     Font & running_font,
1020                                     Font & basefont,
1021                                     Font const & outerfont,
1022                                     bool & open_font,
1023                                     Change & running_change,
1024                                     Layout const & style,
1025                                     pos_type & i,
1026                                     unsigned int & column)
1027 {
1028         Inset * inset = owner_->getInset(i);
1029         LASSERT(inset, /**/);
1030
1031         if (style.pass_thru) {
1032                 inset->plaintext(os.os(), runparams);
1033                 return;
1034         }
1035
1036         // FIXME: move this to InsetNewline::latex
1037         if (inset->lyxCode() == NEWLINE_CODE) {
1038                 // newlines are handled differently here than
1039                 // the default in simpleTeXSpecialChars().
1040                 if (!style.newline_allowed) {
1041                         os << '\n';
1042                 } else {
1043                         if (open_font) {
1044                                 column += running_font.latexWriteEndChanges(
1045                                         os, bparams, runparams,
1046                                         basefont, basefont);
1047                                 open_font = false;
1048                         }
1049
1050                         if (running_font.fontInfo().family() == TYPEWRITER_FAMILY)
1051                                 os << '~';
1052
1053                         basefont = owner_->getLayoutFont(bparams, outerfont);
1054                         running_font = basefont;
1055
1056                         if (runparams.moving_arg)
1057                                 os << "\\protect ";
1058
1059                 }
1060                 os.texrow().start(owner_->id(), i + 1);
1061                 column = 0;
1062         }
1063
1064         if (owner_->isDeleted(i)) {
1065                 if( ++runparams.inDeletedInset == 1)
1066                         runparams.changeOfDeletedInset = owner_->lookupChange(i);
1067         }
1068
1069         if (inset->canTrackChanges()) {
1070                 column += Changes::latexMarkChange(os, bparams, running_change,
1071                         Change(Change::UNCHANGED), runparams);
1072                 running_change = Change(Change::UNCHANGED);
1073         }
1074
1075         bool close = false;
1076         odocstream::pos_type const len = os.os().tellp();
1077
1078         if (inset->forceLTR()
1079             && running_font.isRightToLeft()
1080             // ERT is an exception, it should be output with no
1081             // decorations at all
1082             && inset->lyxCode() != ERT_CODE) {
1083                 if (running_font.language()->lang() == "farsi")
1084                         os << "\\beginL{}";
1085                 else
1086                         os << "\\L{";
1087                 close = true;
1088         }
1089
1090         // FIXME: Bug: we can have an empty font change here!
1091         // if there has just been a font change, we are going to close it
1092         // right now, which means stupid latex code like \textsf{}. AFAIK,
1093         // this does not harm dvi output. A minor bug, thus (JMarc)
1094
1095         // Some insets cannot be inside a font change command.
1096         // However, even such insets *can* be placed in \L or \R
1097         // or their equivalents (for RTL language switches), so we don't
1098         // close the language in those cases.
1099         // ArabTeX, though, cannot handle this special behavior, it seems.
1100         bool arabtex = basefont.language()->lang() == "arabic_arabtex"
1101                 || running_font.language()->lang() == "arabic_arabtex";
1102         if (open_font && !inset->inheritFont()) {
1103                 bool closeLanguage = arabtex
1104                         || basefont.isRightToLeft() == running_font.isRightToLeft();
1105                 unsigned int count = running_font.latexWriteEndChanges(os,
1106                         bparams, runparams, basefont, basefont, closeLanguage);
1107                 column += count;
1108                 // if any font properties were closed, update the running_font,
1109                 // making sure, however, to leave the language as it was
1110                 if (count > 0) {
1111                         // FIXME: probably a better way to keep track of the old
1112                         // language, than copying the entire font?
1113                         Font const copy_font(running_font);
1114                         basefont = owner_->getLayoutFont(bparams, outerfont);
1115                         running_font = basefont;
1116                         if (!closeLanguage)
1117                                 running_font.setLanguage(copy_font.language());
1118                         // leave font open if language is still open
1119                         open_font = (running_font.language() == basefont.language());
1120                         if (closeLanguage)
1121                                 runparams.local_font = &basefont;
1122                 }
1123         }
1124
1125         int prev_rows = os.texrow().rows();
1126
1127         try {
1128                 runparams.lastid = id_;
1129                 runparams.lastpos = i;
1130                 inset->latex(os, runparams);
1131         } catch (EncodingException & e) {
1132                 // add location information and throw again.
1133                 e.par_id = id_;
1134                 e.pos = i;
1135                 throw(e);
1136         }
1137
1138         if (close) {
1139                 if (running_font.language()->lang() == "farsi")
1140                                 os << "\\endL{}";
1141                         else
1142                                 os << '}';
1143         }
1144
1145         if (os.texrow().rows() > prev_rows) {
1146                 os.texrow().start(owner_->id(), i + 1);
1147                 column = 0;
1148         } else {
1149                 column += (unsigned int)(os.os().tellp() - len);
1150         }
1151
1152         if (owner_->isDeleted(i))
1153                 --runparams.inDeletedInset;
1154 }
1155
1156
1157 void Paragraph::Private::latexSpecialChar(otexstream & os,
1158                                           OutputParams const & runparams,
1159                                           Font const & running_font,
1160                                           Change const & running_change,
1161                                           Layout const & style,
1162                                           pos_type & i,
1163                                           pos_type end_pos,
1164                                           unsigned int & column)
1165 {
1166         char_type const c = text_[i];
1167
1168         if (style.pass_thru || runparams.pass_thru) {
1169                 if (c != '\0') {
1170                         Encoding const * const enc = runparams.encoding;
1171                         if (enc && !enc->encodable(c))
1172                                 throw EncodingException(c);
1173                         os.put(c);
1174                 }
1175                 return;
1176         }
1177
1178         // TIPA uses its own T3 encoding
1179         if (runparams.inIPA && latexSpecialT3(c, os, i, column))
1180                 return;
1181         // If T1 font encoding is used, use the special
1182         // characters it provides.
1183         // NOTE: some languages reset the font encoding
1184         // internally
1185         if (!runparams.inIPA && !running_font.language()->internalFontEncoding()
1186             && lyxrc.fontenc == "T1" && latexSpecialT1(c, os, i, column))
1187                 return;
1188
1189         // \tt font needs special treatment
1190         if (!runparams.inIPA
1191              && running_font.fontInfo().family() == TYPEWRITER_FAMILY
1192              && latexSpecialTypewriter(c, os, i, column))
1193                 return;
1194
1195         // Otherwise, we use what LaTeX provides us.
1196         switch (c) {
1197         case '\\':
1198                 os << "\\textbackslash{}";
1199                 column += 15;
1200                 break;
1201         case '<':
1202                 os << "\\textless{}";
1203                 column += 10;
1204                 break;
1205         case '>':
1206                 os << "\\textgreater{}";
1207                 column += 13;
1208                 break;
1209         case '|':
1210                 os << "\\textbar{}";
1211                 column += 9;
1212                 break;
1213         case '-':
1214                 os << '-';
1215                 break;
1216         case '\"':
1217                 os << "\\char`\\\"{}";
1218                 column += 9;
1219                 break;
1220
1221         case '$': case '&':
1222         case '%': case '#': case '{':
1223         case '}': case '_':
1224                 os << '\\';
1225                 os.put(c);
1226                 column += 1;
1227                 break;
1228
1229         case '~':
1230                 os << "\\textasciitilde{}";
1231                 column += 16;
1232                 break;
1233
1234         case '^':
1235                 os << "\\textasciicircum{}";
1236                 column += 17;
1237                 break;
1238
1239         case '*':
1240         case '[':
1241         case ']':
1242                 // avoid being mistaken for optional arguments
1243                 os << '{';
1244                 os.put(c);
1245                 os << '}';
1246                 column += 2;
1247                 break;
1248
1249         case ' ':
1250                 // Blanks are printed before font switching.
1251                 // Sure? I am not! (try nice-latex)
1252                 // I am sure it's correct. LyX might be smarter
1253                 // in the future, but for now, nothing wrong is
1254                 // written. (Asger)
1255                 break;
1256
1257         default:
1258                 // LyX, LaTeX etc.
1259                 if (latexSpecialPhrase(os, i, end_pos, column, runparams))
1260                         return;
1261
1262                 if (c == '\0')
1263                         return;
1264
1265                 Encoding const & encoding = *(runparams.encoding);
1266                 char_type next = '\0';
1267                 if (i + 1 < int(text_.size())) {
1268                         next = text_[i + 1];
1269                         if (Encodings::isCombiningChar(next)) {
1270                                 column += latexSurrogatePair(os, c, next, runparams) - 1;
1271                                 ++i;
1272                                 break;
1273                         }
1274                 }
1275                 string script;
1276                 pair<docstring, bool> latex = encoding.latexChar(c);
1277                 docstring nextlatex;
1278                 bool nexttipas = false;
1279                 string nexttipashortcut;
1280                 if (next != '\0' && next != META_INSET) {
1281                         nextlatex = encoding.latexChar(next).first;
1282                         if (runparams.inIPA) {
1283                                 nexttipashortcut = Encodings::TIPAShortcut(next);
1284                                 nexttipas = !nexttipashortcut.empty();
1285                         }
1286                 }
1287                 bool tipas = false;
1288                 if (runparams.inIPA) {
1289                         string const tipashortcut = Encodings::TIPAShortcut(c);
1290                         if (!tipashortcut.empty()) {
1291                                 latex.first = from_ascii(tipashortcut);
1292                                 latex.second = false;
1293                                 tipas = true;
1294                         }
1295                 }
1296                 if (Encodings::isKnownScriptChar(c, script)
1297                     && prefixIs(latex.first, from_ascii("\\" + script)))
1298                         column += writeScriptChars(os, latex.first,
1299                                         running_change, encoding, i) - 1;
1300                 else if (latex.second
1301                          && ((!prefixIs(nextlatex, '\\')
1302                                && !prefixIs(nextlatex, '{')
1303                                && !prefixIs(nextlatex, '}'))
1304                              || (nexttipas
1305                                  && !prefixIs(from_ascii(nexttipashortcut), '\\')))
1306                          && !tipas) {
1307                         // Prevent eating of a following
1308                         // space or command corruption by
1309                         // following characters
1310                         if (next == ' ' || next == '\0') {
1311                                 column += latex.first.length() + 1;
1312                                 os << latex.first << "{}";
1313                         } else {
1314                                 column += latex.first.length();
1315                                 os << latex.first << " ";
1316                         }
1317                 } else {
1318                         column += latex.first.length() - 1;
1319                         os << latex.first;
1320                 }
1321                 break;
1322         }
1323 }
1324
1325
1326 bool Paragraph::Private::latexSpecialT1(char_type const c, otexstream & os,
1327         pos_type i, unsigned int & column)
1328 {
1329         switch (c) {
1330         case '>':
1331         case '<':
1332                 os.put(c);
1333                 // In T1 encoding, these characters exist
1334                 // but we should avoid ligatures
1335                 if (i + 1 >= int(text_.size()) || text_[i + 1] != c)
1336                         return true;
1337                 os << "\\textcompwordmark{}";
1338                 column += 19;
1339                 return true;
1340         case '|':
1341                 os.put(c);
1342                 return true;
1343         case '\"':
1344                 // soul.sty breaks with \char`\"
1345                 os << "\\textquotedbl{}";
1346                 column += 14;
1347                 return true;
1348         default:
1349                 return false;
1350         }
1351 }
1352
1353
1354 bool Paragraph::Private::latexSpecialT3(char_type const c, otexstream & os,
1355         pos_type /*i*/, unsigned int & /*column*/)
1356 {
1357         switch (c) {
1358         case '*':
1359         case '[':
1360         case ']':
1361         case '|':
1362         case '\"':
1363                 os.put(c);
1364                 return true;
1365         default:
1366                 return false;
1367         }
1368 }
1369
1370
1371 bool Paragraph::Private::latexSpecialTypewriter(char_type const c, otexstream & os,
1372         pos_type i, unsigned int & column)
1373 {
1374         switch (c) {
1375         case '-':
1376                 // within \ttfamily, "--" is merged to "-" (no endash)
1377                 // so we avoid this rather irritating ligature
1378                 if (i + 1 < int(text_.size()) && text_[i + 1] == '-') {
1379                         os << "-{}";
1380                         column += 2;
1381                 } else
1382                         os << '-';
1383                 return true;
1384
1385         // everything else has to be checked separately
1386         // (depending on the encoding)
1387         default:
1388                 return false;
1389         }
1390 }
1391
1392
1393 /// \param end_pos
1394 ///   If [start_pos, end_pos) does not include entirely the special phrase, then
1395 ///   do not apply the macro transformation.
