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 "BufferEncodings.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
  73 /// Inset identifier (above 0x10ffff, for ucs-4)
  74 char_type const META_INSET = 0x200001;
  75
  76 }
  77
  78
  79 /////////////////////////////////////////////////////////////////////
  80 //
  81 // SpellResultRange
  82 //
  83 /////////////////////////////////////////////////////////////////////
  84
  85 class SpellResultRange {
  86 public:
  87         SpellResultRange(FontSpan range, SpellChecker::Result result)
  88         : range_(range), result_(result)
  89         {}
  90         ///
  91         FontSpan const & range() const { return range_; }
  92         ///
  93         void range(FontSpan const & r) { range_ = r; }
  94         ///
  95         SpellChecker::Result result() const { return result_; }
  96         ///
  97         void result(SpellChecker::Result r) { result_ = r; }
  98         ///
  99         bool contains(pos_type pos) const { return range_.contains(pos); }
 100         ///
 101         bool covered(FontSpan const & r) const
 102         {
 103                 // 1. first of new range inside current range or
 104                 // 2. last of new range inside current range or
 105                 // 3. first of current range inside new range or
 106                 // 4. last of current range inside new range
 107                 //FIXME: is this the same as !range_.intersect(r).empty() ?
 108                 return range_.contains(r.first) || range_.contains(r.last) ||
 109                         r.contains(range_.first) || r.contains(range_.last);
 110         }
 111         ///
 112         void shift(pos_type pos, int offset)
 113         {
 114                 if (range_.first > pos) {
 115                         range_.first += offset;
 116                         range_.last += offset;
 117                 } else if (range_.last >= pos) {
 118                         range_.last += offset;
 119                 }
 120         }
 121 private:
 122         FontSpan range_ ;
 123         SpellChecker::Result result_ ;
 124 };
 125
 126
 127 /////////////////////////////////////////////////////////////////////
 128 //
 129 // SpellCheckerState
 130 //
 131 /////////////////////////////////////////////////////////////////////
 132
 133 class SpellCheckerState {
 134 public:
 135         SpellCheckerState()
 136         {
 137                 needs_refresh_ = true;
 138                 current_change_number_ = 0;
 139         }
 140
 141         void setRange(FontSpan const & fp, SpellChecker::Result state)
 142         {
 143                 Ranges result;
 144                 RangesIterator et = ranges_.end();
 145                 RangesIterator it = ranges_.begin();
 146                 for (; it != et; ++it) {
 147                         if (!it->covered(fp))
 148                                 result.push_back(SpellResultRange(it->range(), it->result()));
 149                         else if (state == SpellChecker::WORD_OK) {
 150                                 // trim or split the current misspelled range
 151                                 // store misspelled ranges only
 152                                 FontSpan range = it->range();
 153                                 if (fp.first > range.first) {
 154                                         // misspelled area in front of WORD_OK
 155                                         range.last = fp.first - 1;
 156                                         result.push_back(SpellResultRange(range, it->result()));
 157                                         range = it->range();
 158                                 }
 159                                 if (fp.last < range.last) {
 160                                         // misspelled area after WORD_OK range
 161                                         range.first = fp.last + 1;
 162                                         result.push_back(SpellResultRange(range, it->result()));
 163                                 }
 164                         }
 165                 }
 166                 ranges_ = result;
 167                 if (state != SpellChecker::WORD_OK)
 168                         ranges_.push_back(SpellResultRange(fp, state));
 169         }
 170
 171         void increasePosAfterPos(pos_type pos)
 172         {
 173                 correctRangesAfterPos(pos, 1);
 174                 needsRefresh(pos);
 175         }
 176
 177         void decreasePosAfterPos(pos_type pos)
 178         {
 179                 correctRangesAfterPos(pos, -1);
 180                 needsRefresh(pos);
 181         }
 182
 183         void refreshLast(pos_type pos)
 184         {
 185                 if (pos < refresh_.last)
 186                         refresh_.last = pos;
 187         }
 188
 189         SpellChecker::Result getState(pos_type pos) const
 190         {
 191                 SpellChecker::Result result = SpellChecker::WORD_OK;
 192                 RangesIterator et = ranges_.end();
 193                 RangesIterator it = ranges_.begin();
 194                 for (; it != et; ++it) {
 195                         if(it->contains(pos)) {
 196                                 return it->result();
 197                         }
 198                 }
 199                 return result;
 200         }
 201
 202         FontSpan const & getRange(pos_type pos) const
 203         {
 204                 /// empty span to indicate mismatch
 205                 static FontSpan empty_;
 206                 RangesIterator et = ranges_.end();
 207                 RangesIterator it = ranges_.begin();
 208                 for (; it != et; ++it) {
 209                         if(it->contains(pos)) {
 210                                 return it->range();
 211                         }
 212                 }
 213                 return empty_;
 214         }
 215
 216         bool needsRefresh() const
 217         {
 218                 return needs_refresh_;
 219         }
 220
 221         SpellChecker::ChangeNumber currentChangeNumber() const
 222         {
 223                 return current_change_number_;
 224         }
 225
 226         void refreshRange(pos_type & first, pos_type & last) const
 227         {
 228                 first = refresh_.first;
 229                 last = refresh_.last;
 230         }
 231
 232         void needsRefresh(pos_type pos)
 233         {
 234                 if (needs_refresh_ && pos != -1) {
 235                         if (pos < refresh_.first)
 236                                 refresh_.first = pos;
 237                         if (pos > refresh_.last)
 238                                 refresh_.last = pos;
 239                 } else if (pos != -1) {
 240                         // init request check for neighbour positions too
 241                         refresh_.first = pos > 0 ? pos - 1 : 0;
 242                         // no need for special end of paragraph check
 243                         refresh_.last = pos + 1;
 244                 }
 245                 needs_refresh_ = pos != -1;
 246         }
 247
 248         void needsCompleteRefresh(SpellChecker::ChangeNumber change_number)
 249         {
 250                 needs_refresh_ = true;
 251                 refresh_.first = 0;
 252                 refresh_.last = -1;
 253                 current_change_number_ = change_number;
 254         }
 255 private:
 256         typedef vector<SpellResultRange> Ranges;
 257         typedef Ranges::const_iterator RangesIterator;
 258         Ranges ranges_;
 259         /// the area of the paragraph with pending spell check
 260         FontSpan refresh_;
 261         bool needs_refresh_;
 262         /// spell state cache version number
 263         SpellChecker::ChangeNumber current_change_number_;
 264
 265
 266         void correctRangesAfterPos(pos_type pos, int offset)
 267         {
 268                 RangesIterator et = ranges_.end();
 269                 Ranges::iterator it = ranges_.begin();
 270                 for (; it != et; ++it) {
 271                         it->shift(pos, offset);
 272                 }
 273         }
 274
 275 };
 276
 277 /////////////////////////////////////////////////////////////////////
 278 //
 279 // Paragraph::Private
 280 //
 281 /////////////////////////////////////////////////////////////////////
 282
 283 class Paragraph::Private
 284 {
 285         // Enforce our own "copy" constructor by declaring the standard one and
 286         // the assignment operator private without implementing them.
 287         Private(Private const &);
 288         Private & operator=(Private const &);
 289 public:
 290         ///
 291         Private(Paragraph * owner, Layout const & layout);
 292         /// "Copy constructor"
 293         Private(Private const &, Paragraph * owner);
 294         /// Copy constructor from \p beg  to \p end
 295         Private(Private const &, Paragraph * owner, pos_type beg, pos_type end);
 296
 297         ///
 298         void insertChar(pos_type pos, char_type c, Change const & change);
 299
 300         /// Output the surrogate pair formed by \p c and \p next to \p os.
 301         /// \return the number of characters written.
 302         int latexSurrogatePair(otexstream & os, char_type c, char_type next,
 303                                OutputParams const &);
 304
 305         /// Output a space in appropriate formatting (or a surrogate pair
 306         /// if the next character is a combining character).
 307         /// \return whether a surrogate pair was output.
 308         bool simpleTeXBlanks(OutputParams const &,
 309                              otexstream &,
 310                              pos_type i,
 311                              unsigned int & column,
 312                              Font const & font,
 313                              Layout const & style);
 314
 315         /// Output consecutive unicode chars, belonging to the same script as
 316         /// specified by the latex macro \p ltx, to \p os starting from \p i.
 317         /// \return the number of characters written.
 318         int writeScriptChars(otexstream & os, docstring const & ltx,
 319                            Change const &, Encoding const &, pos_type & i);
 320
 321         /// This could go to ParagraphParameters if we want to.
 322         int startTeXParParams(BufferParams const &, otexstream &,
 323                               OutputParams const &) const;
 324
 325         /// This could go to ParagraphParameters if we want to.
 326         bool endTeXParParams(BufferParams const &, otexstream &,
 327                              OutputParams const &) const;
 328
 329         ///
 330         void latexInset(BufferParams const &,
 331                                    otexstream &,
 332                                    OutputParams &,
 333                                    Font & running_font,
 334                                    Font & basefont,
 335                                    Font const & outerfont,
 336                                    bool & open_font,
 337                                    Change & running_change,
 338                                    Layout const & style,
 339                                    pos_type & i,
 340                                    unsigned int & column);
 341
 342         ///
 343         void latexSpecialChar(
 344                                    otexstream & os,
 345                                    BufferParams const & bparams,
 346                                    OutputParams const & runparams,
 347                                    Font const & running_font,
 348                                    Change const & running_change,
 349                                    Layout const & style,
 350                                    pos_type & i,
 351                                    pos_type end_pos,
 352                                    unsigned int & column);
 353
 354         ///
 355         bool latexSpecialT1(
 356                 char_type const c,
 357                 otexstream & os,
 358                 pos_type i,
 359                 unsigned int & column);
 360         ///
 361         bool latexSpecialT3(
 362                 char_type const c,
 363                 otexstream & os,
 364                 pos_type i,
 365                 unsigned int & column);
 366
 367         ///
 368         void validate(LaTeXFeatures & features) const;
 369
 370         /// Checks if the paragraph contains only text and no inset or font change.
 371         bool onlyText(Buffer const & buf, Font const & outerfont,
 372                       pos_type initial) const;
 373
 374         /// a vector of speller skip positions
 375         typedef vector<FontSpan> SkipPositions;
 376         typedef SkipPositions::const_iterator SkipPositionsIterator;
 377
 378         void appendSkipPosition(SkipPositions & skips, pos_type const pos) const;
 379
 380         Language * getSpellLanguage(pos_type const from) const;
 381
 382         Language * locateSpellRange(pos_type & from, pos_type & to,
 383                                     SkipPositions & skips) const;
 384
 385         bool hasSpellerChange() const
 386         {
 387                 SpellChecker::ChangeNumber speller_change_number = 0;
 388                 if (theSpellChecker())
 389                         speller_change_number = theSpellChecker()->changeNumber();
 390                 return speller_change_number > speller_state_.currentChangeNumber();
 391         }
 392
 393         bool ignoreWord(docstring const & word) const ;
 394
 395         void setMisspelled(pos_type from, pos_type to, SpellChecker::Result state)
 396         {
 397                 pos_type textsize = owner_->size();
 398                 // check for sane arguments
 399                 if (to <= from || from >= textsize)
 400                         return;
 401                 FontSpan fp = FontSpan(from, to - 1);
 402                 speller_state_.setRange(fp, state);
 403         }
 404
 405         void requestSpellCheck(pos_type pos)
 406         {
 407                 if (pos == -1)
 408                         speller_state_.needsCompleteRefresh(speller_state_.currentChangeNumber());
 409                 else
 410                         speller_state_.needsRefresh(pos);
 411         }
 412
 413         void readySpellCheck()
 414         {
 415                 speller_state_.needsRefresh(-1);
 416         }
 417
 418         bool needsSpellCheck() const
 419         {
 420                 return speller_state_.needsRefresh();
 421         }
 422
 423         void rangeOfSpellCheck(pos_type & first, pos_type & last) const
 424         {
 425                 speller_state_.refreshRange(first, last);
 426                 if (last == -1) {
 427                         last = owner_->size();
 428                         return;
 429                 }
 430                 pos_type endpos = last;
 431                 owner_->locateWord(first, endpos, WHOLE_WORD);
 432                 if (endpos < last) {
 433                         endpos = last;
 434                         owner_->locateWord(last, endpos, WHOLE_WORD);
 435                 }
 436                 last = endpos;
 437         }
 438
 439         int countSkips(SkipPositionsIterator & it, SkipPositionsIterator const et,
 440                             int & start) const
 441         {
 442                 int numskips = 0;
 443                 while (it != et && it->first < start) {
 444                         int skip = it->last - it->first + 1;
 445                         start += skip;
 446                         numskips += skip;
 447                         ++it;
 448                 }
 449                 return numskips;
 450         }
 451
 452         void markMisspelledWords(pos_type const & first, pos_type const & last,
 453                                                          SpellChecker::Result result,
 454                                                          docstring const & word,
 455                                                          SkipPositions const & skips);
 456
 457         InsetCode ownerCode() const
 458         {
 459                 return inset_owner_ ? inset_owner_->lyxCode() : NO_CODE;
 460         }
 461
 462         /// Which Paragraph owns us?
 463         Paragraph * owner_;
 464
 465         /// In which Inset?
 466         Inset const * inset_owner_;
 467
 468         ///
 469         FontList fontlist_;
 470
 471         ///
 472         int id_;
 473
 474         ///
 475         ParagraphParameters params_;
 476
 477         /// for recording and looking up changes
 478         Changes changes_;
 479
 480         ///
 481         InsetList insetlist_;
 482
 483         /// end of label
 484         pos_type begin_of_body_;
 485
 486         typedef docstring TextContainer;
 487         ///
 488         TextContainer text_;
 489
 490         typedef set<docstring> Words;
 491         typedef map<string, Words> LangWordsMap;
 492         ///
 493         LangWordsMap words_;
 494         ///
 495         Layout const * layout_;
 496         ///
 497         SpellCheckerState speller_state_;
 498 };
 499
 500
 501 Paragraph::Private::Private(Paragraph * owner, Layout const & layout)
 502         : owner_(owner), inset_owner_(0), id_(-1), begin_of_body_(0), layout_(&layout)
 503 {
 504         text_.reserve(100);
 505 }
 506
 507
 508 // Initialization of the counter for the paragraph id's,
 509 //
 510 // FIXME: There should be a more intelligent way to generate and use the
 511 // paragraph ids per buffer instead a global static counter for all InsetText
 512 // in the running program.
 513 // However, this per-session id is used in LFUN_PARAGRAPH_GOTO to
 514 // switch to a different buffer, as used in the outliner for instance.
 515 static int paragraph_id = -1;
 516
 517 Paragraph::Private::Private(Private const & p, Paragraph * owner)
 518         : owner_(owner), inset_owner_(p.inset_owner_), fontlist_(p.fontlist_),
 519           params_(p.params_), changes_(p.changes_), insetlist_(p.insetlist_),
 520           begin_of_body_(p.begin_of_body_), text_(p.text_), words_(p.words_),
 521           layout_(p.layout_)
 522 {
 523         id_ = ++paragraph_id;
 524         requestSpellCheck(p.text_.size());
 525 }
 526
 527
 528 Paragraph::Private::Private(Private const & p, Paragraph * owner,
 529         pos_type beg, pos_type end)
 530         : owner_(owner), inset_owner_(p.inset_owner_),
 531           params_(p.params_), changes_(p.changes_),
 532           insetlist_(p.insetlist_, beg, end),
 533           begin_of_body_(p.begin_of_body_), words_(p.words_),
 534           layout_(p.layout_)
 535 {
 536         id_ = ++paragraph_id;
 537         if (beg >= pos_type(p.text_.size()))
 538                 return;
 539         text_ = p.text_.substr(beg, end - beg);
 540
 541         FontList::const_iterator fcit = fontlist_.begin();
 542         FontList::const_iterator fend = fontlist_.end();
 543         for (; fcit != fend; ++fcit) {
 544                 if (fcit->pos() < beg)
 545                         continue;
 546                 if (fcit->pos() >= end) {
 547                         // Add last entry in the fontlist_.
 548                         fontlist_.set(text_.size() - 1, fcit->font());
 549                         break;
 550                 }
 551                 // Add a new entry in the fontlist_.
