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                                    BufferParams const & bparams,
 334                                    OutputParams const & runparams,
 335                                    Font const & running_font,
 336                                    Change const & running_change,
 337                                    Layout const & style,
 338                                    pos_type & i,
 339                                    pos_type end_pos,
 340                                    unsigned int & column);
 341
 342         ///
 343         bool latexSpecialT1(
 344                 char_type const c,
 345                 otexstream & os,
 346                 pos_type i,
 347                 unsigned int & column);
 348         ///
 349         bool latexSpecialT3(
 350                 char_type const c,
 351                 otexstream & os,
 352                 pos_type i,
 353                 unsigned int & column);
 354         ///
 355         bool latexSpecialTypewriter(
 356                 char_type const c,
 357                 otexstream & os,
 358                 pos_type i,
 359                 unsigned int & column);
 360         ///
 361         bool latexSpecialPhrase(
 362                 otexstream & os,
 363                 pos_type & i,
 364                 pos_type end_pos,
 365                 unsigned int & column,
 366                 OutputParams const & runparams);
 367
 368         ///
 369         void validate(LaTeXFeatures & features) const;
 370
 371         /// Checks if the paragraph contains only text and no inset or font change.
 372         bool onlyText(Buffer const & buf, Font const & outerfont,
 373                       pos_type initial) const;
 374
 375         /// match a string against a particular point in the paragraph
 376         bool isTextAt(string const & str, pos_type pos) const;
 377
 378         /// a vector of speller skip positions
 379         typedef vector<FontSpan> SkipPositions;
 380         typedef SkipPositions::const_iterator SkipPositionsIterator;
 381
 382         void appendSkipPosition(SkipPositions & skips, pos_type const pos) const;
 383
 384         Language * getSpellLanguage(pos_type const from) const;
 385
 386         Language * locateSpellRange(pos_type & from, pos_type & to,
 387                                                                 SkipPositions & skips) const;
 388
 389         bool hasSpellerChange() const {
 390                 SpellChecker::ChangeNumber speller_change_number = 0;
 391                 if (theSpellChecker())
 392                         speller_change_number = theSpellChecker()->changeNumber();
 393                 return speller_change_number > speller_state_.currentChangeNumber();
 394         }
 395
 396         bool ignoreWord(docstring const & word) const ;
 397
 398         void setMisspelled(pos_type from, pos_type to, SpellChecker::Result state)
 399         {
 400                 pos_type textsize = owner_->size();
 401                 // check for sane arguments
 402                 if (to <= from || from >= textsize)
 403                         return;
 404                 FontSpan fp = FontSpan(from, to - 1);
 405                 speller_state_.setRange(fp, state);
 406         }
 407
 408         void requestSpellCheck(pos_type pos) {
 409                 if (pos == -1)
 410                         speller_state_.needsCompleteRefresh(speller_state_.currentChangeNumber());
 411                 else
 412                         speller_state_.needsRefresh(pos);
 413         }
 414
 415         void readySpellCheck() {
 416                 speller_state_.needsRefresh(-1);
 417         }
 418
 419         bool needsSpellCheck() const
 420         {
 421                 return speller_state_.needsRefresh();
 422         }
 423
 424         void rangeOfSpellCheck(pos_type & first, pos_type & last) const
 425         {
 426                 speller_state_.refreshRange(first, last);
 427                 if (last == -1) {
 428                         last = owner_->size();
 429                         return;
 430                 }
 431                 pos_type endpos = last;
 432                 owner_->locateWord(first, endpos, WHOLE_WORD);
 433                 if (endpos < last) {
 434                         endpos = last;
 435                         owner_->locateWord(last, endpos, WHOLE_WORD);
 436                 }
 437                 last = endpos;
 438         }
 439
 440         int countSkips(SkipPositionsIterator & it, SkipPositionsIterator const et,
 441                             int & start) const
 442         {
 443                 int numskips = 0;
 444                 while (it != et && it->first < start) {
 445                         int skip = it->last - it->first + 1;
 446                         start += skip;
 447                         numskips += skip;
 448                         ++it;
 449                 }
 450                 return numskips;
 451         }
 452
 453         void markMisspelledWords(pos_type const & first, pos_type const & last,
 454                                                          SpellChecker::Result result,
 455                                                          docstring const & word,
 456                                                          SkipPositions const & skips);
 457
 458         InsetCode ownerCode() const
 459         {
 460                 return inset_owner_ ? inset_owner_->lyxCode() : NO_CODE;
 461         }
 462
 463         /// Which Paragraph owns us?
 464         Paragraph * owner_;
 465
 466         /// In which Inset?
 467         Inset const * inset_owner_;
 468
 469         ///
 470         FontList fontlist_;
 471
 472         ///
 473         int id_;
 474
 475         ///
 476         ParagraphParameters params_;
 477
 478         /// for recording and looking up changes
 479         Changes changes_;
 480
 481         ///
 482         InsetList insetlist_;
 483
 484         /// end of label
 485         pos_type begin_of_body_;
 486
 487         typedef docstring TextContainer;
 488         ///
 489         TextContainer text_;
 490
 491         typedef set<docstring> Words;
 492         typedef map<Language, Words> LangWordsMap;
 493         ///
 494         LangWordsMap words_;
 495         ///
 496         Layout const * layout_;
 497         ///
 498         SpellCheckerState speller_state_;
 499 };
 500
 501
 502 namespace {
 503
 504 struct special_phrase {
 505         string phrase;
 506         docstring macro;
 507         bool builtin;
 508 };
 509
 510 special_phrase const special_phrases[] = {
 511         { "LyX", from_ascii("\\LyX{}"), false },
 512         { "TeX", from_ascii("\\TeX{}"), true },
 513         { "LaTeX2e", from_ascii("\\LaTeXe{}"), true },
 514         { "LaTeX", from_ascii("\\LaTeX{}"), true },
 515 };
 516
 517 size_t const phrases_nr = sizeof(special_phrases)/sizeof(special_phrase);
 518
 519 } // namespace anon
 520
 521
 522 Paragraph::Private::Private(Paragraph * owner, Layout const & layout)
 523         : owner_(owner), inset_owner_(0), id_(-1), begin_of_body_(0), layout_(&layout)
 524 {
 525         text_.reserve(100);
 526 }
 527
 528
 529 // Initialization of the counter for the paragraph id's,
 530 //
 531 // FIXME: There should be a more intelligent way to generate and use the
 532 // paragraph ids per buffer instead a global static counter for all InsetText
 533 // in the running program.
 534 static int paragraph_id = -1;
 535
 536 Paragraph::Private::Private(Private const & p, Paragraph * owner)
 537         : owner_(owner), inset_owner_(p.inset_owner_), fontlist_(p.fontlist_),
 538           params_(p.params_), changes_(p.changes_), insetlist_(p.insetlist_),
 539           begin_of_body_(p.begin_of_body_), text_(p.text_), words_(p.words_),
 540           layout_(p.layout_)
 541 {
 542         id_ = ++paragraph_id;
 543         requestSpellCheck(p.text_.size());
 544 }
 545
 546
 547 Paragraph::Private::Private(Private const & p, Paragraph * owner,
 548         pos_type beg, pos_type end)
 549         : owner_(owner), inset_owner_(p.inset_owner_),
 550           params_(p.params_), changes_(p.changes_),
 551           insetlist_(p.insetlist_, beg, end),
 552           begin_of_body_(p.begin_of_body_), words_(p.words_),
 553           layout_(p.layout_)
 554 {
 555         id_ = ++paragraph_id;
 556         if (beg >= pos_type(p.text_.size()))
 557                 return;
 558         text_ = p.text_.substr(beg, end - beg);
 559
 560         FontList::const_iterator fcit = fontlist_.begin();
 561         FontList::const_iterator fend = fontlist_.end();
 562         for (; fcit != fend; ++fcit) {
 563                 if (fcit->pos() < beg)
 564                         continue;
 565                 if (fcit->pos() >= end) {
 566                         // Add last entry in the fontlist_.
 567                         fontlist_.set(text_.size() - 1, fcit->font());
 568                         break;
 569                 }
 570                 // Add a new entry in the fontlist_.
 571                 fontlist_.set(fcit->pos() - beg, fcit->font());
 572         }
 573         requestSpellCheck(p.text_.size());
 574 }
 575
 576
 577 void Paragraph::addChangesToToc(DocIterator const & cdit,
 578         Buffer const & buf) const
 579 {
 580         d->changes_.addToToc(cdit, buf);
 581 }
 582
 583
 584 bool Paragraph::isDeleted(pos_type start, pos_type end) const
 585 {
 586         LASSERT(start >= 0 && start <= size(), /**/);
 587         LASSERT(end > start && end <= size() + 1, /**/);
 588
 589         return d->changes_.isDeleted(start, end);
 590 }
 591
 592
 593 bool Paragraph::isChanged(pos_type start, pos_type end) const
 594 {
 595         LASSERT(start >= 0 && start <= size(), /**/);
 596         LASSERT(end > start && end <= size() + 1, /**/);
 597
 598         return d->changes_.isChanged(start, end);
 599 }
 600
 601
 602 bool Paragraph::isMergedOnEndOfParDeletion(bool trackChanges) const
 603 {
 604         // keep the logic here in sync with the logic of eraseChars()
 605         if (!trackChanges)
 606                 return true;
 607
 608         Change const change = d->changes_.lookup(size());
 609         return change.inserted() && change.currentAuthor();
 610 }
 611
 612
 613 void Paragraph::setChange(Change const & change)
 614 {
 615         // beware of the imaginary end-of-par character!
 616         d->changes_.set(change, 0, size() + 1);
 617
 618         /*
 619          * Propagate the change recursively - but not in case of DELETED!
 620          *
 621          * Imagine that your co-author makes changes in an existing inset. He
 622          * sends your document to you and you come to the conclusion that the
 623          * inset should go completely. If you erase it, LyX must not delete all
 624          * text within the inset. Otherwise, the change tracked insertions of
 625          * your co-author get lost and there is no way to restore them later.
 626          *
 627          * Conclusion: An inset's content should remain untouched if you delete it
 628          */
 629
 630         if (!change.deleted()) {
 631                 for (pos_type pos = 0; pos < size(); ++pos) {
 632                         if (Inset * inset = getInset(pos))
 633                                 inset->setChange(change);
 634                 }
 635         }
 636 }
 637
 638
 639 void Paragraph::setChange(pos_type pos, Change const & change)
 640 {
 641         LASSERT(pos >= 0 && pos <= size(), /**/);
 642         d->changes_.set(change, pos);
 643
 644         // see comment in setChange(Change const &) above
 645         if (!change.deleted() && pos < size())
 646                         if (Inset * inset = getInset(pos))
 647                                 inset->setChange(change);
 648 }
 649
 650
 651 Change const & Paragraph::lookupChange(pos_type pos) const
 652 {
 653         LASSERT(pos >= 0 && pos <= size(), /**/);
 654         return d->changes_.lookup(pos);
 655 }
 656
 657
 658 void Paragraph::acceptChanges(pos_type start, pos_type end)
 659 {
 660         LASSERT(start >= 0 && start <= size(), /**/);
 661         LASSERT(end > start && end <= size() + 1, /**/);
 662
 663         for (pos_type pos = start; pos < end; ++pos) {
 664                 switch (lookupChange(pos).type) {
 665                         case Change::UNCHANGED:
 666                                 // accept changes in nested inset
 667                                 if (Inset * inset = getInset(pos))
 668                                         inset->acceptChanges();
 669                                 break;
 670
 671                         case Change::INSERTED:
 672                                 d->changes_.set(Change(Change::UNCHANGED), pos);
 673                                 // also accept changes in nested inset
 674                                 if (Inset * inset = getInset(pos))
 675                                         inset->acceptChanges();
 676                                 break;
 677
 678                         case Change::DELETED:
 679                                 // Suppress access to non-existent
 680                                 // "end-of-paragraph char"
 681                                 if (pos < size()) {
 682                                         eraseChar(pos, false);
 683                                         --end;
 684                                         --pos;
 685                                 }
 686                                 break;
 687                 }
 688
 689         }
 690 }
 691
 692
 693 void Paragraph::rejectChanges(pos_type start, pos_type end)
 694 {
 695         LASSERT(start >= 0 && start <= size(), /**/);
 696         LASSERT(end > start && end <= size() + 1, /**/);
 697
 698         for (pos_type pos = start; pos < end; ++pos) {
 699                 switch (lookupChange(pos).type) {
 700                         case Change::UNCHANGED:
 701                                 // reject changes in nested inset
 702                                 if (Inset * inset = getInset(pos))
 703                                                 inset->rejectChanges();
 704                                 break;
 705
 706                         case Change::INSERTED:
 707                                 // Suppress access to non-existent
 708                                 // "end-of-paragraph char"
 709                                 if (pos < size()) {
 710                                         eraseChar(pos, false);
 711                                         --end;
 712                                         --pos;
 713                                 }
 714                                 break;
 715
 716                         case Change::DELETED:
 717                                 d->changes_.set(Change(Change::UNCHANGED), pos);
 718
 719                                 // Do NOT reject changes within a deleted inset!
 720                                 // There may be insertions of a co-author inside of it!
 721
 722                                 break;
 723                 }
 724         }
 725 }
 726
 727
 728 void Paragraph::Private::insertChar(pos_type pos, char_type c,
 729                 Change const & change)
 730 {
 731         LASSERT(pos >= 0 && pos <= int(text_.size()), /**/);
 732
 733         // track change
 734         changes_.insert(change, pos);
 735
 736         // This is actually very common when parsing buffers (and
 737         // maybe inserting ascii text)
 738         if (pos == pos_type(text_.size())) {
 739                 // when appending characters, no need to update tables
 740                 text_.push_back(c);
 741                 // but we want spell checking
 742                 requestSpellCheck(pos);
 743                 return;
 744         }
 745
 746         text_.insert(text_.begin() + pos, c);
 747
 748         // Update the font table.
 749         fontlist_.increasePosAfterPos(pos);
 750
 751         // Update the insets
 752         insetlist_.increasePosAfterPos(pos);
 753
 754         // Update list of misspelled positions
 755         speller_state_.increasePosAfterPos(pos);
 756 }
 757
 758
 759 bool Paragraph::insertInset(pos_type pos, Inset * inset,
 760                                    Change const & change)
 761 {
 762         LASSERT(inset, /**/);
 763         LASSERT(pos >= 0 && pos <= size(), /**/);
 764
 765         // Paragraph::insertInset() can be used in cut/copy/paste operation where
 766         // d->inset_owner_ is not set yet.
 767         if (d->inset_owner_ && !d->inset_owner_->insetAllowed(inset->lyxCode()))
 768                 return false;
 769
 770         d->insertChar(pos, META_INSET, change);
 771         LASSERT(d->text_[pos] == META_INSET, /**/);
 772
 773         // Add a new entry in the insetlist_.