1396 bool Paragraph::Private::latexSpecialPhrase(otexstream & os, pos_type & i, pos_type end_pos,
1397         unsigned int & column, OutputParams const & runparams)
1398 {
1399         // FIXME: if we have "LaTeX" with a font
1400         // change in the middle (before the 'T', then
1401         // the "TeX" part is still special cased.
1402         // Really we should only operate this on
1403         // "words" for some definition of word
1404
1405         for (size_t pnr = 0; pnr < phrases_nr; ++pnr) {
1406                 if (!isTextAt(special_phrases[pnr].phrase, i)
1407                     || (end_pos != -1 && i + int(special_phrases[pnr].phrase.size()) > end_pos))
1408                         continue;
1409                 if (runparams.moving_arg)
1410                         os << "\\protect";
1411                 os << special_phrases[pnr].macro;
1412                 i += special_phrases[pnr].phrase.length() - 1;
1413                 column += special_phrases[pnr].macro.length() - 1;
1414                 return true;
1415         }
1416         return false;
1417 }
1418
1419
1420 void Paragraph::Private::validate(LaTeXFeatures & features) const
1421 {
1422         if (layout_->inpreamble && inset_owner_) {
1423                 bool const is_command = layout_->latextype == LATEX_COMMAND;
1424                 Buffer const & buf = inset_owner_->buffer();
1425                 BufferParams const & bp = buf.params();
1426                 Font f;
1427                 TexRow texrow;
1428                 // Using a string stream here circumvents the encoding
1429                 // switching machinery of odocstream. Therefore the
1430                 // output is wrong if this paragraph contains content
1431                 // that needs to switch encoding.
1432                 odocstringstream ods;
1433                 otexstream os(ods, texrow);
1434                 if (is_command) {
1435                         os << '\\' << from_ascii(layout_->latexname());
1436                         // we have to provide all the optional arguments here, even though
1437                         // the last one is the only one we care about.
1438                         // Separate handling of optional argument inset.
1439                         if (layout_->optargs != 0 || layout_->reqargs != 0)
1440                                 latexArgInsets(*owner_, os, features.runparams(),
1441                                         layout_->reqargs, layout_->optargs);
1442                         else
1443                                 os << from_ascii(layout_->latexparam());
1444                 }
1445                 docstring::size_type const length = ods.str().length();
1446                 // this will output "{" at the beginning, but not at the end
1447                 owner_->latex(bp, f, os, features.runparams(), 0, -1, true);
1448                 if (ods.str().length() > length) {
1449                         if (is_command)
1450                                 ods << '}';
1451                         string const snippet = to_utf8(ods.str());
1452                         features.addPreambleSnippet(snippet);
1453                 }
1454         }
1455
1456         if (features.runparams().flavor == OutputParams::HTML
1457             && layout_->htmltitle()) {
1458                 features.setHTMLTitle(owner_->asString(AS_STR_INSETS | AS_STR_SKIPDELETE));
1459         }
1460
1461         // check the params.
1462         if (!params_.spacing().isDefault())
1463                 features.require("setspace");
1464
1465         // then the layouts
1466         features.useLayout(layout_->name());
1467
1468         // then the fonts
1469         fontlist_.validate(features);
1470
1471         // then the indentation
1472         if (!params_.leftIndent().zero())
1473                 features.require("ParagraphLeftIndent");
1474
1475         // then the insets
1476         InsetList::const_iterator icit = insetlist_.begin();
1477         InsetList::const_iterator iend = insetlist_.end();
1478         for (; icit != iend; ++icit) {
1479                 if (icit->inset) {
1480                         icit->inset->validate(features);
1481                         if (layout_->needprotect &&
1482                             icit->inset->lyxCode() == FOOT_CODE)
1483                                 features.require("NeedLyXFootnoteCode");
1484                 }
1485         }
1486
1487         // then the contents
1488         for (pos_type i = 0; i < int(text_.size()) ; ++i) {
1489                 for (size_t pnr = 0; pnr < phrases_nr; ++pnr) {
1490                         if (!special_phrases[pnr].builtin
1491                             && isTextAt(special_phrases[pnr].phrase, i)) {
1492                                 features.require(special_phrases[pnr].phrase);
1493                                 break;
1494                         }
1495                 }
1496                 Encodings::validate(text_[i], features);
1497         }
1498 }
1499
1500 /////////////////////////////////////////////////////////////////////
1501 //
1502 // Paragraph
1503 //
1504 /////////////////////////////////////////////////////////////////////
1505
1506 namespace {
1507         Layout const emptyParagraphLayout;
1508 }
1509
1510 Paragraph::Paragraph()
1511         : d(new Paragraph::Private(this, emptyParagraphLayout))
1512 {
1513         itemdepth = 0;
1514         d->params_.clear();
1515 }
1516
1517
1518 Paragraph::Paragraph(Paragraph const & par)
1519         : itemdepth(par.itemdepth),
1520         d(new Paragraph::Private(*par.d, this))
1521 {
1522         registerWords();
1523 }
1524
1525
1526 Paragraph::Paragraph(Paragraph const & par, pos_type beg, pos_type end)
1527         : itemdepth(par.itemdepth),
1528         d(new Paragraph::Private(*par.d, this, beg, end))
1529 {
1530         registerWords();
1531 }
1532
1533
1534 Paragraph & Paragraph::operator=(Paragraph const & par)
1535 {
1536         // needed as we will destroy the private part before copying it
1537         if (&par != this) {
1538                 itemdepth = par.itemdepth;
1539
1540                 deregisterWords();
1541                 delete d;
1542                 d = new Private(*par.d, this);
1543                 registerWords();
1544         }
1545         return *this;
1546 }
1547
1548
1549 Paragraph::~Paragraph()
1550 {
1551         deregisterWords();
1552         delete d;
1553 }
1554
1555
1556 namespace {
1557
1558 // this shall be called just before every "os << ..." action.
1559 void flushString(ostream & os, docstring & s)
1560 {
1561         os << to_utf8(s);
1562         s.erase();
1563 }
1564
1565 }
1566
1567
1568 void Paragraph::write(ostream & os, BufferParams const & bparams,
1569         depth_type & dth) const
1570 {
1571         // The beginning or end of a deeper (i.e. nested) area?
1572         if (dth != d->params_.depth()) {
1573                 if (d->params_.depth() > dth) {
1574                         while (d->params_.depth() > dth) {
1575                                 os << "\n\\begin_deeper";
1576                                 ++dth;
1577                         }
1578                 } else {
1579                         while (d->params_.depth() < dth) {
1580                                 os << "\n\\end_deeper";
1581                                 --dth;
1582                         }
1583                 }
1584         }
1585
1586         // First write the layout
1587         os << "\n\\begin_layout " << to_utf8(d->layout_->name()) << '\n';
1588
1589         d->params_.write(os);
1590
1591         Font font1(inherit_font, bparams.language);
1592
1593         Change running_change = Change(Change::UNCHANGED);
1594
1595         // this string is used as a buffer to avoid repetitive calls
1596         // to to_utf8(), which turn out to be expensive (JMarc)
1597         docstring write_buffer;
1598
1599         int column = 0;
1600         for (pos_type i = 0; i <= size(); ++i) {
1601
1602                 Change const change = lookupChange(i);
1603                 if (change != running_change)
1604                         flushString(os, write_buffer);
1605                 Changes::lyxMarkChange(os, bparams, column, running_change, change);
1606                 running_change = change;
1607
1608                 if (i == size())
1609                         break;
1610
1611                 // Write font changes
1612                 Font font2 = getFontSettings(bparams, i);
1613                 if (font2 != font1) {
1614                         flushString(os, write_buffer);
1615                         font2.lyxWriteChanges(font1, os);
1616                         column = 0;
1617                         font1 = font2;
1618                 }
1619
1620                 char_type const c = d->text_[i];
1621                 switch (c) {
1622                 case META_INSET:
1623                         if (Inset const * inset = getInset(i)) {
1624                                 flushString(os, write_buffer);
1625                                 if (inset->directWrite()) {
1626                                         // international char, let it write
1627                                         // code directly so it's shorter in
1628                                         // the file
1629                                         inset->write(os);
1630                                 } else {
1631                                         if (i)
1632                                                 os << '\n';
1633                                         os << "\\begin_inset ";
1634                                         inset->write(os);
1635                                         os << "\n\\end_inset\n\n";
1636                                         column = 0;
1637                                 }
1638                         }
1639                         break;
1640                 case '\\':
1641                         flushString(os, write_buffer);
1642                         os << "\n\\backslash\n";
1643                         column = 0;
1644                         break;
1645                 case '.':
1646                         flushString(os, write_buffer);
1647                         if (i + 1 < size() && d->text_[i + 1] == ' ') {
1648                                 os << ".\n";
1649                                 column = 0;
1650                         } else
1651                                 os << '.';
1652                         break;
1653                 default:
1654                         if ((column > 70 && c == ' ')
1655                             || column > 79) {
1656                                 flushString(os, write_buffer);
1657                                 os << '\n';
1658                                 column = 0;
1659                         }
1660                         // this check is to amend a bug. LyX sometimes
1661                         // inserts '\0' this could cause problems.
1662                         if (c != '\0')
1663                                 write_buffer.push_back(c);
1664                         else
1665                                 LYXERR0("NUL char in structure.");
1666                         ++column;
1667                         break;
1668                 }
1669         }
1670
1671         flushString(os, write_buffer);
1672         os << "\n\\end_layout\n";
1673 }
1674
1675
1676 void Paragraph::validate(LaTeXFeatures & features) const
1677 {
1678         d->validate(features);
1679 }
1680
1681
1682 void Paragraph::insert(pos_type start, docstring const & str,
1683                        Font const & font, Change const & change)
1684 {
1685         for (size_t i = 0, n = str.size(); i != n ; ++i)
1686                 insertChar(start + i, str[i], font, change);
1687 }
1688
1689
1690 void Paragraph::appendChar(char_type c, Font const & font,
1691                 Change const & change)
1692 {
1693         // track change
1694         d->changes_.insert(change, d->text_.size());
1695         // when appending characters, no need to update tables
1696         d->text_.push_back(c);
1697         setFont(d->text_.size() - 1, font);
1698         d->requestSpellCheck(d->text_.size() - 1);
1699 }
1700
1701
1702 void Paragraph::appendString(docstring const & s, Font const & font,
1703                 Change const & change)
1704 {
1705         pos_type end = s.size();
1706         size_t oldsize = d->text_.size();
1707         size_t newsize = oldsize + end;
1708         size_t capacity = d->text_.capacity();
1709         if (newsize >= capacity)
1710                 d->text_.reserve(max(capacity + 100, newsize));
1711
1712         // when appending characters, no need to update tables
1713         d->text_.append(s);
1714
1715         // FIXME: Optimize this!
1716         for (size_t i = oldsize; i != newsize; ++i) {
1717                 // track change
1718                 d->changes_.insert(change, i);
1719                 d->requestSpellCheck(i);
1720         }
1721         d->fontlist_.set(oldsize, font);
1722         d->fontlist_.set(newsize - 1, font);
1723 }
1724
1725
1726 void Paragraph::insertChar(pos_type pos, char_type c,
1727                            bool trackChanges)
1728 {
1729         d->insertChar(pos, c, Change(trackChanges ?
1730                            Change::INSERTED : Change::UNCHANGED));
1731 }
1732
1733
1734 void Paragraph::insertChar(pos_type pos, char_type c,
1735                            Font const & font, bool trackChanges)
1736 {
1737         d->insertChar(pos, c, Change(trackChanges ?