 552                 fontlist_.set(fcit->pos() - beg, fcit->font());
 553         }
 554         requestSpellCheck(p.text_.size());
 555 }
 556
 557
 558 void Paragraph::addChangesToToc(DocIterator const & cdit,
 559         Buffer const & buf, bool output_active) const
 560 {
 561         d->changes_.addToToc(cdit, buf, output_active);
 562 }
 563
 564
 565 void Paragraph::addChangesToBuffer(Buffer const & buf) const
 566 {
 567         d->changes_.updateBuffer(buf);
 568 }
 569
 570
 571 bool Paragraph::isChangeUpdateRequired() const
 572 {
 573         return d->changes_.isUpdateRequired();
 574 }
 575
 576
 577 bool Paragraph::isDeleted(pos_type start, pos_type end) const
 578 {
 579         LASSERT(start >= 0 && start <= size(), return false);
 580         LASSERT(end > start && end <= size() + 1, return false);
 581
 582         return d->changes_.isDeleted(start, end);
 583 }
 584
 585
 586 bool Paragraph::isChanged(pos_type start, pos_type end) const
 587 {
 588         LASSERT(start >= 0 && start <= size(), return false);
 589         LASSERT(end > start && end <= size() + 1, return false);
 590
 591         return d->changes_.isChanged(start, end);
 592 }
 593
 594
 595 bool Paragraph::isMergedOnEndOfParDeletion(bool trackChanges) const
 596 {
 597         // keep the logic here in sync with the logic of eraseChars()
 598         if (!trackChanges)
 599                 return true;
 600
 601         Change const & change = d->changes_.lookup(size());
 602         return change.inserted() && change.currentAuthor();
 603 }
 604
 605
 606 void Paragraph::setChange(Change const & change)
 607 {
 608         // beware of the imaginary end-of-par character!
 609         d->changes_.set(change, 0, size() + 1);
 610
 611         /*
 612          * Propagate the change recursively - but not in case of DELETED!
 613          *
 614          * Imagine that your co-author makes changes in an existing inset. He
 615          * sends your document to you and you come to the conclusion that the
 616          * inset should go completely. If you erase it, LyX must not delete all
 617          * text within the inset. Otherwise, the change tracked insertions of
 618          * your co-author get lost and there is no way to restore them later.
 619          *
 620          * Conclusion: An inset's content should remain untouched if you delete it
 621          */
 622
 623         if (!change.deleted()) {
 624                 for (pos_type pos = 0; pos < size(); ++pos) {
 625                         if (Inset * inset = getInset(pos))
 626                                 inset->setChange(change);
 627                 }
 628         }
 629 }
 630
 631
 632 void Paragraph::setChange(pos_type pos, Change const & change)
 633 {
 634         LASSERT(pos >= 0 && pos <= size(), return);
 635         d->changes_.set(change, pos);
 636
 637         // see comment in setChange(Change const &) above
 638         if (!change.deleted() && pos < size())
 639                         if (Inset * inset = getInset(pos))
 640                                 inset->setChange(change);
 641 }
 642
 643
 644 Change const & Paragraph::lookupChange(pos_type pos) const
 645 {
 646         LBUFERR(pos >= 0 && pos <= size());
 647         return d->changes_.lookup(pos);
 648 }
 649
 650
 651 void Paragraph::acceptChanges(pos_type start, pos_type end)
 652 {
 653         LASSERT(start >= 0 && start <= size(), return);
 654         LASSERT(end > start && end <= size() + 1, return);
 655
 656         for (pos_type pos = start; pos < end; ++pos) {
 657                 switch (lookupChange(pos).type) {
 658                         case Change::UNCHANGED:
 659                                 // accept changes in nested inset
 660                                 if (Inset * inset = getInset(pos))
 661                                         inset->acceptChanges();
 662                                 break;
 663
 664                         case Change::INSERTED:
 665                                 d->changes_.set(Change(Change::UNCHANGED), pos);
 666                                 // also accept changes in nested inset
 667                                 if (Inset * inset = getInset(pos))
 668                                         inset->acceptChanges();
 669                                 break;
 670
 671                         case Change::DELETED:
 672                                 // Suppress access to non-existent
 673                                 // "end-of-paragraph char"
 674                                 if (pos < size()) {
 675                                         eraseChar(pos, false);
 676                                         --end;
 677                                         --pos;
 678                                 }
 679                                 break;
 680                 }
 681
 682         }
 683 }
 684
 685
 686 void Paragraph::rejectChanges(pos_type start, pos_type end)
 687 {
 688         LASSERT(start >= 0 && start <= size(), return);
 689         LASSERT(end > start && end <= size() + 1, return);
 690
 691         for (pos_type pos = start; pos < end; ++pos) {
 692                 switch (lookupChange(pos).type) {
 693                         case Change::UNCHANGED:
 694                                 // reject changes in nested inset
 695                                 if (Inset * inset = getInset(pos))
 696                                                 inset->rejectChanges();
 697                                 break;
 698
 699                         case Change::INSERTED:
 700                                 // Suppress access to non-existent
 701                                 // "end-of-paragraph char"
 702                                 if (pos < size()) {
 703                                         eraseChar(pos, false);
 704                                         --end;
 705                                         --pos;
 706                                 }
 707                                 break;
 708
 709                         case Change::DELETED:
 710                                 d->changes_.set(Change(Change::UNCHANGED), pos);
 711
 712                                 // Do NOT reject changes within a deleted inset!
 713                                 // There may be insertions of a co-author inside of it!
 714
 715                                 break;
 716                 }
 717         }
 718 }
 719
 720
 721 void Paragraph::Private::insertChar(pos_type pos, char_type c,
 722                 Change const & change)
 723 {
 724         LASSERT(pos >= 0 && pos <= int(text_.size()), return);
 725
 726         // track change
 727         changes_.insert(change, pos);
 728
 729         // This is actually very common when parsing buffers (and
 730         // maybe inserting ascii text)
 731         if (pos == pos_type(text_.size())) {
 732                 // when appending characters, no need to update tables
 733                 text_.push_back(c);
 734                 // but we want spell checking
 735                 requestSpellCheck(pos);
 736                 return;
 737         }
 738
 739         text_.insert(text_.begin() + pos, c);
 740
 741         // Update the font table.
 742         fontlist_.increasePosAfterPos(pos);
 743
 744         // Update the insets
 745         insetlist_.increasePosAfterPos(pos);
 746
 747         // Update list of misspelled positions
 748         speller_state_.increasePosAfterPos(pos);
 749 }
 750
 751
 752 bool Paragraph::insertInset(pos_type pos, Inset * inset,
 753                                    Font const & font, Change const & change)
 754 {
 755         LASSERT(inset, return false);
 756         LASSERT(pos >= 0 && pos <= size(), return false);
 757
 758         // Paragraph::insertInset() can be used in cut/copy/paste operation where
 759         // d->inset_owner_ is not set yet.
 760         if (d->inset_owner_ && !d->inset_owner_->insetAllowed(inset->lyxCode()))
 761                 return false;
 762
 763         d->insertChar(pos, META_INSET, change);
 764         LASSERT(d->text_[pos] == META_INSET, return false);
 765
 766         // Add a new entry in the insetlist_.
 767         d->insetlist_.insert(inset, pos);
 768
 769         // Some insets require run of spell checker
 770         requestSpellCheck(pos);
 771         setFont(pos, font);
 772         return true;
 773 }
 774
 775
 776 bool Paragraph::eraseChar(pos_type pos, bool trackChanges)
 777 {
 778         LASSERT(pos >= 0 && pos <= size(), return false);
 779
 780         // keep the logic here in sync with the logic of isMergedOnEndOfParDeletion()
 781
 782         if (trackChanges) {
 783                 Change change = d->changes_.lookup(pos);
 784
 785                 // set the character to DELETED if
 786                 //  a) it was previously unchanged or
 787                 //  b) it was inserted by a co-author
 788
 789                 if (!change.changed() ||
 790                       (change.inserted() && !change.currentAuthor())) {
 791                         setChange(pos, Change(Change::DELETED));
 792                         // request run of spell checker
 793                         requestSpellCheck(pos);
 794                         return false;
 795                 }
 796
 797                 if (change.deleted())
 798                         return false;
 799         }
 800
 801         // Don't physically access the imaginary end-of-paragraph character.
 802         // eraseChar() can only mark it as DELETED. A physical deletion of
 803         // end-of-par must be handled externally.
 804         if (pos == size()) {
 805                 return false;
 806         }
 807
 808         // track change
 809         d->changes_.erase(pos);
 810
 811         // if it is an inset, delete the inset entry
 812         if (d->text_[pos] == META_INSET)
 813                 d->insetlist_.erase(pos);
 814
 815         d->text_.erase(d->text_.begin() + pos);
 816
 817         // Update the fontlist_
 818         d->fontlist_.erase(pos);
 819
 820         // Update the insetlist_
 821         d->insetlist_.decreasePosAfterPos(pos);
 822
 823         // Update list of misspelled positions
 824         d->speller_state_.decreasePosAfterPos(pos);
 825         d->speller_state_.refreshLast(size());
 826
 827         return true;
 828 }
 829
 830
 831 int Paragraph::eraseChars(pos_type start, pos_type end, bool trackChanges)
 832 {
 833         LASSERT(start >= 0 && start <= size(), return 0);
 834         LASSERT(end >= start && end <= size() + 1, return 0);
 835
 836         pos_type i = start;
 837         for (pos_type count = end - start; count; --count) {
 838                 if (!eraseChar(i, trackChanges))
 839                         ++i;
 840         }
 841         return end - i;
 842 }
 843
 844
 845 int Paragraph::Private::latexSurrogatePair(otexstream & os, char_type c,
 846                 char_type next, OutputParams const & runparams)
 847 {
 848         // Writing next here may circumvent a possible font change between
 849         // c and next. Since next is only output if it forms a surrogate pair
 850         // with c we can ignore this:
 851         // A font change inside a surrogate pair does not make sense and is
 852         // hopefully impossible to input.
 853         // FIXME: change tracking
 854         // Is this correct WRT change tracking?
 855         Encoding const & encoding = *(runparams.encoding);
 856         docstring latex1 = encoding.latexChar(next).first;
 857         if (runparams.inIPA) {
 858                 string const tipashortcut = Encodings::TIPAShortcut(next);
 859                 if (!tipashortcut.empty()) {
 860                         latex1 = from_ascii(tipashortcut);
 861                 }
 862         }
 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 character text_[i] could not be translated
 933         // into the current latex encoding (or its latex translation has been forced,)
 934         // and it belongs to a known script.
 935         // TODO: We need \textcyr and \textgreek wrappers also for characters
 936         //       that can be encoded in the "LaTeX encoding" but not in the
 937         //       current *font encoding*.
 938         //       (See #9681 for details and test)
 939         // Parameter ltx contains the latex translation of text_[i] as specified
 940         // in the unicodesymbols file and is something like "\textXXX{<spec>}".
 941         // The latex macro name "textXXX" specifies the script to which text_[i]
 942         // belongs and we use it in order to check whether characters from the
 943         // same script immediately follow, such that we can collect them in a
 944         // single "\textXXX" macro. So, we have to retain "\textXXX{<spec>"
 945         // for the first char but only "<spec>" for all subsequent chars.
 946         docstring::size_type const brace1 = ltx.find_first_of(from_ascii("{"));
 947         docstring::size_type const brace2 = ltx.find_last_of(from_ascii("}"));
 948         string script = to_ascii(ltx.substr(1, brace1 - 1));
 949         int pos = 0;
 950         int length = brace2;
 951         bool closing_brace = true;
 952         if (script == "textgreek" && encoding.latexName() == "iso-8859-7") {
 953                 // Correct encoding is being used, so we can avoid \textgreek.
 954                 // TODO: wrong test: we need to check the *font encoding*
 955                 //       (i.e. the active language and its FontEncoding tag)
 956                 //       instead of the LaTeX *input encoding*!
 957                 //       See #9637 for details and test-cases.
 958                 pos = brace1 + 1;
 959                 length -= pos;
 960                 closing_brace = false;
 961         }
 962         os << ltx.substr(pos, length);
 963         int size = text_.size();
 964         while (i + 1 < size) {
 965                 char_type const next = text_[i + 1];
 966                 // Stop here if next character belongs to another script
 967                 // or there is a change in change tracking status.
 968                 if (!Encodings::isKnownScriptChar(next, script) ||
 969                     runningChange != owner_->lookupChange(i + 1))
 970                         break;
 971                 Font prev_font;
 972                 bool found = false;
 973                 FontList::const_iterator cit = fontlist_.begin();
 974                 FontList::const_iterator end = fontlist_.end();
 975                 for (; cit != end; ++cit) {
 976                         if (cit->pos() >= i && !found) {
 977                                 prev_font = cit->font();
 978                                 found = true;
 979                         }
 980                         if (cit->pos() >= i + 1)
 981                                 break;
 982                 }
 983                 // Stop here if there is a font attribute or encoding change.
 984                 if (found && cit != end && prev_font != cit->font())
 985                         break;
 986                 docstring const latex = encoding.latexChar(next).first;
 987                 docstring::size_type const b1 =
 988                                         latex.find_first_of(from_ascii("{"));
 989                 docstring::size_type const b2 =
 990                                         latex.find_last_of(from_ascii("}"));
 991                 int const len = b2 - b1 - 1;
 992                 os << latex.substr(b1 + 1, len);
 993                 length += len;
 994                 ++i;
 995         }
 996         if (closing_brace) {
 997                 os << '}';
 998                 ++length;
 999         }
1000         return length;
1001 }
1002
1003
1004 void Paragraph::Private::latexInset(BufferParams const & bparams,
1005                                     otexstream & os,
1006                                     OutputParams & runparams,
1007                                     Font & running_font,
1008                                     Font & basefont,
1009                                     Font const & outerfont,
1010                                     bool & open_font,
1011                                     Change & running_change,
1012                                     Layout const & style,
1013                                     pos_type & i,
1014                                     unsigned int & column)
1015 {
1016         Inset * inset = owner_->getInset(i);
1017         LBUFERR(inset);
1018
1019         if (style.pass_thru) {
1020                 odocstringstream ods;
1021                 inset->plaintext(ods, runparams);
1022                 os << ods.str();
1023                 return;
1024         }
1025
1026         // FIXME: move this to InsetNewline::latex
1027         if (inset->lyxCode() == NEWLINE_CODE || inset->lyxCode() == SEPARATOR_CODE) {
1028                 // newlines are handled differently here than
1029                 // the default in simpleTeXSpecialChars().
1030                 if (!style.newline_allowed) {
1031                         os << '\n';
1032                 } else {
1033                         if (open_font) {
1034                                 column += running_font.latexWriteEndChanges(
1035                                         os, bparams, runparams,
1036                                         basefont, basefont);
1037                                 open_font = false;
1038                         }
1039
1040                         if (running_font.fontInfo().family() == TYPEWRITER_FAMILY)
1041                                 os << '~';
1042
1043                         basefont = owner_->getLayoutFont(bparams, outerfont);
1044                         running_font = basefont;
1045
1046                         if (runparams.moving_arg)
1047                                 os << "\\protect ";
1048
1049                 }
1050                 os.texrow().start(owner_->id(), i + 1);
1051                 column = 0;
1052         }
1053
1054         if (owner_->isDeleted(i)) {
1055                 if( ++runparams.inDeletedInset == 1)
1056                         runparams.changeOfDeletedInset = owner_->lookupChange(i);
1057         }
1058
1059         if (inset->canTrackChanges()) {
1060                 column += Changes::latexMarkChange(os, bparams, running_change,
1061                         Change(Change::UNCHANGED), runparams);
1062                 running_change = Change(Change::UNCHANGED);
1063         }
1064
1065         bool close = false;
1066         odocstream::pos_type const len = os.os().tellp();
1067
1068         if (inset->forceLTR()
1069             && !runparams.use_polyglossia
1070             && running_font.isRightToLeft()
1071             // ERT is an exception, it should be output with no
1072             // decorations at all
1073             && inset->lyxCode() != ERT_CODE) {
1074                 if (running_font.language()->lang() == "farsi")
1075                         os << "\\beginL{}";
1076                 else
1077                         os << "\\L{";
1078                 close = true;
1079         }
1080
1081         // FIXME: Bug: we can have an empty font change here!
1082         // if there has just been a font change, we are going to close it
1083         // right now, which means stupid latex code like \textsf{}. AFAIK,
1084         // this does not harm dvi output. A minor bug, thus (JMarc)
1085
1086         // Some insets cannot be inside a font change command.
1087         // However, even such insets *can* be placed in \L or \R
1088         // or their equivalents (for RTL language switches), so we don't
1089         // close the language in those cases.
1090         // ArabTeX, though, cannot handle this special behavior, it seems.
1091         bool arabtex = basefont.language()->lang() == "arabic_arabtex"
1092                 || running_font.language()->lang() == "arabic_arabtex";
1093         if (open_font && !inset->inheritFont()) {
1094                 bool closeLanguage = arabtex
1095                         || basefont.isRightToLeft() == running_font.isRightToLeft();
1096                 unsigned int count = running_font.latexWriteEndChanges(os,
1097                         bparams, runparams, basefont, basefont, closeLanguage);
1098                 column += count;
1099                 // if any font properties were closed, update the running_font,
1100                 // making sure, however, to leave the language as it was
1101                 if (count > 0) {
1102                         // FIXME: probably a better way to keep track of the old
1103                         // language, than copying the entire font?