 774         d->insetlist_.insert(inset, pos);
 775
 776         // Some insets require run of spell checker
 777         requestSpellCheck(pos);
 778         return true;
 779 }
 780
 781
 782 bool Paragraph::eraseChar(pos_type pos, bool trackChanges)
 783 {
 784         LASSERT(pos >= 0 && pos <= size(), return false);
 785
 786         // keep the logic here in sync with the logic of isMergedOnEndOfParDeletion()
 787
 788         if (trackChanges) {
 789                 Change change = d->changes_.lookup(pos);
 790
 791                 // set the character to DELETED if
 792                 //  a) it was previously unchanged or
 793                 //  b) it was inserted by a co-author
 794
 795                 if (!change.changed() ||
 796                       (change.inserted() && !change.currentAuthor())) {
 797                         setChange(pos, Change(Change::DELETED));
 798                         // request run of spell checker
 799                         requestSpellCheck(pos);
 800                         return false;
 801                 }
 802
 803                 if (change.deleted())
 804                         return false;
 805         }
 806
 807         // Don't physically access the imaginary end-of-paragraph character.
 808         // eraseChar() can only mark it as DELETED. A physical deletion of
 809         // end-of-par must be handled externally.
 810         if (pos == size()) {
 811                 return false;
 812         }
 813
 814         // track change
 815         d->changes_.erase(pos);
 816
 817         // if it is an inset, delete the inset entry
 818         if (d->text_[pos] == META_INSET)
 819                 d->insetlist_.erase(pos);
 820
 821         d->text_.erase(d->text_.begin() + pos);
 822
 823         // Update the fontlist_
 824         d->fontlist_.erase(pos);
 825
 826         // Update the insetlist_
 827         d->insetlist_.decreasePosAfterPos(pos);
 828
 829         // Update list of misspelled positions
 830         d->speller_state_.decreasePosAfterPos(pos);
 831         d->speller_state_.refreshLast(size());
 832
 833         return true;
 834 }
 835
 836
 837 int Paragraph::eraseChars(pos_type start, pos_type end, bool trackChanges)
 838 {
 839         LASSERT(start >= 0 && start <= size(), /**/);
 840         LASSERT(end >= start && end <= size() + 1, /**/);
 841
 842         pos_type i = start;
 843         for (pos_type count = end - start; count; --count) {
 844                 if (!eraseChar(i, trackChanges))
 845                         ++i;
 846         }
 847         return end - i;
 848 }
 849
 850
 851 int Paragraph::Private::latexSurrogatePair(otexstream & os, char_type c,
 852                 char_type next, OutputParams const & runparams)
 853 {
 854         // Writing next here may circumvent a possible font change between
 855         // c and next. Since next is only output if it forms a surrogate pair
 856         // with c we can ignore this:
 857         // A font change inside a surrogate pair does not make sense and is
 858         // hopefully impossible to input.
 859         // FIXME: change tracking
 860         // Is this correct WRT change tracking?
 861         Encoding const & encoding = *(runparams.encoding);
 862         docstring const latex1 = encoding.latexChar(next).first;
 863         docstring const latex2 = encoding.latexChar(c).first;
 864         if (docstring(1, next) == latex1) {
 865                 // the encoding supports the combination
 866                 os << latex2 << latex1;
 867                 return latex1.length() + latex2.length();
 868         } else if (runparams.local_font &&
 869                    runparams.local_font->language()->lang() == "polutonikogreek") {
 870                 // polutonikogreek only works without the brackets
 871                 os << latex1 << latex2;
 872                 return latex1.length() + latex2.length();
 873         } else
 874                 os << latex1 << '{' << latex2 << '}';
 875         return latex1.length() + latex2.length() + 2;
 876 }
 877
 878
 879 bool Paragraph::Private::simpleTeXBlanks(OutputParams const & runparams,
 880                                        otexstream & os,
 881                                        pos_type i,
 882                                        unsigned int & column,
 883                                        Font const & font,
 884                                        Layout const & style)
 885 {
 886         if (style.pass_thru || runparams.pass_thru)
 887                 return false;
 888
 889         if (i + 1 < int(text_.size())) {
 890                 char_type next = text_[i + 1];
 891                 if (Encodings::isCombiningChar(next)) {
 892                         // This space has an accent, so we must always output it.
 893                         column += latexSurrogatePair(os, ' ', next, runparams) - 1;
 894                         return true;
 895                 }
 896         }
 897
 898         if (runparams.linelen > 0
 899             && column > runparams.linelen
 900             && i
 901             && text_[i - 1] != ' '
 902             && (i + 1 < int(text_.size()))
 903             // same in FreeSpacing mode
 904             && !owner_->isFreeSpacing()
 905             // In typewriter mode, we want to avoid
 906             // ! . ? : at the end of a line
 907             && !(font.fontInfo().family() == TYPEWRITER_FAMILY
 908                  && (text_[i - 1] == '.'
 909                      || text_[i - 1] == '?'
 910                      || text_[i - 1] == ':'
 911                      || text_[i - 1] == '!'))) {
 912                 os << '\n';
 913                 os.texrow().start(owner_->id(), i + 1);
 914                 column = 0;
 915         } else if (style.free_spacing) {
 916                 os << '~';
 917         } else {
 918                 os << ' ';
 919         }
 920         return false;
 921 }
 922
 923
 924 int Paragraph::Private::writeScriptChars(otexstream & os,
 925                                          docstring const & ltx,
 926                                          Change const & runningChange,
 927                                          Encoding const & encoding,
 928                                          pos_type & i)
 929 {
 930         // FIXME: modifying i here is not very nice...
 931
 932         // We only arrive here when a proper language for character text_[i] has
 933         // not been specified (i.e., it could not be translated in the current
 934         // latex encoding) or its latex translation has been forced, and it
 935         // belongs to a known script.
 936         // Parameter ltx contains the latex translation of text_[i] as specified
 937         // in the unicodesymbols file and is something like "\textXXX{<spec>}".
 938         // The latex macro name "textXXX" specifies the script to which text_[i]
 939         // belongs and we use it in order to check whether characters from the
 940         // same script immediately follow, such that we can collect them in a
 941         // single "\textXXX" macro. So, we have to retain "\textXXX{<spec>"
 942         // for the first char but only "<spec>" for all subsequent chars.
 943         docstring::size_type const brace1 = ltx.find_first_of(from_ascii("{"));
 944         docstring::size_type const brace2 = ltx.find_last_of(from_ascii("}"));
 945         string script = to_ascii(ltx.substr(1, brace1 - 1));
 946         int pos = 0;
 947         int length = brace2;
 948         bool closing_brace = true;
 949         if (script == "textgreek" && encoding.latexName() == "iso-8859-7") {
 950                 // Correct encoding is being used, so we can avoid \textgreek.
 951                 pos = brace1 + 1;
 952                 length -= pos;
 953                 closing_brace = false;
 954         }
 955         os << ltx.substr(pos, length);
 956         int size = text_.size();
 957         while (i + 1 < size) {
 958                 char_type const next = text_[i + 1];
 959                 // Stop here if next character belongs to another script
 960                 // or there is a change in change tracking status.
 961                 if (!Encodings::isKnownScriptChar(next, script) ||
 962                     runningChange != owner_->lookupChange(i + 1))
 963                         break;
 964                 Font prev_font;
 965                 bool found = false;
 966                 FontList::const_iterator cit = fontlist_.begin();
 967                 FontList::const_iterator end = fontlist_.end();
 968                 for (; cit != end; ++cit) {
 969                         if (cit->pos() >= i && !found) {
 970                                 prev_font = cit->font();
 971                                 found = true;
 972                         }
 973                         if (cit->pos() >= i + 1)
 974                                 break;
 975                 }
 976                 // Stop here if there is a font attribute or encoding change.
 977                 if (found && cit != end && prev_font != cit->font())
 978                         break;
 979                 docstring const latex = encoding.latexChar(next).first;
 980                 docstring::size_type const b1 =
 981                                         latex.find_first_of(from_ascii("{"));
 982                 docstring::size_type const b2 =
 983                                         latex.find_last_of(from_ascii("}"));
 984                 int const len = b2 - b1 - 1;
 985                 os << latex.substr(b1 + 1, len);
 986                 length += len;
 987                 ++i;
 988         }
 989         if (closing_brace) {
 990                 os << '}';
 991                 ++length;
 992         }
 993         return length;
 994 }
 995
 996
 997 bool Paragraph::Private::isTextAt(string const & str, pos_type pos) const
 998 {
 999         pos_type const len = str.length();
1000
1001         // is the paragraph large enough?
1002         if (pos + len > int(text_.size()))
1003                 return false;
1004
1005         // does the wanted text start at point?
1006         for (string::size_type i = 0; i < str.length(); ++i) {
1007                 // Caution: direct comparison of characters works only
1008                 // because str is pure ASCII.
1009                 if (str[i] != text_[pos + i])
1010                         return false;
1011         }
1012
1013         return fontlist_.hasChangeInRange(pos, len);
1014 }
1015
1016
1017 void Paragraph::Private::latexInset(BufferParams const & bparams,
1018                                     otexstream & os,
1019                                     OutputParams & runparams,
1020                                     Font & running_font,
1021                                     Font & basefont,
1022                                     Font const & outerfont,
1023                                     bool & open_font,
1024                                     Change & running_change,
1025                                     Layout const & style,
1026                                     pos_type & i,
1027                                     unsigned int & column)
1028 {
1029         Inset * inset = owner_->getInset(i);
1030         LASSERT(inset, /**/);
1031
1032         if (style.pass_thru) {
1033                 inset->plaintext(os.os(), runparams);
1034                 return;
1035         }
1036
1037         // FIXME: move this to InsetNewline::latex
1038         if (inset->lyxCode() == NEWLINE_CODE) {
1039                 // newlines are handled differently here than
1040                 // the default in simpleTeXSpecialChars().
1041                 if (!style.newline_allowed) {
1042                         os << '\n';
1043                 } else {
1044                         if (open_font) {
1045                                 column += running_font.latexWriteEndChanges(
1046                                         os, bparams, runparams,
1047                                         basefont, basefont);
1048                                 open_font = false;
1049                         }
1050
1051                         if (running_font.fontInfo().family() == TYPEWRITER_FAMILY)
1052                                 os << '~';
1053
1054                         basefont = owner_->getLayoutFont(bparams, outerfont);
1055                         running_font = basefont;
1056
1057                         if (runparams.moving_arg)
1058                                 os << "\\protect ";
1059
1060                 }
1061                 os.texrow().start(owner_->id(), i + 1);
1062                 column = 0;
1063         }
1064
1065         if (owner_->isDeleted(i)) {
1066                 if( ++runparams.inDeletedInset == 1)
1067                         runparams.changeOfDeletedInset = owner_->lookupChange(i);
1068         }
1069
1070         if (inset->canTrackChanges()) {
1071                 column += Changes::latexMarkChange(os, bparams, running_change,
1072                         Change(Change::UNCHANGED), runparams);
1073                 running_change = Change(Change::UNCHANGED);
1074         }
1075
1076         bool close = false;
1077         odocstream::pos_type const len = os.os().tellp();
1078
1079         if (inset->forceLTR()
1080             && !runparams.use_polyglossia
1081             && running_font.isRightToLeft()
1082             // ERT is an exception, it should be output with no
1083             // decorations at all
1084             && inset->lyxCode() != ERT_CODE) {
1085                 if (running_font.language()->lang() == "farsi")
1086                         os << "\\beginL{}";
1087                 else
1088                         os << "\\L{";
1089                 close = true;
1090         }
1091
1092         // FIXME: Bug: we can have an empty font change here!
1093         // if there has just been a font change, we are going to close it
1094         // right now, which means stupid latex code like \textsf{}. AFAIK,
1095         // this does not harm dvi output. A minor bug, thus (JMarc)
1096
1097         // Some insets cannot be inside a font change command.
1098         // However, even such insets *can* be placed in \L or \R
1099         // or their equivalents (for RTL language switches), so we don't
1100         // close the language in those cases.
1101         // ArabTeX, though, cannot handle this special behavior, it seems.
1102         bool arabtex = basefont.language()->lang() == "arabic_arabtex"
1103                 || running_font.language()->lang() == "arabic_arabtex";
1104         if (open_font && !inset->inheritFont()) {
1105                 bool closeLanguage = arabtex
1106                         || basefont.isRightToLeft() == running_font.isRightToLeft();
1107                 unsigned int count = running_font.latexWriteEndChanges(os,
1108                         bparams, runparams, basefont, basefont, closeLanguage);
1109                 column += count;
1110                 // if any font properties were closed, update the running_font,
1111                 // making sure, however, to leave the language as it was
1112                 if (count > 0) {
1113                         // FIXME: probably a better way to keep track of the old
1114                         // language, than copying the entire font?
1115                         Font const copy_font(running_font);
1116                         basefont = owner_->getLayoutFont(bparams, outerfont);
1117                         running_font = basefont;
1118                         if (!closeLanguage)
1119                                 running_font.setLanguage(copy_font.language());
1120                         // leave font open if language is still open
1121                         open_font = (running_font.language() == basefont.language());
1122                         if (closeLanguage)
1123                                 runparams.local_font = &basefont;
1124                 }
1125         }
1126
1127         int prev_rows = os.texrow().rows();
1128
1129         try {
1130                 runparams.lastid = id_;
1131                 runparams.lastpos = i;
1132                 inset->latex(os, runparams);
1133         } catch (EncodingException & e) {
1134                 // add location information and throw again.
1135                 e.par_id = id_;
1136                 e.pos = i;
1137                 throw(e);
1138         }
1139
1140         if (close) {
1141                 if (running_font.language()->lang() == "farsi")
1142                                 os << "\\endL{}";
1143                         else
1144                                 os << '}';
1145         }
1146
1147         if (os.texrow().rows() > prev_rows) {
1148                 os.texrow().start(owner_->id(), i + 1);
1149                 column = 0;
1150         } else {
1151                 column += (unsigned int)(os.os().tellp() - len);
1152         }
1153
1154         if (owner_->isDeleted(i))
1155                 --runparams.inDeletedInset;
1156 }
1157
1158
1159 void Paragraph::Private::latexSpecialChar(otexstream & os,
1160                                           BufferParams const & bparams,
1161                                           OutputParams const & runparams,
1162                                           Font const & running_font,
1163                                           Change const & running_change,
1164                                           Layout const & style,
1165                                           pos_type & i,
1166                                           pos_type end_pos,
1167                                           unsigned int & column)
1168 {
1169         // With polyglossia, brackets and stuff need not be reversed
1170         // in RTL scripts (see bug #8251)
1171         char_type const c = (runparams.use_polyglossia) ?
1172                 owner_->getUChar(bparams, i) : text_[i];
1173
1174         if (style.pass_thru || runparams.pass_thru) {
1175                 if (c != '\0') {
1176                         Encoding const * const enc = runparams.encoding;
1177                         if (enc && !enc->encodable(c))
1178                                 throw EncodingException(c);
1179                         os.put(c);
1180                 }
1181                 return;
1182         }
1183
1184         // TIPA uses its own T3 encoding
1185         if (runparams.inIPA && latexSpecialT3(c, os, i, column))
1186                 return;
1187         // If T1 font encoding is used, use the special
1188         // characters it provides.
1189         // NOTE: some languages reset the font encoding
1190         // internally
1191         if (!runparams.inIPA && !running_font.language()->internalFontEncoding()
1192             && lyxrc.fontenc == "T1" && latexSpecialT1(c, os, i, column))
1193                 return;
1194
1195         // \tt font needs special treatment
1196         if (!runparams.inIPA
1197              && running_font.fontInfo().family() == TYPEWRITER_FAMILY
1198              && latexSpecialTypewriter(c, os, i, column))
1199                 return;
1200
1201         // Otherwise, we use what LaTeX provides us.