1738                            Change::INSERTED : Change::UNCHANGED));
1739         setFont(pos, font);
1740 }
1741
1742
1743 void Paragraph::insertChar(pos_type pos, char_type c,
1744                            Font const & font, Change const & change)
1745 {
1746         d->insertChar(pos, c, change);
1747         setFont(pos, font);
1748 }
1749
1750
1751 bool Paragraph::insertInset(pos_type pos, Inset * inset,
1752                             Font const & font, Change const & change)
1753 {
1754         bool const success = insertInset(pos, inset, change);
1755         // Set the font/language of the inset...
1756         setFont(pos, font);
1757         return success;
1758 }
1759
1760
1761 void Paragraph::resetFonts(Font const & font)
1762 {
1763         d->fontlist_.clear();
1764         d->fontlist_.set(0, font);
1765         d->fontlist_.set(d->text_.size() - 1, font);
1766 }
1767
1768 // Gets uninstantiated font setting at position.
1769 Font const & Paragraph::getFontSettings(BufferParams const & bparams,
1770                                          pos_type pos) const
1771 {
1772         if (pos > size()) {
1773                 LYXERR0("pos: " << pos << " size: " << size());
1774                 LASSERT(pos <= size(), /**/);
1775         }
1776
1777         FontList::const_iterator cit = d->fontlist_.fontIterator(pos);
1778         if (cit != d->fontlist_.end())
1779                 return cit->font();
1780
1781         if (pos == size() && !empty())
1782                 return getFontSettings(bparams, pos - 1);
1783
1784         // Optimisation: avoid a full font instantiation if there is no
1785         // language change from previous call.
1786         static Font previous_font;
1787         static Language const * previous_lang = 0;
1788         Language const * lang = getParLanguage(bparams);
1789         if (lang != previous_lang) {
1790                 previous_lang = lang;
1791                 previous_font = Font(inherit_font, lang);
1792         }
1793         return previous_font;
1794 }
1795
1796
1797 FontSpan Paragraph::fontSpan(pos_type pos) const
1798 {
1799         LASSERT(pos <= size(), /**/);
1800         pos_type start = 0;
1801
1802         FontList::const_iterator cit = d->fontlist_.begin();
1803         FontList::const_iterator end = d->fontlist_.end();
1804         for (; cit != end; ++cit) {
1805                 if (cit->pos() >= pos) {
1806                         if (pos >= beginOfBody())
1807                                 return FontSpan(max(start, beginOfBody()),
1808                                                 cit->pos());
1809                         else
1810                                 return FontSpan(start,
1811                                                 min(beginOfBody() - 1,
1812                                                          cit->pos()));
1813                 }
1814                 start = cit->pos() + 1;
1815         }
1816
1817         // This should not happen, but if so, we take no chances.
1818         // LYXERR0("Paragraph::getEndPosOfFontSpan: This should not happen!");
1819         return FontSpan(pos, pos);
1820 }
1821
1822
1823 // Gets uninstantiated font setting at position 0
1824 Font const & Paragraph::getFirstFontSettings(BufferParams const & bparams) const
1825 {
1826         if (!empty() && !d->fontlist_.empty())
1827                 return d->fontlist_.begin()->font();
1828
1829         // Optimisation: avoid a full font instantiation if there is no
1830         // language change from previous call.
1831         static Font previous_font;
1832         static Language const * previous_lang = 0;
1833         if (bparams.language != previous_lang) {
1834                 previous_lang = bparams.language;
1835                 previous_font = Font(inherit_font, bparams.language);
1836         }
1837
1838         return previous_font;
1839 }
1840
1841
1842 // Gets the fully instantiated font at a given position in a paragraph
1843 // This is basically the same function as Text::GetFont() in text2.cpp.
1844 // The difference is that this one is used for generating the LaTeX file,
1845 // and thus cosmetic "improvements" are disallowed: This has to deliver
1846 // the true picture of the buffer. (Asger)
1847 Font const Paragraph::getFont(BufferParams const & bparams, pos_type pos,
1848                                  Font const & outerfont) const
1849 {
1850         LASSERT(pos >= 0, /**/);
1851
1852         Font font = getFontSettings(bparams, pos);
1853
1854         pos_type const body_pos = beginOfBody();
1855         FontInfo & fi = font.fontInfo();
1856         if (pos < body_pos)
1857                 fi.realize(d->layout_->labelfont);
1858         else
1859                 fi.realize(d->layout_->font);
1860
1861         fi.realize(outerfont.fontInfo());
1862         fi.realize(bparams.getFont().fontInfo());
1863
1864         return font;
1865 }
1866
1867
1868 Font const Paragraph::getLabelFont
1869         (BufferParams const & bparams, Font const & outerfont) const
1870 {
1871         FontInfo tmpfont = d->layout_->labelfont;
1872         tmpfont.realize(outerfont.fontInfo());
1873         tmpfont.realize(bparams.getFont().fontInfo());
1874         return Font(tmpfont, getParLanguage(bparams));
1875 }
1876
1877
1878 Font const Paragraph::getLayoutFont
1879         (BufferParams const & bparams, Font const & outerfont) const
1880 {
1881         FontInfo tmpfont = d->layout_->font;
1882         tmpfont.realize(outerfont.fontInfo());
1883         tmpfont.realize(bparams.getFont().fontInfo());
1884         return Font(tmpfont, getParLanguage(bparams));
1885 }
1886
1887
1888 /// Returns the height of the highest font in range
1889 FontSize Paragraph::highestFontInRange
1890         (pos_type startpos, pos_type endpos, FontSize def_size) const
1891 {
1892         return d->fontlist_.highestInRange(startpos, endpos, def_size);
1893 }
1894
1895
1896 char_type Paragraph::getUChar(BufferParams const & bparams, pos_type pos) const
1897 {
1898         char_type c = d->text_[pos];
1899         if (!lyxrc.rtl_support)
1900                 return c;
1901
1902         char_type uc = c;
1903         switch (c) {
1904         case '(':
1905                 uc = ')';
1906                 break;
1907         case ')':
1908                 uc = '(';
1909                 break;
1910         case '[':
1911                 uc = ']';
1912                 break;
1913         case ']':
1914                 uc = '[';
1915                 break;
1916         case '{':
1917                 uc = '}';
1918                 break;
1919         case '}':
1920                 uc = '{';
1921                 break;
1922         case '<':
1923                 uc = '>';
1924                 break;
1925         case '>':
1926                 uc = '<';
1927                 break;
1928         }
1929         if (uc != c && getFontSettings(bparams, pos).isRightToLeft())
1930                 return uc;
1931         return c;
1932 }
1933
1934
1935 void Paragraph::setFont(pos_type pos, Font const & font)
1936 {
1937         LASSERT(pos <= size(), /**/);
1938
1939         // First, reduce font against layout/label font
1940         // Update: The setCharFont() routine in text2.cpp already
1941         // reduces font, so we don't need to do that here. (Asger)
1942
1943         d->fontlist_.set(pos, font);
1944 }
1945
1946
1947 void Paragraph::makeSameLayout(Paragraph const & par)
1948 {
1949         d->layout_ = par.d->layout_;
1950         d->params_ = par.d->params_;
1951 }
1952
1953
1954 bool Paragraph::stripLeadingSpaces(bool trackChanges)
1955 {
1956         if (isFreeSpacing())
1957                 return false;
1958
1959         int pos = 0;
1960         int count = 0;
1961
1962         while (pos < size() && (isNewline(pos) || isLineSeparator(pos))) {
1963                 if (eraseChar(pos, trackChanges))
1964                         ++count;
1965                 else
1966                         ++pos;
1967         }
1968
1969         return count > 0 || pos > 0;
1970 }
1971
1972
1973 bool Paragraph::hasSameLayout(Paragraph const & par) const
1974 {
1975         return par.d->layout_ == d->layout_
1976                 && d->params_.sameLayout(par.d->params_);
1977 }
1978
1979
1980 depth_type Paragraph::getDepth() const
1981 {
1982         return d->params_.depth();
1983 }
1984
1985
1986 depth_type Paragraph::getMaxDepthAfter() const
1987 {
1988         if (d->layout_->isEnvironment())
1989                 return d->params_.depth() + 1;
1990         else
1991                 return d->params_.depth();
1992 }
1993
1994
1995 char Paragraph::getAlign() const
1996 {
1997         if (d->params_.align() == LYX_ALIGN_LAYOUT)
1998                 return d->layout_->align;
1999         else
2000                 return d->params_.align();
2001 }
2002
2003
2004 docstring const & Paragraph::labelString() const
2005 {
2006         return d->params_.labelString();
2007 }
2008
2009
2010 // the next two functions are for the manual labels
2011 docstring const Paragraph::getLabelWidthString() const
2012 {
2013         if (d->layout_->margintype == MARGIN_MANUAL
2014             || d->layout_->latextype == LATEX_BIB_ENVIRONMENT)
2015                 return d->params_.labelWidthString();
2016         else
2017                 return _("Senseless with this layout!");
2018 }
2019
2020
2021 void Paragraph::setLabelWidthString(docstring const & s)
2022 {
2023         d->params_.labelWidthString(s);
2024 }
2025
2026
2027 docstring Paragraph::expandLabel(Layout const & layout,
2028                 BufferParams const & bparams) const
2029 {
2030         return expandParagraphLabel(layout, bparams, true);
2031 }
2032
2033
2034 docstring Paragraph::expandDocBookLabel(Layout const & layout,
2035                 BufferParams const & bparams) const
2036 {
2037         return expandParagraphLabel(layout, bparams, false);
2038 }
2039
2040
2041 docstring Paragraph::expandParagraphLabel(Layout const & layout,
2042                 BufferParams const & bparams, bool process_appendix) const
2043 {
2044         DocumentClass const & tclass = bparams.documentClass();
2045         string const & lang = getParLanguage(bparams)->code();
2046         bool const in_appendix = process_appendix && d->params_.appendix();
2047         docstring fmt = translateIfPossible(layout.labelstring(in_appendix), lang);
2048
2049         if (fmt.empty() && layout.labeltype == LABEL_COUNTER
2050             && !layout.counter.empty())
2051                 return tclass.counters().theCounter(layout.counter, lang);
2052
2053         // handle 'inherited level parts' in 'fmt',
2054         // i.e. the stuff between '@' in   '@Section@.\arabic{subsection}'
2055         size_t const i = fmt.find('@', 0);
2056         if (i != docstring::npos) {
2057                 size_t const j = fmt.find('@', i + 1);
2058                 if (j != docstring::npos) {
2059                         docstring parent(fmt, i + 1, j - i - 1);
2060                         docstring label = from_ascii("??");
2061                         if (tclass.hasLayout(parent))
2062                                 docstring label = expandParagraphLabel(tclass[parent], bparams,
2063                                                       process_appendix);
2064                         fmt = docstring(fmt, 0, i) + label
2065                                 + docstring(fmt, j + 1, docstring::npos);
2066                 }
2067         }
2068
2069         return tclass.counters().counterLabel(fmt, lang);
2070 }
2071
2072
2073 void Paragraph::applyLayout(Layout const & new_layout)
2074 {
2075         d->layout_ = &new_layout;
2076         LyXAlignment const oldAlign = d->params_.align();
2077
2078         if (!(oldAlign & d->layout_->alignpossible)) {
2079                 frontend::Alert::warning(_("Alignment not permitted"),
2080                         _("The new layout does not permit the alignment previously used.\nSetting to default."));
2081                 d->params_.align(LYX_ALIGN_LAYOUT);
2082         }
2083 }
2084
2085
2086 pos_type Paragraph::beginOfBody() const
2087 {
2088         return d->begin_of_body_;
2089 }
2090
2091
2092 void Paragraph::setBeginOfBody()
2093 {
2094         if (d->layout_->labeltype != LABEL_MANUAL) {
2095                 d->begin_of_body_ = 0;
2096                 return;
2097         }
2098
2099         // Unroll the first two cycles of the loop
2100         // and remember the previous character to
2101         // remove unnecessary getChar() calls
2102         pos_type i = 0;
2103         pos_type end = size();
2104         if (i < end && !isNewline(i)) {
2105                 ++i;
2106                 char_type previous_char = 0;
2107                 char_type temp = 0;
2108                 if (i < end) {
2109                         previous_char = d->text_[i];
2110                         if (!isNewline(i)) {
2111                                 ++i;
2112                                 while (i < end && previous_char != ' ') {
2113                                         temp = d->text_[i];
2114                                         if (isNewline(i))
2115                                                 break;
2116                                         ++i;
2117                                         previous_char = temp;
2118                                 }
2119                         }
2120                 }
2121         }
2122
2123         d->begin_of_body_ = i;
2124 }
2125
2126
2127 bool Paragraph::allowParagraphCustomization() const
2128 {
2129         return inInset().allowParagraphCustomization();
2130 }
2131
2132
2133 bool Paragraph::usePlainLayout() const
2134 {