1104                         Font const copy_font(running_font);
1105                         basefont = owner_->getLayoutFont(bparams, outerfont);
1106                         running_font = basefont;
1107                         if (!closeLanguage)
1108                                 running_font.setLanguage(copy_font.language());
1109                         // leave font open if language is still open
1110                         open_font = (running_font.language() == basefont.language());
1111                         if (closeLanguage)
1112                                 runparams.local_font = &basefont;
1113                 }
1114         }
1115
1116         int prev_rows = os.texrow().rows();
1117
1118         try {
1119                 runparams.lastid = id_;
1120                 runparams.lastpos = i;
1121                 inset->latex(os, runparams);
1122         } catch (EncodingException & e) {
1123                 // add location information and throw again.
1124                 e.par_id = id_;
1125                 e.pos = i;
1126                 throw(e);
1127         }
1128
1129         if (close) {
1130                 if (running_font.language()->lang() == "farsi")
1131                                 os << "\\endL{}";
1132                         else
1133                                 os << '}';
1134         }
1135
1136         if (os.texrow().rows() > prev_rows) {
1137                 os.texrow().start(owner_->id(), i + 1);
1138                 column = 0;
1139         } else {
1140                 column += (unsigned int)(os.os().tellp() - len);
1141         }
1142
1143         if (owner_->isDeleted(i))
1144                 --runparams.inDeletedInset;
1145 }
1146
1147
1148 void Paragraph::Private::latexSpecialChar(otexstream & os,
1149                                           BufferParams const & bparams,
1150                                           OutputParams const & runparams,
1151                                           Font const & running_font,
1152                                           Change const & running_change,
1153                                           Layout const & style,
1154                                           pos_type & i,
1155                                           pos_type end_pos,
1156                                           unsigned int & column)
1157 {
1158         // With polyglossia, brackets and stuff need not be reversed
1159         // in RTL scripts (see bug #8251)
1160         char_type const c = (runparams.use_polyglossia) ?
1161                 owner_->getUChar(bparams, i) : text_[i];
1162
1163         if (style.pass_thru || runparams.pass_thru
1164             || contains(style.pass_thru_chars, c)
1165             || contains(runparams.pass_thru_chars, c)) {
1166                 if (c != '\0') {
1167                         Encoding const * const enc = runparams.encoding;
1168                         if (enc && !enc->encodable(c))
1169                                 throw EncodingException(c);
1170                         os.put(c);
1171                 }
1172                 return;
1173         }
1174
1175         // TIPA uses its own T3 encoding
1176         if (runparams.inIPA && latexSpecialT3(c, os, i, column))
1177                 return;
1178         // If T1 font encoding is used, use the special
1179         // characters it provides.
1180         // NOTE: Some languages reset the font encoding internally to a
1181         //       non-standard font encoding. If we are using such a language,
1182         //       we do not output special T1 chars.
1183         if (!runparams.inIPA && !running_font.language()->internalFontEncoding()
1184             && bparams.font_encoding() == "T1" && latexSpecialT1(c, os, i, column))
1185                 return;
1186
1187         // Otherwise, we use what LaTeX provides us.
1188         switch (c) {
1189         case '\\':
1190                 os << "\\textbackslash{}";
1191                 column += 15;
1192                 break;
1193         case '<':
1194                 os << "\\textless{}";
1195                 column += 10;
1196                 break;
1197         case '>':
1198                 os << "\\textgreater{}";
1199                 column += 13;
1200                 break;
1201         case '|':
1202                 os << "\\textbar{}";
1203                 column += 9;
1204                 break;
1205         case '-':
1206                 os << '-';
1207                 if (i + 1 < static_cast<pos_type>(text_.size()) &&
1208                     (end_pos == -1 || i + 1 < end_pos) &&
1209                     text_[i+1] == '-') {
1210                         // Prevent "--" becoming an endash and "---" becoming
1211                         // an emdash.
1212                         // Within \ttfamily, "--" is merged to "-" (no endash)
1213                         // so we avoid this rather irritating ligature as well
1214                         os << "{}";
1215                         column += 2;
1216                 }
1217                 break;
1218         case '\"':
1219                 os << "\\char`\\\"{}";
1220                 column += 9;
1221                 break;
1222
1223         case '$': case '&':
1224         case '%': case '#': case '{':
1225         case '}': case '_':
1226                 os << '\\';
1227                 os.put(c);
1228                 column += 1;
1229                 break;
1230
1231         case '~':
1232                 os << "\\textasciitilde{}";
1233                 column += 16;
1234                 break;
1235
1236         case '^':
1237                 os << "\\textasciicircum{}";
1238                 column += 17;
1239                 break;
1240
1241         case '*':
1242         case '[':
1243         case ']':
1244                 // avoid being mistaken for optional arguments
1245                 os << '{';
1246                 os.put(c);
1247                 os << '}';
1248                 column += 2;
1249                 break;
1250
1251         case ' ':
1252                 // Blanks are printed before font switching.
1253                 // Sure? I am not! (try nice-latex)
1254                 // I am sure it's correct. LyX might be smarter
1255                 // in the future, but for now, nothing wrong is
1256                 // written. (Asger)
1257                 break;
1258
1259         default:
1260                 if (c == '\0')
1261                         return;
1262
1263                 Encoding const & encoding = *(runparams.encoding);
1264                 char_type next = '\0';
1265                 if (i + 1 < int(text_.size())) {
1266                         next = text_[i + 1];
1267                         if (Encodings::isCombiningChar(next)) {
1268                                 column += latexSurrogatePair(os, c, next, runparams) - 1;
1269                                 ++i;
1270                                 break;
1271                         }
1272                 }
1273                 string script;
1274                 pair<docstring, bool> latex = encoding.latexChar(c);
1275                 docstring nextlatex;
1276                 bool nexttipas = false;
1277                 string nexttipashortcut;
1278                 if (next != '\0' && next != META_INSET && encoding.encodable(next)) {
1279                         nextlatex = encoding.latexChar(next).first;
1280                         if (runparams.inIPA) {
1281                                 nexttipashortcut = Encodings::TIPAShortcut(next);
1282                                 nexttipas = !nexttipashortcut.empty();
1283                         }
1284                 }
1285                 bool tipas = false;
1286                 if (runparams.inIPA) {
1287                         string const tipashortcut = Encodings::TIPAShortcut(c);
1288                         if (!tipashortcut.empty()) {
1289                                 latex.first = from_ascii(tipashortcut);
1290                                 latex.second = false;
1291                                 tipas = true;
1292                         }
1293                 }
1294                 if (Encodings::isKnownScriptChar(c, script)
1295                     && prefixIs(latex.first, from_ascii("\\" + script)))
1296                         column += writeScriptChars(os, latex.first,
1297                                         running_change, encoding, i) - 1;
1298                 else if (latex.second
1299                          && ((!prefixIs(nextlatex, '\\')
1300                                && !prefixIs(nextlatex, '{')
1301                                && !prefixIs(nextlatex, '}'))
1302                              || (nexttipas
1303                                  && !prefixIs(from_ascii(nexttipashortcut), '\\')))
1304                          && !tipas) {
1305                         // Prevent eating of a following
1306                         // space or command corruption by
1307                         // following characters
1308                         if (next == ' ' || next == '\0') {
1309                                 column += latex.first.length() + 1;
1310                                 os << latex.first << "{}";
1311                         } else {
1312                                 column += latex.first.length();
1313                                 os << latex.first << " ";
1314                         }
1315                 } else {
1316                         column += latex.first.length() - 1;
1317                         os << latex.first;
1318                 }
1319                 break;
1320         }
1321 }
1322
1323
1324 bool Paragraph::Private::latexSpecialT1(char_type const c, otexstream & os,
1325         pos_type i, unsigned int & column)
1326 {
1327         switch (c) {
1328         case '>':
1329         case '<':
1330                 os.put(c);
1331                 // In T1 encoding, these characters exist
1332                 // but we should avoid ligatures
1333                 if (i + 1 >= int(text_.size()) || text_[i + 1] != c)
1334                         return true;
1335                 os << "\\textcompwordmark{}";
1336                 column += 19;
1337                 return true;
1338         case '|':
1339                 os.put(c);
1340                 return true;
1341         case '\"':
1342                 // soul.sty breaks with \char`\"
1343                 os << "\\textquotedbl{}";
1344                 column += 14;
1345                 return true;
1346         default:
1347                 return false;
1348         }
1349 }
1350
1351
1352 bool Paragraph::Private::latexSpecialT3(char_type const c, otexstream & os,
1353         pos_type /*i*/, unsigned int & column)
1354 {
1355         switch (c) {
1356         case '*':
1357         case '[':
1358         case ']':
1359         case '\"':
1360                 os.put(c);
1361                 return true;
1362         case '|':
1363                 os << "\\textvertline{}";
1364                 column += 14;
1365                 return true;
1366         default:
1367                 return false;
1368         }
1369 }
1370
1371
1372 void Paragraph::Private::validate(LaTeXFeatures & features) const
1373 {
1374         if (layout_->inpreamble && inset_owner_) {
1375                 bool const is_command = layout_->latextype == LATEX_COMMAND;
1376                 Buffer const & buf = inset_owner_->buffer();
1377                 BufferParams const & bp = features.runparams().is_child
1378                         ? buf.masterParams() : buf.params();
1379                 Font f;
1380                 TexRow texrow;
1381                 // Using a string stream here circumvents the encoding
1382                 // switching machinery of odocstream. Therefore the
1383                 // output is wrong if this paragraph contains content
1384                 // that needs to switch encoding.
1385                 odocstringstream ods;
1386                 otexstream os(ods, texrow);
1387                 if (is_command) {
1388                         os << '\\' << from_ascii(layout_->latexname());
1389                         // we have to provide all the optional arguments here, even though
1390                         // the last one is the only one we care about.
1391                         // Separate handling of optional argument inset.
1392                         if (!layout_->latexargs().empty()) {
1393                                 OutputParams rp = features.runparams();
1394                                 rp.local_font = &owner_->getFirstFontSettings(bp);
1395                                 latexArgInsets(*owner_, os, rp, layout_->latexargs());
1396                         }
1397                         os << from_ascii(layout_->latexparam());
1398                 }
1399                 docstring::size_type const length = ods.str().length();
1400                 // this will output "{" at the beginning, but not at the end
1401                 owner_->latex(bp, f, os, features.runparams(), 0, -1, true);
1402                 if (ods.str().length() > length) {
1403                         if (is_command) {
1404                                 ods << '}';
1405                                 if (!layout_->postcommandargs().empty()) {
1406                                         OutputParams rp = features.runparams();
1407                                         rp.local_font = &owner_->getFirstFontSettings(bp);
1408                                         latexArgInsets(*owner_, os, rp, layout_->postcommandargs(), "post:");
1409                                 }
1410                         }
1411                         string const snippet = to_utf8(ods.str());
1412                         features.addPreambleSnippet(snippet, true);
1413                 }
1414         }
1415
1416         if (features.runparams().flavor == OutputParams::HTML
1417             && layout_->htmltitle()) {
1418                 features.setHTMLTitle(owner_->asString(AS_STR_INSETS | AS_STR_SKIPDELETE));
1419         }
1420
1421         // check the params.
1422         if (!params_.spacing().isDefault())
1423                 features.require("setspace");
1424
1425         // then the layouts
1426         features.useLayout(layout_->name());
1427
1428         // then the fonts
1429         fontlist_.validate(features);
1430
1431         // then the indentation
1432         if (!params_.leftIndent().zero())
1433                 features.require("ParagraphLeftIndent");
1434
1435         // then the insets
1436         InsetList::const_iterator icit = insetlist_.begin();
1437         InsetList::const_iterator iend = insetlist_.end();
1438         for (; icit != iend; ++icit) {
1439                 if (icit->inset) {
1440                         icit->inset->validate(features);
1441                         if (layout_->needprotect &&
1442                             icit->inset->lyxCode() == FOOT_CODE)
1443                                 features.require("NeedLyXFootnoteCode");
1444                 }
1445         }
1446
1447         // then the contents
1448         for (pos_type i = 0; i < int(text_.size()) ; ++i) {
1449                 BufferEncodings::validate(text_[i], features);
1450         }
1451 }
1452
1453 /////////////////////////////////////////////////////////////////////
1454 //
1455 // Paragraph
1456 //
1457 /////////////////////////////////////////////////////////////////////
1458
1459 namespace {
1460         Layout const emptyParagraphLayout;
1461 }
1462
1463 Paragraph::Paragraph()
1464         : d(new Paragraph::Private(this, emptyParagraphLayout))
1465 {
1466         itemdepth = 0;
1467         d->params_.clear();
1468 }
1469
1470
1471 Paragraph::Paragraph(Paragraph const & par)
1472         : itemdepth(par.itemdepth),
1473         d(new Paragraph::Private(*par.d, this))
1474 {
1475         registerWords();
1476 }
1477
1478
1479 Paragraph::Paragraph(Paragraph const & par, pos_type beg, pos_type end)
1480         : itemdepth(par.itemdepth),
1481         d(new Paragraph::Private(*par.d, this, beg, end))
1482 {
1483         registerWords();
1484 }
1485
1486
1487 Paragraph & Paragraph::operator=(Paragraph const & par)
1488 {
1489         // needed as we will destroy the private part before copying it
1490         if (&par != this) {
1491                 itemdepth = par.itemdepth;
1492
1493                 deregisterWords();
1494                 delete d;
1495                 d = new Private(*par.d, this);
1496                 registerWords();
1497         }
1498         return *this;
1499 }
1500
1501
1502 Paragraph::~Paragraph()
1503 {
1504         deregisterWords();
1505         delete d;
1506 }
1507
1508
1509 namespace {
1510
1511 // this shall be called just before every "os << ..." action.
1512 void flushString(ostream & os, docstring & s)
1513 {
1514         os << to_utf8(s);
1515         s.erase();
1516 }
1517
1518 }
1519
1520
1521 void Paragraph::write(ostream & os, BufferParams const & bparams,
1522         depth_type & dth) const
1523 {
1524         // The beginning or end of a deeper (i.e. nested) area?
1525         if (dth != d->params_.depth()) {
1526                 if (d->params_.depth() > dth) {
1527                         while (d->params_.depth() > dth) {
1528                                 os << "\n\\begin_deeper";
1529                                 ++dth;
1530                         }
1531                 } else {
1532                         while (d->params_.depth() < dth) {
1533                                 os << "\n\\end_deeper";
1534                                 --dth;
1535                         }
1536                 }
1537         }
1538
1539         // First write the layout
1540         os << "\n\\begin_layout " << to_utf8(d->layout_->name()) << '\n';
1541
1542         d->params_.write(os);
1543
1544         Font font1(inherit_font, bparams.language);
1545
1546         Change running_change = Change(Change::UNCHANGED);
1547
1548         // this string is used as a buffer to avoid repetitive calls
1549         // to to_utf8(), which turn out to be expensive (JMarc)
1550         docstring write_buffer;
1551
1552         int column = 0;
1553         for (pos_type i = 0; i <= size(); ++i) {
1554
1555                 Change const & change = lookupChange(i);
1556                 if (change != running_change)
1557                         flushString(os, write_buffer);
1558                 Changes::lyxMarkChange(os, bparams, column, running_change, change);
1559                 running_change = change;
1560
1561                 if (i == size())
1562                         break;
1563
1564                 // Write font changes
1565                 Font font2 = getFontSettings(bparams, i);
1566                 if (font2 != font1) {
1567                         flushString(os, write_buffer);
1568                         font2.lyxWriteChanges(font1, os);
1569                         column = 0;
1570                         font1 = font2;
1571                 }
1572
1573                 char_type const c = d->text_[i];
1574                 switch (c) {
1575                 case META_INSET:
1576                         if (Inset const * inset = getInset(i)) {
1577                                 flushString(os, write_buffer);
1578                                 if (inset->directWrite()) {
1579                                         // international char, let it write
1580                                         // code directly so it's shorter in
1581                                         // the file
1582                                         inset->write(os);
1583                                 } else {
1584                                         if (i)
1585                                                 os << '\n';
1586                                         os << "\\begin_inset ";
1587                                         inset->write(os);
1588                                         os << "\n\\end_inset\n\n";
1589                                         column = 0;
1590                                 }
1591                                 // FIXME This can be removed again once the mystery
1592                                 // crash has been resolved.
1593                                 os << flush;
1594                         }
1595                         break;
1596                 case '\\':
1597                         flushString(os, write_buffer);
1598                         os << "\n\\backslash\n";
1599                         column = 0;
1600                         break;
1601                 case '.':
1602                         flushString(os, write_buffer);
1603                         if (i + 1 < size() && d->text_[i + 1] == ' ') {
1604                                 os << ".\n";
1605                                 column = 0;
1606                         } else
1607                                 os << '.';
1608                         break;
1609                 default:
1610                         if ((column > 70 && c == ' ')
1611                             || column > 79) {
1612                                 flushString(os, write_buffer);
1613                                 os << '\n';
1614                                 column = 0;
1615                         }
1616                         // this check is to amend a bug. LyX sometimes
1617                         // inserts '\0' this could cause problems.