1202         switch (c) {
1203         case '\\':
1204                 os << "\\textbackslash{}";
1205                 column += 15;
1206                 break;
1207         case '<':
1208                 os << "\\textless{}";
1209                 column += 10;
1210                 break;
1211         case '>':
1212                 os << "\\textgreater{}";
1213                 column += 13;
1214                 break;
1215         case '|':
1216                 os << "\\textbar{}";
1217                 column += 9;
1218                 break;
1219         case '-':
1220                 os << '-';
1221                 break;
1222         case '\"':
1223                 os << "\\char`\\\"{}";
1224                 column += 9;
1225                 break;
1226
1227         case '$': case '&':
1228         case '%': case '#': case '{':
1229         case '}': case '_':
1230                 os << '\\';
1231                 os.put(c);
1232                 column += 1;
1233                 break;
1234
1235         case '~':
1236                 os << "\\textasciitilde{}";
1237                 column += 16;
1238                 break;
1239
1240         case '^':
1241                 os << "\\textasciicircum{}";
1242                 column += 17;
1243                 break;
1244
1245         case '*':
1246         case '[':
1247         case ']':
1248                 // avoid being mistaken for optional arguments
1249                 os << '{';
1250                 os.put(c);
1251                 os << '}';
1252                 column += 2;
1253                 break;
1254
1255         case ' ':
1256                 // Blanks are printed before font switching.
1257                 // Sure? I am not! (try nice-latex)
1258                 // I am sure it's correct. LyX might be smarter
1259                 // in the future, but for now, nothing wrong is
1260                 // written. (Asger)
1261                 break;
1262
1263         default:
1264                 // LyX, LaTeX etc.
1265                 if (latexSpecialPhrase(os, i, end_pos, column, runparams))
1266                         return;
1267
1268                 if (c == '\0')
1269                         return;
1270
1271                 Encoding const & encoding = *(runparams.encoding);
1272                 char_type next = '\0';
1273                 if (i + 1 < int(text_.size())) {
1274                         next = text_[i + 1];
1275                         if (Encodings::isCombiningChar(next)) {
1276                                 column += latexSurrogatePair(os, c, next, runparams) - 1;
1277                                 ++i;
1278                                 break;
1279                         }
1280                 }
1281                 string script;
1282                 pair<docstring, bool> latex = encoding.latexChar(c);
1283                 docstring nextlatex;
1284                 bool nexttipas = false;
1285                 string nexttipashortcut;
1286                 if (next != '\0' && next != META_INSET) {
1287                         nextlatex = encoding.latexChar(next).first;
1288                         if (runparams.inIPA) {
1289                                 nexttipashortcut = Encodings::TIPAShortcut(next);
1290                                 nexttipas = !nexttipashortcut.empty();
1291                         }
1292                 }
1293                 bool tipas = false;
1294                 if (runparams.inIPA) {
1295                         string const tipashortcut = Encodings::TIPAShortcut(c);
1296                         if (!tipashortcut.empty()) {
1297                                 latex.first = from_ascii(tipashortcut);
1298                                 latex.second = false;
1299                                 tipas = true;
1300                         }
1301                 }
1302                 if (Encodings::isKnownScriptChar(c, script)
1303                     && prefixIs(latex.first, from_ascii("\\" + script)))
1304                         column += writeScriptChars(os, latex.first,
1305                                         running_change, encoding, i) - 1;
1306                 else if (latex.second
1307                          && ((!prefixIs(nextlatex, '\\')
1308                                && !prefixIs(nextlatex, '{')
1309                                && !prefixIs(nextlatex, '}'))
1310                              || (nexttipas
1311                                  && !prefixIs(from_ascii(nexttipashortcut), '\\')))
1312                          && !tipas) {
1313                         // Prevent eating of a following
1314                         // space or command corruption by
1315                         // following characters
1316                         if (next == ' ' || next == '\0') {
1317                                 column += latex.first.length() + 1;
1318                                 os << latex.first << "{}";
1319                         } else {
1320                                 column += latex.first.length();
1321                                 os << latex.first << " ";
1322                         }
1323                 } else {
1324                         column += latex.first.length() - 1;
1325                         os << latex.first;
1326                 }
1327                 break;
1328         }
1329 }
1330
1331
1332 bool Paragraph::Private::latexSpecialT1(char_type const c, otexstream & os,
1333         pos_type i, unsigned int & column)
1334 {
1335         switch (c) {
1336         case '>':
1337         case '<':
1338                 os.put(c);
1339                 // In T1 encoding, these characters exist
1340                 // but we should avoid ligatures
1341                 if (i + 1 >= int(text_.size()) || text_[i + 1] != c)
1342                         return true;
1343                 os << "\\textcompwordmark{}";
1344                 column += 19;
1345                 return true;
1346         case '|':
1347                 os.put(c);
1348                 return true;
1349         case '\"':
1350                 // soul.sty breaks with \char`\"
1351                 os << "\\textquotedbl{}";
1352                 column += 14;
1353                 return true;
1354         default:
1355                 return false;
1356         }
1357 }
1358
1359
1360 bool Paragraph::Private::latexSpecialT3(char_type const c, otexstream & os,
1361         pos_type /*i*/, unsigned int & /*column*/)
1362 {
1363         switch (c) {
1364         case '*':
1365         case '[':
1366         case ']':
1367         case '|':
1368         case '\"':
1369                 os.put(c);
1370                 return true;
1371         default:
1372                 return false;
1373         }
1374 }
1375
1376
1377 bool Paragraph::Private::latexSpecialTypewriter(char_type const c, otexstream & os,
1378         pos_type i, unsigned int & column)
1379 {
1380         switch (c) {
1381         case '-':
1382                 // within \ttfamily, "--" is merged to "-" (no endash)
1383                 // so we avoid this rather irritating ligature
1384                 if (i + 1 < int(text_.size()) && text_[i + 1] == '-') {
1385                         os << "-{}";
1386                         column += 2;
1387                 } else
1388                         os << '-';
1389                 return true;
1390
1391         // everything else has to be checked separately
1392         // (depending on the encoding)
1393         default:
1394                 return false;
1395         }
1396 }
1397
1398
1399 /// \param end_pos
1400 ///   If [start_pos, end_pos) does not include entirely the special phrase, then
1401 ///   do not apply the macro transformation.
1402 bool Paragraph::Private::latexSpecialPhrase(otexstream & os, pos_type & i, pos_type end_pos,
1403         unsigned int & column, OutputParams const & runparams)
1404 {
1405         // FIXME: if we have "LaTeX" with a font
1406         // change in the middle (before the 'T', then
1407         // the "TeX" part is still special cased.
1408         // Really we should only operate this on
1409         // "words" for some definition of word
1410
1411         for (size_t pnr = 0; pnr < phrases_nr; ++pnr) {
1412                 if (!isTextAt(special_phrases[pnr].phrase, i)
1413                     || (end_pos != -1 && i + int(special_phrases[pnr].phrase.size()) > end_pos))
1414                         continue;
1415                 if (runparams.moving_arg)
1416                         os << "\\protect";
1417                 os << special_phrases[pnr].macro;
1418                 i += special_phrases[pnr].phrase.length() - 1;
1419                 column += special_phrases[pnr].macro.length() - 1;
1420                 return true;
1421         }
1422         return false;
1423 }
1424
1425
1426 void Paragraph::Private::validate(LaTeXFeatures & features) const
1427 {
1428         if (layout_->inpreamble && inset_owner_) {
1429                 bool const is_command = layout_->latextype == LATEX_COMMAND;
1430                 Buffer const & buf = inset_owner_->buffer();
1431                 BufferParams const & bp = buf.params();
1432                 Font f;
1433                 TexRow texrow;
1434                 // Using a string stream here circumvents the encoding
1435                 // switching machinery of odocstream. Therefore the
1436                 // output is wrong if this paragraph contains content
1437                 // that needs to switch encoding.
1438                 odocstringstream ods;
1439                 otexstream os(ods, texrow);
1440                 if (is_command) {
1441                         os << '\\' << from_ascii(layout_->latexname());
1442                         // we have to provide all the optional arguments here, even though
1443                         // the last one is the only one we care about.
1444                         // Separate handling of optional argument inset.
1445                         if (layout_->optargs != 0 || layout_->reqargs != 0)
1446                                 latexArgInsets(*owner_, os, features.runparams(),
1447                                         layout_->reqargs, layout_->optargs);
1448                         else
1449                                 os << from_ascii(layout_->latexparam());
1450                 }
1451                 docstring::size_type const length = ods.str().length();
1452                 // this will output "{" at the beginning, but not at the end
1453                 owner_->latex(bp, f, os, features.runparams(), 0, -1, true);
1454                 if (ods.str().length() > length) {
1455                         if (is_command)
1456                                 ods << '}';
1457                         string const snippet = to_utf8(ods.str());
1458                         features.addPreambleSnippet(snippet);
1459                 }
1460         }
1461
1462         if (features.runparams().flavor == OutputParams::HTML
1463             && layout_->htmltitle()) {
1464                 features.setHTMLTitle(owner_->asString(AS_STR_INSETS | AS_STR_SKIPDELETE));
1465         }
1466
1467         // check the params.
1468         if (!params_.spacing().isDefault())
1469                 features.require("setspace");
1470
1471         // then the layouts
1472         features.useLayout(layout_->name());
1473
1474         // then the fonts
1475         fontlist_.validate(features);
1476
1477         // then the indentation
1478         if (!params_.leftIndent().zero())
1479                 features.require("ParagraphLeftIndent");
1480
1481         // then the insets
1482         InsetList::const_iterator icit = insetlist_.begin();
1483         InsetList::const_iterator iend = insetlist_.end();
1484         for (; icit != iend; ++icit) {
1485                 if (icit->inset) {
1486                         icit->inset->validate(features);
1487                         if (layout_->needprotect &&
1488                             icit->inset->lyxCode() == FOOT_CODE)
1489                                 features.require("NeedLyXFootnoteCode");
1490                 }
1491         }
1492
1493         // then the contents
1494         for (pos_type i = 0; i < int(text_.size()) ; ++i) {
1495                 for (size_t pnr = 0; pnr < phrases_nr; ++pnr) {
1496                         if (!special_phrases[pnr].builtin
1497                             && isTextAt(special_phrases[pnr].phrase, i)) {
1498                                 features.require(special_phrases[pnr].phrase);
1499                                 break;
1500                         }
1501                 }
1502                 Encodings::validate(text_[i], features);
1503         }
1504 }
1505
1506 /////////////////////////////////////////////////////////////////////
1507 //
1508 // Paragraph
1509 //
1510 /////////////////////////////////////////////////////////////////////
1511
1512 namespace {
1513         Layout const emptyParagraphLayout;
1514 }
1515
1516 Paragraph::Paragraph()
1517         : d(new Paragraph::Private(this, emptyParagraphLayout))
1518 {
1519         itemdepth = 0;
1520         d->params_.clear();
1521 }
1522
1523
1524 Paragraph::Paragraph(Paragraph const & par)
1525         : itemdepth(par.itemdepth),
1526         d(new Paragraph::Private(*par.d, this))
1527 {
1528         registerWords();
1529 }
1530
1531
1532 Paragraph::Paragraph(Paragraph const & par, pos_type beg, pos_type end)
1533         : itemdepth(par.itemdepth),
1534         d(new Paragraph::Private(*par.d, this, beg, end))
1535 {
1536         registerWords();
1537 }
1538
1539
1540 Paragraph & Paragraph::operator=(Paragraph const & par)
1541 {
1542         // needed as we will destroy the private part before copying it
1543         if (&par != this) {
1544                 itemdepth = par.itemdepth;
1545
1546                 deregisterWords();
1547                 delete d;
1548                 d = new Private(*par.d, this);
1549                 registerWords();
1550         }
1551         return *this;
1552 }
1553
1554
1555 Paragraph::~Paragraph()
1556 {
1557         deregisterWords();
1558         delete d;
1559 }
1560
1561
1562 namespace {
1563
1564 // this shall be called just before every "os << ..." action.
1565 void flushString(ostream & os, docstring & s)
1566 {
1567         os << to_utf8(s);
1568         s.erase();
1569 }
1570
1571 }
1572
1573
1574 void Paragraph::write(ostream & os, BufferParams const & bparams,
1575         depth_type & dth) const
1576 {
1577         // The beginning or end of a deeper (i.e. nested) area?
1578         if (dth != d->params_.depth()) {
1579                 if (d->params_.depth() > dth) {
1580                         while (d->params_.depth() > dth) {
1581                                 os << "\n\\begin_deeper";
1582                                 ++dth;
1583                         }
1584                 } else {
1585                         while (d->params_.depth() < dth) {
1586                                 os << "\n\\end_deeper";
1587                                 --dth;
1588                         }
1589                 }
1590         }
1591
1592         // First write the layout
1593         os << "\n\\begin_layout " << to_utf8(d->layout_->name()) << '\n';
1594
1595         d->params_.write(os);
1596
1597         Font font1(inherit_font, bparams.language);
1598
1599         Change running_change = Change(Change::UNCHANGED);
1600
1601         // this string is used as a buffer to avoid repetitive calls
1602         // to to_utf8(), which turn out to be expensive (JMarc)
1603         docstring write_buffer;
1604
1605         int column = 0;
1606         for (pos_type i = 0; i <= size(); ++i) {
1607
1608                 Change const change = lookupChange(i);
1609                 if (change != running_change)
1610                         flushString(os, write_buffer);
1611                 Changes::lyxMarkChange(os, bparams, column, running_change, change);
1612                 running_change = change;
1613
1614                 if (i == size())
1615                         break;
1616
1617                 // Write font changes
1618                 Font font2 = getFontSettings(bparams, i);
1619                 if (font2 != font1) {
1620                         flushString(os, write_buffer);
1621                         font2.lyxWriteChanges(font1, os);
1622                         column = 0;
1623                         font1 = font2;
1624                 }
1625
1626                 char_type const c = d->text_[i];
1627                 switch (c) {
1628                 case META_INSET:
1629                         if (Inset const * inset = getInset(i)) {
1630                                 flushString(os, write_buffer);
1631                                 if (inset->directWrite()) {
1632                                         // international char, let it write
1633                                         // code directly so it's shorter in
1634                                         // the file
1635                                         inset->write(os);
1636                                 } else {
1637                                         if (i)
1638                                                 os << '\n';
1639                                         os << "\\begin_inset ";
1640                                         inset->write(os);
1641                                         os << "\n\\end_inset\n\n";
1642                                         column = 0;
1643                                 }
1644                         }
1645                         break;
1646                 case '\\':
1647                         flushString(os, write_buffer);
1648                         os << "\n\\backslash\n";
1649                         column = 0;
1650                         break;
1651                 case '.':
1652                         flushString(os, write_buffer);
1653                         if (i + 1 < size() && d->text_[i + 1] == ' ') {
1654                                 os << ".\n";
1655                                 column = 0;
1656                         } else
1657                                 os << '.';
1658                         break;
1659                 default:
1660                         if ((column > 70 && c == ' ')
1661                             || column > 79) {
1662                                 flushString(os, write_buffer);
1663                                 os << '\n';
1664                                 column = 0;
1665                         }
1666                         // this check is to amend a bug. LyX sometimes
1667                         // inserts '\0' this could cause problems.