2135         return inInset().usePlainLayout();
2136 }
2137
2138
2139 bool Paragraph::isPassThru() const
2140 {
2141         return inInset().getLayout().isPassThru() || d->layout_->pass_thru;
2142 }
2143
2144 namespace {
2145
2146 // paragraphs inside floats need different alignment tags to avoid
2147 // unwanted space
2148
2149 bool noTrivlistCentering(InsetCode code)
2150 {
2151         return code == FLOAT_CODE
2152                || code == WRAP_CODE
2153                || code == CELL_CODE;
2154 }
2155
2156
2157 string correction(string const & orig)
2158 {
2159         if (orig == "flushleft")
2160                 return "raggedright";
2161         if (orig == "flushright")
2162                 return "raggedleft";
2163         if (orig == "center")
2164                 return "centering";
2165         return orig;
2166 }
2167
2168
2169 string const corrected_env(string const & suffix, string const & env,
2170         InsetCode code, bool const lastpar)
2171 {
2172         string output = suffix + "{";
2173         if (noTrivlistCentering(code)) {
2174                 if (lastpar) {
2175                         // the last paragraph in non-trivlist-aligned
2176                         // context is special (to avoid unwanted whitespace)
2177                         if (suffix == "\\begin")
2178                                 return "\\" + correction(env) + "{}";
2179                         return string();
2180                 }
2181                 output += correction(env);
2182         } else
2183                 output += env;
2184         output += "}";
2185         if (suffix == "\\begin")
2186                 output += "\n";
2187         return output;
2188 }
2189
2190
2191 void adjust_column(string const & str, int & column)
2192 {
2193         if (!contains(str, "\n"))
2194                 column += str.size();
2195         else {
2196                 string tmp;
2197                 column = rsplit(str, tmp, '\n').size();
2198         }
2199 }
2200
2201 } // namespace anon
2202
2203
2204 int Paragraph::Private::startTeXParParams(BufferParams const & bparams,
2205                         otexstream & os, OutputParams const & runparams) const
2206 {
2207         int column = 0;
2208
2209         if (params_.noindent() && !layout_->pass_thru) {
2210                 os << "\\noindent ";
2211                 column += 10;
2212         }
2213
2214         LyXAlignment const curAlign = params_.align();
2215
2216         if (curAlign == layout_->align)
2217                 return column;
2218
2219         switch (curAlign) {
2220         case LYX_ALIGN_NONE:
2221         case LYX_ALIGN_BLOCK:
2222         case LYX_ALIGN_LAYOUT:
2223         case LYX_ALIGN_SPECIAL:
2224         case LYX_ALIGN_DECIMAL:
2225                 break;
2226         case LYX_ALIGN_LEFT:
2227         case LYX_ALIGN_RIGHT:
2228         case LYX_ALIGN_CENTER:
2229                 if (runparams.moving_arg) {
2230                         os << "\\protect";
2231                         column += 8;
2232                 }
2233                 break;
2234         }
2235
2236         string const begin_tag = "\\begin";
2237         InsetCode code = ownerCode();
2238         bool const lastpar = runparams.isLastPar;
2239
2240         switch (curAlign) {
2241         case LYX_ALIGN_NONE:
2242         case LYX_ALIGN_BLOCK:
2243         case LYX_ALIGN_LAYOUT:
2244         case LYX_ALIGN_SPECIAL:
2245         case LYX_ALIGN_DECIMAL:
2246                 break;
2247         case LYX_ALIGN_LEFT: {
2248                 string output;
2249                 if (owner_->getParLanguage(bparams)->babel() != "hebrew")
2250                         output = corrected_env(begin_tag, "flushleft", code, lastpar);
2251                 else
2252                         output = corrected_env(begin_tag, "flushright", code, lastpar);
2253                 os << from_ascii(output);
2254                 adjust_column(output, column);
2255                 break;
2256         } case LYX_ALIGN_RIGHT: {
2257                 string output;
2258                 if (owner_->getParLanguage(bparams)->babel() != "hebrew")
2259                         output = corrected_env(begin_tag, "flushright", code, lastpar);
2260                 else
2261                         output = corrected_env(begin_tag, "flushleft", code, lastpar);
2262                 os << from_ascii(output);
2263                 adjust_column(output, column);
2264                 break;
2265         } case LYX_ALIGN_CENTER: {
2266                 string output;
2267                 output = corrected_env(begin_tag, "center", code, lastpar);
2268                 os << from_ascii(output);
2269                 adjust_column(output, column);
2270                 break;
2271         }
2272         }
2273
2274         return column;
2275 }
2276
2277
2278 bool Paragraph::Private::endTeXParParams(BufferParams const & bparams,
2279                         otexstream & os, OutputParams const & runparams) const
2280 {
2281         LyXAlignment const curAlign = params_.align();
2282
2283         if (curAlign == layout_->align)
2284                 return false;
2285
2286         switch (curAlign) {
2287         case LYX_ALIGN_NONE:
2288         case LYX_ALIGN_BLOCK:
2289         case LYX_ALIGN_LAYOUT:
2290         case LYX_ALIGN_SPECIAL:
2291         case LYX_ALIGN_DECIMAL:
2292                 break;
2293         case LYX_ALIGN_LEFT:
2294         case LYX_ALIGN_RIGHT:
2295         case LYX_ALIGN_CENTER:
2296                 if (runparams.moving_arg)
2297                         os << "\\protect";
2298                 break;
2299         }
2300
2301         string output;
2302         string const end_tag = "\n\\par\\end";
2303         InsetCode code = ownerCode();
2304         bool const lastpar = runparams.isLastPar;
2305
2306         switch (curAlign) {
2307         case LYX_ALIGN_NONE:
2308         case LYX_ALIGN_BLOCK:
2309         case LYX_ALIGN_LAYOUT:
2310         case LYX_ALIGN_SPECIAL:
2311         case LYX_ALIGN_DECIMAL:
2312                 break;
2313         case LYX_ALIGN_LEFT: {
2314                 if (owner_->getParLanguage(bparams)->babel() != "hebrew")
2315                         output = corrected_env(end_tag, "flushleft", code, lastpar);
2316                 else
2317                         output = corrected_env(end_tag, "flushright", code, lastpar);
2318                 os << from_ascii(output);
2319                 break;
2320         } case LYX_ALIGN_RIGHT: {
2321                 if (owner_->getParLanguage(bparams)->babel() != "hebrew")
2322                         output = corrected_env(end_tag, "flushright", code, lastpar);
2323                 else
2324                         output = corrected_env(end_tag, "flushleft", code, lastpar);
2325                 os << from_ascii(output);
2326                 break;
2327         } case LYX_ALIGN_CENTER: {
2328                 output = corrected_env(end_tag, "center", code, lastpar);
2329                 os << from_ascii(output);
2330                 break;
2331         }
2332         }
2333
2334         return !output.empty() || lastpar;
2335 }
2336
2337
2338 // This one spits out the text of the paragraph
2339 void Paragraph::latex(BufferParams const & bparams,
2340         Font const & outerfont,
2341         otexstream & os,
2342         OutputParams const & runparams,
2343         int start_pos, int end_pos, bool force) const
2344 {
2345         LYXERR(Debug::LATEX, "Paragraph::latex...     " << this);
2346
2347         // FIXME This check should not be needed. Perhaps issue an
2348         // error if it triggers.
2349         Layout const & style = inInset().forcePlainLayout() ?
2350                 bparams.documentClass().plainLayout() : *d->layout_;
2351
2352         if (!force && style.inpreamble)
2353                 return;
2354
2355         bool const allowcust = allowParagraphCustomization();
2356
2357         // Current base font for all inherited font changes, without any
2358         // change caused by an individual character, except for the language:
2359         // It is set to the language of the first character.
2360         // As long as we are in the label, this font is the base font of the
2361         // label. Before the first body character it is set to the base font
2362         // of the body.
2363         Font basefont;
2364
2365         // Maybe we have to create a optional argument.
2366         pos_type body_pos = beginOfBody();
2367         unsigned int column = 0;
2368
2369         if (body_pos > 0) {
2370                 // the optional argument is kept in curly brackets in
2371                 // case it contains a ']'
2372                 // This is not strictly needed, but if this is changed it
2373                 // would be a file format change, and tex2lyx would need
2374                 // to be adjusted, since it unconditionally removes the
2375                 // braces when it parses \item.
2376                 os << "[{";
2377                 column += 2;
2378                 basefont = getLabelFont(bparams, outerfont);
2379         } else {
2380                 basefont = getLayoutFont(bparams, outerfont);
2381         }
2382
2383         // Which font is currently active?
2384         Font running_font(basefont);
2385         // Do we have an open font change?
2386         bool open_font = false;
2387
2388         Change runningChange = Change(Change::UNCHANGED);
2389
2390         Encoding const * const prev_encoding = runparams.encoding;
2391
2392         os.texrow().start(id(), 0);
2393
2394         // if the paragraph is empty, the loop will not be entered at all
2395         if (empty()) {
2396                 if (style.isCommand()) {
2397                         os << '{';
2398                         ++column;
2399                 }
2400                 if (allowcust)
2401                         column += d->startTeXParParams(bparams, os, runparams);
2402         }
2403
2404         for (pos_type i = 0; i < size(); ++i) {
2405                 // First char in paragraph or after label?
2406                 if (i == body_pos) {
2407                         if (body_pos > 0) {
2408                                 if (open_font) {
2409                                         column += running_font.latexWriteEndChanges(
2410                                                 os, bparams, runparams,
2411                                                 basefont, basefont);
2412                                         open_font = false;
2413                                 }
2414                                 basefont = getLayoutFont(bparams, outerfont);
2415                                 running_font = basefont;
2416
2417                                 column += Changes::latexMarkChange(os, bparams,
2418                                                 runningChange, Change(Change::UNCHANGED),
2419                                                 runparams);
2420                                 runningChange = Change(Change::UNCHANGED);
2421
2422                                 os << "}] ";
2423                                 column +=3;
2424                         }
2425                         if (style.isCommand()) {
2426                                 os << '{';
2427                                 ++column;
2428                         }
2429
2430                         if (allowcust)
2431                                 column += d->startTeXParParams(bparams, os,
2432                                                             runparams);
2433                 }
2434
2435                 Change const & change = runparams.inDeletedInset
2436                         ? runparams.changeOfDeletedInset : lookupChange(i);
2437
2438                 if (bparams.outputChanges && runningChange != change) {
2439                         if (open_font) {
2440                                 column += running_font.latexWriteEndChanges(
2441                                                 os, bparams, runparams, basefont, basefont);
2442                                 open_font = false;
2443                         }
2444                         basefont = getLayoutFont(bparams, outerfont);
2445                         running_font = basefont;
2446
2447                         column += Changes::latexMarkChange(os, bparams, runningChange,
2448                                                            change, runparams);
2449                         runningChange = change;
2450                 }
2451
2452                 // do not output text which is marked deleted
2453                 // if change tracking output is disabled
2454                 if (!bparams.outputChanges && change.deleted()) {
2455                         continue;
2456                 }
2457
2458                 ++column;
2459
2460                 // Fully instantiated font
2461                 Font const font = getFont(bparams, i, outerfont);
2462
2463                 Font const last_font = running_font;
2464
2465                 // Do we need to close the previous font?
2466                 if (open_font &&
2467                     (font != running_font ||
2468                      font.language() != running_font.language()))
2469                 {
2470                         column += running_font.latexWriteEndChanges(
2471                                         os, bparams, runparams, basefont,
2472                                         (i == body_pos-1) ? basefont : font);
2473                         running_font = basefont;
2474                         open_font = false;
2475                 }
2476
2477                 string const running_lang = runparams.use_polyglossia ?