1618                         if (c != '\0')
1619                                 write_buffer.push_back(c);
1620                         else
1621                                 LYXERR0("NUL char in structure.");
1622                         ++column;
1623                         break;
1624                 }
1625         }
1626
1627         flushString(os, write_buffer);
1628         os << "\n\\end_layout\n";
1629         // FIXME This can be removed again once the mystery
1630         // crash has been resolved.
1631         os << flush;
1632 }
1633
1634
1635 void Paragraph::validate(LaTeXFeatures & features) const
1636 {
1637         d->validate(features);
1638 }
1639
1640
1641 void Paragraph::insert(pos_type start, docstring const & str,
1642                        Font const & font, Change const & change)
1643 {
1644         for (size_t i = 0, n = str.size(); i != n ; ++i)
1645                 insertChar(start + i, str[i], font, change);
1646 }
1647
1648
1649 void Paragraph::appendChar(char_type c, Font const & font,
1650                 Change const & change)
1651 {
1652         // track change
1653         d->changes_.insert(change, d->text_.size());
1654         // when appending characters, no need to update tables
1655         d->text_.push_back(c);
1656         setFont(d->text_.size() - 1, font);
1657         d->requestSpellCheck(d->text_.size() - 1);
1658 }
1659
1660
1661 void Paragraph::appendString(docstring const & s, Font const & font,
1662                 Change const & change)
1663 {
1664         pos_type end = s.size();
1665         size_t oldsize = d->text_.size();
1666         size_t newsize = oldsize + end;
1667         size_t capacity = d->text_.capacity();
1668         if (newsize >= capacity)
1669                 d->text_.reserve(max(capacity + 100, newsize));
1670
1671         // when appending characters, no need to update tables
1672         d->text_.append(s);
1673
1674         // FIXME: Optimize this!
1675         for (size_t i = oldsize; i != newsize; ++i) {
1676                 // track change
1677                 d->changes_.insert(change, i);
1678                 d->requestSpellCheck(i);
1679         }
1680         d->fontlist_.set(oldsize, font);
1681         d->fontlist_.set(newsize - 1, font);
1682 }
1683
1684
1685 void Paragraph::insertChar(pos_type pos, char_type c,
1686                            bool trackChanges)
1687 {
1688         d->insertChar(pos, c, Change(trackChanges ?
1689                            Change::INSERTED : Change::UNCHANGED));
1690 }
1691
1692
1693 void Paragraph::insertChar(pos_type pos, char_type c,
1694                            Font const & font, bool trackChanges)
1695 {
1696         d->insertChar(pos, c, Change(trackChanges ?
1697                            Change::INSERTED : Change::UNCHANGED));
1698         setFont(pos, font);
1699 }
1700
1701
1702 void Paragraph::insertChar(pos_type pos, char_type c,
1703                            Font const & font, Change const & change)
1704 {
1705         d->insertChar(pos, c, change);
1706         setFont(pos, font);
1707 }
1708
1709
1710 void Paragraph::resetFonts(Font const & font)
1711 {
1712         d->fontlist_.clear();
1713         d->fontlist_.set(0, font);
1714         d->fontlist_.set(d->text_.size() - 1, font);
1715 }
1716
1717 // Gets uninstantiated font setting at position.
1718 Font const & Paragraph::getFontSettings(BufferParams const & bparams,
1719                                          pos_type pos) const
1720 {
1721         if (pos > size()) {
1722                 LYXERR0("pos: " << pos << " size: " << size());
1723                 LBUFERR(false);
1724         }
1725
1726         FontList::const_iterator cit = d->fontlist_.fontIterator(pos);
1727         if (cit != d->fontlist_.end())
1728                 return cit->font();
1729
1730         if (pos == size() && !empty())
1731                 return getFontSettings(bparams, pos - 1);
1732
1733         // Optimisation: avoid a full font instantiation if there is no
1734         // language change from previous call.
1735         static Font previous_font;
1736         static Language const * previous_lang = 0;
1737         Language const * lang = getParLanguage(bparams);
1738         if (lang != previous_lang) {
1739                 previous_lang = lang;
1740                 previous_font = Font(inherit_font, lang);
1741         }
1742         return previous_font;
1743 }
1744
1745
1746 FontSpan Paragraph::fontSpan(pos_type pos) const
1747 {
1748         LBUFERR(pos <= size());
1749
1750         if (pos == size())
1751                 return FontSpan(pos, pos);
1752
1753         pos_type start = 0;
1754         FontList::const_iterator cit = d->fontlist_.begin();
1755         FontList::const_iterator end = d->fontlist_.end();
1756         for (; cit != end; ++cit) {
1757                 if (cit->pos() >= pos) {
1758                         if (pos >= beginOfBody())
1759                                 return FontSpan(max(start, beginOfBody()),
1760                                                 cit->pos());
1761                         else
1762                                 return FontSpan(start,
1763                                                 min(beginOfBody() - 1,
1764                                                          cit->pos()));
1765                 }
1766                 start = cit->pos() + 1;
1767         }
1768
1769         // This should not happen, but if so, we take no chances.
1770         LYXERR0("Paragraph::fontSpan: position not found in fontinfo table!");
1771         LASSERT(false, return FontSpan(pos, pos));
1772 }
1773
1774
1775 // Gets uninstantiated font setting at position 0
1776 Font const & Paragraph::getFirstFontSettings(BufferParams const & bparams) const
1777 {
1778         if (!empty() && !d->fontlist_.empty())
1779                 return d->fontlist_.begin()->font();
1780
1781         // Optimisation: avoid a full font instantiation if there is no
1782         // language change from previous call.
1783         static Font previous_font;
1784         static Language const * previous_lang = 0;
1785         if (bparams.language != previous_lang) {
1786                 previous_lang = bparams.language;
1787                 previous_font = Font(inherit_font, bparams.language);
1788         }
1789
1790         return previous_font;
1791 }
1792
1793
1794 // Gets the fully instantiated font at a given position in a paragraph
1795 // This is basically the same function as Text::GetFont() in text2.cpp.
1796 // The difference is that this one is used for generating the LaTeX file,
1797 // and thus cosmetic "improvements" are disallowed: This has to deliver
1798 // the true picture of the buffer. (Asger)
1799 Font const Paragraph::getFont(BufferParams const & bparams, pos_type pos,
1800                                  Font const & outerfont) const
1801 {
1802         LBUFERR(pos >= 0);
1803
1804         Font font = getFontSettings(bparams, pos);
1805
1806         pos_type const body_pos = beginOfBody();
1807         FontInfo & fi = font.fontInfo();
1808         if (pos < body_pos)
1809                 fi.realize(d->layout_->labelfont);
1810         else
1811                 fi.realize(d->layout_->font);
1812
1813         fi.realize(outerfont.fontInfo());
1814         fi.realize(bparams.getFont().fontInfo());
1815
1816         return font;
1817 }
1818
1819
1820 Font const Paragraph::getLabelFont
1821         (BufferParams const & bparams, Font const & outerfont) const
1822 {
1823         FontInfo tmpfont = d->layout_->labelfont;
1824         tmpfont.realize(outerfont.fontInfo());
1825         tmpfont.realize(bparams.getFont().fontInfo());
1826         return Font(tmpfont, getParLanguage(bparams));
1827 }
1828
1829
1830 Font const Paragraph::getLayoutFont
1831         (BufferParams const & bparams, Font const & outerfont) const
1832 {
1833         FontInfo tmpfont = d->layout_->font;
1834         tmpfont.realize(outerfont.fontInfo());
1835         tmpfont.realize(bparams.getFont().fontInfo());
1836         return Font(tmpfont, getParLanguage(bparams));
1837 }
1838
1839
1840 char_type Paragraph::getUChar(BufferParams const & bparams, pos_type pos) const
1841 {
1842         char_type c = d->text_[pos];
1843         if (!getFontSettings(bparams, pos).isRightToLeft())
1844                 return c;
1845
1846         // FIXME: The arabic special casing is due to the difference of arabic
1847         // round brackets input introduced in r18599. Check if this should be
1848         // unified with Hebrew or at least if all bracket types should be
1849         // handled the same (file format change in either case).
1850         string const & lang = getFontSettings(bparams, pos).language()->lang();
1851         bool const arabic = lang == "arabic_arabtex" || lang == "arabic_arabi"
1852                 || lang == "farsi";
1853         char_type uc = c;
1854         switch (c) {
1855         case '(':
1856                 uc = arabic ? c : ')';
1857                 break;
1858         case ')':
1859                 uc = arabic ? c : '(';
1860                 break;
1861         case '[':
1862                 uc = ']';
1863                 break;
1864         case ']':
1865                 uc = '[';
1866                 break;
1867         case '{':
1868                 uc = '}';
1869                 break;
1870         case '}':
1871                 uc = '{';
1872                 break;
1873         case '<':
1874                 uc = '>';
1875                 break;
1876         case '>':
1877                 uc = '<';
1878                 break;
1879         }
1880
1881         return uc;
1882 }
1883
1884
1885 void Paragraph::setFont(pos_type pos, Font const & font)
1886 {
1887         LASSERT(pos <= size(), return);
1888
1889         // First, reduce font against layout/label font
1890         // Update: The setCharFont() routine in text2.cpp already
1891         // reduces font, so we don't need to do that here. (Asger)
1892
1893         d->fontlist_.set(pos, font);
1894 }
1895
1896
1897 void Paragraph::makeSameLayout(Paragraph const & par)
1898 {
1899         d->layout_ = par.d->layout_;
1900         d->params_ = par.d->params_;
1901 }
1902
1903
1904 bool Paragraph::stripLeadingSpaces(bool trackChanges)
1905 {
1906         if (isFreeSpacing())
1907                 return false;
1908
1909         int pos = 0;
1910         int count = 0;
1911
1912         while (pos < size() && (isNewline(pos) || isLineSeparator(pos))) {
1913                 if (eraseChar(pos, trackChanges))
1914                         ++count;
1915                 else
1916                         ++pos;
1917         }
1918
1919         return count > 0 || pos > 0;
1920 }
1921
1922
1923 bool Paragraph::hasSameLayout(Paragraph const & par) const
1924 {
1925         return par.d->layout_ == d->layout_
1926                 && d->params_.sameLayout(par.d->params_);
1927 }
1928
1929
1930 depth_type Paragraph::getDepth() const
1931 {
1932         return d->params_.depth();
1933 }
1934
1935
1936 depth_type Paragraph::getMaxDepthAfter() const
1937 {
1938         if (d->layout_->isEnvironment())
1939                 return d->params_.depth() + 1;
1940         else
1941                 return d->params_.depth();
1942 }
1943
1944
1945 char Paragraph::getAlign() const
1946 {
1947         if (d->params_.align() == LYX_ALIGN_LAYOUT)
1948                 return d->layout_->align;
1949         else
1950                 return d->params_.align();
1951 }
1952
1953
1954 docstring const & Paragraph::labelString() const
1955 {
1956         return d->params_.labelString();
1957 }
1958
1959
1960 // the next two functions are for the manual labels
1961 docstring const Paragraph::getLabelWidthString() const
1962 {
1963         if (d->layout_->margintype == MARGIN_MANUAL
1964             || d->layout_->latextype == LATEX_BIB_ENVIRONMENT)
1965                 return d->params_.labelWidthString();
1966         else
1967                 return _("Senseless with this layout!");
1968 }
1969
1970
1971 void Paragraph::setLabelWidthString(docstring const & s)
1972 {
1973         d->params_.labelWidthString(s);
1974 }
1975
1976
1977 docstring Paragraph::expandLabel(Layout const & layout,
1978                 BufferParams const & bparams) const
1979 {
1980         return expandParagraphLabel(layout, bparams, true);
1981 }
1982
1983
1984 docstring Paragraph::expandDocBookLabel(Layout const & layout,
1985                 BufferParams const & bparams) const
1986 {
1987         return expandParagraphLabel(layout, bparams, false);
1988 }
1989
1990
1991 docstring Paragraph::expandParagraphLabel(Layout const & layout,
1992                 BufferParams const & bparams, bool process_appendix) const
1993 {
1994         DocumentClass const & tclass = bparams.documentClass();
1995         string const & lang = getParLanguage(bparams)->code();
1996         bool const in_appendix = process_appendix && d->params_.appendix();
1997         docstring fmt = translateIfPossible(layout.labelstring(in_appendix), lang);
1998
1999         if (fmt.empty() && !layout.counter.empty())
2000                 return tclass.counters().theCounter(layout.counter, lang);
2001
2002         // handle 'inherited level parts' in 'fmt',
2003         // i.e. the stuff between '@' in   '@Section@.\arabic{subsection}'
2004         size_t const i = fmt.find('@', 0);
2005         if (i != docstring::npos) {
2006                 size_t const j = fmt.find('@', i + 1);
2007                 if (j != docstring::npos) {
2008                         docstring parent(fmt, i + 1, j - i - 1);
2009                         docstring label = from_ascii("??");
2010                         if (tclass.hasLayout(parent))
2011                                 label = expandParagraphLabel(tclass[parent], bparams,
2012                                                       process_appendix);
2013                         fmt = docstring(fmt, 0, i) + label
2014                                 + docstring(fmt, j + 1, docstring::npos);
2015                 }
2016         }
2017
2018         return tclass.counters().counterLabel(fmt, lang);
2019 }
2020
2021
2022 void Paragraph::applyLayout(Layout const & new_layout)
2023 {
2024         d->layout_ = &new_layout;
2025         LyXAlignment const oldAlign = d->params_.align();
2026
2027         if (!(oldAlign & d->layout_->alignpossible)) {
2028                 frontend::Alert::warning(_("Alignment not permitted"),
2029                         _("The new layout does not permit the alignment previously used.\nSetting to default."));
2030                 d->params_.align(LYX_ALIGN_LAYOUT);
2031         }
2032 }
2033
2034
2035 pos_type Paragraph::beginOfBody() const
2036 {
2037         return d->begin_of_body_;
2038 }
2039
2040
2041 void Paragraph::setBeginOfBody()
2042 {
2043         if (d->layout_->labeltype != LABEL_MANUAL) {
2044                 d->begin_of_body_ = 0;
2045                 return;
2046         }
2047
2048         // Unroll the first two cycles of the loop
2049         // and remember the previous character to
2050         // remove unnecessary getChar() calls
2051         pos_type i = 0;
2052         pos_type end = size();
2053         if (i < end && !(isNewline(i) || isEnvSeparator(i))) {
2054                 ++i;
2055                 if (i < end) {
2056                         char_type previous_char = d->text_[i];
2057                         if (!(isNewline(i) || isEnvSeparator(i))) {
2058                                 ++i;
2059                                 while (i < end && previous_char != ' ') {
2060                                         char_type temp = d->text_[i];
2061                                         if (isNewline(i) || isEnvSeparator(i))
2062                                                 break;
2063                                         ++i;
2064                                         previous_char = temp;
2065                                 }
2066                         }
2067                 }
2068         }
2069
2070         d->begin_of_body_ = i;
2071 }
2072
2073
2074 bool Paragraph::allowParagraphCustomization() const
2075 {
2076         return inInset().allowParagraphCustomization();
2077 }
2078
2079
2080 bool Paragraph::usePlainLayout() const
2081 {
2082         return inInset().usePlainLayout();
2083 }
2084
2085
2086 bool Paragraph::isPassThru() const
2087 {
2088         return inInset().isPassThru() || d->layout_->pass_thru;
2089 }
2090
2091 namespace {
2092
2093 // paragraphs inside floats need different alignment tags to avoid
2094 // unwanted space
2095
2096 bool noTrivlistCentering(InsetCode code)
2097 {
2098         return code == FLOAT_CODE
2099                || code == WRAP_CODE
2100                || code == CELL_CODE;
2101 }
2102
2103
2104 string correction(string const & orig)
2105 {
2106         if (orig == "flushleft")
2107                 return "raggedright";
2108         if (orig == "flushright")
2109                 return "raggedleft";
2110         if (orig == "center")
2111                 return "centering";
2112         return orig;
2113 }
2114
2115
2116 bool corrected_env(otexstream & os, string const & suffix, string const & env,
2117         InsetCode code, bool const lastpar, int & col)
2118 {
2119         string macro = suffix + "{";
2120         if (noTrivlistCentering(code)) {
2121                 if (lastpar) {
2122                         // the last paragraph in non-trivlist-aligned
2123                         // context is special (to avoid unwanted whitespace)
2124                         if (suffix == "\\begin") {
2125                                 macro = "\\" + correction(env) + "{}";
2126                                 os << from_ascii(macro);
2127                                 col += macro.size();
2128                                 return true;
2129                         }
2130                         return false;
2131                 }
2132                 macro += correction(env);
2133         } else
2134                 macro += env;
2135         macro += "}";
2136         if (suffix == "\\par\\end") {
2137                 os << breakln;
2138                 col = 0;
2139         }
2140         os << from_ascii(macro);
2141         col += macro.size();
2142         if (suffix == "\\begin") {
2143                 os << breakln;
2144                 col = 0;
2145         }
2146         return true;
2147 }
2148
2149 } // namespace anon
2150
2151
2152 int Paragraph::Private::startTeXParParams(BufferParams const & bparams,
2153                         otexstream & os, OutputParams const & runparams) const
2154 {
2155         int column = 0;
2156
2157         bool canindent =
2158                 (bparams.paragraph_separation == BufferParams::ParagraphIndentSeparation) ?