1668                         if (c != '\0')
1669                                 write_buffer.push_back(c);
1670                         else
1671                                 LYXERR0("NUL char in structure.");
1672                         ++column;
1673                         break;
1674                 }
1675         }
1676
1677         flushString(os, write_buffer);
1678         os << "\n\\end_layout\n";
1679 }
1680
1681
1682 void Paragraph::validate(LaTeXFeatures & features) const
1683 {
1684         d->validate(features);
1685 }
1686
1687
1688 void Paragraph::insert(pos_type start, docstring const & str,
1689                        Font const & font, Change const & change)
1690 {
1691         for (size_t i = 0, n = str.size(); i != n ; ++i)
1692                 insertChar(start + i, str[i], font, change);
1693 }
1694
1695
1696 void Paragraph::appendChar(char_type c, Font const & font,
1697                 Change const & change)
1698 {
1699         // track change
1700         d->changes_.insert(change, d->text_.size());
1701         // when appending characters, no need to update tables
1702         d->text_.push_back(c);
1703         setFont(d->text_.size() - 1, font);
1704         d->requestSpellCheck(d->text_.size() - 1);
1705 }
1706
1707
1708 void Paragraph::appendString(docstring const & s, Font const & font,
1709                 Change const & change)
1710 {
1711         pos_type end = s.size();
1712         size_t oldsize = d->text_.size();
1713         size_t newsize = oldsize + end;
1714         size_t capacity = d->text_.capacity();
1715         if (newsize >= capacity)
1716                 d->text_.reserve(max(capacity + 100, newsize));
1717
1718         // when appending characters, no need to update tables
1719         d->text_.append(s);
1720
1721         // FIXME: Optimize this!
1722         for (size_t i = oldsize; i != newsize; ++i) {
1723                 // track change
1724                 d->changes_.insert(change, i);
1725                 d->requestSpellCheck(i);
1726         }
1727         d->fontlist_.set(oldsize, font);
1728         d->fontlist_.set(newsize - 1, font);
1729 }
1730
1731
1732 void Paragraph::insertChar(pos_type pos, char_type c,
1733                            bool trackChanges)
1734 {
1735         d->insertChar(pos, c, Change(trackChanges ?
1736                            Change::INSERTED : Change::UNCHANGED));
1737 }
1738
1739
1740 void Paragraph::insertChar(pos_type pos, char_type c,
1741                            Font const & font, bool trackChanges)
1742 {
1743         d->insertChar(pos, c, Change(trackChanges ?
1744                            Change::INSERTED : Change::UNCHANGED));
1745         setFont(pos, font);
1746 }
1747
1748
1749 void Paragraph::insertChar(pos_type pos, char_type c,
1750                            Font const & font, Change const & change)
1751 {
1752         d->insertChar(pos, c, change);
1753         setFont(pos, font);
1754 }
1755
1756
1757 bool Paragraph::insertInset(pos_type pos, Inset * inset,
1758                             Font const & font, Change const & change)
1759 {
1760         bool const success = insertInset(pos, inset, change);
1761         // Set the font/language of the inset...
1762         setFont(pos, font);
1763         return success;
1764 }
1765
1766
1767 void Paragraph::resetFonts(Font const & font)
1768 {
1769         d->fontlist_.clear();
1770         d->fontlist_.set(0, font);
1771         d->fontlist_.set(d->text_.size() - 1, font);
1772 }
1773
1774 // Gets uninstantiated font setting at position.
1775 Font const & Paragraph::getFontSettings(BufferParams const & bparams,
1776                                          pos_type pos) const
1777 {
1778         if (pos > size()) {
1779                 LYXERR0("pos: " << pos << " size: " << size());
1780                 LASSERT(pos <= size(), /**/);
1781         }
1782
1783         FontList::const_iterator cit = d->fontlist_.fontIterator(pos);
1784         if (cit != d->fontlist_.end())
1785                 return cit->font();
1786
1787         if (pos == size() && !empty())
1788                 return getFontSettings(bparams, pos - 1);
1789
1790         // Optimisation: avoid a full font instantiation if there is no
1791         // language change from previous call.
1792         static Font previous_font;
1793         static Language const * previous_lang = 0;
1794         Language const * lang = getParLanguage(bparams);
1795         if (lang != previous_lang) {
1796                 previous_lang = lang;
1797                 previous_font = Font(inherit_font, lang);
1798         }
1799         return previous_font;
1800 }
1801
1802
1803 FontSpan Paragraph::fontSpan(pos_type pos) const
1804 {
1805         LASSERT(pos <= size(), /**/);
1806         pos_type start = 0;
1807
1808         FontList::const_iterator cit = d->fontlist_.begin();
1809         FontList::const_iterator end = d->fontlist_.end();
1810         for (; cit != end; ++cit) {
1811                 if (cit->pos() >= pos) {
1812                         if (pos >= beginOfBody())
1813                                 return FontSpan(max(start, beginOfBody()),
1814                                                 cit->pos());
1815                         else
1816                                 return FontSpan(start,
1817                                                 min(beginOfBody() - 1,
1818                                                          cit->pos()));
1819                 }
1820                 start = cit->pos() + 1;
1821         }
1822
1823         // This should not happen, but if so, we take no chances.
1824         // LYXERR0("Paragraph::getEndPosOfFontSpan: This should not happen!");
1825         return FontSpan(pos, pos);
1826 }
1827
1828
1829 // Gets uninstantiated font setting at position 0
1830 Font const & Paragraph::getFirstFontSettings(BufferParams const & bparams) const
1831 {
1832         if (!empty() && !d->fontlist_.empty())
1833                 return d->fontlist_.begin()->font();
1834
1835         // Optimisation: avoid a full font instantiation if there is no
1836         // language change from previous call.
1837         static Font previous_font;
1838         static Language const * previous_lang = 0;
1839         if (bparams.language != previous_lang) {
1840                 previous_lang = bparams.language;
1841                 previous_font = Font(inherit_font, bparams.language);
1842         }
1843
1844         return previous_font;
1845 }
1846
1847
1848 // Gets the fully instantiated font at a given position in a paragraph
1849 // This is basically the same function as Text::GetFont() in text2.cpp.
1850 // The difference is that this one is used for generating the LaTeX file,
1851 // and thus cosmetic "improvements" are disallowed: This has to deliver
1852 // the true picture of the buffer. (Asger)
1853 Font const Paragraph::getFont(BufferParams const & bparams, pos_type pos,
1854                                  Font const & outerfont) const
1855 {
1856         LASSERT(pos >= 0, /**/);
1857
1858         Font font = getFontSettings(bparams, pos);
1859
1860         pos_type const body_pos = beginOfBody();
1861         FontInfo & fi = font.fontInfo();
1862         if (pos < body_pos)
1863                 fi.realize(d->layout_->labelfont);
1864         else
1865                 fi.realize(d->layout_->font);
1866
1867         fi.realize(outerfont.fontInfo());
1868         fi.realize(bparams.getFont().fontInfo());
1869
1870         return font;
1871 }
1872
1873
1874 Font const Paragraph::getLabelFont
1875         (BufferParams const & bparams, Font const & outerfont) const
1876 {
1877         FontInfo tmpfont = d->layout_->labelfont;
1878         tmpfont.realize(outerfont.fontInfo());
1879         tmpfont.realize(bparams.getFont().fontInfo());
1880         return Font(tmpfont, getParLanguage(bparams));
1881 }
1882
1883
1884 Font const Paragraph::getLayoutFont
1885         (BufferParams const & bparams, Font const & outerfont) const
1886 {
1887         FontInfo tmpfont = d->layout_->font;
1888         tmpfont.realize(outerfont.fontInfo());
1889         tmpfont.realize(bparams.getFont().fontInfo());
1890         return Font(tmpfont, getParLanguage(bparams));
1891 }
1892
1893
1894 /// Returns the height of the highest font in range
1895 FontSize Paragraph::highestFontInRange
1896         (pos_type startpos, pos_type endpos, FontSize def_size) const
1897 {
1898         return d->fontlist_.highestInRange(startpos, endpos, def_size);
1899 }
1900
1901
1902 char_type Paragraph::getUChar(BufferParams const & bparams, pos_type pos) const
1903 {
1904         char_type c = d->text_[pos];
1905         if (!lyxrc.rtl_support || !getFontSettings(bparams, pos).isRightToLeft())
1906                 return c;
1907
1908         // FIXME: The arabic special casing is due to the difference of arabic
1909         // round brackets input introduced in r18599. Check if this should be
1910         // unified with Hebrew or at least if all bracket types should be
1911         // handled the same (file format change in either case).
1912         string const & lang = getFontSettings(bparams, pos).language()->lang();
1913         bool const arabic = lang == "arabic_arabtex" || lang == "arabic_arabi"
1914                 || lang == "farsi";
1915         char_type uc = c;
1916         switch (c) {
1917         case '(':
1918                 uc = arabic ? c : ')';
1919                 break;
1920         case ')':
1921                 uc = arabic ? c : '(';
1922                 break;
1923         case '[':
1924                 uc = ']';
1925                 break;
1926         case ']':
1927                 uc = '[';
1928                 break;
1929         case '{':
1930                 uc = '}';
1931                 break;
1932         case '}':
1933                 uc = '{';
1934                 break;
1935         case '<':
1936                 uc = '>';
1937                 break;
1938         case '>':
1939                 uc = '<';
1940                 break;
1941         }
1942
1943         return uc;
1944 }
1945
1946
1947 void Paragraph::setFont(pos_type pos, Font const & font)
1948 {
1949         LASSERT(pos <= size(), /**/);
1950
1951         // First, reduce font against layout/label font
1952         // Update: The setCharFont() routine in text2.cpp already
1953         // reduces font, so we don't need to do that here. (Asger)
1954
1955         d->fontlist_.set(pos, font);
1956 }
1957
1958
1959 void Paragraph::makeSameLayout(Paragraph const & par)
1960 {
1961         d->layout_ = par.d->layout_;
1962         d->params_ = par.d->params_;
1963 }
1964
1965
1966 bool Paragraph::stripLeadingSpaces(bool trackChanges)
1967 {
1968         if (isFreeSpacing())
1969                 return false;
1970
1971         int pos = 0;
1972         int count = 0;
1973
1974         while (pos < size() && (isNewline(pos) || isLineSeparator(pos))) {
1975                 if (eraseChar(pos, trackChanges))
1976                         ++count;
1977                 else
1978                         ++pos;
1979         }
1980
1981         return count > 0 || pos > 0;
1982 }
1983
1984
1985 bool Paragraph::hasSameLayout(Paragraph const & par) const
1986 {
1987         return par.d->layout_ == d->layout_
1988                 && d->params_.sameLayout(par.d->params_);
1989 }
1990
1991
1992 depth_type Paragraph::getDepth() const
1993 {
1994         return d->params_.depth();
1995 }
1996
1997
1998 depth_type Paragraph::getMaxDepthAfter() const
1999 {
2000         if (d->layout_->isEnvironment())
2001                 return d->params_.depth() + 1;
2002         else
2003                 return d->params_.depth();
2004 }
2005
2006
2007 char Paragraph::getAlign() const
2008 {
2009         if (d->params_.align() == LYX_ALIGN_LAYOUT)
2010                 return d->layout_->align;
2011         else
2012                 return d->params_.align();
2013 }
2014
2015
2016 docstring const & Paragraph::labelString() const
2017 {
2018         return d->params_.labelString();
2019 }
2020
2021
2022 // the next two functions are for the manual labels
2023 docstring const Paragraph::getLabelWidthString() const
2024 {
2025         if (d->layout_->margintype == MARGIN_MANUAL
2026             || d->layout_->latextype == LATEX_BIB_ENVIRONMENT)
2027                 return d->params_.labelWidthString();
2028         else
2029                 return _("Senseless with this layout!");
2030 }
2031
2032
2033 void Paragraph::setLabelWidthString(docstring const & s)
2034 {
2035         d->params_.labelWidthString(s);
2036 }
2037
2038
2039 docstring Paragraph::expandLabel(Layout const & layout,
2040                 BufferParams const & bparams) const
2041 {
2042         return expandParagraphLabel(layout, bparams, true);
2043 }
2044
2045
2046 docstring Paragraph::expandDocBookLabel(Layout const & layout,
2047                 BufferParams const & bparams) const
2048 {
2049         return expandParagraphLabel(layout, bparams, false);
2050 }
2051
2052
2053 docstring Paragraph::expandParagraphLabel(Layout const & layout,
2054                 BufferParams const & bparams, bool process_appendix) const
2055 {
2056         DocumentClass const & tclass = bparams.documentClass();
2057         string const & lang = getParLanguage(bparams)->code();
2058         bool const in_appendix = process_appendix && d->params_.appendix();
2059         docstring fmt = translateIfPossible(layout.labelstring(in_appendix), lang);
2060
2061         if (fmt.empty() && layout.labeltype == LABEL_COUNTER
2062             && !layout.counter.empty())
2063                 return tclass.counters().theCounter(layout.counter, lang);
2064
2065         // handle 'inherited level parts' in 'fmt',
2066         // i.e. the stuff between '@' in   '@Section@.\arabic{subsection}'
2067         size_t const i = fmt.find('@', 0);
2068         if (i != docstring::npos) {
2069                 size_t const j = fmt.find('@', i + 1);
2070                 if (j != docstring::npos) {
2071                         docstring parent(fmt, i + 1, j - i - 1);
2072                         docstring label = from_ascii("??");
2073                         if (tclass.hasLayout(parent))
2074                                 docstring label = expandParagraphLabel(tclass[parent], bparams,
2075                                                       process_appendix);
2076                         fmt = docstring(fmt, 0, i) + label
2077                                 + docstring(fmt, j + 1, docstring::npos);
2078                 }
2079         }
2080
2081         return tclass.counters().counterLabel(fmt, lang);
2082 }
2083
2084
2085 void Paragraph::applyLayout(Layout const & new_layout)
2086 {
2087         d->layout_ = &new_layout;
2088         LyXAlignment const oldAlign = d->params_.align();
2089
2090         if (!(oldAlign & d->layout_->alignpossible)) {
2091                 frontend::Alert::warning(_("Alignment not permitted"),
2092                         _("The new layout does not permit the alignment previously used.\nSetting to default."));
2093                 d->params_.align(LYX_ALIGN_LAYOUT);
2094         }
2095 }
2096
2097
2098 pos_type Paragraph::beginOfBody() const
2099 {
2100         return d->begin_of_body_;
2101 }
2102
2103
2104 void Paragraph::setBeginOfBody()
2105 {
2106         if (d->layout_->labeltype != LABEL_MANUAL) {
2107                 d->begin_of_body_ = 0;
2108                 return;
2109         }
2110
2111         // Unroll the first two cycles of the loop
2112         // and remember the previous character to
2113         // remove unnecessary getChar() calls
2114         pos_type i = 0;
2115         pos_type end = size();
2116         if (i < end && !isNewline(i)) {
2117                 ++i;
2118                 char_type previous_char = 0;
2119                 char_type temp = 0;
2120                 if (i < end) {
2121                         previous_char = d->text_[i];
2122                         if (!isNewline(i)) {
2123                                 ++i;
2124                                 while (i < end && previous_char != ' ') {
2125                                         temp = d->text_[i];
2126                                         if (isNewline(i))
2127                                                 break;
2128                                         ++i;
2129                                         previous_char = temp;
2130                                 }
2131                         }
2132                 }
2133         }
2134
2135         d->begin_of_body_ = i;
2136 }
2137
2138
2139 bool Paragraph::allowParagraphCustomization() const
2140 {
2141         return inInset().allowParagraphCustomization();
2142 }
2143
2144
2145 bool Paragraph::usePlainLayout() const
2146 {
2147         return inInset().usePlainLayout();
2148 }
2149
2150
2151 bool Paragraph::isPassThru() const
2152 {
2153         return inInset().getLayout().isPassThru() || d->layout_->pass_thru;
2154 }
2155
2156 namespace {
2157
2158 // paragraphs inside floats need different alignment tags to avoid
2159 // unwanted space
2160
2161 bool noTrivlistCentering(InsetCode code)