2478                         running_font.language()->polyglossia() : running_font.language()->babel();
2479                 // close babel's font environment before opening CJK.
2480                 string const lang_end_command = runparams.use_polyglossia ?
2481                         "\\end{$$lang}" : lyxrc.language_command_end;
2482                 if (!running_lang.empty() &&
2483                     font.language()->encoding()->package() == Encoding::CJK) {
2484                                 string end_tag = subst(lang_end_command,
2485                                                         "$$lang",
2486                                                         running_lang);
2487                                 os << from_ascii(end_tag);
2488                                 column += end_tag.length();
2489                 }
2490
2491                 // Switch file encoding if necessary (and allowed)
2492                 if (!runparams.pass_thru && !style.pass_thru &&
2493                     runparams.encoding->package() != Encoding::none &&
2494                     font.language()->encoding()->package() != Encoding::none) {
2495                         pair<bool, int> const enc_switch =
2496                                 switchEncoding(os.os(), bparams, runparams,
2497                                         *(font.language()->encoding()));
2498                         if (enc_switch.first) {
2499                                 column += enc_switch.second;
2500                                 runparams.encoding = font.language()->encoding();
2501                         }
2502                 }
2503
2504                 char_type const c = d->text_[i];
2505
2506                 // Do we need to change font?
2507                 if ((font != running_font ||
2508                      font.language() != running_font.language()) &&
2509                         i != body_pos - 1)
2510                 {
2511                         odocstringstream ods;
2512                         column += font.latexWriteStartChanges(ods, bparams,
2513                                                               runparams, basefont,
2514                                                               last_font);
2515                         running_font = font;
2516                         open_font = true;
2517                         docstring fontchange = ods.str();
2518                         // check whether the fontchange ends with a \\textcolor
2519                         // modifier and the text starts with a space (bug 4473)
2520                         docstring const last_modifier = rsplit(fontchange, '\\');
2521                         if (prefixIs(last_modifier, from_ascii("textcolor")) && c == ' ')
2522                                 os << fontchange << from_ascii("{}");
2523                         // check if the fontchange ends with a trailing blank
2524                         // (like "\small " (see bug 3382)
2525                         else if (suffixIs(fontchange, ' ') && c == ' ')
2526                                 os << fontchange.substr(0, fontchange.size() - 1)
2527                                    << from_ascii("{}");
2528                         else
2529                                 os << fontchange;
2530                 }
2531
2532                 // FIXME: think about end_pos implementation...
2533                 if (c == ' ' && i >= start_pos && (end_pos == -1 || i < end_pos)) {
2534                         // FIXME: integrate this case in latexSpecialChar
2535                         // Do not print the separation of the optional argument
2536                         // if style.pass_thru is false. This works because
2537                         // latexSpecialChar ignores spaces if
2538                         // style.pass_thru is false.
2539                         if (i != body_pos - 1) {
2540                                 if (d->simpleTeXBlanks(runparams, os,
2541                                                 i, column, font, style)) {
2542                                         // A surrogate pair was output. We
2543                                         // must not call latexSpecialChar
2544                                         // in this iteration, since it would output
2545                                         // the combining character again.
2546                                         ++i;
2547                                         continue;
2548                                 }
2549                         }
2550                 }
2551
2552                 OutputParams rp = runparams;
2553                 rp.free_spacing = style.free_spacing;
2554                 rp.local_font = &font;
2555                 rp.intitle = style.intitle;
2556
2557                 // Two major modes:  LaTeX or plain
2558                 // Handle here those cases common to both modes
2559                 // and then split to handle the two modes separately.
2560                 if (c == META_INSET) {
2561                         if (i >= start_pos && (end_pos == -1 || i < end_pos)) {
2562                                 d->latexInset(bparams, os, rp, running_font,
2563                                                 basefont, outerfont, open_font,
2564                                                 runningChange, style, i, column);
2565                         }
2566                 } else {
2567                         if (i >= start_pos && (end_pos == -1 || i < end_pos)) {
2568                                 try {
2569                                         d->latexSpecialChar(os, rp, running_font, runningChange,
2570                                                             style, i, end_pos, column);
2571                                 } catch (EncodingException & e) {
2572                                 if (runparams.dryrun) {
2573                                         os << "<" << _("LyX Warning: ")
2574                                            << _("uncodable character") << " '";
2575                                         os.put(c);
2576                                         os << "'>";
2577                                 } else {
2578                                         // add location information and throw again.
2579                                         e.par_id = id();
2580                                         e.pos = i;
2581                                         throw(e);
2582                                 }
2583                         }
2584                 }
2585                 }
2586
2587                 // Set the encoding to that returned from latexSpecialChar (see
2588                 // comment for encoding member in OutputParams.h)
2589                 runparams.encoding = rp.encoding;
2590         }
2591
2592         // If we have an open font definition, we have to close it
2593         if (open_font) {
2594 #ifdef FIXED_LANGUAGE_END_DETECTION
2595                 if (next_) {
2596                         running_font.latexWriteEndChanges(os, bparams,
2597                                         runparams, basefont,
2598                                         next_->getFont(bparams, 0, outerfont));
2599                 } else {
2600                         running_font.latexWriteEndChanges(os, bparams,
2601                                         runparams, basefont, basefont);
2602                 }
2603 #else
2604 //FIXME: For now we ALWAYS have to close the foreign font settings if they are
2605 //FIXME: there as we start another \selectlanguage with the next paragraph if
2606 //FIXME: we are in need of this. This should be fixed sometime (Jug)
2607                 running_font.latexWriteEndChanges(os, bparams, runparams,
2608                                 basefont, basefont);
2609 #endif
2610         }
2611
2612         column += Changes::latexMarkChange(os, bparams, runningChange,
2613                                            Change(Change::UNCHANGED), runparams);
2614
2615         // Needed if there is an optional argument but no contents.
2616         if (body_pos > 0 && body_pos == size()) {
2617                 os << "}]~";
2618         }
2619
2620         if (allowcust && d->endTeXParParams(bparams, os, runparams)
2621             && runparams.encoding != prev_encoding) {
2622                 runparams.encoding = prev_encoding;
2623                 if (!runparams.isFullUnicode())
2624                         os << setEncoding(prev_encoding->iconvName());
2625         }
2626
2627         LYXERR(Debug::LATEX, "Paragraph::latex... done " << this);
2628 }
2629
2630
2631 bool Paragraph::emptyTag() const
2632 {
2633         for (pos_type i = 0; i < size(); ++i) {
2634                 if (Inset const * inset = getInset(i)) {
2635                         InsetCode lyx_code = inset->lyxCode();
2636                         // FIXME testing like that is wrong. What is
2637                         // the intent?
2638                         if (lyx_code != TOC_CODE &&
2639                             lyx_code != INCLUDE_CODE &&
2640                             lyx_code != GRAPHICS_CODE &&
2641                             lyx_code != ERT_CODE &&
2642                             lyx_code != LISTINGS_CODE &&
2643                             lyx_code != FLOAT_CODE &&
2644                             lyx_code != TABULAR_CODE) {
2645                                 return false;
2646                         }
2647                 } else {
2648                         char_type c = d->text_[i];
2649                         if (c != ' ' && c != '\t')
2650                                 return false;
2651                 }
2652         }
2653         return true;
2654 }
2655
2656
2657 string Paragraph::getID(Buffer const & buf, OutputParams const & runparams)
2658         const
2659 {
2660         for (pos_type i = 0; i < size(); ++i) {
2661                 if (Inset const * inset = getInset(i)) {
2662                         InsetCode lyx_code = inset->lyxCode();
2663                         if (lyx_code == LABEL_CODE) {
2664                                 InsetLabel const * const il = static_cast<InsetLabel const *>(inset);
2665                                 docstring const & id = il->getParam("name");
2666                                 return "id='" + to_utf8(sgml::cleanID(buf, runparams, id)) + "'";
2667                         }
2668                 }
2669         }
2670         return string();
2671 }
2672
2673
2674 pos_type Paragraph::firstWordDocBook(odocstream & os, OutputParams const & runparams)
2675         const
2676 {
2677         pos_type i;
2678         for (i = 0; i < size(); ++i) {
2679                 if (Inset const * inset = getInset(i)) {
2680                         inset->docbook(os, runparams);
2681                 } else {
2682                         char_type c = d->text_[i];
2683                         if (c == ' ')
2684                                 break;
2685                         os << sgml::escapeChar(c);
2686                 }
2687         }
2688         return i;
2689 }
2690
2691
2692 pos_type Paragraph::firstWordLyXHTML(XHTMLStream & xs, OutputParams const & runparams)
2693         const
2694 {
2695         pos_type i;
2696         for (i = 0; i < size(); ++i) {
2697                 if (Inset const * inset = getInset(i)) {
2698                         inset->xhtml(xs, runparams);
2699                 } else {
2700                         char_type c = d->text_[i];
2701                         if (c == ' ')
2702                                 break;
2703                         xs << c;
2704                 }
2705         }
2706         return i;
2707 }
2708
2709
2710 bool Paragraph::Private::onlyText(Buffer const & buf, Font const & outerfont, pos_type initial) const
2711 {
2712         Font font_old;
2713         pos_type size = text_.size();
2714         for (pos_type i = initial; i < size; ++i) {
2715                 Font font = owner_->getFont(buf.params(), i, outerfont);
2716                 if (text_[i] == META_INSET)
2717                         return false;
2718                 if (i != initial && font != font_old)
2719                         return false;
2720                 font_old = font;
2721         }
2722
2723         return true;
2724 }
2725
2726
2727 void Paragraph::simpleDocBookOnePar(Buffer const & buf,
2728                                     odocstream & os,
2729                                     OutputParams const & runparams,
2730                                     Font const & outerfont,
2731                                     pos_type initial) const
2732 {
2733         bool emph_flag = false;
2734
2735         Layout const & style = *d->layout_;
2736         FontInfo font_old =
2737                 style.labeltype == LABEL_MANUAL ? style.labelfont : style.font;
2738
2739         if (style.pass_thru && !d->onlyText(buf, outerfont, initial))
2740                 os << "]]>";
2741
2742         // parsing main loop
2743         for (pos_type i = initial; i < size(); ++i) {
2744                 Font font = getFont(buf.params(), i, outerfont);
2745
2746                 // handle <emphasis> tag
2747                 if (font_old.emph() != font.fontInfo().emph()) {
2748                         if (font.fontInfo().emph() == FONT_ON) {
2749                                 os << "<emphasis>";
2750                                 emph_flag = true;
2751                         } else if (i != initial) {
2752                                 os << "</emphasis>";
2753                                 emph_flag = false;
2754                         }
2755                 }
2756
2757                 if (Inset const * inset = getInset(i)) {
2758                         inset->docbook(os, runparams);
2759                 } else {
2760                         char_type c = d->text_[i];
2761
2762                         if (style.pass_thru)
2763                                 os.put(c);
2764                         else
2765                                 os << sgml::escapeChar(c);
2766                 }
2767                 font_old = font.fontInfo();
2768         }
2769
2770         if (emph_flag) {
2771                 os << "</emphasis>";
2772         }
2773
2774         if (style.free_spacing)
2775                 os << '\n';
2776         if (style.pass_thru && !d->onlyText(buf, outerfont, initial))
2777                 os << "<![CDATA[";
2778 }
2779
2780
2781 docstring Paragraph::simpleLyXHTMLOnePar(Buffer const & buf,
2782                                     XHTMLStream & xs,
2783                                     OutputParams const & runparams,
2784                                     Font const & outerfont,
2785                                     pos_type initial) const
2786 {
2787         docstring retval;
2788
2789         bool emph_flag = false;
2790         bool bold_flag = false;
2791
2792         Layout const & style = *d->layout_;
2793
2794         xs.startParagraph(allowEmpty());
2795
2796         if (!runparams.for_toc && runparams.html_make_pars) {
2797                 // generate a magic label for this paragraph
2798                 string const attr = "id='" + magicLabel() + "'";
2799                 xs << html::CompTag("a", attr);
2800         }
2801
2802         FontInfo font_old =
2803                 style.labeltype == LABEL_MANUAL ? style.labelfont : style.font;
2804
2805         // parsing main loop
2806         for (pos_type i = initial; i < size(); ++i) {
2807                 // let's not show deleted material in the output
2808                 if (isDeleted(i))
2809                         continue;
2810
2811                 Font font = getFont(buf.params(), i, outerfont);
2812
2813                 // emphasis
2814                 if (font_old.emph() != font.fontInfo().emph()) {
2815                         if (font.fontInfo().emph() == FONT_ON) {
2816                                 xs << html::StartTag("em");
2817                                 emph_flag = true;
2818                         } else if (emph_flag && i != initial) {
2819                                 xs << html::EndTag("em");
2820                                 emph_flag = false;
2821                         }
2822                 }
2823                 // bold
2824                 if (font_old.series() != font.fontInfo().series()) {
2825                         if (font.fontInfo().series() == BOLD_SERIES) {
2826                                 xs << html::StartTag("strong");
2827                                 bold_flag = true;
2828                         } else if (bold_flag && i != initial) {
2829                                 xs << html::EndTag("strong");
2830                                 bold_flag = false;
2831                         }
2832                 }
2833                 // FIXME XHTML
2834                 // Other such tags? What about the other text ranges?