2159                         (layout_->toggle_indent != ITOGGLE_NEVER) :
2160                         (layout_->toggle_indent == ITOGGLE_ALWAYS);
2161
2162         if (canindent && params_.noindent() && !layout_->pass_thru) {
2163                 os << "\\noindent ";
2164                 column += 10;
2165         }
2166
2167         LyXAlignment const curAlign = params_.align();
2168
2169         if (curAlign == layout_->align)
2170                 return column;
2171
2172         switch (curAlign) {
2173         case LYX_ALIGN_NONE:
2174         case LYX_ALIGN_BLOCK:
2175         case LYX_ALIGN_LAYOUT:
2176         case LYX_ALIGN_SPECIAL:
2177         case LYX_ALIGN_DECIMAL:
2178                 break;
2179         case LYX_ALIGN_LEFT:
2180         case LYX_ALIGN_RIGHT:
2181         case LYX_ALIGN_CENTER:
2182                 if (runparams.moving_arg) {
2183                         os << "\\protect";
2184                         column += 8;
2185                 }
2186                 break;
2187         }
2188
2189         string const begin_tag = "\\begin";
2190         InsetCode code = ownerCode();
2191         bool const lastpar = runparams.isLastPar;
2192
2193         switch (curAlign) {
2194         case LYX_ALIGN_NONE:
2195         case LYX_ALIGN_BLOCK:
2196         case LYX_ALIGN_LAYOUT:
2197         case LYX_ALIGN_SPECIAL:
2198         case LYX_ALIGN_DECIMAL:
2199                 break;
2200         case LYX_ALIGN_LEFT: {
2201                 if (owner_->getParLanguage(bparams)->babel() != "hebrew")
2202                         corrected_env(os, begin_tag, "flushleft", code, lastpar, column);
2203                 else
2204                         corrected_env(os, begin_tag, "flushright", code, lastpar, column);
2205                 break;
2206         } case LYX_ALIGN_RIGHT: {
2207                 if (owner_->getParLanguage(bparams)->babel() != "hebrew")
2208                         corrected_env(os, begin_tag, "flushright", code, lastpar, column);
2209                 else
2210                         corrected_env(os, begin_tag, "flushleft", code, lastpar, column);
2211                 break;
2212         } case LYX_ALIGN_CENTER: {
2213                 corrected_env(os, begin_tag, "center", code, lastpar, column);
2214                 break;
2215         }
2216         }
2217
2218         return column;
2219 }
2220
2221
2222 bool Paragraph::Private::endTeXParParams(BufferParams const & bparams,
2223                         otexstream & os, OutputParams const & runparams) const
2224 {
2225         LyXAlignment const curAlign = params_.align();
2226
2227         if (curAlign == layout_->align)
2228                 return false;
2229
2230         switch (curAlign) {
2231         case LYX_ALIGN_NONE:
2232         case LYX_ALIGN_BLOCK:
2233         case LYX_ALIGN_LAYOUT:
2234         case LYX_ALIGN_SPECIAL:
2235         case LYX_ALIGN_DECIMAL:
2236                 break;
2237         case LYX_ALIGN_LEFT:
2238         case LYX_ALIGN_RIGHT:
2239         case LYX_ALIGN_CENTER:
2240                 if (runparams.moving_arg)
2241                         os << "\\protect";
2242                 break;
2243         }
2244
2245         bool output = false;
2246         int col = 0;
2247         string const end_tag = "\\par\\end";
2248         InsetCode code = ownerCode();
2249         bool const lastpar = runparams.isLastPar;
2250
2251         switch (curAlign) {
2252         case LYX_ALIGN_NONE:
2253         case LYX_ALIGN_BLOCK:
2254         case LYX_ALIGN_LAYOUT:
2255         case LYX_ALIGN_SPECIAL:
2256         case LYX_ALIGN_DECIMAL:
2257                 break;
2258         case LYX_ALIGN_LEFT: {
2259                 if (owner_->getParLanguage(bparams)->babel() != "hebrew")
2260                         output = corrected_env(os, end_tag, "flushleft", code, lastpar, col);
2261                 else
2262                         output = corrected_env(os, end_tag, "flushright", code, lastpar, col);
2263                 break;
2264         } case LYX_ALIGN_RIGHT: {
2265                 if (owner_->getParLanguage(bparams)->babel() != "hebrew")
2266                         output = corrected_env(os, end_tag, "flushright", code, lastpar, col);
2267                 else
2268                         output = corrected_env(os, end_tag, "flushleft", code, lastpar, col);
2269                 break;
2270         } case LYX_ALIGN_CENTER: {
2271                 corrected_env(os, end_tag, "center", code, lastpar, col);
2272                 break;
2273         }
2274         }
2275
2276         return output || lastpar;
2277 }
2278
2279
2280 // This one spits out the text of the paragraph
2281 void Paragraph::latex(BufferParams const & bparams,
2282         Font const & outerfont,
2283         otexstream & os,
2284         OutputParams const & runparams,
2285         int start_pos, int end_pos, bool force) const
2286 {
2287         LYXERR(Debug::LATEX, "Paragraph::latex...     " << this);
2288
2289         // FIXME This check should not be needed. Perhaps issue an
2290         // error if it triggers.
2291         Layout const & style = inInset().forcePlainLayout() ?
2292                 bparams.documentClass().plainLayout() : *d->layout_;
2293
2294         if (!force && style.inpreamble)
2295                 return;
2296
2297         bool const allowcust = allowParagraphCustomization();
2298
2299         // Current base font for all inherited font changes, without any
2300         // change caused by an individual character, except for the language:
2301         // It is set to the language of the first character.
2302         // As long as we are in the label, this font is the base font of the
2303         // label. Before the first body character it is set to the base font
2304         // of the body.
2305         Font basefont;
2306
2307         // Maybe we have to create a optional argument.
2308         pos_type body_pos = beginOfBody();
2309         unsigned int column = 0;
2310
2311         if (body_pos > 0) {
2312                 // the optional argument is kept in curly brackets in
2313                 // case it contains a ']'
2314                 // This is not strictly needed, but if this is changed it
2315                 // would be a file format change, and tex2lyx would need
2316                 // to be adjusted, since it unconditionally removes the
2317                 // braces when it parses \item.
2318                 os << "[{";
2319                 column += 2;
2320                 basefont = getLabelFont(bparams, outerfont);
2321         } else {
2322                 basefont = getLayoutFont(bparams, outerfont);
2323         }
2324
2325         // Which font is currently active?
2326         Font running_font(basefont);
2327         // Do we have an open font change?
2328         bool open_font = false;
2329
2330         Change runningChange = Change(Change::UNCHANGED);
2331
2332         Encoding const * const prev_encoding = runparams.encoding;
2333
2334         os.texrow().start(id(), 0);
2335
2336         // if the paragraph is empty, the loop will not be entered at all
2337         if (empty()) {
2338                 if (style.isCommand()) {
2339                         os << '{';
2340                         ++column;
2341                 }
2342                 if (!style.leftdelim().empty()) {
2343                         os << style.leftdelim();
2344                         column += style.leftdelim().size();
2345                 }
2346                 if (allowcust)
2347                         column += d->startTeXParParams(bparams, os, runparams);
2348         }
2349
2350         for (pos_type i = 0; i < size(); ++i) {
2351                 // First char in paragraph or after label?
2352                 if (i == body_pos) {
2353                         if (body_pos > 0) {
2354                                 if (open_font) {
2355                                         column += running_font.latexWriteEndChanges(
2356                                                 os, bparams, runparams,
2357                                                 basefont, basefont);
2358                                         open_font = false;
2359                                 }
2360                                 basefont = getLayoutFont(bparams, outerfont);
2361                                 running_font = basefont;
2362
2363                                 column += Changes::latexMarkChange(os, bparams,
2364                                                 runningChange, Change(Change::UNCHANGED),
2365                                                 runparams);
2366                                 runningChange = Change(Change::UNCHANGED);
2367
2368                                 os << "}] ";
2369                                 column +=3;
2370                         }
2371                         if (style.isCommand()) {
2372                                 os << '{';
2373                                 ++column;
2374                         }
2375
2376                         if (!style.leftdelim().empty()) {
2377                                 os << style.leftdelim();
2378                                 column += style.leftdelim().size();
2379                         }
2380
2381                         if (allowcust)
2382                                 column += d->startTeXParParams(bparams, os,
2383                                                             runparams);
2384                 }
2385
2386                 Change const & change = runparams.inDeletedInset
2387                         ? runparams.changeOfDeletedInset : lookupChange(i);
2388
2389                 if (bparams.output_changes && runningChange != change) {
2390                         if (open_font) {
2391                                 column += running_font.latexWriteEndChanges(
2392                                                 os, bparams, runparams, basefont, basefont);
2393                                 open_font = false;
2394                         }
2395                         basefont = getLayoutFont(bparams, outerfont);
2396                         running_font = basefont;
2397
2398                         column += Changes::latexMarkChange(os, bparams, runningChange,
2399                                                            change, runparams);
2400                         runningChange = change;
2401                 }
2402
2403                 // do not output text which is marked deleted
2404                 // if change tracking output is disabled
2405                 if (!bparams.output_changes && change.deleted()) {
2406                         continue;
2407                 }
2408
2409                 ++column;
2410
2411                 // Fully instantiated font
2412                 Font const font = getFont(bparams, i, outerfont);
2413
2414                 Font const last_font = running_font;
2415
2416                 // Do we need to close the previous font?
2417                 if (open_font &&
2418                     (font != running_font ||
2419                      font.language() != running_font.language()))
2420                 {
2421                         column += running_font.latexWriteEndChanges(
2422                                         os, bparams, runparams, basefont,
2423                                         (i == body_pos-1) ? basefont : font);
2424                         running_font = basefont;
2425                         open_font = false;
2426                 }
2427
2428                 string const running_lang = runparams.use_polyglossia ?
2429                         running_font.language()->polyglossia() : running_font.language()->babel();
2430                 // close babel's font environment before opening CJK.
2431                 string const lang_end_command = runparams.use_polyglossia ?
2432                         "\\end{$$lang}" : lyxrc.language_command_end;
2433                 if (!running_lang.empty() &&
2434                     font.language()->encoding()->package() == Encoding::CJK) {
2435                                 string end_tag = subst(lang_end_command,
2436                                                         "$$lang",
2437                                                         running_lang);
2438                                 os << from_ascii(end_tag);
2439                                 column += end_tag.length();
2440                 }
2441
2442                 // Switch file encoding if necessary (and allowed)
2443                 if (!runparams.pass_thru && !style.pass_thru &&
2444                     runparams.encoding->package() != Encoding::none &&
2445                     font.language()->encoding()->package() != Encoding::none) {
2446                         pair<bool, int> const enc_switch =
2447                                 switchEncoding(os.os(), bparams, runparams,
2448                                         *(font.language()->encoding()));
2449                         if (enc_switch.first) {
2450                                 column += enc_switch.second;
2451                                 runparams.encoding = font.language()->encoding();
2452                         }
2453                 }
2454
2455                 char_type const c = d->text_[i];
2456
2457                 // Do we need to change font?
2458                 if ((font != running_font ||
2459                      font.language() != running_font.language()) &&
2460                         i != body_pos - 1)
2461                 {
2462                         odocstringstream ods;
2463                         column += font.latexWriteStartChanges(ods, bparams,
2464                                                               runparams, basefont,
2465                                                               last_font);
2466                         running_font = font;
2467                         open_font = true;
2468                         docstring fontchange = ods.str();
2469                         // check whether the fontchange ends with a \\textcolor
2470                         // modifier and the text starts with a space (bug 4473)
2471                         docstring const last_modifier = rsplit(fontchange, '\\');
2472                         if (prefixIs(last_modifier, from_ascii("textcolor")) && c == ' ')
2473                                 os << fontchange << from_ascii("{}");
2474                         // check if the fontchange ends with a trailing blank
2475                         // (like "\small " (see bug 3382)
2476                         else if (suffixIs(fontchange, ' ') && c == ' ')
2477                                 os << fontchange.substr(0, fontchange.size() - 1)
2478                                    << from_ascii("{}");
2479                         else
2480                                 os << fontchange;
2481                 }
2482
2483                 // FIXME: think about end_pos implementation...
2484                 if (c == ' ' && i >= start_pos && (end_pos == -1 || i < end_pos)) {
2485                         // FIXME: integrate this case in latexSpecialChar
2486                         // Do not print the separation of the optional argument
2487                         // if style.pass_thru is false. This works because
2488                         // latexSpecialChar ignores spaces if
2489                         // style.pass_thru is false.
2490                         if (i != body_pos - 1) {
2491                                 if (d->simpleTeXBlanks(runparams, os,
2492                                                 i, column, font, style)) {
2493                                         // A surrogate pair was output. We
2494                                         // must not call latexSpecialChar
2495                                         // in this iteration, since it would output
2496                                         // the combining character again.
2497                                         ++i;
2498                                         continue;
2499                                 }
2500                         }
2501                 }
2502
2503                 OutputParams rp = runparams;
2504                 rp.free_spacing = style.free_spacing;
2505                 rp.local_font = &font;
2506                 rp.intitle = style.intitle;
2507
2508                 // Two major modes:  LaTeX or plain
2509                 // Handle here those cases common to both modes
2510                 // and then split to handle the two modes separately.
2511                 if (c == META_INSET) {
2512                         if (i >= start_pos && (end_pos == -1 || i < end_pos)) {
2513                                 d->latexInset(bparams, os, rp, running_font,
2514                                                 basefont, outerfont, open_font,
2515                                                 runningChange, style, i, column);
2516                         }
2517                 } else {
2518                         if (i >= start_pos && (end_pos == -1 || i < end_pos)) {
2519                                 try {
2520                                         d->latexSpecialChar(os, bparams, rp, running_font, runningChange,
2521                                                             style, i, end_pos, column);
2522                                 } catch (EncodingException & e) {
2523                                 if (runparams.dryrun) {
2524                                         os << "<" << _("LyX Warning: ")
2525                                            << _("uncodable character") << " '";
2526                                         os.put(c);
2527                                         os << "'>";
2528                                 } else {
2529                                         // add location information and throw again.
2530                                         e.par_id = id();
2531                                         e.pos = i;
2532                                         throw(e);
2533                                 }
2534                         }
2535                 }
2536                 }
2537
2538                 // Set the encoding to that returned from latexSpecialChar (see
2539                 // comment for encoding member in OutputParams.h)
2540                 runparams.encoding = rp.encoding;
2541         }
2542
2543         // If we have an open font definition, we have to close it
2544         if (open_font) {
2545 #ifdef FIXED_LANGUAGE_END_DETECTION
2546                 if (next_) {
2547                         running_font.latexWriteEndChanges(os, bparams,
2548                                         runparams, basefont,
2549                                         next_->getFont(bparams, 0, outerfont));
2550                 } else {
2551                         running_font.latexWriteEndChanges(os, bparams,
2552                                         runparams, basefont, basefont);
2553                 }
2554 #else
2555 //FIXME: For now we ALWAYS have to close the foreign font settings if they are
2556 //FIXME: there as we start another \selectlanguage with the next paragraph if
2557 //FIXME: we are in need of this. This should be fixed sometime (Jug)
2558                 running_font.latexWriteEndChanges(os, bparams, runparams,
2559                                 basefont, basefont);
2560 #endif
2561         }
2562
2563         column += Changes::latexMarkChange(os, bparams, runningChange,
2564                                            Change(Change::UNCHANGED), runparams);
2565
2566         // Needed if there is an optional argument but no contents.
2567         if (body_pos > 0 && body_pos == size()) {
2568                 os << "}]~";
2569         }
2570
2571         if (!style.rightdelim().empty()) {
2572                 os << style.rightdelim();
2573                 column += style.rightdelim().size();
2574         }
2575
2576         if (allowcust && d->endTeXParParams(bparams, os, runparams)
2577             && runparams.encoding != prev_encoding) {
2578                 runparams.encoding = prev_encoding;
2579                 os << setEncoding(prev_encoding->iconvName());
2580         }
2581
2582         LYXERR(Debug::LATEX, "Paragraph::latex... done " << this);
2583 }
2584
2585
2586 bool Paragraph::emptyTag() const
2587 {
2588         for (pos_type i = 0; i < size(); ++i) {
2589                 if (Inset const * inset = getInset(i)) {
2590                         InsetCode lyx_code = inset->lyxCode();
2591                         // FIXME testing like that is wrong. What is
2592                         // the intent?