2162 {
2163         return code == FLOAT_CODE
2164                || code == WRAP_CODE
2165                || code == CELL_CODE;
2166 }
2167
2168
2169 string correction(string const & orig)
2170 {
2171         if (orig == "flushleft")
2172                 return "raggedright";
2173         if (orig == "flushright")
2174                 return "raggedleft";
2175         if (orig == "center")
2176                 return "centering";
2177         return orig;
2178 }
2179
2180
2181 string const corrected_env(string const & suffix, string const & env,
2182         InsetCode code, bool const lastpar)
2183 {
2184         string output = suffix + "{";
2185         if (noTrivlistCentering(code)) {
2186                 if (lastpar) {
2187                         // the last paragraph in non-trivlist-aligned
2188                         // context is special (to avoid unwanted whitespace)
2189                         if (suffix == "\\begin")
2190                                 return "\\" + correction(env) + "{}";
2191                         return string();
2192                 }
2193                 output += correction(env);
2194         } else
2195                 output += env;
2196         output += "}";
2197         if (suffix == "\\begin")
2198                 output += "\n";
2199         return output;
2200 }
2201
2202
2203 void adjust_column(string const & str, int & column)
2204 {
2205         if (!contains(str, "\n"))
2206                 column += str.size();
2207         else {
2208                 string tmp;
2209                 column = rsplit(str, tmp, '\n').size();
2210         }
2211 }
2212
2213 } // namespace anon
2214
2215
2216 int Paragraph::Private::startTeXParParams(BufferParams const & bparams,
2217                         otexstream & os, OutputParams const & runparams) const
2218 {
2219         int column = 0;
2220
2221         if (params_.noindent() && !layout_->pass_thru) {
2222                 os << "\\noindent ";
2223                 column += 10;
2224         }
2225
2226         LyXAlignment const curAlign = params_.align();
2227
2228         if (curAlign == layout_->align)
2229                 return column;
2230
2231         switch (curAlign) {
2232         case LYX_ALIGN_NONE:
2233         case LYX_ALIGN_BLOCK:
2234         case LYX_ALIGN_LAYOUT:
2235         case LYX_ALIGN_SPECIAL:
2236         case LYX_ALIGN_DECIMAL:
2237                 break;
2238         case LYX_ALIGN_LEFT:
2239         case LYX_ALIGN_RIGHT:
2240         case LYX_ALIGN_CENTER:
2241                 if (runparams.moving_arg) {
2242                         os << "\\protect";
2243                         column += 8;
2244                 }
2245                 break;
2246         }
2247
2248         string const begin_tag = "\\begin";
2249         InsetCode code = ownerCode();
2250         bool const lastpar = runparams.isLastPar;
2251
2252         switch (curAlign) {
2253         case LYX_ALIGN_NONE:
2254         case LYX_ALIGN_BLOCK:
2255         case LYX_ALIGN_LAYOUT:
2256         case LYX_ALIGN_SPECIAL:
2257         case LYX_ALIGN_DECIMAL:
2258                 break;
2259         case LYX_ALIGN_LEFT: {
2260                 string output;
2261                 if (owner_->getParLanguage(bparams)->babel() != "hebrew")
2262                         output = corrected_env(begin_tag, "flushleft", code, lastpar);
2263                 else
2264                         output = corrected_env(begin_tag, "flushright", code, lastpar);
2265                 os << from_ascii(output);
2266                 adjust_column(output, column);
2267                 break;
2268         } case LYX_ALIGN_RIGHT: {
2269                 string output;
2270                 if (owner_->getParLanguage(bparams)->babel() != "hebrew")
2271                         output = corrected_env(begin_tag, "flushright", code, lastpar);
2272                 else
2273                         output = corrected_env(begin_tag, "flushleft", code, lastpar);
2274                 os << from_ascii(output);
2275                 adjust_column(output, column);
2276                 break;
2277         } case LYX_ALIGN_CENTER: {
2278                 string output;
2279                 output = corrected_env(begin_tag, "center", code, lastpar);
2280                 os << from_ascii(output);
2281                 adjust_column(output, column);
2282                 break;
2283         }
2284         }
2285
2286         return column;
2287 }
2288
2289
2290 bool Paragraph::Private::endTeXParParams(BufferParams const & bparams,
2291                         otexstream & os, OutputParams const & runparams) const
2292 {
2293         LyXAlignment const curAlign = params_.align();
2294
2295         if (curAlign == layout_->align)
2296                 return false;
2297
2298         switch (curAlign) {
2299         case LYX_ALIGN_NONE:
2300         case LYX_ALIGN_BLOCK:
2301         case LYX_ALIGN_LAYOUT:
2302         case LYX_ALIGN_SPECIAL:
2303         case LYX_ALIGN_DECIMAL:
2304                 break;
2305         case LYX_ALIGN_LEFT:
2306         case LYX_ALIGN_RIGHT:
2307         case LYX_ALIGN_CENTER:
2308                 if (runparams.moving_arg)
2309                         os << "\\protect";
2310                 break;
2311         }
2312
2313         string output;
2314         string const end_tag = "\n\\par\\end";
2315         InsetCode code = ownerCode();
2316         bool const lastpar = runparams.isLastPar;
2317
2318         switch (curAlign) {
2319         case LYX_ALIGN_NONE:
2320         case LYX_ALIGN_BLOCK:
2321         case LYX_ALIGN_LAYOUT:
2322         case LYX_ALIGN_SPECIAL:
2323         case LYX_ALIGN_DECIMAL:
2324                 break;
2325         case LYX_ALIGN_LEFT: {
2326                 if (owner_->getParLanguage(bparams)->babel() != "hebrew")
2327                         output = corrected_env(end_tag, "flushleft", code, lastpar);
2328                 else
2329                         output = corrected_env(end_tag, "flushright", code, lastpar);
2330                 os << from_ascii(output);
2331                 break;
2332         } case LYX_ALIGN_RIGHT: {
2333                 if (owner_->getParLanguage(bparams)->babel() != "hebrew")
2334                         output = corrected_env(end_tag, "flushright", code, lastpar);
2335                 else
2336                         output = corrected_env(end_tag, "flushleft", code, lastpar);
2337                 os << from_ascii(output);
2338                 break;
2339         } case LYX_ALIGN_CENTER: {
2340                 output = corrected_env(end_tag, "center", code, lastpar);
2341                 os << from_ascii(output);
2342                 break;
2343         }
2344         }
2345
2346         return !output.empty() || lastpar;
2347 }
2348
2349
2350 // This one spits out the text of the paragraph
2351 void Paragraph::latex(BufferParams const & bparams,
2352         Font const & outerfont,
2353         otexstream & os,
2354         OutputParams const & runparams,
2355         int start_pos, int end_pos, bool force) const
2356 {
2357         LYXERR(Debug::LATEX, "Paragraph::latex...     " << this);
2358
2359         // FIXME This check should not be needed. Perhaps issue an
2360         // error if it triggers.
2361         Layout const & style = inInset().forcePlainLayout() ?
2362                 bparams.documentClass().plainLayout() : *d->layout_;
2363
2364         if (!force && style.inpreamble)
2365                 return;
2366
2367         bool const allowcust = allowParagraphCustomization();
2368
2369         // Current base font for all inherited font changes, without any
2370         // change caused by an individual character, except for the language:
2371         // It is set to the language of the first character.
2372         // As long as we are in the label, this font is the base font of the
2373         // label. Before the first body character it is set to the base font
2374         // of the body.
2375         Font basefont;
2376
2377         // Maybe we have to create a optional argument.
2378         pos_type body_pos = beginOfBody();
2379         unsigned int column = 0;
2380
2381         if (body_pos > 0) {
2382                 // the optional argument is kept in curly brackets in
2383                 // case it contains a ']'
2384                 // This is not strictly needed, but if this is changed it
2385                 // would be a file format change, and tex2lyx would need
2386                 // to be adjusted, since it unconditionally removes the
2387                 // braces when it parses \item.
2388                 os << "[{";
2389                 column += 2;
2390                 basefont = getLabelFont(bparams, outerfont);
2391         } else {
2392                 basefont = getLayoutFont(bparams, outerfont);
2393         }
2394
2395         // Which font is currently active?
2396         Font running_font(basefont);
2397         // Do we have an open font change?
2398         bool open_font = false;
2399
2400         Change runningChange = Change(Change::UNCHANGED);
2401
2402         Encoding const * const prev_encoding = runparams.encoding;
2403
2404         os.texrow().start(id(), 0);
2405
2406         // if the paragraph is empty, the loop will not be entered at all
2407         if (empty()) {
2408                 if (style.isCommand()) {
2409                         os << '{';
2410                         ++column;
2411                 }
2412                 if (allowcust)
2413                         column += d->startTeXParParams(bparams, os, runparams);
2414         }
2415
2416         for (pos_type i = 0; i < size(); ++i) {
2417                 // First char in paragraph or after label?
2418                 if (i == body_pos) {
2419                         if (body_pos > 0) {
2420                                 if (open_font) {
2421                                         column += running_font.latexWriteEndChanges(
2422                                                 os, bparams, runparams,
2423                                                 basefont, basefont);
2424                                         open_font = false;
2425                                 }
2426                                 basefont = getLayoutFont(bparams, outerfont);
2427                                 running_font = basefont;
2428
2429                                 column += Changes::latexMarkChange(os, bparams,
2430                                                 runningChange, Change(Change::UNCHANGED),
2431                                                 runparams);
2432                                 runningChange = Change(Change::UNCHANGED);
2433
2434                                 os << "}] ";
2435                                 column +=3;
2436                         }
2437                         if (style.isCommand()) {
2438                                 os << '{';
2439                                 ++column;
2440                         }
2441
2442                         if (allowcust)
2443                                 column += d->startTeXParParams(bparams, os,
2444                                                             runparams);
2445                 }
2446
2447                 Change const & change = runparams.inDeletedInset
2448                         ? runparams.changeOfDeletedInset : lookupChange(i);
2449
2450                 if (bparams.outputChanges && runningChange != change) {
2451                         if (open_font) {
2452                                 column += running_font.latexWriteEndChanges(
2453                                                 os, bparams, runparams, basefont, basefont);
2454                                 open_font = false;
2455                         }
2456                         basefont = getLayoutFont(bparams, outerfont);
2457                         running_font = basefont;
2458
2459                         column += Changes::latexMarkChange(os, bparams, runningChange,
2460                                                            change, runparams);
2461                         runningChange = change;
2462                 }
2463
2464                 // do not output text which is marked deleted
2465                 // if change tracking output is disabled
2466                 if (!bparams.outputChanges && change.deleted()) {
2467                         continue;
2468                 }
2469
2470                 ++column;
2471
2472                 // Fully instantiated font
2473                 Font const font = getFont(bparams, i, outerfont);
2474
2475                 Font const last_font = running_font;
2476
2477                 // Do we need to close the previous font?
2478                 if (open_font &&
2479                     (font != running_font ||
2480                      font.language() != running_font.language()))
2481                 {
2482                         column += running_font.latexWriteEndChanges(
2483                                         os, bparams, runparams, basefont,
2484                                         (i == body_pos-1) ? basefont : font);
2485                         running_font = basefont;
2486                         open_font = false;
2487                 }
2488
2489                 string const running_lang = runparams.use_polyglossia ?
2490                         running_font.language()->polyglossia() : running_font.language()->babel();
2491                 // close babel's font environment before opening CJK.
2492                 string const lang_end_command = runparams.use_polyglossia ?
2493                         "\\end{$$lang}" : lyxrc.language_command_end;
2494                 if (!running_lang.empty() &&
2495                     font.language()->encoding()->package() == Encoding::CJK) {
2496                                 string end_tag = subst(lang_end_command,
2497                                                         "$$lang",
2498                                                         running_lang);
2499                                 os << from_ascii(end_tag);
2500                                 column += end_tag.length();
2501                 }
2502
2503                 // Switch file encoding if necessary (and allowed)
2504                 if (!runparams.pass_thru && !style.pass_thru &&
2505                     runparams.encoding->package() != Encoding::none &&
2506                     font.language()->encoding()->package() != Encoding::none) {
2507                         pair<bool, int> const enc_switch =
2508                                 switchEncoding(os.os(), bparams, runparams,
2509                                         *(font.language()->encoding()));
2510                         if (enc_switch.first) {
2511                                 column += enc_switch.second;
2512                                 runparams.encoding = font.language()->encoding();
2513                         }
2514                 }
2515
2516                 char_type const c = d->text_[i];
2517
2518                 // Do we need to change font?
2519                 if ((font != running_font ||
2520                      font.language() != running_font.language()) &&
2521                         i != body_pos - 1)
2522                 {
2523                         odocstringstream ods;
2524                         column += font.latexWriteStartChanges(ods, bparams,
2525                                                               runparams, basefont,
2526                                                               last_font);
2527                         running_font = font;
2528                         open_font = true;
2529                         docstring fontchange = ods.str();
2530                         // check whether the fontchange ends with a \\textcolor
2531                         // modifier and the text starts with a space (bug 4473)
2532                         docstring const last_modifier = rsplit(fontchange, '\\');
2533                         if (prefixIs(last_modifier, from_ascii("textcolor")) && c == ' ')
2534                                 os << fontchange << from_ascii("{}");
2535                         // check if the fontchange ends with a trailing blank
2536                         // (like "\small " (see bug 3382)
2537                         else if (suffixIs(fontchange, ' ') && c == ' ')
2538                                 os << fontchange.substr(0, fontchange.size() - 1)
2539                                    << from_ascii("{}");
2540                         else
2541                                 os << fontchange;
2542                 }
2543
2544                 // FIXME: think about end_pos implementation...
2545                 if (c == ' ' && i >= start_pos && (end_pos == -1 || i < end_pos)) {
2546                         // FIXME: integrate this case in latexSpecialChar
2547                         // Do not print the separation of the optional argument
2548                         // if style.pass_thru is false. This works because
2549                         // latexSpecialChar ignores spaces if
2550                         // style.pass_thru is false.