2835
2836                 Inset const * inset = getInset(i);
2837                 if (inset) {
2838                         if (!runparams.for_toc || inset->isInToc()) {
2839                                 OutputParams np = runparams;
2840                                 if (!inset->getLayout().htmlisblock())
2841                                         np.html_in_par = true;
2842                                 retval += inset->xhtml(xs, np);
2843                         }
2844                 } else {
2845                         char_type c = d->text_[i];
2846
2847                         if (style.pass_thru)
2848                                 xs << c;
2849                         else if (c == '-') {
2850                                 docstring str;
2851                                 int j = i + 1;
2852                                 if (j < size() && d->text_[j] == '-') {
2853                                         j += 1;
2854                                         if (j < size() && d->text_[j] == '-') {
2855                                                 str += from_ascii("&mdash;");
2856                                                 i += 2;
2857                                         } else {
2858                                                 str += from_ascii("&ndash;");
2859                                                 i += 1;
2860                                         }
2861                                 }
2862                                 else
2863                                         str += c;
2864                                 // We don't want to escape the entities. Note that
2865                                 // it is safe to do this, since str can otherwise
2866                                 // only be "-". E.g., it can't be "<".
2867                                 xs << XHTMLStream::ESCAPE_NONE << str;
2868                         } else
2869                                 xs << c;
2870                 }
2871                 font_old = font.fontInfo();
2872         }
2873
2874         xs.closeFontTags();
2875         xs.endParagraph();
2876         return retval;
2877 }
2878
2879
2880 bool Paragraph::isHfill(pos_type pos) const
2881 {
2882         Inset const * inset = getInset(pos);
2883         return inset && (inset->lyxCode() == SPACE_CODE &&
2884                          inset->isStretchableSpace());
2885 }
2886
2887
2888 bool Paragraph::isNewline(pos_type pos) const
2889 {
2890         Inset const * inset = getInset(pos);
2891         return inset && inset->lyxCode() == NEWLINE_CODE;
2892 }
2893
2894
2895 bool Paragraph::isLineSeparator(pos_type pos) const
2896 {
2897         char_type const c = d->text_[pos];
2898         if (isLineSeparatorChar(c))
2899                 return true;
2900         Inset const * inset = getInset(pos);
2901         return inset && inset->isLineSeparator();
2902 }
2903
2904
2905 bool Paragraph::isWordSeparator(pos_type pos) const
2906 {
2907         if (pos == size())
2908                 return true;
2909         if (Inset const * inset = getInset(pos))
2910                 return !inset->isLetter();
2911         // if we have a hard hyphen (no en- or emdash) or apostrophe
2912         // we pass this to the spell checker
2913         // FIXME: this method is subject to change, visit
2914         // https://bugzilla.mozilla.org/show_bug.cgi?id=355178
2915         // to get an impression how complex this is.
2916         if (isHardHyphenOrApostrophe(pos))
2917                 return false;
2918         char_type const c = d->text_[pos];
2919         // We want to pass the escape chars to the spellchecker
2920         docstring const escape_chars = from_utf8(lyxrc.spellchecker_esc_chars);
2921         return !isLetterChar(c) && !isDigitASCII(c) && !contains(escape_chars, c);
2922 }
2923
2924
2925 bool Paragraph::isHardHyphenOrApostrophe(pos_type pos) const
2926 {
2927         pos_type const psize = size();
2928         if (pos >= psize)
2929                 return false;
2930         char_type const c = d->text_[pos];
2931         if (c != '-' && c != '\'')
2932                 return false;
2933         int nextpos = pos + 1;
2934         int prevpos = pos > 0 ? pos - 1 : 0;
2935         if ((nextpos == psize || isSpace(nextpos))
2936                 && (pos == 0 || isSpace(prevpos)))
2937                 return false;
2938         return c == '\''
2939                 || ((nextpos == psize || d->text_[nextpos] != '-')
2940                 && (pos == 0 || d->text_[prevpos] != '-'));
2941 }
2942
2943
2944 bool Paragraph::isSameSpellRange(pos_type pos1, pos_type pos2) const
2945 {
2946         return pos1 == pos2
2947                 || d->speller_state_.getRange(pos1) == d->speller_state_.getRange(pos2);
2948 }
2949
2950
2951 bool Paragraph::isChar(pos_type pos) const
2952 {
2953         if (Inset const * inset = getInset(pos))
2954                 return inset->isChar();
2955         char_type const c = d->text_[pos];
2956         return !isLetterChar(c) && !isDigitASCII(c) && !lyx::isSpace(c);
2957 }
2958
2959
2960 bool Paragraph::isSpace(pos_type pos) const
2961 {
2962         if (Inset const * inset = getInset(pos))
2963                 return inset->isSpace();
2964         char_type const c = d->text_[pos];
2965         return lyx::isSpace(c);
2966 }
2967
2968
2969 Language const *
2970 Paragraph::getParLanguage(BufferParams const & bparams) const
2971 {
2972         if (!empty())
2973                 return getFirstFontSettings(bparams).language();
2974         // FIXME: we should check the prev par as well (Lgb)
2975         return bparams.language;
2976 }
2977
2978
2979 bool Paragraph::isRTL(BufferParams const & bparams) const
2980 {
2981         return lyxrc.rtl_support
2982                 && getParLanguage(bparams)->rightToLeft()
2983                 && !inInset().getLayout().forceLTR();
2984 }
2985
2986
2987 void Paragraph::changeLanguage(BufferParams const & bparams,
2988                                Language const * from, Language const * to)
2989 {
2990         // change language including dummy font change at the end
2991         for (pos_type i = 0; i <= size(); ++i) {
2992                 Font font = getFontSettings(bparams, i);
2993                 if (font.language() == from) {
2994                         font.setLanguage(to);
2995                         setFont(i, font);
2996                         d->requestSpellCheck(i);
2997                 }
2998         }
2999 }
3000
3001
3002 bool Paragraph::isMultiLingual(BufferParams const & bparams) const
3003 {
3004         Language const * doc_language = bparams.language;
3005         FontList::const_iterator cit = d->fontlist_.begin();
3006         FontList::const_iterator end = d->fontlist_.end();
3007
3008         for (; cit != end; ++cit)
3009                 if (cit->font().language() != ignore_language &&
3010                     cit->font().language() != latex_language &&
3011                     cit->font().language() != doc_language)
3012                         return true;
3013         return false;
3014 }
3015
3016
3017 void Paragraph::getLanguages(std::set<Language const *> & languages) const
3018 {
3019         FontList::const_iterator cit = d->fontlist_.begin();
3020         FontList::const_iterator end = d->fontlist_.end();
3021
3022         for (; cit != end; ++cit) {
3023                 Language const * lang = cit->font().language();
3024                 if (lang != ignore_language &&
3025                     lang != latex_language)
3026                         languages.insert(lang);
3027         }
3028 }
3029
3030
3031 docstring Paragraph::asString(int options) const
3032 {
3033         return asString(0, size(), options);
3034 }
3035
3036
3037 docstring Paragraph::asString(pos_type beg, pos_type end, int options) const
3038 {
3039         odocstringstream os;
3040
3041         if (beg == 0
3042             && options & AS_STR_LABEL
3043             && !d->params_.labelString().empty())
3044                 os << d->params_.labelString() << ' ';
3045
3046         for (pos_type i = beg; i < end; ++i) {
3047                 if ((options & AS_STR_SKIPDELETE) && isDeleted(i))
3048                         continue;
3049                 char_type const c = d->text_[i];
3050                 if (isPrintable(c) || c == '\t'
3051                     || (c == '\n' && (options & AS_STR_NEWLINES)))
3052                         os.put(c);
3053                 else if (c == META_INSET && (options & AS_STR_INSETS)) {
3054                         getInset(i)->toString(os);
3055                         if (getInset(i)->asInsetMath())
3056                                 os << " ";
3057                 }
3058         }
3059
3060         return os.str();
3061 }
3062
3063
3064 void Paragraph::forToc(docstring & os, size_t maxlen) const
3065 {
3066         if (!d->params_.labelString().empty())
3067                 os += d->params_.labelString() + ' ';
3068         for (pos_type i = 0; i < size() && os.length() < maxlen; ++i) {
3069                 if (isDeleted(i))
3070                         continue;
3071                 char_type const c = d->text_[i];
3072                 if (isPrintable(c))
3073                         os += c;
3074                 else if (c == '\t' || c == '\n')
3075                         os += ' ';
3076                 else if (c == META_INSET)
3077                         getInset(i)->forToc(os, maxlen);
3078         }
3079 }
3080
3081
3082 docstring Paragraph::stringify(pos_type beg, pos_type end, int options, OutputParams & runparams) const
3083 {
3084         odocstringstream os;
3085
3086         if (beg == 0
3087                 && options & AS_STR_LABEL
3088                 && !d->params_.labelString().empty())
3089                 os << d->params_.labelString() << ' ';
3090
3091         for (pos_type i = beg; i < end; ++i) {
3092                 char_type const c = d->text_[i];
3093                 if (isPrintable(c) || c == '\t'
3094                     || (c == '\n' && (options & AS_STR_NEWLINES)))
3095                         os.put(c);
3096                 else if (c == META_INSET && (options & AS_STR_INSETS)) {
3097                         getInset(i)->plaintext(os, runparams);
3098                 }
3099         }
3100
3101         return os.str();
3102 }
3103
3104
3105 void Paragraph::setInsetOwner(Inset const * inset)
3106 {
3107         d->inset_owner_ = inset;
3108 }
3109
3110
3111 int Paragraph::id() const
3112 {
3113         return d->id_;
3114 }
3115
3116
3117 void Paragraph::setId(int id)
3118 {
3119         d->id_ = id;
3120 }
3121
3122
3123 Layout const & Paragraph::layout() const
3124 {
3125         return *d->layout_;
3126 }
3127
3128
3129 void Paragraph::setLayout(Layout const & layout)
3130 {
3131         d->layout_ = &layout;
3132 }
3133
3134
3135 void Paragraph::setDefaultLayout(DocumentClass const & tc)
3136 {
3137         setLayout(tc.defaultLayout());
3138 }
3139
3140
3141 void Paragraph::setPlainLayout(DocumentClass const & tc)
3142 {
3143         setLayout(tc.plainLayout());
3144 }
3145
3146
3147 void Paragraph::setPlainOrDefaultLayout(DocumentClass const & tclass)
3148 {
3149         if (usePlainLayout())
3150                 setPlainLayout(tclass);
3151         else
3152                 setDefaultLayout(tclass);
3153 }
3154
3155
3156 Inset const & Paragraph::inInset() const
3157 {
3158         LASSERT(d->inset_owner_, throw ExceptionMessage(BufferException,
3159                 _("Memory problem"), _("Paragraph not properly initialized")));
3160         return *d->inset_owner_;
3161 }
3162
3163
3164 ParagraphParameters & Paragraph::params()
3165 {
3166         return d->params_;
3167 }
3168
3169
3170 ParagraphParameters const & Paragraph::params() const
3171 {
3172         return d->params_;
3173 }
3174
3175
3176 bool Paragraph::isFreeSpacing() const
3177 {
3178         if (d->layout_->free_spacing)
3179                 return true;
3180         return d->inset_owner_ && d->inset_owner_->isFreeSpacing();
3181 }
3182
3183
3184 bool Paragraph::allowEmpty() const
3185 {
3186         if (d->layout_->keepempty)
3187                 return true;
3188         return d->inset_owner_ && d->inset_owner_->allowEmpty();
3189 }
3190
3191
3192 char_type Paragraph::transformChar(char_type c, pos_type pos) const
3193 {
3194         if (!Encodings::isArabicChar(c))
3195                 return c;
3196
3197         char_type prev_char = ' ';
3198         char_type next_char = ' ';
3199
3200         for (pos_type i = pos - 1; i >= 0; --i) {
3201                 char_type const par_char = d->text_[i];
3202                 if (!Encodings::isArabicComposeChar(par_char)) {
3203                         prev_char = par_char;
3204                         break;
3205                 }
3206         }
3207
3208         for (pos_type i = pos + 1, end = size(); i < end; ++i) {
3209                 char_type const par_char = d->text_[i];
3210                 if (!Encodings::isArabicComposeChar(par_char)) {
3211                         next_char = par_char;
3212                         break;
3213                 }
3214         }
3215
3216         if (Encodings::isArabicChar(next_char)) {
3217                 if (Encodings::isArabicChar(prev_char) &&
3218                         !Encodings::isArabicSpecialChar(prev_char))
3219                         return Encodings::transformChar(c, Encodings::FORM_MEDIAL);
3220                 else
3221                         return Encodings::transformChar(c, Encodings::FORM_INITIAL);
3222         } else {
3223                 if (Encodings::isArabicChar(prev_char) &&
3224                         !Encodings::isArabicSpecialChar(prev_char))
3225                         return Encodings::transformChar(c, Encodings::FORM_FINAL);
3226                 else
3227                         return Encodings::transformChar(c, Encodings::FORM_ISOLATED);
3228         }
3229 }
3230
3231
3232 int Paragraph::checkBiblio(Buffer const & buffer)
3233 {
3234         // FIXME From JS:
3235         // This is getting more and more a mess. ...We really should clean
3236         // up this bibitem issue for 1.6.
