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 "texstream.h"
  44 #include "TextClass.h"
  45 #include "TexRow.h"
  46 #include "Text.h"
  47 #include "WordLangTuple.h"
  48 #include "WordList.h"
  49
  50 #include "frontends/alert.h"
  51
  52 #include "insets/InsetBibitem.h"
  53 #include "insets/InsetLabel.h"
  54 #include "insets/InsetSpecialChar.h"
  55
  56 #include "mathed/InsetMathHull.h"
  57
  58 #include "support/debug.h"
  59 #include "support/docstring_list.h"
  60 #include "support/ExceptionMessage.h"
  61 #include "support/gettext.h"
  62 #include "support/lassert.h"
  63 #include "support/lstrings.h"
  64 #include "support/textutils.h"
  65
  66 #include <atomic>
  67 #include <sstream>
  68 #include <vector>
  69
  70 using namespace std;
  71 using namespace lyx::support;
  72
  73 namespace lyx {
  74
  75 namespace {
  76
  77 /// Inset identifier (above 0x10ffff, for ucs-4)
  78 char_type const META_INSET = 0x200001;
  79
  80 }
  81
  82
  83 /////////////////////////////////////////////////////////////////////
  84 //
  85 // SpellResultRange
  86 //
  87 /////////////////////////////////////////////////////////////////////
  88
  89 class SpellResultRange {
  90 public:
  91         SpellResultRange(FontSpan range, SpellChecker::Result result)
  92         : range_(range), result_(result)
  93         {}
  94         ///
  95         FontSpan const & range() const { return range_; }
  96         ///
  97         void range(FontSpan const & r) { range_ = r; }
  98         ///
  99         SpellChecker::Result result() const { return result_; }
 100         ///
 101         void result(SpellChecker::Result r) { result_ = r; }
 102         ///
 103         bool contains(pos_type pos) const { return range_.contains(pos); }
 104         ///
 105         bool covered(FontSpan const & r) const
 106         {
 107                 // 1. first of new range inside current range or
 108                 // 2. last of new range inside current range or
 109                 // 3. first of current range inside new range or
 110                 // 4. last of current range inside new range
 111                 //FIXME: is this the same as !range_.intersect(r).empty() ?
 112                 return range_.contains(r.first) || range_.contains(r.last) ||
 113                         r.contains(range_.first) || r.contains(range_.last);
 114         }
 115         ///
 116         void shift(pos_type pos, int offset)
 117         {
 118                 if (range_.first > pos) {
 119                         range_.first += offset;
 120                         range_.last += offset;
 121                 } else if (range_.last >= pos) {
 122                         range_.last += offset;
 123                 }
 124         }
 125 private:
 126         FontSpan range_ ;
 127         SpellChecker::Result result_ ;
 128 };
 129
 130
 131 /////////////////////////////////////////////////////////////////////
 132 //
 133 // SpellCheckerState
 134 //
 135 /////////////////////////////////////////////////////////////////////
 136
 137 class SpellCheckerState {
 138 public:
 139         SpellCheckerState()
 140         {
 141                 needs_refresh_ = true;
 142                 current_change_number_ = 0;
 143         }
 144
 145         void setRange(FontSpan const & fp, SpellChecker::Result state)
 146         {
 147                 Ranges result;
 148                 RangesIterator et = ranges_.end();
 149                 RangesIterator it = ranges_.begin();
 150                 for (; it != et; ++it) {
 151                         if (!it->covered(fp))
 152                                 result.push_back(SpellResultRange(it->range(), it->result()));
 153                         else if (state == SpellChecker::WORD_OK) {
 154                                 // trim or split the current misspelled range
 155                                 // store misspelled ranges only
 156                                 FontSpan range = it->range();
 157                                 if (fp.first > range.first) {
 158                                         // misspelled area in front of WORD_OK
 159                                         range.last = fp.first - 1;
 160                                         result.push_back(SpellResultRange(range, it->result()));
 161                                         range = it->range();
 162                                 }
 163                                 if (fp.last < range.last) {
 164                                         // misspelled area after WORD_OK range
 165                                         range.first = fp.last + 1;
 166                                         result.push_back(SpellResultRange(range, it->result()));
 167                                 }
 168                         }
 169                 }
 170                 ranges_ = result;
 171                 if (state != SpellChecker::WORD_OK)
 172                         ranges_.push_back(SpellResultRange(fp, state));
 173         }
 174
 175         void increasePosAfterPos(pos_type pos)
 176         {
 177                 correctRangesAfterPos(pos, 1);
 178                 needsRefresh(pos);
 179         }
 180
 181         void decreasePosAfterPos(pos_type pos)
 182         {
 183                 correctRangesAfterPos(pos, -1);
 184                 needsRefresh(pos);
 185         }
 186