2551                         if (i != body_pos - 1) {
2552                                 if (d->simpleTeXBlanks(runparams, os,
2553                                                 i, column, font, style)) {
2554                                         // A surrogate pair was output. We
2555                                         // must not call latexSpecialChar
2556                                         // in this iteration, since it would output
2557                                         // the combining character again.
2558                                         ++i;
2559                                         continue;
2560                                 }
2561                         }
2562                 }
2563
2564                 OutputParams rp = runparams;
2565                 rp.free_spacing = style.free_spacing;
2566                 rp.local_font = &font;
2567                 rp.intitle = style.intitle;
2568
2569                 // Two major modes:  LaTeX or plain
2570                 // Handle here those cases common to both modes
2571                 // and then split to handle the two modes separately.
2572                 if (c == META_INSET) {
2573                         if (i >= start_pos && (end_pos == -1 || i < end_pos)) {
2574                                 d->latexInset(bparams, os, rp, running_font,
2575                                                 basefont, outerfont, open_font,
2576                                                 runningChange, style, i, column);
2577                         }
2578                 } else {
2579                         if (i >= start_pos && (end_pos == -1 || i < end_pos)) {
2580                                 try {
2581                                         d->latexSpecialChar(os, bparams, rp, running_font, runningChange,
2582                                                             style, i, end_pos, column);
2583                                 } catch (EncodingException & e) {
2584                                 if (runparams.dryrun) {
2585                                         os << "<" << _("LyX Warning: ")
2586                                            << _("uncodable character") << " '";
2587                                         os.put(c);
2588                                         os << "'>";
2589                                 } else {
2590                                         // add location information and throw again.
2591                                         e.par_id = id();
2592                                         e.pos = i;
2593                                         throw(e);
2594                                 }
2595                         }
2596                 }
2597                 }
2598
2599                 // Set the encoding to that returned from latexSpecialChar (see
2600                 // comment for encoding member in OutputParams.h)
2601                 runparams.encoding = rp.encoding;
2602         }
2603
2604         // If we have an open font definition, we have to close it
2605         if (open_font) {
2606 #ifdef FIXED_LANGUAGE_END_DETECTION
2607                 if (next_) {
2608                         running_font.latexWriteEndChanges(os, bparams,
2609                                         runparams, basefont,
2610                                         next_->getFont(bparams, 0, outerfont));
2611                 } else {
2612                         running_font.latexWriteEndChanges(os, bparams,
2613                                         runparams, basefont, basefont);
2614                 }
2615 #else
2616 //FIXME: For now we ALWAYS have to close the foreign font settings if they are
2617 //FIXME: there as we start another \selectlanguage with the next paragraph if
2618 //FIXME: we are in need of this. This should be fixed sometime (Jug)
2619                 running_font.latexWriteEndChanges(os, bparams, runparams,
2620                                 basefont, basefont);
2621 #endif
2622         }
2623
2624         column += Changes::latexMarkChange(os, bparams, runningChange,
2625                                            Change(Change::UNCHANGED), runparams);
2626
2627         // Needed if there is an optional argument but no contents.
2628         if (body_pos > 0 && body_pos == size()) {
2629                 os << "}]~";
2630         }
2631
2632         if (allowcust && d->endTeXParParams(bparams, os, runparams)
2633             && runparams.encoding != prev_encoding) {
2634                 runparams.encoding = prev_encoding;
2635                 if (!runparams.isFullUnicode())
2636                         os << setEncoding(prev_encoding->iconvName());
2637         }
2638
2639         LYXERR(Debug::LATEX, "Paragraph::latex... done " << this);
2640 }
2641
2642
2643 bool Paragraph::emptyTag() const
2644 {
2645         for (pos_type i = 0; i < size(); ++i) {
2646                 if (Inset const * inset = getInset(i)) {
2647                         InsetCode lyx_code = inset->lyxCode();
2648                         // FIXME testing like that is wrong. What is
2649                         // the intent?
2650                         if (lyx_code != TOC_CODE &&
2651                             lyx_code != INCLUDE_CODE &&
2652                             lyx_code != GRAPHICS_CODE &&
2653                             lyx_code != ERT_CODE &&
2654                             lyx_code != LISTINGS_CODE &&
2655                             lyx_code != FLOAT_CODE &&
2656                             lyx_code != TABULAR_CODE) {
2657                                 return false;
2658                         }
2659                 } else {
2660                         char_type c = d->text_[i];
2661                         if (c != ' ' && c != '\t')
2662                                 return false;
2663                 }
2664         }
2665         return true;
2666 }
2667
2668
2669 string Paragraph::getID(Buffer const & buf, OutputParams const & runparams)
2670         const
2671 {
2672         for (pos_type i = 0; i < size(); ++i) {
2673                 if (Inset const * inset = getInset(i)) {
2674                         InsetCode lyx_code = inset->lyxCode();
2675                         if (lyx_code == LABEL_CODE) {
2676                                 InsetLabel const * const il = static_cast<InsetLabel const *>(inset);
2677                                 docstring const & id = il->getParam("name");
2678                                 return "id='" + to_utf8(sgml::cleanID(buf, runparams, id)) + "'";
2679                         }
2680                 }
2681         }
2682         return string();
2683 }
2684
2685
2686 pos_type Paragraph::firstWordDocBook(odocstream & os, OutputParams const & runparams)
2687         const
2688 {
2689         pos_type i;
2690         for (i = 0; i < size(); ++i) {
2691                 if (Inset const * inset = getInset(i)) {
2692                         inset->docbook(os, runparams);
2693                 } else {
2694                         char_type c = d->text_[i];
2695                         if (c == ' ')
2696                                 break;
2697                         os << sgml::escapeChar(c);
2698                 }
2699         }
2700         return i;
2701 }
2702
2703
2704 pos_type Paragraph::firstWordLyXHTML(XHTMLStream & xs, OutputParams const & runparams)
2705         const
2706 {
2707         pos_type i;
2708         for (i = 0; i < size(); ++i) {
2709                 if (Inset const * inset = getInset(i)) {
2710                         inset->xhtml(xs, runparams);
2711                 } else {
2712                         char_type c = d->text_[i];
2713                         if (c == ' ')
2714                                 break;
2715                         xs << c;
2716                 }
2717         }
2718         return i;
2719 }
2720
2721
2722 bool Paragraph::Private::onlyText(Buffer const & buf, Font const & outerfont, pos_type initial) const
2723 {
2724         Font font_old;
2725         pos_type size = text_.size();
2726         for (pos_type i = initial; i < size; ++i) {
2727                 Font font = owner_->getFont(buf.params(), i, outerfont);
2728                 if (text_[i] == META_INSET)
2729                         return false;
2730                 if (i != initial && font != font_old)
2731                         return false;
2732                 font_old = font;
2733         }
2734
2735         return true;
2736 }
2737
2738
2739 void Paragraph::simpleDocBookOnePar(Buffer const & buf,
2740                                     odocstream & os,
2741                                     OutputParams const & runparams,
2742                                     Font const & outerfont,
2743                                     pos_type initial) const
2744 {
2745         bool emph_flag = false;
2746
2747         Layout const & style = *d->layout_;
2748         FontInfo font_old =
2749                 style.labeltype == LABEL_MANUAL ? style.labelfont : style.font;
2750
2751         if (style.pass_thru && !d->onlyText(buf, outerfont, initial))
2752                 os << "]]>";
2753
2754         // parsing main loop
2755         for (pos_type i = initial; i < size(); ++i) {
2756                 Font font = getFont(buf.params(), i, outerfont);
2757
2758                 // handle <emphasis> tag
2759                 if (font_old.emph() != font.fontInfo().emph()) {
2760                         if (font.fontInfo().emph() == FONT_ON) {
2761                                 os << "<emphasis>";
2762                                 emph_flag = true;
2763                         } else if (i != initial) {
2764                                 os << "</emphasis>";
2765                                 emph_flag = false;
2766                         }
2767                 }
2768
2769                 if (Inset const * inset = getInset(i)) {
2770                         inset->docbook(os, runparams);
2771                 } else {
2772                         char_type c = d->text_[i];
2773
2774                         if (style.pass_thru)
2775                                 os.put(c);
2776                         else
2777                                 os << sgml::escapeChar(c);
2778                 }
2779                 font_old = font.fontInfo();
2780         }
2781
2782         if (emph_flag) {
2783                 os << "</emphasis>";
2784         }
2785
2786         if (style.free_spacing)
2787                 os << '\n';
2788         if (style.pass_thru && !d->onlyText(buf, outerfont, initial))
2789                 os << "<![CDATA[";
2790 }
2791
2792
2793 docstring Paragraph::simpleLyXHTMLOnePar(Buffer const & buf,
2794                                     XHTMLStream & xs,
2795                                     OutputParams const & runparams,
2796                                     Font const & outerfont,
2797                                     pos_type initial) const
2798 {
2799         docstring retval;
2800
2801         bool emph_flag = false;
2802         bool bold_flag = false;
2803
2804         Layout const & style = *d->layout_;
2805
2806         xs.startParagraph(allowEmpty());
2807
2808         if (!runparams.for_toc && runparams.html_make_pars) {
2809                 // generate a magic label for this paragraph
2810                 string const attr = "id='" + magicLabel() + "'";
2811                 xs << html::CompTag("a", attr);
2812         }
2813
2814         FontInfo font_old =
2815                 style.labeltype == LABEL_MANUAL ? style.labelfont : style.font;
2816
2817         // parsing main loop
2818         for (pos_type i = initial; i < size(); ++i) {
2819                 // let's not show deleted material in the output
2820                 if (isDeleted(i))
2821                         continue;
2822
2823                 Font font = getFont(buf.params(), i, outerfont);
2824
2825                 // emphasis
2826                 if (font_old.emph() != font.fontInfo().emph()) {
2827                         if (font.fontInfo().emph() == FONT_ON) {
2828                                 xs << html::StartTag("em");
2829                                 emph_flag = true;
2830                         } else if (emph_flag && i != initial) {
2831                                 xs << html::EndTag("em");
2832                                 emph_flag = false;
2833                         }
2834                 }
2835                 // bold
2836                 if (font_old.series() != font.fontInfo().series()) {
2837                         if (font.fontInfo().series() == BOLD_SERIES) {
2838                                 xs << html::StartTag("strong");
2839                                 bold_flag = true;
2840                         } else if (bold_flag && i != initial) {
2841                                 xs << html::EndTag("strong");
2842                                 bold_flag = false;
2843                         }
2844                 }
2845                 // FIXME XHTML
2846                 // Other such tags? What about the other text ranges?
2847
2848                 Inset const * inset = getInset(i);
2849                 if (inset) {
2850                         if (!runparams.for_toc || inset->isInToc()) {
2851                                 OutputParams np = runparams;
2852                                 if (!inset->getLayout().htmlisblock())
2853                                         np.html_in_par = true;
2854                                 retval += inset->xhtml(xs, np);
2855                         }
2856                 } else {
2857                         char_type c = d->text_[getUChar(buf.params(), i)];
2858
2859                         if (style.pass_thru)
2860                                 xs << c;
2861                         else if (c == '-') {
2862                                 docstring str;
2863                                 int j = i + 1;
2864                                 if (j < size() && d->text_[j] == '-') {
2865                                         j += 1;
2866                                         if (j < size() && d->text_[j] == '-') {
2867                                                 str += from_ascii("&mdash;");
2868                                                 i += 2;
2869                                         } else {
2870                                                 str += from_ascii("&ndash;");
2871                                                 i += 1;
2872                                         }
2873                                 }
2874                                 else
2875                                         str += c;
2876                                 // We don't want to escape the entities. Note that
2877                                 // it is safe to do this, since str can otherwise
2878                                 // only be "-". E.g., it can't be "<".
2879                                 xs << XHTMLStream::ESCAPE_NONE << str;
2880                         } else
2881                                 xs << c;
2882                 }
2883                 font_old = font.fontInfo();
2884         }
2885
2886         xs.closeFontTags();
2887         xs.endParagraph();
2888         return retval;
2889 }
2890
2891
2892 bool Paragraph::isHfill(pos_type pos) const
2893 {
2894         Inset const * inset = getInset(pos);
2895         return inset && (inset->lyxCode() == SPACE_CODE &&
2896                          inset->isStretchableSpace());
2897 }
2898
2899
2900 bool Paragraph::isNewline(pos_type pos) const
2901 {
2902         Inset const * inset = getInset(pos);
2903         return inset && inset->lyxCode() == NEWLINE_CODE;
2904 }
2905
2906
2907 bool Paragraph::isLineSeparator(pos_type pos) const
2908 {
2909         char_type const c = d->text_[pos];
2910         if (isLineSeparatorChar(c))
2911                 return true;
2912         Inset const * inset = getInset(pos);
2913         return inset && inset->isLineSeparator();
2914 }
2915
2916
2917 bool Paragraph::isWordSeparator(pos_type pos) const
2918 {
2919         if (pos == size())
2920                 return true;
2921         if (Inset const * inset = getInset(pos))
2922                 return !inset->isLetter();
2923         // if we have a hard hyphen (no en- or emdash) or apostrophe
2924         // we pass this to the spell checker
2925         // FIXME: this method is subject to change, visit
2926         // https://bugzilla.mozilla.org/show_bug.cgi?id=355178
2927         // to get an impression how complex this is.