2593                         if (lyx_code != TOC_CODE &&
2594                             lyx_code != INCLUDE_CODE &&
2595                             lyx_code != GRAPHICS_CODE &&
2596                             lyx_code != ERT_CODE &&
2597                             lyx_code != LISTINGS_CODE &&
2598                             lyx_code != FLOAT_CODE &&
2599                             lyx_code != TABULAR_CODE) {
2600                                 return false;
2601                         }
2602                 } else {
2603                         char_type c = d->text_[i];
2604                         if (c != ' ' && c != '\t')
2605                                 return false;
2606                 }
2607         }
2608         return true;
2609 }
2610
2611
2612 string Paragraph::getID(Buffer const & buf, OutputParams const & runparams)
2613         const
2614 {
2615         for (pos_type i = 0; i < size(); ++i) {
2616                 if (Inset const * inset = getInset(i)) {
2617                         InsetCode lyx_code = inset->lyxCode();
2618                         if (lyx_code == LABEL_CODE) {
2619                                 InsetLabel const * const il = static_cast<InsetLabel const *>(inset);
2620                                 docstring const & id = il->getParam("name");
2621                                 return "id='" + to_utf8(sgml::cleanID(buf, runparams, id)) + "'";
2622                         }
2623                 }
2624         }
2625         return string();
2626 }
2627
2628
2629 pos_type Paragraph::firstWordDocBook(odocstream & os, OutputParams const & runparams)
2630         const
2631 {
2632         pos_type i;
2633         for (i = 0; i < size(); ++i) {
2634                 if (Inset const * inset = getInset(i)) {
2635                         inset->docbook(os, runparams);
2636                 } else {
2637                         char_type c = d->text_[i];
2638                         if (c == ' ')
2639                                 break;
2640                         os << sgml::escapeChar(c);
2641                 }
2642         }
2643         return i;
2644 }
2645
2646
2647 pos_type Paragraph::firstWordLyXHTML(XHTMLStream & xs, OutputParams const & runparams)
2648         const
2649 {
2650         pos_type i;
2651         for (i = 0; i < size(); ++i) {
2652                 if (Inset const * inset = getInset(i)) {
2653                         inset->xhtml(xs, runparams);
2654                 } else {
2655                         char_type c = d->text_[i];
2656                         if (c == ' ')
2657                                 break;
2658                         xs << c;
2659                 }
2660         }
2661         return i;
2662 }
2663
2664
2665 bool Paragraph::Private::onlyText(Buffer const & buf, Font const & outerfont, pos_type initial) const
2666 {
2667         Font font_old;
2668         pos_type size = text_.size();
2669         for (pos_type i = initial; i < size; ++i) {
2670                 Font font = owner_->getFont(buf.params(), i, outerfont);
2671                 if (text_[i] == META_INSET)
2672                         return false;
2673                 if (i != initial && font != font_old)
2674                         return false;
2675                 font_old = font;
2676         }
2677
2678         return true;
2679 }
2680
2681
2682 void Paragraph::simpleDocBookOnePar(Buffer const & buf,
2683                                     odocstream & os,
2684                                     OutputParams const & runparams,
2685                                     Font const & outerfont,
2686                                     pos_type initial) const
2687 {
2688         bool emph_flag = false;
2689
2690         Layout const & style = *d->layout_;
2691         FontInfo font_old =
2692                 style.labeltype == LABEL_MANUAL ? style.labelfont : style.font;
2693
2694         if (style.pass_thru && !d->onlyText(buf, outerfont, initial))
2695                 os << "]]>";
2696
2697         // parsing main loop
2698         for (pos_type i = initial; i < size(); ++i) {
2699                 Font font = getFont(buf.params(), i, outerfont);
2700
2701                 // handle <emphasis> tag
2702                 if (font_old.emph() != font.fontInfo().emph()) {
2703                         if (font.fontInfo().emph() == FONT_ON) {
2704                                 os << "<emphasis>";
2705                                 emph_flag = true;
2706                         } else if (i != initial) {
2707                                 os << "</emphasis>";
2708                                 emph_flag = false;
2709                         }
2710                 }
2711
2712                 if (Inset const * inset = getInset(i)) {
2713                         inset->docbook(os, runparams);
2714                 } else {
2715                         char_type c = d->text_[i];
2716
2717                         if (style.pass_thru)
2718                                 os.put(c);
2719                         else
2720                                 os << sgml::escapeChar(c);
2721                 }
2722                 font_old = font.fontInfo();
2723         }
2724
2725         if (emph_flag) {
2726                 os << "</emphasis>";
2727         }
2728
2729         if (style.free_spacing)
2730                 os << '\n';
2731         if (style.pass_thru && !d->onlyText(buf, outerfont, initial))
2732                 os << "<![CDATA[";
2733 }
2734
2735
2736 namespace {
2737 void doFontSwitch(vector<html::FontTag> & tagsToOpen,
2738                   vector<html::EndFontTag> & tagsToClose,
2739                   bool & flag, FontState curstate, html::FontTypes type)
2740 {
2741         if (curstate == FONT_ON) {
2742                 tagsToOpen.push_back(html::FontTag(type));
2743                 flag = true;
2744         } else if (flag) {
2745                 tagsToClose.push_back(html::EndFontTag(type));
2746                 flag = false;
2747         }
2748 }
2749 }
2750
2751
2752 docstring Paragraph::simpleLyXHTMLOnePar(Buffer const & buf,
2753                                     XHTMLStream & xs,
2754                                     OutputParams const & runparams,
2755                                     Font const & outerfont,
2756                                     pos_type initial) const
2757 {
2758         docstring retval;
2759
2760         // track whether we have opened these tags
2761         bool emph_flag = false;
2762         bool bold_flag = false;
2763         bool noun_flag = false;
2764         bool ubar_flag = false;
2765         bool dbar_flag = false;
2766         bool sout_flag = false;
2767         bool wave_flag = false;
2768         // shape tags
2769         bool shap_flag = false;
2770         // family tags
2771         bool faml_flag = false;
2772         // size tags
2773         bool size_flag = false;
2774
2775         Layout const & style = *d->layout_;
2776
2777         xs.startParagraph(allowEmpty());
2778
2779         FontInfo font_old =
2780                 style.labeltype == LABEL_MANUAL ? style.labelfont : style.font;
2781
2782         FontShape  curr_fs   = INHERIT_SHAPE;
2783         FontFamily curr_fam  = INHERIT_FAMILY;
2784         FontSize   curr_size = FONT_SIZE_INHERIT;
2785
2786         string const default_family =
2787                 buf.masterBuffer()->params().fonts_default_family;
2788
2789         vector<html::FontTag> tagsToOpen;
2790         vector<html::EndFontTag> tagsToClose;
2791
2792         // parsing main loop
2793         for (pos_type i = initial; i < size(); ++i) {
2794                 // let's not show deleted material in the output
2795                 if (isDeleted(i))
2796                         continue;
2797
2798                 Font const font = getFont(buf.masterBuffer()->params(), i, outerfont);
2799
2800                 // emphasis
2801                 FontState curstate = font.fontInfo().emph();
2802                 if (font_old.emph() != curstate)
2803                         doFontSwitch(tagsToOpen, tagsToClose, emph_flag, curstate, html::FT_EMPH);
2804
2805                 // noun
2806                 curstate = font.fontInfo().noun();
2807                 if (font_old.noun() != curstate)
2808                         doFontSwitch(tagsToOpen, tagsToClose, noun_flag, curstate, html::FT_NOUN);
2809
2810                 // underbar
2811                 curstate = font.fontInfo().underbar();
2812                 if (font_old.underbar() != curstate)
2813                         doFontSwitch(tagsToOpen, tagsToClose, ubar_flag, curstate, html::FT_UBAR);
2814
2815                 // strikeout
2816                 curstate = font.fontInfo().strikeout();
2817                 if (font_old.strikeout() != curstate)
2818                         doFontSwitch(tagsToOpen, tagsToClose, sout_flag, curstate, html::FT_SOUT);
2819
2820                 // double underbar
2821                 curstate = font.fontInfo().uuline();
2822                 if (font_old.uuline() != curstate)
2823                         doFontSwitch(tagsToOpen, tagsToClose, dbar_flag, curstate, html::FT_DBAR);
2824
2825                 // wavy line
2826                 curstate = font.fontInfo().uwave();
2827                 if (font_old.uwave() != curstate)
2828                         doFontSwitch(tagsToOpen, tagsToClose, wave_flag, curstate, html::FT_WAVE);
2829
2830                 // bold
2831                 // a little hackish, but allows us to reuse what we have.
2832                 curstate = (font.fontInfo().series() == BOLD_SERIES ? FONT_ON : FONT_OFF);
2833                 if (font_old.series() != font.fontInfo().series())
2834                         doFontSwitch(tagsToOpen, tagsToClose, bold_flag, curstate, html::FT_BOLD);
2835
2836                 // Font shape
2837                 curr_fs = font.fontInfo().shape();
2838                 FontShape old_fs = font_old.shape();
2839                 if (old_fs != curr_fs) {
2840                         if (shap_flag) {
2841                                 switch (old_fs) {
2842                                 case ITALIC_SHAPE:
2843                                         tagsToClose.push_back(html::EndFontTag(html::FT_ITALIC));
2844                                         break;
2845                                 case SLANTED_SHAPE:
2846                                         tagsToClose.push_back(html::EndFontTag(html::FT_SLANTED));
2847                                         break;
2848                                 case SMALLCAPS_SHAPE:
2849                                         tagsToClose.push_back(html::EndFontTag(html::FT_SMALLCAPS));
2850                                         break;
2851                                 case UP_SHAPE:
2852                                 case INHERIT_SHAPE:
2853                                         break;
2854                                 default:
2855                                         // the other tags are for internal use
2856                                         LATTEST(false);
2857                                         break;
2858                                 }
2859                                 shap_flag = false;
2860                         }
2861                         switch (curr_fs) {
2862                         case ITALIC_SHAPE:
2863                                 tagsToOpen.push_back(html::FontTag(html::FT_ITALIC));
2864                                 shap_flag = true;
2865                                 break;
2866                         case SLANTED_SHAPE:
2867                                 tagsToOpen.push_back(html::FontTag(html::FT_SLANTED));
2868                                 shap_flag = true;
2869                                 break;
2870                         case SMALLCAPS_SHAPE:
2871                                 tagsToOpen.push_back(html::FontTag(html::FT_SMALLCAPS));
2872                                 shap_flag = true;
2873                                 break;
2874                         case UP_SHAPE:
2875                         case INHERIT_SHAPE:
2876                                 break;
2877                         default:
2878                                 // the other tags are for internal use
2879                                 LATTEST(false);
2880                                 break;
2881                         }
2882                 }
2883
2884                 // Font family
2885                 curr_fam = font.fontInfo().family();
2886                 FontFamily old_fam = font_old.family();
2887                 if (old_fam != curr_fam) {
2888                         if (faml_flag) {
2889                                 switch (old_fam) {
2890                                 case ROMAN_FAMILY:
2891                                         tagsToClose.push_back(html::EndFontTag(html::FT_ROMAN));
2892                                         break;
2893                                 case SANS_FAMILY:
2894                                         tagsToClose.push_back(html::EndFontTag(html::FT_SANS));
2895                                         break;
2896                                 case TYPEWRITER_FAMILY:
2897                                         tagsToClose.push_back(html::EndFontTag(html::FT_TYPE));
2898                                         break;
2899                                 case INHERIT_FAMILY:
2900                                         break;
2901                                 default:
2902                                         // the other tags are for internal use
2903                                         LATTEST(false);
2904                                         break;
2905                                 }
2906                                 faml_flag = false;
2907                         }
2908                         switch (curr_fam) {
2909                         case ROMAN_FAMILY:
2910                                 // we will treat a "default" font family as roman, since we have
2911                                 // no other idea what to do.
2912                                 if (default_family != "rmdefault" && default_family != "default") {
2913                                         tagsToOpen.push_back(html::FontTag(html::FT_ROMAN));
2914                                         faml_flag = true;
2915                                 }
2916                                 break;
2917                         case SANS_FAMILY:
2918                                 if (default_family != "sfdefault") {
2919                                         tagsToOpen.push_back(html::FontTag(html::FT_SANS));
2920                                         faml_flag = true;
2921                                 }
2922                                 break;
2923                         case TYPEWRITER_FAMILY:
2924                                 if (default_family != "ttdefault") {
2925                                         tagsToOpen.push_back(html::FontTag(html::FT_TYPE));
2926                                         faml_flag = true;
2927                                 }
2928                                 break;
2929                         case INHERIT_FAMILY:
2930                                 break;
2931                         default:
2932                                 // the other tags are for internal use
2933                                 LATTEST(false);
2934                                 break;
2935                         }
2936                 }
2937
2938                 // Font size
2939                 curr_size = font.fontInfo().size();
2940                 FontSize old_size = font_old.size();
2941                 if (old_size != curr_size) {
2942                         if (size_flag) {
2943                                 switch (old_size) {
2944                                 case FONT_SIZE_TINY:
2945                                         tagsToClose.push_back(html::EndFontTag(html::FT_SIZE_TINY));
2946                                         break;
2947                                 case FONT_SIZE_SCRIPT:
2948                                         tagsToClose.push_back(html::EndFontTag(html::FT_SIZE_SCRIPT));
2949                                         break;
2950                                 case FONT_SIZE_FOOTNOTE:
2951                                         tagsToClose.push_back(html::EndFontTag(html::FT_SIZE_FOOTNOTE));
2952                                         break;
2953                                 case FONT_SIZE_SMALL:
2954                                         tagsToClose.push_back(html::EndFontTag(html::FT_SIZE_SMALL));
2955                                         break;
2956                                 case FONT_SIZE_LARGE:
2957                                         tagsToClose.push_back(html::EndFontTag(html::FT_SIZE_LARGE));
2958                                         break;
2959                                 case FONT_SIZE_LARGER:
2960                                         tagsToClose.push_back(html::EndFontTag(html::FT_SIZE_LARGER));
2961                                         break;
2962                                 case FONT_SIZE_LARGEST:
2963                                         tagsToClose.push_back(html::EndFontTag(html::FT_SIZE_LARGEST));
2964                                         break;
2965                                 case FONT_SIZE_HUGE:
2966                                         tagsToClose.push_back(html::EndFontTag(html::FT_SIZE_HUGE));
2967                                         break;
2968                                 case FONT_SIZE_HUGER:
2969                                         tagsToClose.push_back(html::EndFontTag(html::FT_SIZE_HUGER));
2970                                         break;
2971                                 case FONT_SIZE_INCREASE:
2972                                         tagsToClose.push_back(html::EndFontTag(html::FT_SIZE_INCREASE));
2973                                         break;
2974                                 case FONT_SIZE_DECREASE:
2975                                         tagsToClose.push_back(html::EndFontTag(html::FT_SIZE_DECREASE));
2976                                         break;
2977                                 case FONT_SIZE_INHERIT:
2978                                 case FONT_SIZE_NORMAL:
2979                                         break;
2980                                 default:
2981                                         // the other tags are for internal use
2982                                         LATTEST(false);
2983                                         break;
2984                                 }
2985                                 size_flag = false;
2986                         }
2987                         switch (curr_size) {
2988                         case FONT_SIZE_TINY:
2989                                 tagsToOpen.push_back(html::FontTag(html::FT_SIZE_TINY));
2990                                 size_flag = true;
2991                                 break;
2992                         case FONT_SIZE_SCRIPT:
2993                                 tagsToOpen.push_back(html::FontTag(html::FT_SIZE_SCRIPT));
2994                                 size_flag = true;
2995                                 break;
2996                         case FONT_SIZE_FOOTNOTE:
2997                                 tagsToOpen.push_back(html::FontTag(html::FT_SIZE_FOOTNOTE));
2998                                 size_flag = true;
2999                                 break;
3000                         case FONT_SIZE_SMALL:
3001                                 tagsToOpen.push_back(html::FontTag(html::FT_SIZE_SMALL));
3002                                 size_flag = true;
3003                                 break;
3004                         case FONT_SIZE_LARGE:
3005                                 tagsToOpen.push_back(html::FontTag(html::FT_SIZE_LARGE));
3006                                 size_flag = true;
3007                                 break;
3008                         case FONT_SIZE_LARGER:
3009                                 tagsToOpen.push_back(html::FontTag(html::FT_SIZE_LARGER));
3010                                 size_flag = true;
3011                                 break;
3012                         case FONT_SIZE_LARGEST:
3013                                 tagsToOpen.push_back(html::FontTag(html::FT_SIZE_LARGEST));
3014                                 size_flag = true;
3015                                 break;
3016                         case FONT_SIZE_HUGE:
3017                                 tagsToOpen.push_back(html::FontTag(html::FT_SIZE_HUGE));
3018                                 size_flag = true;
3019                                 break;
3020                         case FONT_SIZE_HUGER:
3021                                 tagsToOpen.push_back(html::FontTag(html::FT_SIZE_HUGER));
3022                                 size_flag = true;
3023                                 break;
3024                         case FONT_SIZE_INCREASE:
3025                                 tagsToOpen.push_back(html::FontTag(html::FT_SIZE_INCREASE));
3026                                 size_flag = true;
3027                                 break;
3028                         case FONT_SIZE_DECREASE:
3029                                 tagsToOpen.push_back(html::FontTag(html::FT_SIZE_DECREASE));
3030                                 size_flag = true;
3031                                 break;
3032                         case FONT_SIZE_NORMAL:
3033                         case FONT_SIZE_INHERIT:
3034                                 break;
3035                         default:
3036                                 // the other tags are for internal use
3037                                 LATTEST(false);
3038                                 break;
3039                         }
3040                 }
3041
3042                 // FIXME XHTML
3043                 // Other such tags? What about the other text ranges?