 187         void refreshLast(pos_type pos)
 188         {
 189                 if (pos < refresh_.last)
 190                         refresh_.last = pos;
 191         }
 192
 193         SpellChecker::Result getState(pos_type pos) const
 194         {
 195                 SpellChecker::Result result = SpellChecker::WORD_OK;
 196                 RangesIterator et = ranges_.end();
 197                 RangesIterator it = ranges_.begin();
 198                 for (; it != et; ++it) {
 199                         if(it->contains(pos)) {
 200                                 return it->result();
 201                         }
 202                 }
 203                 return result;
 204         }
 205
 206         FontSpan const & getRange(pos_type pos) const
 207         {
 208                 /// empty span to indicate mismatch
 209                 static FontSpan empty_;
 210                 RangesIterator et = ranges_.end();
 211                 RangesIterator it = ranges_.begin();
 212                 for (; it != et; ++it) {
 213                         if(it->contains(pos)) {
 214                                 return it->range();
 215                         }
 216                 }
 217                 return empty_;
 218         }
 219
 220         bool needsRefresh() const
 221         {
 222                 return needs_refresh_;
 223         }
 224
 225         SpellChecker::ChangeNumber currentChangeNumber() const
 226         {
 227                 return current_change_number_;
 228         }
 229
 230         void refreshRange(pos_type & first, pos_type & last) const
 231         {
 232                 first = refresh_.first;
 233                 last = refresh_.last;
 234         }
 235
 236         void needsRefresh(pos_type pos)
 237         {
 238                 if (needs_refresh_ && pos != -1) {
 239                         if (pos < refresh_.first)
 240                                 refresh_.first = pos;
 241                         if (pos > refresh_.last)
 242                                 refresh_.last = pos;
 243                 } else if (pos != -1) {
 244                         // init request check for neighbour positions too
 245                         refresh_.first = pos > 0 ? pos - 1 : 0;
 246                         // no need for special end of paragraph check
 247                         refresh_.last = pos + 1;
 248                 }
 249                 needs_refresh_ = pos != -1;
 250         }
 251
 252         void needsCompleteRefresh(SpellChecker::ChangeNumber change_number)
 253         {
 254                 needs_refresh_ = true;
 255                 refresh_.first = 0;
 256                 refresh_.last = -1;
 257                 current_change_number_ = change_number;
 258         }
 259 private:
 260         typedef vector<SpellResultRange> Ranges;
 261         typedef Ranges::const_iterator RangesIterator;
 262         Ranges ranges_;
 263         /// the area of the paragraph with pending spell check
 264         FontSpan refresh_;
 265         bool needs_refresh_;
 266         /// spell state cache version number
 267         SpellChecker::ChangeNumber current_change_number_;
 268
 269
 270         void correctRangesAfterPos(pos_type pos, int offset)
 271         {
 272                 RangesIterator et = ranges_.end();
 273                 Ranges::iterator it = ranges_.begin();
 274                 for (; it != et; ++it) {
 275                         it->shift(pos, offset);
 276                 }
 277         }
 278
 279 };
 280
 281 /////////////////////////////////////////////////////////////////////
 282 //
 283 // Paragraph::Private
 284 //
 285 /////////////////////////////////////////////////////////////////////
 286
 287 class Paragraph::Private
 288 {
 289         // Enforce our own "copy" constructor
 290         Private(Private const &) = delete;
 291         Private & operator=(Private const &) = delete;
 292         // Unique ID generator
 293         static int make_id();
 294 public:
 295         ///
 296         Private(Paragraph * owner, Layout const & layout);
 297         /// "Copy constructor"
 298         Private(Private const &, Paragraph * owner);
 299         /// Copy constructor from \p beg  to \p end
 300         Private(Private const &, Paragraph * owner, pos_type beg, pos_type end);
 301
 302         ///
 303         void insertChar(pos_type pos, char_type c, Change const & change);
 304
 305         /// Output the surrogate pair formed by \p c and \p next to \p os.
 306         /// \return the number of characters written.
 307         int latexSurrogatePair(otexstream & os, char_type c, char_type next,
 308                                OutputParams const &);
 309
 310         /// Output a space in appropriate formatting (or a surrogate pair
 311         /// if the next character is a combining character).
 312         /// \return whether a surrogate pair was output.
 313         bool simpleTeXBlanks(OutputParams const &,
 314                              otexstream &,
 315                              pos_type i,
 316                              unsigned int & column,
 317                              Font const & font,
 318                              Layout const & style);
 319
 320         /// Output consecutive unicode chars, belonging to the same script as
 321         /// specified by the latex macro \p ltx, to \p os starting from \p i.