3237
3238         // Add bibitem insets if necessary
3239         if (d->layout_->labeltype != LABEL_BIBLIO)
3240                 return 0;
3241
3242         bool hasbibitem = !d->insetlist_.empty()
3243                 // Insist on it being in pos 0
3244                 && d->text_[0] == META_INSET
3245                 && d->insetlist_.begin()->inset->lyxCode() == BIBITEM_CODE;
3246
3247         bool track_changes = buffer.params().trackChanges;
3248
3249         docstring oldkey;
3250         docstring oldlabel;
3251
3252         // remove a bibitem in pos != 0
3253         // restore it later in pos 0 if necessary
3254         // (e.g. if a user inserts contents _before_ the item)
3255         // we're assuming there's only one of these, which there
3256         // should be.
3257         int erasedInsetPosition = -1;
3258         InsetList::iterator it = d->insetlist_.begin();
3259         InsetList::iterator end = d->insetlist_.end();
3260         for (; it != end; ++it)
3261                 if (it->inset->lyxCode() == BIBITEM_CODE
3262                       && it->pos > 0) {
3263                         InsetCommand * olditem = it->inset->asInsetCommand();
3264                         oldkey = olditem->getParam("key");
3265                         oldlabel = olditem->getParam("label");
3266                         erasedInsetPosition = it->pos;
3267                         eraseChar(erasedInsetPosition, track_changes);
3268                         break;
3269         }
3270
3271         // There was an InsetBibitem at the beginning, and we didn't
3272         // have to erase one.
3273         if (hasbibitem && erasedInsetPosition < 0)
3274                         return 0;
3275
3276         // There was an InsetBibitem at the beginning and we did have to
3277         // erase one. So we give its properties to the beginning inset.
3278         if (hasbibitem) {
3279                 InsetCommand * inset = d->insetlist_.begin()->inset->asInsetCommand();
3280                 if (!oldkey.empty())
3281                         inset->setParam("key", oldkey);
3282                 inset->setParam("label", oldlabel);
3283                 return -erasedInsetPosition;
3284         }
3285
3286         // There was no inset at the beginning, so we need to create one with
3287         // the key and label of the one we erased.
3288         InsetBibitem * inset =
3289                 new InsetBibitem(const_cast<Buffer *>(&buffer), InsetCommandParams(BIBITEM_CODE));
3290         // restore values of previously deleted item in this par.
3291         if (!oldkey.empty())
3292                 inset->setParam("key", oldkey);
3293         inset->setParam("label", oldlabel);
3294         insertInset(0, inset,
3295                     Change(track_changes ? Change::INSERTED : Change::UNCHANGED));
3296
3297         return 1;
3298 }
3299
3300
3301 void Paragraph::checkAuthors(AuthorList const & authorList)
3302 {
3303         d->changes_.checkAuthors(authorList);
3304 }
3305
3306
3307 bool Paragraph::isChanged(pos_type pos) const
3308 {
3309         return lookupChange(pos).changed();
3310 }
3311
3312
3313 bool Paragraph::isInserted(pos_type pos) const
3314 {
3315         return lookupChange(pos).inserted();
3316 }
3317
3318
3319 bool Paragraph::isDeleted(pos_type pos) const
3320 {
3321         return lookupChange(pos).deleted();
3322 }
3323
3324
3325 InsetList const & Paragraph::insetList() const
3326 {
3327         return d->insetlist_;
3328 }
3329
3330
3331 void Paragraph::setBuffer(Buffer & b)
3332 {
3333         d->insetlist_.setBuffer(b);
3334 }
3335
3336
3337 Inset * Paragraph::releaseInset(pos_type pos)
3338 {
3339         Inset * inset = d->insetlist_.release(pos);
3340         /// does not honour change tracking!
3341         eraseChar(pos, false);
3342         return inset;
3343 }
3344
3345
3346 Inset * Paragraph::getInset(pos_type pos)
3347 {
3348         return (pos < pos_type(d->text_.size()) && d->text_[pos] == META_INSET)
3349                  ? d->insetlist_.get(pos) : 0;
3350 }
3351
3352
3353 Inset const * Paragraph::getInset(pos_type pos) const
3354 {
3355         return (pos < pos_type(d->text_.size()) && d->text_[pos] == META_INSET)
3356                  ? d->insetlist_.get(pos) : 0;
3357 }
3358
3359
3360 void Paragraph::changeCase(BufferParams const & bparams, pos_type pos,
3361                 pos_type & right, TextCase action)
3362 {
3363         // process sequences of modified characters; in change
3364         // tracking mode, this approach results in much better
3365         // usability than changing case on a char-by-char basis
3366         docstring changes;
3367
3368         bool const trackChanges = bparams.trackChanges;
3369
3370         bool capitalize = true;
3371
3372         for (; pos < right; ++pos) {
3373                 char_type oldChar = d->text_[pos];
3374                 char_type newChar = oldChar;
3375
3376                 // ignore insets and don't play with deleted text!
3377                 if (oldChar != META_INSET && !isDeleted(pos)) {
3378                         switch (action) {
3379                                 case text_lowercase:
3380                                         newChar = lowercase(oldChar);
3381                                         break;
3382                                 case text_capitalization:
3383                                         if (capitalize) {
3384                                                 newChar = uppercase(oldChar);
3385                                                 capitalize = false;
3386                                         }
3387                                         break;
3388                                 case text_uppercase:
3389                                         newChar = uppercase(oldChar);
3390                                         break;
3391                         }
3392                 }
3393
3394                 if (isWordSeparator(pos) || isDeleted(pos)) {
3395                         // permit capitalization again
3396                         capitalize = true;
3397                 }
3398
3399                 if (oldChar != newChar) {
3400                         changes += newChar;
3401                         if (pos != right - 1)
3402                                 continue;
3403                         // step behind the changing area
3404                         pos++;
3405                 }
3406
3407                 int erasePos = pos - changes.size();
3408                 for (size_t i = 0; i < changes.size(); i++) {
3409                         insertChar(pos, changes[i],
3410                                    getFontSettings(bparams,
3411                                                    erasePos),
3412                                    trackChanges);
3413                         if (!eraseChar(erasePos, trackChanges)) {
3414                                 ++erasePos;
3415                                 ++pos; // advance
3416                                 ++right; // expand selection
3417                         }
3418                 }
3419                 changes.clear();
3420         }
3421 }
3422
3423
3424 int Paragraph::find(docstring const & str, bool cs, bool mw,
3425                 pos_type start_pos, bool del) const
3426 {
3427         pos_type pos = start_pos;
3428         int const strsize = str.length();
3429         int i = 0;
3430         pos_type const parsize = d->text_.size();
3431         for (i = 0; i < strsize && pos < parsize; ++i, ++pos) {
3432                 // Ignore "invisible" letters such as ligature breaks
3433                 // and hyphenation chars while searching
3434                 while (pos < parsize - 1 && isInset(pos)) {
3435                         odocstringstream os;
3436                         getInset(pos)->toString(os);
3437                         if (!getInset(pos)->isLetter() || !os.str().empty())
3438                                 break;
3439                         pos++;
3440                 }
3441                 if (cs && str[i] != d->text_[pos])
3442                         break;
3443                 if (!cs && uppercase(str[i]) != uppercase(d->text_[pos]))
3444                         break;
3445                 if (!del && isDeleted(pos))
3446                         break;
3447         }
3448
3449         if (i != strsize)
3450                 return 0;
3451
3452         // if necessary, check whether string matches word
3453         if (mw) {
3454                 if (start_pos > 0 && !isWordSeparator(start_pos - 1))
3455                         return 0;
3456                 if (pos < parsize
3457                         && !isWordSeparator(pos))
3458                         return 0;
3459         }
3460
3461         return pos - start_pos;
3462 }
3463
3464
3465 char_type Paragraph::getChar(pos_type pos) const
3466 {
3467         return d->text_[pos];
3468 }
3469
3470
3471 pos_type Paragraph::size() const
3472 {
3473         return d->text_.size();
3474 }
3475
3476
3477 bool Paragraph::empty() const
3478 {
3479         return d->text_.empty();
3480 }
3481
3482
3483 bool Paragraph::isInset(pos_type pos) const
3484 {
3485         return d->text_[pos] == META_INSET;
3486 }
3487
3488
3489 bool Paragraph::isSeparator(pos_type pos) const
3490 {
3491         //FIXME: Are we sure this can be the only separator?