3044
3045                 vector<html::EndFontTag>::const_iterator cit = tagsToClose.begin();
3046                 vector<html::EndFontTag>::const_iterator cen = tagsToClose.end();
3047                 for (; cit != cen; ++cit)
3048                         xs << *cit;
3049
3050                 vector<html::FontTag>::const_iterator sit = tagsToOpen.begin();
3051                 vector<html::FontTag>::const_iterator sen = tagsToOpen.end();
3052                 for (; sit != sen; ++sit)
3053                         xs << *sit;
3054
3055                 tagsToClose.clear();
3056                 tagsToOpen.clear();
3057
3058                 Inset const * inset = getInset(i);
3059                 if (inset) {
3060                         if (!runparams.for_toc || inset->isInToc()) {
3061                                 OutputParams np = runparams;
3062                                 np.local_font = &font;
3063                                 if (!inset->getLayout().htmlisblock())
3064                                         np.html_in_par = true;
3065                                 retval += inset->xhtml(xs, np);
3066                         }
3067                 } else {
3068                         char_type c = getUChar(buf.masterBuffer()->params(), i);
3069                         xs << c;
3070                 }
3071                 font_old = font.fontInfo();
3072         }
3073
3074         xs.closeFontTags();
3075         xs.endParagraph();
3076         return retval;
3077 }
3078
3079
3080 bool Paragraph::isHfill(pos_type pos) const
3081 {
3082         Inset const * inset = getInset(pos);
3083         return inset && inset->isHfill();
3084 }
3085
3086
3087 bool Paragraph::isNewline(pos_type pos) const
3088 {
3089         Inset const * inset = getInset(pos);
3090         return inset && inset->lyxCode() == NEWLINE_CODE;
3091 }
3092
3093
3094 bool Paragraph::isEnvSeparator(pos_type pos) const
3095 {
3096         Inset const * inset = getInset(pos);
3097         return inset && inset->lyxCode() == SEPARATOR_CODE;
3098 }
3099
3100
3101 bool Paragraph::isLineSeparator(pos_type pos) const
3102 {
3103         char_type const c = d->text_[pos];
3104         if (isLineSeparatorChar(c))
3105                 return true;
3106         Inset const * inset = getInset(pos);
3107         return inset && inset->isLineSeparator();
3108 }
3109
3110
3111 bool Paragraph::isWordSeparator(pos_type pos) const
3112 {
3113         if (pos == size())
3114                 return true;
3115         if (Inset const * inset = getInset(pos))
3116                 return !inset->isLetter();
3117         // if we have a hard hyphen (no en- or emdash) or apostrophe
3118         // we pass this to the spell checker
3119         // FIXME: this method is subject to change, visit
3120         // https://bugzilla.mozilla.org/show_bug.cgi?id=355178
3121         // to get an impression how complex this is.
3122         if (isHardHyphenOrApostrophe(pos))
3123                 return false;
3124         char_type const c = d->text_[pos];
3125         // We want to pass the escape chars to the spellchecker
3126         docstring const escape_chars = from_utf8(lyxrc.spellchecker_esc_chars);
3127         return !isLetterChar(c) && !isDigitASCII(c) && !contains(escape_chars, c);
3128 }
3129
3130
3131 bool Paragraph::isHardHyphenOrApostrophe(pos_type pos) const
3132 {
3133         pos_type const psize = size();
3134         if (pos >= psize)
3135                 return false;
3136         char_type const c = d->text_[pos];
3137         if (c != '-' && c != '\'')
3138                 return false;
3139         int nextpos = pos + 1;
3140         int prevpos = pos > 0 ? pos - 1 : 0;
3141         if ((nextpos == psize || isSpace(nextpos))
3142                 && (pos == 0 || isSpace(prevpos)))
3143                 return false;
3144         return true;
3145 }
3146
3147
3148 FontSpan const & Paragraph::getSpellRange(pos_type pos) const
3149 {
3150         return d->speller_state_.getRange(pos);
3151 }
3152
3153
3154 bool Paragraph::isChar(pos_type pos) const
3155 {
3156         if (Inset const * inset = getInset(pos))
3157                 return inset->isChar();
3158         char_type const c = d->text_[pos];
3159         return !isLetterChar(c) && !isDigitASCII(c) && !lyx::isSpace(c);
3160 }
3161
3162
3163 bool Paragraph::isSpace(pos_type pos) const
3164 {
3165         if (Inset const * inset = getInset(pos))
3166                 return inset->isSpace();
3167         char_type const c = d->text_[pos];
3168         return lyx::isSpace(c);
3169 }
3170
3171
3172 Language const *
3173 Paragraph::getParLanguage(BufferParams const & bparams) const
3174 {
3175         if (!empty())
3176                 return getFirstFontSettings(bparams).language();
3177         // FIXME: we should check the prev par as well (Lgb)
3178         return bparams.language;
3179 }
3180
3181
3182 bool Paragraph::isRTL(BufferParams const & bparams) const
3183 {
3184         return getParLanguage(bparams)->rightToLeft()
3185                 && !inInset().getLayout().forceLTR();
3186 }
3187
3188
3189 void Paragraph::changeLanguage(BufferParams const & bparams,
3190                                Language const * from, Language const * to)
3191 {
3192         // change language including dummy font change at the end
3193         for (pos_type i = 0; i <= size(); ++i) {
3194                 Font font = getFontSettings(bparams, i);
3195                 if (font.language() == from) {
3196                         font.setLanguage(to);
3197                         setFont(i, font);
3198                         d->requestSpellCheck(i);
3199                 }
3200         }
3201 }
3202
3203
3204 bool Paragraph::isMultiLingual(BufferParams const & bparams) const
3205 {
3206         Language const * doc_language = bparams.language;
3207         FontList::const_iterator cit = d->fontlist_.begin();
3208         FontList::const_iterator end = d->fontlist_.end();
3209
3210         for (; cit != end; ++cit)
3211                 if (cit->font().language() != ignore_language &&
3212                     cit->font().language() != latex_language &&
3213                     cit->font().language() != doc_language)
3214                         return true;
3215         return false;
3216 }
3217
3218
3219 void Paragraph::getLanguages(std::set<Language const *> & languages) const
3220 {
3221         FontList::const_iterator cit = d->fontlist_.begin();
3222         FontList::const_iterator end = d->fontlist_.end();
3223
3224         for (; cit != end; ++cit) {
3225                 Language const * lang = cit->font().language();
3226                 if (lang != ignore_language &&
3227                     lang != latex_language)
3228                         languages.insert(lang);
3229         }
3230 }
3231
3232
3233 docstring Paragraph::asString(int options) const
3234 {
3235         return asString(0, size(), options);
3236 }
3237
3238
3239 docstring Paragraph::asString(pos_type beg, pos_type end, int options, const OutputParams *runparams) const
3240 {
3241         odocstringstream os;
3242
3243         if (beg == 0
3244             && options & AS_STR_LABEL
3245             && !d->params_.labelString().empty())
3246                 os << d->params_.labelString() << ' ';
3247
3248         for (pos_type i = beg; i < end; ++i) {
3249                 if ((options & AS_STR_SKIPDELETE) && isDeleted(i))
3250                         continue;
3251                 char_type const c = d->text_[i];
3252                 if (isPrintable(c) || c == '\t'
3253                     || (c == '\n' && (options & AS_STR_NEWLINES)))
3254                         os.put(c);
3255                 else if (c == META_INSET && (options & AS_STR_INSETS)) {
3256                         if (c == META_INSET && (options & AS_STR_PLAINTEXT)) {
3257                                 LASSERT(runparams != 0, return docstring());
3258                                 getInset(i)->plaintext(os, *runparams);
3259                         } else {
3260                                 getInset(i)->toString(os);
3261                         }
3262                 }
3263         }
3264
3265         return os.str();
3266 }
3267
3268
3269 void Paragraph::forOutliner(docstring & os, size_t const maxlen,
3270                                                         bool const shorten) const
3271 {
3272         size_t tmplen = shorten ? maxlen + 1 : maxlen;
3273         if (!d->params_.labelString().empty())
3274                 os += d->params_.labelString() + ' ';
3275         for (pos_type i = 0; i < size() && os.length() < tmplen; ++i) {
3276                 if (isDeleted(i))
3277                         continue;
3278                 char_type const c = d->text_[i];
3279                 if (isPrintable(c))
3280                         os += c;
3281                 else if (c == META_INSET)
3282                         getInset(i)->forOutliner(os, tmplen, false);
3283         }
3284         if (shorten)
3285                 Text::shortenForOutliner(os, maxlen);
3286 }
3287
3288
3289 void Paragraph::setInsetOwner(Inset const * inset)
3290 {
3291         d->inset_owner_ = inset;
3292 }
3293
3294
3295 int Paragraph::id() const
3296 {
3297         return d->id_;
3298 }
3299
3300
3301 void Paragraph::setId(int id)
3302 {
3303         d->id_ = id;
3304 }
3305
3306
3307 Layout const & Paragraph::layout() const
3308 {
3309         return *d->layout_;
3310 }
3311
3312
3313 void Paragraph::setLayout(Layout const & layout)
3314 {
3315         d->layout_ = &layout;
3316 }
3317
3318
3319 void Paragraph::setDefaultLayout(DocumentClass const & tc)
3320 {
3321         setLayout(tc.defaultLayout());
3322 }
3323
3324
3325 void Paragraph::setPlainLayout(DocumentClass const & tc)
3326 {
3327         setLayout(tc.plainLayout());
3328 }
3329
3330
3331 void Paragraph::setPlainOrDefaultLayout(DocumentClass const & tclass)
3332 {
3333         if (usePlainLayout())
3334                 setPlainLayout(tclass);
3335         else
3336                 setDefaultLayout(tclass);
3337 }
3338
3339
3340 Inset const & Paragraph::inInset() const
3341 {
3342         LBUFERR(d->inset_owner_);
3343         return *d->inset_owner_;
3344 }
3345
3346
3347 ParagraphParameters & Paragraph::params()
3348 {
3349         return d->params_;
3350 }
3351
3352
3353 ParagraphParameters const & Paragraph::params() const
3354 {
3355         return d->params_;
3356 }
3357
3358
3359 bool Paragraph::isFreeSpacing() const
3360 {
3361         if (d->layout_->free_spacing)
3362                 return true;
3363         return d->inset_owner_ && d->inset_owner_->isFreeSpacing();
3364 }
3365
3366
3367 bool Paragraph::allowEmpty() const
3368 {
3369         if (d->layout_->keepempty)
3370                 return true;
3371         return d->inset_owner_ && d->inset_owner_->allowEmpty();
3372 }
3373
3374
3375 bool Paragraph::brokenBiblio() const
3376 {
3377         // there is a problem if there is no bibitem at position 0 or
3378         // if there is another bibitem in the paragraph.
3379         return d->layout_->labeltype == LABEL_BIBLIO
3380                 && (d->insetlist_.find(BIBITEM_CODE) != 0
3381                     || d->insetlist_.find(BIBITEM_CODE, 1) > 0);
3382 }
3383
3384
3385 int Paragraph::fixBiblio(Buffer const & buffer)
3386 {
3387         // FIXME: What about the case where paragraph is not BIBLIO
3388         // but there is an InsetBibitem?
3389         // FIXME: when there was already an inset at 0, the return value is 1,
3390         // which does not tell whether another inset has been remove; the
3391         // cursor cannot be correctly updated.
3392
3393         if (d->layout_->labeltype != LABEL_BIBLIO)
3394                 return 0;
3395
3396         bool const track_changes = buffer.params().track_changes;
3397         int bibitem_pos = d->insetlist_.find(BIBITEM_CODE);
3398         bool const hasbibitem0 = bibitem_pos == 0;
3399
3400         if (hasbibitem0) {
3401                 bibitem_pos = d->insetlist_.find(BIBITEM_CODE, 1);
3402                 // There was an InsetBibitem at pos 0, and no other one => OK
3403                 if (bibitem_pos == -1)
3404                         return 0;
3405                 // there is a bibitem at the 0 position, but since
3406                 // there is a second one, we copy the second on the
3407                 // first. We're assuming there are at most two of
3408                 // these, which there should be.
3409                 // FIXME: why does it make sense to do that rather
3410                 // than keep the first? (JMarc)
3411                 Inset * inset = releaseInset(bibitem_pos);
3412                 d->insetlist_.begin()->inset = inset;
3413                 return -bibitem_pos;
3414         }
3415
3416         // We need to create an inset at the beginning
3417         Inset * inset = 0;
3418         if (bibitem_pos > 0) {
3419                 // there was one somewhere in the paragraph, let's move it
3420                 inset = d->insetlist_.release(bibitem_pos);
3421                 eraseChar(bibitem_pos, track_changes);
3422         } else
3423                 // make a fresh one
3424                 inset = new InsetBibitem(const_cast<Buffer *>(&buffer),
3425                                          InsetCommandParams(BIBITEM_CODE));
3426
3427         Font font(inherit_font, buffer.params().language);
3428         insertInset(0, inset, font, Change(track_changes ? Change::INSERTED
3429                                                    : Change::UNCHANGED));
3430
3431         return 1;
3432 }
3433
3434
3435 void Paragraph::checkAuthors(AuthorList const & authorList)
3436 {
3437         d->changes_.checkAuthors(authorList);
3438 }
3439
3440
3441 bool Paragraph::isChanged(pos_type pos) const
3442 {
3443         return lookupChange(pos).changed();
3444 }
3445
3446
3447 bool Paragraph::isInserted(pos_type pos) const
3448 {
3449         return lookupChange(pos).inserted();
3450 }
3451
3452
3453 bool Paragraph::isDeleted(pos_type pos) const
3454 {
3455         return lookupChange(pos).deleted();
3456 }
3457
3458
3459 InsetList const & Paragraph::insetList() const
3460 {
3461         return d->insetlist_;
3462 }
3463
3464
3465 void Paragraph::setBuffer(Buffer & b)
3466 {
3467         d->insetlist_.setBuffer(b);
3468 }
3469
3470
3471 Inset * Paragraph::releaseInset(pos_type pos)
3472 {
3473         Inset * inset = d->insetlist_.release(pos);
3474         /// does not honour change tracking!
3475         eraseChar(pos, false);
3476         return inset;
3477 }
3478
3479
3480 Inset * Paragraph::getInset(pos_type pos)
3481 {
3482         return (pos < pos_type(d->text_.size()) && d->text_[pos] == META_INSET)
3483                  ? d->insetlist_.get(pos) : 0;
3484 }
3485
3486
3487 Inset const * Paragraph::getInset(pos_type pos) const
3488 {
3489         return (pos < pos_type(d->text_.size()) && d->text_[pos] == META_INSET)
3490                  ? d->insetlist_.get(pos) : 0;
3491 }
3492
3493
3494 void Paragraph::changeCase(BufferParams const & bparams, pos_type pos,
3495                 pos_type & right, TextCase action)
3496 {
3497         // process sequences of modified characters; in change
3498         // tracking mode, this approach results in much better
3499         // usability than changing case on a char-by-char basis
3500         // We also need to track the current font, since font
3501         // changes within sequences can occur.
3502         vector<pair<char_type, Font> > changes;
3503
3504         bool const trackChanges = bparams.track_changes;
3505
3506         bool capitalize = true;
3507
3508         for (; pos < right; ++pos) {
3509                 char_type oldChar = d->text_[pos];
3510                 char_type newChar = oldChar;
3511
3512                 // ignore insets and don't play with deleted text!