2928         if (isHardHyphenOrApostrophe(pos))
2929                 return false;
2930         char_type const c = d->text_[pos];
2931         // We want to pass the escape chars to the spellchecker
2932         docstring const escape_chars = from_utf8(lyxrc.spellchecker_esc_chars);
2933         return !isLetterChar(c) && !isDigitASCII(c) && !contains(escape_chars, c);
2934 }
2935
2936
2937 bool Paragraph::isHardHyphenOrApostrophe(pos_type pos) const
2938 {
2939         pos_type const psize = size();
2940         if (pos >= psize)
2941                 return false;
2942         char_type const c = d->text_[pos];
2943         if (c != '-' && c != '\'')
2944                 return false;
2945         int nextpos = pos + 1;
2946         int prevpos = pos > 0 ? pos - 1 : 0;
2947         if ((nextpos == psize || isSpace(nextpos))
2948                 && (pos == 0 || isSpace(prevpos)))
2949                 return false;
2950         return c == '\''
2951                 || ((nextpos == psize || d->text_[nextpos] != '-')
2952                 && (pos == 0 || d->text_[prevpos] != '-'));
2953 }
2954
2955
2956 bool Paragraph::isSameSpellRange(pos_type pos1, pos_type pos2) const
2957 {
2958         return pos1 == pos2
2959                 || d->speller_state_.getRange(pos1) == d->speller_state_.getRange(pos2);
2960 }
2961
2962
2963 bool Paragraph::isChar(pos_type pos) const
2964 {
2965         if (Inset const * inset = getInset(pos))
2966                 return inset->isChar();
2967         char_type const c = d->text_[pos];
2968         return !isLetterChar(c) && !isDigitASCII(c) && !lyx::isSpace(c);
2969 }
2970
2971
2972 bool Paragraph::isSpace(pos_type pos) const
2973 {
2974         if (Inset const * inset = getInset(pos))
2975                 return inset->isSpace();
2976         char_type const c = d->text_[pos];
2977         return lyx::isSpace(c);
2978 }
2979
2980
2981 Language const *
2982 Paragraph::getParLanguage(BufferParams const & bparams) const
2983 {
2984         if (!empty())
2985                 return getFirstFontSettings(bparams).language();
2986         // FIXME: we should check the prev par as well (Lgb)
2987         return bparams.language;
2988 }
2989
2990
2991 bool Paragraph::isRTL(BufferParams const & bparams) const
2992 {
2993         return lyxrc.rtl_support
2994                 && getParLanguage(bparams)->rightToLeft()
2995                 && !inInset().getLayout().forceLTR();
2996 }
2997
2998
2999 void Paragraph::changeLanguage(BufferParams const & bparams,
3000                                Language const * from, Language const * to)
3001 {
3002         // change language including dummy font change at the end
3003         for (pos_type i = 0; i <= size(); ++i) {
3004                 Font font = getFontSettings(bparams, i);
3005                 if (font.language() == from) {
3006                         font.setLanguage(to);
3007                         setFont(i, font);
3008                         d->requestSpellCheck(i);
3009                 }
3010         }
3011 }
3012
3013
3014 bool Paragraph::isMultiLingual(BufferParams const & bparams) const
3015 {
3016         Language const * doc_language = bparams.language;
3017         FontList::const_iterator cit = d->fontlist_.begin();
3018         FontList::const_iterator end = d->fontlist_.end();
3019
3020         for (; cit != end; ++cit)
3021                 if (cit->font().language() != ignore_language &&
3022                     cit->font().language() != latex_language &&
3023                     cit->font().language() != doc_language)
3024                         return true;
3025         return false;
3026 }
3027
3028
3029 void Paragraph::getLanguages(std::set<Language const *> & languages) const
3030 {
3031         FontList::const_iterator cit = d->fontlist_.begin();
3032         FontList::const_iterator end = d->fontlist_.end();
3033
3034         for (; cit != end; ++cit) {
3035                 Language const * lang = cit->font().language();
3036                 if (lang != ignore_language &&
3037                     lang != latex_language)
3038                         languages.insert(lang);
3039         }
3040 }
3041
3042
3043 docstring Paragraph::asString(int options) const
3044 {
3045         return asString(0, size(), options);
3046 }
3047
3048
3049 docstring Paragraph::asString(pos_type beg, pos_type end, int options) const
3050 {
3051         odocstringstream os;
3052
3053         if (beg == 0
3054             && options & AS_STR_LABEL
3055             && !d->params_.labelString().empty())
3056                 os << d->params_.labelString() << ' ';
3057
3058         for (pos_type i = beg; i < end; ++i) {
3059                 if ((options & AS_STR_SKIPDELETE) && isDeleted(i))
3060                         continue;
3061                 char_type const c = d->text_[i];
3062                 if (isPrintable(c) || c == '\t'
3063                     || (c == '\n' && (options & AS_STR_NEWLINES)))
3064                         os.put(c);
3065                 else if (c == META_INSET && (options & AS_STR_INSETS)) {
3066                         getInset(i)->toString(os);
3067                         if (getInset(i)->asInsetMath())
3068                                 os << " ";
3069                 }
3070         }
3071
3072         return os.str();
3073 }
3074
3075
3076 void Paragraph::forToc(docstring & os, size_t maxlen) const
3077 {
3078         if (!d->params_.labelString().empty())
3079                 os += d->params_.labelString() + ' ';
3080         for (pos_type i = 0; i < size() && os.length() < maxlen; ++i) {
3081                 if (isDeleted(i))
3082                         continue;
3083                 char_type const c = d->text_[i];
3084                 if (isPrintable(c))
3085                         os += c;
3086                 else if (c == '\t' || c == '\n')
3087                         os += ' ';
3088                 else if (c == META_INSET)
3089                         getInset(i)->forToc(os, maxlen);
3090         }
3091 }
3092
3093
3094 docstring Paragraph::stringify(pos_type beg, pos_type end, int options, OutputParams & runparams) const
3095 {
3096         odocstringstream os;
3097
3098         if (beg == 0
3099                 && options & AS_STR_LABEL
3100                 && !d->params_.labelString().empty())
3101                 os << d->params_.labelString() << ' ';
3102
3103         for (pos_type i = beg; i < end; ++i) {
3104                 char_type const c = d->text_[i];
3105                 if (isPrintable(c) || c == '\t'
3106                     || (c == '\n' && (options & AS_STR_NEWLINES)))
3107                         os.put(c);
3108                 else if (c == META_INSET && (options & AS_STR_INSETS)) {
3109                         getInset(i)->plaintext(os, runparams);
3110                 }
3111         }
3112
3113         return os.str();
3114 }
3115
3116
3117 void Paragraph::setInsetOwner(Inset const * inset)
3118 {
3119         d->inset_owner_ = inset;
3120 }
3121
3122
3123 int Paragraph::id() const
3124 {
3125         return d->id_;
3126 }
3127
3128
3129 void Paragraph::setId(int id)
3130 {
3131         d->id_ = id;
3132 }
3133
3134
3135 Layout const & Paragraph::layout() const
3136 {
3137         return *d->layout_;
3138 }
3139
3140
3141 void Paragraph::setLayout(Layout const & layout)
3142 {
3143         d->layout_ = &layout;
3144 }
3145
3146
3147 void Paragraph::setDefaultLayout(DocumentClass const & tc)
3148 {
3149         setLayout(tc.defaultLayout());
3150 }
3151
3152
3153 void Paragraph::setPlainLayout(DocumentClass const & tc)
3154 {
3155         setLayout(tc.plainLayout());
3156 }
3157
3158
3159 void Paragraph::setPlainOrDefaultLayout(DocumentClass const & tclass)
3160 {
3161         if (usePlainLayout())
3162                 setPlainLayout(tclass);
3163         else
3164                 setDefaultLayout(tclass);
3165 }
3166
3167
3168 Inset const & Paragraph::inInset() const
3169 {
3170         LASSERT(d->inset_owner_, throw ExceptionMessage(BufferException,
3171                 _("Memory problem"), _("Paragraph not properly initialized")));
3172         return *d->inset_owner_;
3173 }
3174
3175
3176 ParagraphParameters & Paragraph::params()
3177 {
3178         return d->params_;
3179 }
3180
3181
3182 ParagraphParameters const & Paragraph::params() const
3183 {
3184         return d->params_;
3185 }
3186
3187
3188 bool Paragraph::isFreeSpacing() const
3189 {
3190         if (d->layout_->free_spacing)
3191                 return true;
3192         return d->inset_owner_ && d->inset_owner_->isFreeSpacing();
3193 }
3194
3195
3196 bool Paragraph::allowEmpty() const
3197 {
3198         if (d->layout_->keepempty)
3199                 return true;
3200         return d->inset_owner_ && d->inset_owner_->allowEmpty();
3201 }
3202
3203
3204 char_type Paragraph::transformChar(char_type c, pos_type pos) const
3205 {
3206         if (!Encodings::isArabicChar(c))
3207                 return c;
3208
3209         char_type prev_char = ' ';
3210         char_type next_char = ' ';
3211
3212         for (pos_type i = pos - 1; i >= 0; --i) {
3213                 char_type const par_char = d->text_[i];
3214                 if (!Encodings::isArabicComposeChar(par_char)) {
3215                         prev_char = par_char;
3216                         break;
3217                 }
3218         }
3219
3220         for (pos_type i = pos + 1, end = size(); i < end; ++i) {
3221                 char_type const par_char = d->text_[i];
3222                 if (!Encodings::isArabicComposeChar(par_char)) {
3223                         next_char = par_char;
3224                         break;
3225                 }
3226         }
3227
3228         if (Encodings::isArabicChar(next_char)) {
3229                 if (Encodings::isArabicChar(prev_char) &&
3230                         !Encodings::isArabicSpecialChar(prev_char))
3231                         return Encodings::transformChar(c, Encodings::FORM_MEDIAL);
3232                 else
3233                         return Encodings::transformChar(c, Encodings::FORM_INITIAL);
3234         } else {
3235                 if (Encodings::isArabicChar(prev_char) &&
3236                         !Encodings::isArabicSpecialChar(prev_char))
3237                         return Encodings::transformChar(c, Encodings::FORM_FINAL);
3238                 else
3239                         return Encodings::transformChar(c, Encodings::FORM_ISOLATED);
3240         }
3241 }
3242
3243
3244 bool Paragraph::brokenBiblio() const
3245 {
3246         // there is a problem if there is no bibitem at position 0 or
3247         // if there is another bibitem in the paragraph.
3248         return d->layout_->labeltype == LABEL_BIBLIO
3249                 && (d->insetlist_.find(BIBITEM_CODE) != 0
3250                     || d->insetlist_.find(BIBITEM_CODE, 1) > 0);
3251 }
3252
3253
3254 int Paragraph::fixBiblio(Buffer const & buffer)
3255 {
3256         // FIXME: What about the case where paragraph is not BIBLIO
3257         // but there is an InsetBibitem?
3258         // FIXME: when there was already an inset at 0, the return value is 1,
3259         // which does not tell whether another inset has been remove; the
3260         // cursor cannot be correctly updated.
3261
3262         if (d->layout_->labeltype != LABEL_BIBLIO)
3263                 return 0;
3264
3265         bool const track_changes = buffer.params().trackChanges;
3266         int bibitem_pos = d->insetlist_.find(BIBITEM_CODE);
3267         bool const hasbibitem0 = bibitem_pos == 0;
3268
3269         if (hasbibitem0) {
3270                 bibitem_pos = d->insetlist_.find(BIBITEM_CODE, 1);
3271                 // There was an InsetBibitem at pos 0, and no other one => OK
3272                 if (bibitem_pos == -1)
3273                         return 0;
3274                 // there is a bibitem at the 0 position, but since
3275                 // there is a second one, we copy the second on the
3276                 // first. We're assuming there are at most two of
3277                 // these, which there should be.
3278                 // FIXME: why does it make sense to do that rather
3279                 // than keep the first? (JMarc)
3280                 Inset * inset = d->insetlist_.release(bibitem_pos);
3281                 eraseChar(bibitem_pos, track_changes);
3282                 d->insetlist_.begin()->inset = inset;
3283                 return -bibitem_pos;
3284         }
3285
3286         // We need to create an inset at the beginning
3287         Inset * inset = 0;
3288         if (bibitem_pos > 0) {
3289                 // there was one somewhere in the paragraph, let's move it
3290                 inset = d->insetlist_.release(bibitem_pos);
3291                 eraseChar(bibitem_pos, track_changes);
3292         } else
3293                 // make a fresh one
3294                 inset = new InsetBibitem(const_cast<Buffer *>(&buffer),
3295                                          InsetCommandParams(BIBITEM_CODE));
3296
3297         insertInset(0, inset, Change(track_changes ? Change::INSERTED
3298                                                    : Change::UNCHANGED));
3299
3300         return 1;
3301 }
3302
3303
3304 void Paragraph::checkAuthors(AuthorList const & authorList)
3305 {
3306         d->changes_.checkAuthors(authorList);
3307 }
3308
3309
3310 bool Paragraph::isChanged(pos_type pos) const
3311 {
3312         return lookupChange(pos).changed();
3313 }
3314
3315
3316 bool Paragraph::isInserted(pos_type pos) const
3317 {
3318         return lookupChange(pos).inserted();
3319 }
3320
3321
3322 bool Paragraph::isDeleted(pos_type pos) const
3323 {
3324         return lookupChange(pos).deleted();
3325 }
3326
3327
3328 InsetList const & Paragraph::insetList() const
3329 {
3330         return d->insetlist_;
3331 }
3332
3333
3334 void Paragraph::setBuffer(Buffer & b)
3335 {
3336         d->insetlist_.setBuffer(b);
3337 }
3338
3339
3340 Inset * Paragraph::releaseInset(pos_type pos)
3341 {
3342         Inset * inset = d->insetlist_.release(pos);
3343         /// does not honour change tracking!
3344         eraseChar(pos, false);
3345         return inset;
3346 }
3347
3348
3349 Inset * Paragraph::getInset(pos_type pos)
3350 {
3351         return (pos < pos_type(d->text_.size()) && d->text_[pos] == META_INSET)
3352                  ? d->insetlist_.get(pos) : 0;
3353 }
3354
3355
3356 Inset const * Paragraph::getInset(pos_type pos) const
3357 {
3358         return (pos < pos_type(d->text_.size()) && d->text_[pos] == META_INSET)
3359                  ? d->insetlist_.get(pos) : 0;
3360 }
3361
3362
3363 void Paragraph::changeCase(BufferParams const & bparams, pos_type pos,
3364                 pos_type & right, TextCase action)
3365 {
3366         // process sequences of modified characters; in change
3367         // tracking mode, this approach results in much better
3368         // usability than changing case on a char-by-char basis
3369         docstring changes;
3370
3371         bool const trackChanges = bparams.trackChanges;
3372
3373         bool capitalize = true;
3374
3375         for (; pos < right; ++pos) {
3376                 char_type oldChar = d->text_[pos];
3377                 char_type newChar = oldChar;
3378
3379                 // ignore insets and don't play with deleted text!
3380                 if (oldChar != META_INSET && !isDeleted(pos)) {
3381                         switch (action) {
3382                                 case text_lowercase:
3383                                         newChar = lowercase(oldChar);
3384                                         break;
3385                                 case text_capitalization:
3386                                         if (capitalize) {
3387                                                 newChar = uppercase(oldChar);
3388                                                 capitalize = false;
3389                                         }
3390                                         break;
3391                                 case text_uppercase:
3392                                         newChar = uppercase(oldChar);
3393                                         break;
3394                         }
3395                 }
3396
3397                 if (isWordSeparator(pos) || isDeleted(pos)) {
3398                         // permit capitalization again
3399                         capitalize = true;
3400                 }
3401
3402                 if (oldChar != newChar) {
3403                         changes += newChar;
3404                         if (pos != right - 1)
3405                                 continue;
3406                         // step behind the changing area
3407                         pos++;
3408                 }
3409
3410                 int erasePos = pos - changes.size();
3411                 for (size_t i = 0; i < changes.size(); i++) {
3412                         insertChar(pos, changes[i],
3413                                    getFontSettings(bparams,
3414                                                    erasePos),
3415                                    trackChanges);
3416                         if (!eraseChar(erasePos, trackChanges)) {
3417                                 ++erasePos;
3418                                 ++pos; // advance
3419                                 ++right; // expand selection
3420                         }
3421                 }
3422                 changes.clear();
3423         }
3424 }
3425
3426
3427 int Paragraph::find(docstring const & str, bool cs, bool mw,
3428                 pos_type start_pos, bool del) const
3429 {
3430         pos_type pos = start_pos;
3431         int const strsize = str.length();
3432         int i = 0;
3433         pos_type const parsize = d->text_.size();
3434         for (i = 0; i < strsize && pos < parsize; ++i, ++pos) {
3435                 // Ignore "invisible" letters such as ligature breaks
3436                 // and hyphenation chars while searching
3437                 while (pos < parsize - 1 && isInset(pos)) {
3438                         odocstringstream os;
3439                         getInset(pos)->toString(os);
3440                         if (!getInset(pos)->isLetter() || !os.str().empty())
3441                                 break;
3442                         pos++;
3443                 }
3444                 if (cs && str[i] != d->text_[pos])
3445                         break;
3446                 if (!cs && uppercase(str[i]) != uppercase(d->text_[pos]))
3447                         break;
3448                 if (!del && isDeleted(pos))
3449                         break;
3450         }
3451
3452         if (i != strsize)
3453                 return 0;
3454
3455         // if necessary, check whether string matches word
3456         if (mw) {
3457                 if (start_pos > 0 && !isWordSeparator(start_pos - 1))
3458                         return 0;
3459                 if (pos < parsize
3460                         && !isWordSeparator(pos))
3461                         return 0;
3462         }
3463
3464         return pos - start_pos;
3465 }
3466
3467
3468 char_type Paragraph::getChar(pos_type pos) const
3469 {
3470         return d->text_[pos];
3471 }
3472
3473
3474 pos_type Paragraph::size() const
3475 {
3476         return d->text_.size();
3477 }
3478
3479
3480 bool Paragraph::empty() const
3481 {
3482         return d->text_.empty();
3483 }
3484
3485
3486 bool Paragraph::isInset(pos_type pos) const
3487 {
3488         return d->text_[pos] == META_INSET;
3489 }
3490
3491
3492 bool Paragraph::isSeparator(pos_type pos) const
3493 {
3494         //FIXME: Are we sure this can be the only separator?