3492         return d->text_[pos] == ' ';
3493 }
3494
3495
3496 void Paragraph::deregisterWords()
3497 {
3498         Private::LangWordsMap::const_iterator itl = d->words_.begin();
3499         Private::LangWordsMap::const_iterator ite = d->words_.end();
3500         for (; itl != ite; ++itl) {
3501                 WordList * wl = theWordList(itl->first);
3502                 Private::Words::const_iterator it = (itl->second).begin();
3503                 Private::Words::const_iterator et = (itl->second).end();
3504                 for (; it != et; ++it)
3505                         wl->remove(*it);
3506         }
3507         d->words_.clear();
3508 }
3509
3510
3511 void Paragraph::locateWord(pos_type & from, pos_type & to,
3512         word_location const loc) const
3513 {
3514         switch (loc) {
3515         case WHOLE_WORD_STRICT:
3516                 if (from == 0 || from == size()
3517                     || isWordSeparator(from)
3518                     || isWordSeparator(from - 1)) {
3519                         to = from;
3520                         return;
3521                 }
3522                 // no break here, we go to the next
3523
3524         case WHOLE_WORD:
3525                 // If we are already at the beginning of a word, do nothing
3526                 if (!from || isWordSeparator(from - 1))
3527                         break;
3528                 // no break here, we go to the next
3529
3530         case PREVIOUS_WORD:
3531                 // always move the cursor to the beginning of previous word
3532                 while (from && !isWordSeparator(from - 1))
3533                         --from;
3534                 break;
3535         case NEXT_WORD:
3536                 LYXERR0("Paragraph::locateWord: NEXT_WORD not implemented yet");
3537                 break;
3538         case PARTIAL_WORD:
3539                 // no need to move the 'from' cursor
3540                 break;
3541         }
3542         to = from;
3543         while (to < size() && !isWordSeparator(to))
3544                 ++to;
3545 }
3546
3547
3548 void Paragraph::collectWords()
3549 {
3550         pos_type n = size();
3551         for (pos_type pos = 0; pos < n; ++pos) {
3552                 if (isWordSeparator(pos))
3553                         continue;
3554                 pos_type from = pos;
3555                 locateWord(from, pos, WHOLE_WORD);
3556                 if ((pos - from) >= (int)lyxrc.completion_minlength) {
3557                         docstring word = asString(from, pos, AS_STR_NONE);
3558                         FontList::const_iterator cit = d->fontlist_.fontIterator(pos);
3559                         if (cit == d->fontlist_.end())
3560                                 return;
3561                         Language const * lang = cit->font().language();
3562                         d->words_[*lang].insert(word);
3563                 }
3564         }
3565 }
3566
3567
3568 void Paragraph::registerWords()
3569 {
3570         Private::LangWordsMap::const_iterator itl = d->words_.begin();
3571         Private::LangWordsMap::const_iterator ite = d->words_.end();
3572         for (; itl != ite; ++itl) {
3573                 WordList * wl = theWordList(itl->first);
3574                 Private::Words::const_iterator it = (itl->second).begin();
3575                 Private::Words::const_iterator et = (itl->second).end();
3576                 for (; it != et; ++it)
3577                         wl->insert(*it);
3578         }
3579 }
3580
3581
3582 void Paragraph::updateWords()
3583 {
3584         deregisterWords();
3585         collectWords();
3586         registerWords();
3587 }
3588
3589
3590 void Paragraph::Private::appendSkipPosition(SkipPositions & skips, pos_type const pos) const
3591 {
3592         SkipPositionsIterator begin = skips.begin();
3593         SkipPositions::iterator end = skips.end();
3594         if (pos > 0 && begin < end) {
3595                 --end;
3596                 if (end->last == pos - 1) {
3597                         end->last = pos;
3598                         return;
3599                 }
3600         }
3601         skips.insert(end, FontSpan(pos, pos));
3602 }
3603
3604
3605 Language * Paragraph::Private::locateSpellRange(
3606         pos_type & from, pos_type & to,
3607         SkipPositions & skips) const
3608 {
3609         // skip leading white space
3610         while (from < to && owner_->isWordSeparator(from))
3611                 ++from;
3612         // don't check empty range
3613         if (from >= to)
3614                 return 0;
3615         // get current language
3616         Language * lang = getSpellLanguage(from);
3617         pos_type last = from;
3618         bool samelang = true;
3619         bool sameinset = true;
3620         while (last < to && samelang && sameinset) {
3621                 // hop to end of word
3622                 while (last < to && !owner_->isWordSeparator(last)) {
3623                         if (owner_->getInset(last)) {
3624                                 appendSkipPosition(skips, last);
3625                         } else if (owner_->isDeleted(last)) {
3626                                 appendSkipPosition(skips, last);
3627                         }
3628                         ++last;
3629                 }
3630                 // hop to next word while checking for insets
3631                 while (sameinset && last < to && owner_->isWordSeparator(last)) {
3632                         if (Inset const * inset = owner_->getInset(last))
3633                                 sameinset = inset->isChar() && inset->isLetter();
3634                         if (sameinset && owner_->isDeleted(last)) {
3635                                 appendSkipPosition(skips, last);
3636                         }
3637                         if (sameinset)
3638                                 last++;
3639                 }
3640                 if (sameinset && last < to) {
3641                         // now check for language change
3642                         samelang = lang == getSpellLanguage(last);
3643                 }
3644         }
3645         // if language change detected backstep is needed
3646         if (!samelang)
3647                 --last;
3648         to = last;
3649         return lang;
3650 }
3651
3652
3653 Language * Paragraph::Private::getSpellLanguage(pos_type const from) const
3654 {
3655         Language * lang =
3656                 const_cast<Language *>(owner_->getFontSettings(
3657                         inset_owner_->buffer().params(), from).language());
3658         if (lang == inset_owner_->buffer().params().language
3659                 && !lyxrc.spellchecker_alt_lang.empty()) {
3660                 string lang_code;
3661                 string const lang_variety =
3662                         split(lyxrc.spellchecker_alt_lang, lang_code, '-');
3663                 lang->setCode(lang_code);
3664                 lang->setVariety(lang_variety);
3665         }
3666         return lang;
3667 }
3668
3669
3670 void Paragraph::requestSpellCheck(pos_type pos)
3671 {
3672         d->requestSpellCheck(pos);
3673 }
3674
3675
3676 bool Paragraph::needsSpellCheck() const
3677 {
3678         SpellChecker::ChangeNumber speller_change_number = 0;
3679         if (theSpellChecker())
3680                 speller_change_number = theSpellChecker()->changeNumber();
3681         if (speller_change_number > d->speller_state_.currentChangeNumber()) {
3682                 d->speller_state_.needsCompleteRefresh(speller_change_number);
3683         }
3684         return d->needsSpellCheck();
3685 }
3686
3687
3688 bool Paragraph::Private::ignoreWord(docstring const & word) const
3689 {
3690         // Ignore words with digits
3691         // FIXME: make this customizable
3692         // (note that some checkers ignore words with digits by default)
3693         docstring::const_iterator cit = word.begin();
3694         docstring::const_iterator const end = word.end();
3695         for (; cit != end; ++cit) {
3696                 if (isNumber((*cit)))
3697                         return true;
3698         }
3699         return false;
3700 }
3701
3702
3703 SpellChecker::Result Paragraph::spellCheck(pos_type & from, pos_type & to,
3704         WordLangTuple & wl, docstring_list & suggestions,
3705         bool do_suggestion, bool check_learned) const
3706 {
3707         SpellChecker::Result result = SpellChecker::WORD_OK;
3708         SpellChecker * speller = theSpellChecker();
3709         if (!speller)
3710                 return result;
3711
3712         if (!d->layout_->spellcheck || !inInset().allowSpellCheck())
3713                 return result;
3714
3715         locateWord(from, to, WHOLE_WORD);
3716         if (from == to || from >= size())
3717                 return result;
3718
3719         docstring word = asString(from, to, AS_STR_INSETS | AS_STR_SKIPDELETE);
3720         Language * lang = d->getSpellLanguage(from);
3721
3722         wl = WordLangTuple(word, lang);
3723
3724         if (!word.size())
3725                 return result;
3726
3727         if (needsSpellCheck() || check_learned) {
3728                 pos_type end = to;
3729                 if (!d->ignoreWord(word)) {
3730                         bool const trailing_dot = to < size() && d->text_[to] == '.';
3731                         result = speller->check(wl);
3732                         if (SpellChecker::misspelled(result) && trailing_dot) {
3733                                 wl = WordLangTuple(word.append(from_ascii(".")), lang);
3734                                 result = speller->check(wl);
3735                                 if (!SpellChecker::misspelled(result)) {
3736                                         LYXERR(Debug::GUI, "misspelled word is correct with dot: \"" <<
3737                                            word << "\" [" <<
3738                                            from << ".." << to << "]");
3739                                 } else {
3740                                         // spell check with dot appended failed too
3741                                         // restore original word/lang value
3742                                         word = asString(from, to, AS_STR_INSETS | AS_STR_SKIPDELETE);
3743                                         wl = WordLangTuple(word, lang);
3744                                 }
3745                         }
3746                 }
3747                 if (!SpellChecker::misspelled(result)) {
3748                         // area up to the begin of the next word is not misspelled
3749                         while (end < size() && isWordSeparator(end))
3750                                 ++end;
3751                 }
3752                 d->setMisspelled(from, end, result);
3753         } else {
3754                 result = d->speller_state_.getState(from);
3755         }
3756
3757         if (do_suggestion)
3758                 suggestions.clear();
3759
3760         if (SpellChecker::misspelled(result)) {
3761                 LYXERR(Debug::GUI, "misspelled word: \"" <<
3762                            word << "\" [" <<
3763                            from << ".." << to << "]");
3764                 if (do_suggestion)
3765                         speller->suggest(wl, suggestions);
3766         }
3767         return result;
3768 }
3769
3770
3771 void Paragraph::Private::markMisspelledWords(
3772         pos_type const & first, pos_type const & last,
3773         SpellChecker::Result result,
3774         docstring const & word,
3775         SkipPositions const & skips)
3776 {
3777         if (!SpellChecker::misspelled(result)) {
3778                 setMisspelled(first, last, SpellChecker::WORD_OK);
3779                 return;
3780         }
3781         int snext = first;
3782         SpellChecker * speller = theSpellChecker();
3783         // locate and enumerate the error positions
3784         int nerrors = speller->numMisspelledWords();
3785         int numskipped = 0;
3786         SkipPositionsIterator it = skips.begin();
3787         SkipPositionsIterator et = skips.end();
3788         for (int index = 0; index < nerrors; ++index) {
3789                 int wstart;
3790                 int wlen = 0;
3791                 speller->misspelledWord(index, wstart, wlen);
3792                 /// should not happen if speller supports range checks
3793                 if (!wlen) continue;
3794                 docstring const misspelled = word.substr(wstart, wlen);
3795                 wstart += first + numskipped;
3796                 if (snext < wstart) {
3797                         /// mark the range of correct spelling
3798                         numskipped += countSkips(it, et, wstart);
3799                         setMisspelled(snext,
3800                                 wstart - 1, SpellChecker::WORD_OK);
3801                 }
3802                 snext = wstart + wlen;
3803                 numskipped += countSkips(it, et, snext);
3804                 /// mark the range of misspelling
3805                 setMisspelled(wstart, snext, result);
3806                 LYXERR(Debug::GUI, "misspelled word: \"" <<
3807                            misspelled << "\" [" <<
3808                            wstart << ".." << (snext-1) << "]");
3809                 ++snext;
3810         }
3811         if (snext <= last) {
3812                 /// mark the range of correct spelling at end
3813                 setMisspelled(snext, last, SpellChecker::WORD_OK);
3814         }
3815 }
3816
3817
3818 void Paragraph::spellCheck() const
3819 {
3820         SpellChecker * speller = theSpellChecker();
3821         if (!speller || !size() ||!needsSpellCheck())
3822                 return;
3823         pos_type start;
3824         pos_type endpos;
3825         d->rangeOfSpellCheck(start, endpos);
3826         if (speller->canCheckParagraph()) {
3827                 // loop until we leave the range
3828                 for (pos_type first = start; first < endpos; ) {
3829                         pos_type last = endpos;
3830                         Private::SkipPositions skips;
3831                         Language * lang = d->locateSpellRange(first, last, skips);
3832                         if (first >= endpos)
3833                                 break;
3834                         // start the spell checker on the unit of meaning
3835                         docstring word = asString(first, last, AS_STR_INSETS + AS_STR_SKIPDELETE);
3836                         WordLangTuple wl = WordLangTuple(word, lang);
3837                         SpellChecker::Result result = word.size() ?
3838                                 speller->check(wl) : SpellChecker::WORD_OK;
3839                         d->markMisspelledWords(first, last, result, word, skips);
3840                         first = ++last;
3841                 }
3842         } else {
3843                 static docstring_list suggestions;
3844                 pos_type to = endpos;
3845                 while (start < endpos) {
3846                         WordLangTuple wl;
3847                         spellCheck(start, to, wl, suggestions, false);
3848                         start = to + 1;
3849                 }
3850         }
3851         d->readySpellCheck();
3852 }
3853
3854
3855 bool Paragraph::isMisspelled(pos_type pos, bool check_boundary) const
3856 {
3857         bool result = SpellChecker::misspelled(d->speller_state_.getState(pos));
3858         if (result || pos <= 0 || pos > size())
3859                 return result;
3860         if (check_boundary && (pos == size() || isWordSeparator(pos)))
3861                 result = SpellChecker::misspelled(d->speller_state_.getState(pos - 1));
3862         return result;
3863 }
3864
3865
3866 string Paragraph::magicLabel() const
3867 {
3868         stringstream ss;
3869         ss << "magicparlabel-" << id();
3870         return ss.str();
3871 }
3872
3873
3874 } // namespace lyx