3513                 if (oldChar != META_INSET && !isDeleted(pos)) {
3514                         switch (action) {
3515                                 case text_lowercase:
3516                                         newChar = lowercase(oldChar);
3517                                         break;
3518                                 case text_capitalization:
3519                                         if (capitalize) {
3520                                                 newChar = uppercase(oldChar);
3521                                                 capitalize = false;
3522                                         }
3523                                         break;
3524                                 case text_uppercase:
3525                                         newChar = uppercase(oldChar);
3526                                         break;
3527                         }
3528                 }
3529
3530                 if (isWordSeparator(pos) || isDeleted(pos)) {
3531                         // permit capitalization again
3532                         capitalize = true;
3533                 }
3534
3535                 if (oldChar != newChar) {
3536                         changes.push_back(make_pair(newChar, getFontSettings(bparams, pos)));
3537                         if (pos != right - 1)
3538                                 continue;
3539                         // step behind the changing area
3540                         pos++;
3541                 }
3542
3543                 int erasePos = pos - changes.size();
3544                 for (size_t i = 0; i < changes.size(); i++) {
3545                         insertChar(pos, changes[i].first,
3546                                    changes[i].second,
3547                                    trackChanges);
3548                         if (!eraseChar(erasePos, trackChanges)) {
3549                                 ++erasePos;
3550                                 ++pos; // advance
3551                                 ++right; // expand selection
3552                         }
3553                 }
3554                 changes.clear();
3555         }
3556 }
3557
3558
3559 int Paragraph::find(docstring const & str, bool cs, bool mw,
3560                 pos_type start_pos, bool del) const
3561 {
3562         pos_type pos = start_pos;
3563         int const strsize = str.length();
3564         int i = 0;
3565         pos_type const parsize = d->text_.size();
3566         for (i = 0; i < strsize && pos < parsize; ++i, ++pos) {
3567                 // Ignore "invisible" letters such as ligature breaks
3568                 // and hyphenation chars while searching
3569                 while (pos < parsize - 1 && isInset(pos)) {
3570                         odocstringstream os;
3571                         getInset(pos)->toString(os);
3572                         if (!getInset(pos)->isLetter() || !os.str().empty())
3573                                 break;
3574                         pos++;
3575                 }
3576                 if (cs && str[i] != d->text_[pos])
3577                         break;
3578                 if (!cs && uppercase(str[i]) != uppercase(d->text_[pos]))
3579                         break;
3580                 if (!del && isDeleted(pos))
3581                         break;
3582         }
3583
3584         if (i != strsize)
3585                 return 0;
3586
3587         // if necessary, check whether string matches word
3588         if (mw) {
3589                 if (start_pos > 0 && !isWordSeparator(start_pos - 1))
3590                         return 0;
3591                 if (pos < parsize
3592                         && !isWordSeparator(pos))
3593                         return 0;
3594         }
3595
3596         return pos - start_pos;
3597 }
3598
3599
3600 char_type Paragraph::getChar(pos_type pos) const
3601 {
3602         return d->text_[pos];
3603 }
3604
3605
3606 pos_type Paragraph::size() const
3607 {
3608         return d->text_.size();
3609 }
3610
3611
3612 bool Paragraph::empty() const
3613 {
3614         return d->text_.empty();
3615 }
3616
3617
3618 bool Paragraph::isInset(pos_type pos) const
3619 {
3620         return d->text_[pos] == META_INSET;
3621 }
3622
3623
3624 bool Paragraph::isSeparator(pos_type pos) const
3625 {
3626         //FIXME: Are we sure this can be the only separator?
3627         return d->text_[pos] == ' ';
3628 }
3629
3630
3631 void Paragraph::deregisterWords()
3632 {
3633         Private::LangWordsMap::const_iterator itl = d->words_.begin();
3634         Private::LangWordsMap::const_iterator ite = d->words_.end();
3635         for (; itl != ite; ++itl) {
3636                 WordList * wl = theWordList(itl->first);
3637                 Private::Words::const_iterator it = (itl->second).begin();
3638                 Private::Words::const_iterator et = (itl->second).end();
3639                 for (; it != et; ++it)
3640                         wl->remove(*it);
3641         }
3642         d->words_.clear();
3643 }
3644
3645
3646 void Paragraph::locateWord(pos_type & from, pos_type & to,
3647         word_location const loc) const
3648 {
3649         switch (loc) {
3650         case WHOLE_WORD_STRICT:
3651                 if (from == 0 || from == size()
3652                     || isWordSeparator(from)
3653                     || isWordSeparator(from - 1)) {
3654                         to = from;
3655                         return;
3656                 }
3657                 // no break here, we go to the next
3658
3659         case WHOLE_WORD:
3660                 // If we are already at the beginning of a word, do nothing
3661                 if (!from || isWordSeparator(from - 1))
3662                         break;
3663                 // no break here, we go to the next
3664
3665         case PREVIOUS_WORD:
3666                 // always move the cursor to the beginning of previous word
3667                 while (from && !isWordSeparator(from - 1))
3668                         --from;
3669                 break;
3670         case NEXT_WORD:
3671                 LYXERR0("Paragraph::locateWord: NEXT_WORD not implemented yet");
3672                 break;
3673         case PARTIAL_WORD:
3674                 // no need to move the 'from' cursor
3675                 break;
3676         }
3677         to = from;
3678         while (to < size() && !isWordSeparator(to))
3679                 ++to;
3680 }
3681
3682
3683 void Paragraph::collectWords()
3684 {
3685         for (pos_type pos = 0; pos < size(); ++pos) {
3686                 if (isWordSeparator(pos))
3687                         continue;
3688                 pos_type from = pos;
3689                 locateWord(from, pos, WHOLE_WORD);
3690                 // Work around MSVC warning: The statement
3691                 // if (pos < from + lyxrc.completion_minlength)
3692                 // triggers a signed vs. unsigned warning.
3693                 // I don't know why this happens, it could be a MSVC bug, or
3694                 // related to LLP64 (windows) vs. LP64 (unix) programming
3695                 // model, or the C++ standard might be ambigous in the section
3696                 // defining the "usual arithmetic conversions". However, using
3697                 // a temporary variable is safe and works on all compilers.
3698                 pos_type const endpos = from + lyxrc.completion_minlength;
3699                 if (pos < endpos)
3700                         continue;
3701                 FontList::const_iterator cit = d->fontlist_.fontIterator(from);
3702                 if (cit == d->fontlist_.end())
3703                         return;
3704                 Language const * lang = cit->font().language();
3705                 docstring const word = asString(from, pos, AS_STR_NONE);
3706                 d->words_[lang->lang()].insert(word);
3707         }
3708 }
3709
3710
3711 void Paragraph::registerWords()
3712 {
3713         Private::LangWordsMap::const_iterator itl = d->words_.begin();
3714         Private::LangWordsMap::const_iterator ite = d->words_.end();
3715         for (; itl != ite; ++itl) {
3716                 WordList * wl = theWordList(itl->first);
3717                 Private::Words::const_iterator it = (itl->second).begin();
3718                 Private::Words::const_iterator et = (itl->second).end();
3719                 for (; it != et; ++it)
3720                         wl->insert(*it);
3721         }
3722 }
3723
3724
3725 void Paragraph::updateWords()
3726 {
3727         deregisterWords();
3728         collectWords();
3729         registerWords();
3730 }
3731
3732
3733 void Paragraph::Private::appendSkipPosition(SkipPositions & skips, pos_type const pos) const
3734 {
3735         SkipPositionsIterator begin = skips.begin();
3736         SkipPositions::iterator end = skips.end();
3737         if (pos > 0 && begin < end) {
3738                 --end;
3739                 if (end->last == pos - 1) {
3740                         end->last = pos;
3741                         return;
3742                 }
3743         }
3744         skips.insert(end, FontSpan(pos, pos));
3745 }
3746
3747
3748 Language * Paragraph::Private::locateSpellRange(
3749         pos_type & from, pos_type & to,
3750         SkipPositions & skips) const
3751 {
3752         // skip leading white space
3753         while (from < to && owner_->isWordSeparator(from))
3754                 ++from;
3755         // don't check empty range
3756         if (from >= to)
3757                 return 0;
3758         // get current language
3759         Language * lang = getSpellLanguage(from);
3760         pos_type last = from;
3761         bool samelang = true;
3762         bool sameinset = true;
3763         while (last < to && samelang && sameinset) {
3764                 // hop to end of word
3765                 while (last < to && !owner_->isWordSeparator(last)) {
3766                         if (owner_->getInset(last)) {
3767                                 appendSkipPosition(skips, last);
3768                         } else if (owner_->isDeleted(last)) {
3769                                 appendSkipPosition(skips, last);
3770                         }
3771                         ++last;
3772                 }
3773                 // hop to next word while checking for insets
3774                 while (sameinset && last < to && owner_->isWordSeparator(last)) {
3775                         if (Inset const * inset = owner_->getInset(last))
3776                                 sameinset = inset->isChar() && inset->isLetter();
3777                         if (sameinset && owner_->isDeleted(last)) {
3778                                 appendSkipPosition(skips, last);
3779                         }
3780                         if (sameinset)
3781                                 last++;
3782                 }
3783                 if (sameinset && last < to) {
3784                         // now check for language change
3785                         samelang = lang == getSpellLanguage(last);
3786                 }
3787         }
3788         // if language change detected backstep is needed
3789         if (!samelang)
3790                 --last;
3791         to = last;
3792         return lang;
3793 }
3794
3795
3796 Language * Paragraph::Private::getSpellLanguage(pos_type const from) const
3797 {
3798         Language * lang =
3799                 const_cast<Language *>(owner_->getFontSettings(
3800                         inset_owner_->buffer().params(), from).language());
3801         if (lang == inset_owner_->buffer().params().language
3802                 && !lyxrc.spellchecker_alt_lang.empty()) {
3803                 string lang_code;
3804                 string const lang_variety =
3805                         split(lyxrc.spellchecker_alt_lang, lang_code, '-');
3806                 lang->setCode(lang_code);
3807                 lang->setVariety(lang_variety);
3808         }
3809         return lang;
3810 }
3811
3812
3813 void Paragraph::requestSpellCheck(pos_type pos)
3814 {
3815         d->requestSpellCheck(pos);
3816 }
3817
3818
3819 bool Paragraph::needsSpellCheck() const
3820 {
3821         SpellChecker::ChangeNumber speller_change_number = 0;
3822         if (theSpellChecker())
3823                 speller_change_number = theSpellChecker()->changeNumber();
3824         if (speller_change_number > d->speller_state_.currentChangeNumber()) {
3825                 d->speller_state_.needsCompleteRefresh(speller_change_number);
3826         }
3827         return d->needsSpellCheck();
3828 }
3829
3830
3831 bool Paragraph::Private::ignoreWord(docstring const & word) const
3832 {
3833         // Ignore words with digits
3834         // FIXME: make this customizable
3835         // (note that some checkers ignore words with digits by default)
3836         docstring::const_iterator cit = word.begin();
3837         docstring::const_iterator const end = word.end();
3838         for (; cit != end; ++cit) {
3839                 if (isNumber((*cit)))
3840                         return true;
3841         }
3842         return false;
3843 }
3844
3845
3846 SpellChecker::Result Paragraph::spellCheck(pos_type & from, pos_type & to,
3847         WordLangTuple & wl, docstring_list & suggestions,
3848         bool do_suggestion, bool check_learned) const
3849 {
3850         SpellChecker::Result result = SpellChecker::WORD_OK;
3851         SpellChecker * speller = theSpellChecker();
3852         if (!speller)
3853                 return result;
3854
3855         if (!d->layout_->spellcheck || !inInset().allowSpellCheck())
3856                 return result;
3857
3858         locateWord(from, to, WHOLE_WORD);
3859         if (from == to || from >= size())
3860                 return result;
3861
3862         docstring word = asString(from, to, AS_STR_INSETS | AS_STR_SKIPDELETE);
3863         Language * lang = d->getSpellLanguage(from);
3864
3865         wl = WordLangTuple(word, lang);
3866
3867         if (word.empty())
3868                 return result;
3869
3870         if (needsSpellCheck() || check_learned) {
3871                 pos_type end = to;
3872                 if (!d->ignoreWord(word)) {
3873                         bool const trailing_dot = to < size() && d->text_[to] == '.';
3874                         result = speller->check(wl);
3875                         if (SpellChecker::misspelled(result) && trailing_dot) {
3876                                 wl = WordLangTuple(word.append(from_ascii(".")), lang);
3877                                 result = speller->check(wl);
3878                                 if (!SpellChecker::misspelled(result)) {
3879                                         LYXERR(Debug::GUI, "misspelled word is correct with dot: \"" <<
3880                                            word << "\" [" <<
3881                                            from << ".." << to << "]");
3882                                 } else {
3883                                         // spell check with dot appended failed too
3884                                         // restore original word/lang value
3885                                         word = asString(from, to, AS_STR_INSETS | AS_STR_SKIPDELETE);
3886                                         wl = WordLangTuple(word, lang);
3887                                 }
3888                         }
3889                 }
3890                 if (!SpellChecker::misspelled(result)) {
3891                         // area up to the begin of the next word is not misspelled
3892                         while (end < size() && isWordSeparator(end))
3893                                 ++end;
3894                 }
3895                 d->setMisspelled(from, end, result);
3896         } else {
3897                 result = d->speller_state_.getState(from);
3898         }
3899
3900         if (do_suggestion)
3901                 suggestions.clear();
3902
3903         if (SpellChecker::misspelled(result)) {
3904                 LYXERR(Debug::GUI, "misspelled word: \"" <<
3905                            word << "\" [" <<
3906                            from << ".." << to << "]");
3907                 if (do_suggestion)
3908                         speller->suggest(wl, suggestions);
3909         }
3910         return result;
3911 }
3912
3913
3914 void Paragraph::Private::markMisspelledWords(
3915         pos_type const & first, pos_type const & last,
3916         SpellChecker::Result result,
3917         docstring const & word,
3918         SkipPositions const & skips)
3919 {
3920         if (!SpellChecker::misspelled(result)) {
3921                 setMisspelled(first, last, SpellChecker::WORD_OK);
3922                 return;
3923         }
3924         int snext = first;
3925         SpellChecker * speller = theSpellChecker();
3926         // locate and enumerate the error positions
3927         int nerrors = speller->numMisspelledWords();
3928         int numskipped = 0;
3929         SkipPositionsIterator it = skips.begin();
3930         SkipPositionsIterator et = skips.end();
3931         for (int index = 0; index < nerrors; ++index) {
3932                 int wstart;
3933                 int wlen = 0;
3934                 speller->misspelledWord(index, wstart, wlen);
3935                 /// should not happen if speller supports range checks
3936                 if (!wlen) continue;
3937                 docstring const misspelled = word.substr(wstart, wlen);
3938                 wstart += first + numskipped;
3939                 if (snext < wstart) {
3940                         /// mark the range of correct spelling
3941                         numskipped += countSkips(it, et, wstart);
3942                         setMisspelled(snext,
3943                                 wstart - 1, SpellChecker::WORD_OK);
3944                 }
3945                 snext = wstart + wlen;
3946                 numskipped += countSkips(it, et, snext);
3947                 /// mark the range of misspelling
3948                 setMisspelled(wstart, snext, result);
3949                 LYXERR(Debug::GUI, "misspelled word: \"" <<
3950                            misspelled << "\" [" <<
3951                            wstart << ".." << (snext-1) << "]");
3952                 ++snext;
3953         }
3954         if (snext <= last) {
3955                 /// mark the range of correct spelling at end
3956                 setMisspelled(snext, last, SpellChecker::WORD_OK);
3957         }
3958 }
3959
3960
3961 void Paragraph::spellCheck() const
3962 {
3963         SpellChecker * speller = theSpellChecker();
3964         if (!speller || empty() ||!needsSpellCheck())
3965                 return;
3966         pos_type start;
3967         pos_type endpos;
3968         d->rangeOfSpellCheck(start, endpos);
3969         if (speller->canCheckParagraph()) {
3970                 // loop until we leave the range
3971                 for (pos_type first = start; first < endpos; ) {
3972                         pos_type last = endpos;
3973                         Private::SkipPositions skips;
3974                         Language * lang = d->locateSpellRange(first, last, skips);
3975                         if (first >= endpos)
3976                                 break;
3977                         // start the spell checker on the unit of meaning
3978                         docstring word = asString(first, last, AS_STR_INSETS + AS_STR_SKIPDELETE);
3979                         WordLangTuple wl = WordLangTuple(word, lang);
3980                         SpellChecker::Result result = word.size() ?
3981                                 speller->check(wl) : SpellChecker::WORD_OK;
3982                         d->markMisspelledWords(first, last, result, word, skips);
3983                         first = ++last;
3984                 }
3985         } else {
3986                 static docstring_list suggestions;
3987                 pos_type to = endpos;
3988                 while (start < endpos) {
3989                         WordLangTuple wl;
3990                         spellCheck(start, to, wl, suggestions, false);
3991                         start = to + 1;
3992                 }
3993         }
3994         d->readySpellCheck();
3995 }
3996
3997
3998 bool Paragraph::isMisspelled(pos_type pos, bool check_boundary) const
3999 {
4000         bool result = SpellChecker::misspelled(d->speller_state_.getState(pos));
4001         if (result || pos <= 0 || pos > size())
4002                 return result;
4003         if (check_boundary && (pos == size() || isWordSeparator(pos)))
4004                 result = SpellChecker::misspelled(d->speller_state_.getState(pos - 1));
4005         return result;
4006 }
4007
4008
4009 string Paragraph::magicLabel() const
4010 {
4011         stringstream ss;
4012         ss << "magicparlabel-" << id();
4013         return ss.str();
4014 }
4015
4016
4017 } // namespace lyx