3495         return d->text_[pos] == ' ';
3496 }
3497
3498
3499 void Paragraph::deregisterWords()
3500 {
3501         Private::LangWordsMap::const_iterator itl = d->words_.begin();
3502         Private::LangWordsMap::const_iterator ite = d->words_.end();
3503         for (; itl != ite; ++itl) {
3504                 WordList * wl = theWordList(itl->first);
3505                 Private::Words::const_iterator it = (itl->second).begin();
3506                 Private::Words::const_iterator et = (itl->second).end();
3507                 for (; it != et; ++it)
3508                         wl->remove(*it);
3509         }
3510         d->words_.clear();
3511 }
3512
3513
3514 void Paragraph::locateWord(pos_type & from, pos_type & to,
3515         word_location const loc) const
3516 {
3517         switch (loc) {
3518         case WHOLE_WORD_STRICT:
3519                 if (from == 0 || from == size()
3520                     || isWordSeparator(from)
3521                     || isWordSeparator(from - 1)) {
3522                         to = from;
3523                         return;
3524                 }
3525                 // no break here, we go to the next
3526
3527         case WHOLE_WORD:
3528                 // If we are already at the beginning of a word, do nothing
3529                 if (!from || isWordSeparator(from - 1))
3530                         break;
3531                 // no break here, we go to the next
3532
3533         case PREVIOUS_WORD:
3534                 // always move the cursor to the beginning of previous word
3535                 while (from && !isWordSeparator(from - 1))
3536                         --from;
3537                 break;
3538         case NEXT_WORD:
3539                 LYXERR0("Paragraph::locateWord: NEXT_WORD not implemented yet");
3540                 break;
3541         case PARTIAL_WORD:
3542                 // no need to move the 'from' cursor
3543                 break;
3544         }
3545         to = from;
3546         while (to < size() && !isWordSeparator(to))
3547                 ++to;
3548 }
3549
3550
3551 void Paragraph::collectWords()
3552 {
3553         pos_type n = size();
3554         for (pos_type pos = 0; pos < n; ++pos) {
3555                 if (isWordSeparator(pos))
3556                         continue;
3557                 pos_type from = pos;
3558                 locateWord(from, pos, WHOLE_WORD);
3559                 if ((pos - from) >= (int)lyxrc.completion_minlength) {
3560                         docstring word = asString(from, pos, AS_STR_NONE);
3561                         FontList::const_iterator cit = d->fontlist_.fontIterator(pos);
3562                         if (cit == d->fontlist_.end())
3563                                 return;
3564                         Language const * lang = cit->font().language();
3565                         d->words_[*lang].insert(word);
3566                 }
3567         }
3568 }
3569
3570
3571 void Paragraph::registerWords()
3572 {
3573         Private::LangWordsMap::const_iterator itl = d->words_.begin();
3574         Private::LangWordsMap::const_iterator ite = d->words_.end();
3575         for (; itl != ite; ++itl) {
3576                 WordList * wl = theWordList(itl->first);
3577                 Private::Words::const_iterator it = (itl->second).begin();
3578                 Private::Words::const_iterator et = (itl->second).end();
3579                 for (; it != et; ++it)
3580                         wl->insert(*it);
3581         }
3582 }
3583
3584
3585 void Paragraph::updateWords()
3586 {
3587         deregisterWords();
3588         collectWords();
3589         registerWords();
3590 }
3591
3592
3593 void Paragraph::Private::appendSkipPosition(SkipPositions & skips, pos_type const pos) const
3594 {
3595         SkipPositionsIterator begin = skips.begin();
3596         SkipPositions::iterator end = skips.end();
3597         if (pos > 0 && begin < end) {
3598                 --end;
3599                 if (end->last == pos - 1) {
3600                         end->last = pos;
3601                         return;
3602                 }
3603         }
3604         skips.insert(end, FontSpan(pos, pos));
3605 }
3606
3607
3608 Language * Paragraph::Private::locateSpellRange(
3609         pos_type & from, pos_type & to,
3610         SkipPositions & skips) const
3611 {
3612         // skip leading white space
3613         while (from < to && owner_->isWordSeparator(from))
3614                 ++from;
3615         // don't check empty range
3616         if (from >= to)
3617                 return 0;
3618         // get current language
3619         Language * lang = getSpellLanguage(from);
3620         pos_type last = from;
3621         bool samelang = true;
3622         bool sameinset = true;
3623         while (last < to && samelang && sameinset) {
3624                 // hop to end of word
3625                 while (last < to && !owner_->isWordSeparator(last)) {
3626                         if (owner_->getInset(last)) {
3627                                 appendSkipPosition(skips, last);
3628                         } else if (owner_->isDeleted(last)) {
3629                                 appendSkipPosition(skips, last);
3630                         }
3631                         ++last;
3632                 }
3633                 // hop to next word while checking for insets
3634                 while (sameinset && last < to && owner_->isWordSeparator(last)) {
3635                         if (Inset const * inset = owner_->getInset(last))
3636                                 sameinset = inset->isChar() && inset->isLetter();
3637                         if (sameinset && owner_->isDeleted(last)) {
3638                                 appendSkipPosition(skips, last);
3639                         }
3640                         if (sameinset)
3641                                 last++;
3642                 }
3643                 if (sameinset && last < to) {
3644                         // now check for language change
3645                         samelang = lang == getSpellLanguage(last);
3646                 }
3647         }
3648         // if language change detected backstep is needed
3649         if (!samelang)
3650                 --last;
3651         to = last;
3652         return lang;
3653 }
3654
3655
3656 Language * Paragraph::Private::getSpellLanguage(pos_type const from) const
3657 {
3658         Language * lang =
3659                 const_cast<Language *>(owner_->getFontSettings(
3660                         inset_owner_->buffer().params(), from).language());
3661         if (lang == inset_owner_->buffer().params().language
3662                 && !lyxrc.spellchecker_alt_lang.empty()) {
3663                 string lang_code;
3664                 string const lang_variety =
3665                         split(lyxrc.spellchecker_alt_lang, lang_code, '-');
3666                 lang->setCode(lang_code);
3667                 lang->setVariety(lang_variety);
3668         }
3669         return lang;
3670 }
3671
3672
3673 void Paragraph::requestSpellCheck(pos_type pos)
3674 {
3675         d->requestSpellCheck(pos);
3676 }
3677
3678
3679 bool Paragraph::needsSpellCheck() const
3680 {
3681         SpellChecker::ChangeNumber speller_change_number = 0;
3682         if (theSpellChecker())
3683                 speller_change_number = theSpellChecker()->changeNumber();
3684         if (speller_change_number > d->speller_state_.currentChangeNumber()) {
3685                 d->speller_state_.needsCompleteRefresh(speller_change_number);
3686         }
3687         return d->needsSpellCheck();
3688 }
3689
3690
3691 bool Paragraph::Private::ignoreWord(docstring const & word) const
3692 {
3693         // Ignore words with digits
3694         // FIXME: make this customizable
3695         // (note that some checkers ignore words with digits by default)
3696         docstring::const_iterator cit = word.begin();
3697         docstring::const_iterator const end = word.end();
3698         for (; cit != end; ++cit) {
3699                 if (isNumber((*cit)))
3700                         return true;
3701         }
3702         return false;
3703 }
3704
3705
3706 SpellChecker::Result Paragraph::spellCheck(pos_type & from, pos_type & to,
3707         WordLangTuple & wl, docstring_list & suggestions,
3708         bool do_suggestion, bool check_learned) const
3709 {
3710         SpellChecker::Result result = SpellChecker::WORD_OK;
3711         SpellChecker * speller = theSpellChecker();
3712         if (!speller)
3713                 return result;
3714
3715         if (!d->layout_->spellcheck || !inInset().allowSpellCheck())
3716                 return result;
3717
3718         locateWord(from, to, WHOLE_WORD);
3719         if (from == to || from >= size())
3720                 return result;
3721
3722         docstring word = asString(from, to, AS_STR_INSETS | AS_STR_SKIPDELETE);
3723         Language * lang = d->getSpellLanguage(from);
3724
3725         wl = WordLangTuple(word, lang);
3726
3727         if (!word.size())
3728                 return result;
3729
3730         if (needsSpellCheck() || check_learned) {
3731                 pos_type end = to;
3732                 if (!d->ignoreWord(word)) {
3733                         bool const trailing_dot = to < size() && d->text_[to] == '.';
3734                         result = speller->check(wl);
3735                         if (SpellChecker::misspelled(result) && trailing_dot) {
3736                                 wl = WordLangTuple(word.append(from_ascii(".")), lang);
3737                                 result = speller->check(wl);
3738                                 if (!SpellChecker::misspelled(result)) {
3739                                         LYXERR(Debug::GUI, "misspelled word is correct with dot: \"" <<
3740                                            word << "\" [" <<
3741                                            from << ".." << to << "]");
3742                                 } else {
3743                                         // spell check with dot appended failed too
3744                                         // restore original word/lang value
3745                                         word = asString(from, to, AS_STR_INSETS | AS_STR_SKIPDELETE);
3746                                         wl = WordLangTuple(word, lang);
3747                                 }
3748                         }
3749                 }
3750                 if (!SpellChecker::misspelled(result)) {
3751                         // area up to the begin of the next word is not misspelled
3752                         while (end < size() && isWordSeparator(end))
3753                                 ++end;
3754                 }
3755                 d->setMisspelled(from, end, result);
3756         } else {
3757                 result = d->speller_state_.getState(from);
3758         }
3759
3760         if (do_suggestion)
3761                 suggestions.clear();
3762
3763         if (SpellChecker::misspelled(result)) {
3764                 LYXERR(Debug::GUI, "misspelled word: \"" <<
3765                            word << "\" [" <<
3766                            from << ".." << to << "]");
3767                 if (do_suggestion)
3768                         speller->suggest(wl, suggestions);
3769         }
3770         return result;
3771 }
3772
3773
3774 void Paragraph::Private::markMisspelledWords(
3775         pos_type const & first, pos_type const & last,
3776         SpellChecker::Result result,
3777         docstring const & word,
3778         SkipPositions const & skips)
3779 {
3780         if (!SpellChecker::misspelled(result)) {
3781                 setMisspelled(first, last, SpellChecker::WORD_OK);
3782                 return;
3783         }
3784         int snext = first;
3785         SpellChecker * speller = theSpellChecker();
3786         // locate and enumerate the error positions
3787         int nerrors = speller->numMisspelledWords();
3788         int numskipped = 0;
3789         SkipPositionsIterator it = skips.begin();
3790         SkipPositionsIterator et = skips.end();
3791         for (int index = 0; index < nerrors; ++index) {
3792                 int wstart;
3793                 int wlen = 0;
3794                 speller->misspelledWord(index, wstart, wlen);
3795                 /// should not happen if speller supports range checks
3796                 if (!wlen) continue;
3797                 docstring const misspelled = word.substr(wstart, wlen);
3798                 wstart += first + numskipped;
3799                 if (snext < wstart) {
3800                         /// mark the range of correct spelling
3801                         numskipped += countSkips(it, et, wstart);
3802                         setMisspelled(snext,
3803                                 wstart - 1, SpellChecker::WORD_OK);
3804                 }
3805                 snext = wstart + wlen;
3806                 numskipped += countSkips(it, et, snext);
3807                 /// mark the range of misspelling
3808                 setMisspelled(wstart, snext, result);
3809                 LYXERR(Debug::GUI, "misspelled word: \"" <<
3810                            misspelled << "\" [" <<
3811                            wstart << ".." << (snext-1) << "]");
3812                 ++snext;
3813         }
3814         if (snext <= last) {
3815                 /// mark the range of correct spelling at end
3816                 setMisspelled(snext, last, SpellChecker::WORD_OK);
3817         }
3818 }
3819
3820
3821 void Paragraph::spellCheck() const
3822 {
3823         SpellChecker * speller = theSpellChecker();
3824         if (!speller || !size() ||!needsSpellCheck())
3825                 return;
3826         pos_type start;
3827         pos_type endpos;
3828         d->rangeOfSpellCheck(start, endpos);
3829         if (speller->canCheckParagraph()) {
3830                 // loop until we leave the range
3831                 for (pos_type first = start; first < endpos; ) {
3832                         pos_type last = endpos;
3833                         Private::SkipPositions skips;
3834                         Language * lang = d->locateSpellRange(first, last, skips);
3835                         if (first >= endpos)
3836                                 break;
3837                         // start the spell checker on the unit of meaning
3838                         docstring word = asString(first, last, AS_STR_INSETS + AS_STR_SKIPDELETE);
3839                         WordLangTuple wl = WordLangTuple(word, lang);
3840                         SpellChecker::Result result = word.size() ?
3841                                 speller->check(wl) : SpellChecker::WORD_OK;
3842                         d->markMisspelledWords(first, last, result, word, skips);
3843                         first = ++last;
3844                 }
3845         } else {
3846                 static docstring_list suggestions;
3847                 pos_type to = endpos;
3848                 while (start < endpos) {
3849                         WordLangTuple wl;
3850                         spellCheck(start, to, wl, suggestions, false);
3851                         start = to + 1;
3852                 }
3853         }
3854         d->readySpellCheck();
3855 }
3856
3857
3858 bool Paragraph::isMisspelled(pos_type pos, bool check_boundary) const
3859 {
3860         bool result = SpellChecker::misspelled(d->speller_state_.getState(pos));
3861         if (result || pos <= 0 || pos > size())
3862                 return result;
3863         if (check_boundary && (pos == size() || isWordSeparator(pos)))
3864                 result = SpellChecker::misspelled(d->speller_state_.getState(pos - 1));
3865         return result;
3866 }
3867
3868
3869 string Paragraph::magicLabel() const
3870 {
3871         stringstream ss;
3872         ss << "magicparlabel-" << id();
3873         return ss.str();
3874 }
3875
3876
3877 } // namespace lyx