 322         /// \return the number of characters written.
 323         int writeScriptChars(otexstream & os, docstring const & ltx,
 324                            Change const &, Encoding const &, pos_type & i);
 325
 326         /// This could go to ParagraphParameters if we want to.
 327         int startTeXParParams(BufferParams const &, otexstream &,
 328                               OutputParams const &) const;
 329
 330         /// This could go to ParagraphParameters if we want to.
 331         bool endTeXParParams(BufferParams const &, otexstream &,
 332                              OutputParams const &) const;
 333
 334         ///
 335         void latexInset(BufferParams const &,
 336                                    otexstream &,
 337                                    OutputParams &,
 338                                    Font & running_font,
 339                                    Font & basefont,
 340                                    Font const & outerfont,
 341                                    bool & open_font,
 342                                    Change & running_change,
 343                                    Layout const & style,
 344                                    pos_type & i,
 345                                    unsigned int & column);
 346
 347         ///
 348         void latexSpecialChar(
 349                                    otexstream & os,
 350                                    BufferParams const & bparams,
 351                                    OutputParams const & runparams,
 352                                    Font const & running_font,
 353                                    Change const & running_change,
 354                                    Layout const & style,
 355                                    pos_type & i,
 356                                    pos_type end_pos,
 357                                    unsigned int & column);
 358
 359         ///
 360         bool latexSpecialT1(
 361                 char_type const c,
 362                 otexstream & os,
 363                 pos_type i,
 364                 unsigned int & column);
 365         ///
 366         bool latexSpecialT3(
 367                 char_type const c,
 368                 otexstream & os,
 369                 pos_type i,
 370                 unsigned int & column);
 371
 372         ///
 373         void validate(LaTeXFeatures & features) const;
 374
 375         /// Checks if the paragraph contains only text and no inset or font change.
 376         bool onlyText(Buffer const & buf, Font const & outerfont,
 377                       pos_type initial) const;
 378
 379         /// a vector of speller skip positions
 380         typedef vector<FontSpan> SkipPositions;
 381         typedef SkipPositions::const_iterator SkipPositionsIterator;
 382
 383         void appendSkipPosition(SkipPositions & skips, pos_type const pos) const;
 384
 385         Language * getSpellLanguage(pos_type const from) const;
 386
 387         Language * locateSpellRange(pos_type & from, pos_type & to,
 388                                     SkipPositions & skips) const;
 389
 390         bool hasSpellerChange() const
 391         {
 392                 SpellChecker::ChangeNumber speller_change_number = 0;
 393                 if (theSpellChecker())
 394                         speller_change_number = theSpellChecker()->changeNumber();
 395                 return speller_change_number > speller_state_.currentChangeNumber();
 396         }
 397
 398         bool ignoreWord(docstring const & word) const ;
 399
 400         void setMisspelled(pos_type from, pos_type to, SpellChecker::Result state)
 401         {
 402                 pos_type textsize = owner_->size();
 403                 // check for sane arguments
 404                 if (to <= from || from >= textsize)
 405                         return;
 406                 FontSpan fp = FontSpan(from, to - 1);
 407                 speller_state_.setRange(fp, state);
 408         }
 409
 410         void requestSpellCheck(pos_type pos)
 411         {
 412                 if (pos == -1)
 413                         speller_state_.needsCompleteRefresh(speller_state_.currentChangeNumber());
 414                 else
 415                         speller_state_.needsRefresh(pos);
 416         }
 417
 418         void readySpellCheck()
 419         {
 420                 speller_state_.needsRefresh(-1);
 421         }
 422
 423         bool needsSpellCheck() const
 424         {
 425                 return speller_state_.needsRefresh();
 426         }
 427
 428         void rangeOfSpellCheck(pos_type & first, pos_type & last) const
 429         {
 430                 speller_state_.refreshRange(first, last);
 431                 if (last == -1) {
 432                         last = owner_->size();
 433                         return;
 434                 }
 435                 pos_type endpos = last;
 436                 owner_->locateWord(first, endpos, WHOLE_WORD);
 437                 if (endpos < last) {
 438                         endpos = last;
 439                         owner_->locateWord(last, endpos, WHOLE_WORD);
 440                 }
 441                 last = endpos;
 442         }
 443
 444         int countSkips(SkipPositionsIterator & it, SkipPositionsIterator const et,
 445                             int & start) const
 446         {
 447                 int numskips = 0;
 448                 while (it != et && it->first < start) {
 449                         int skip = it->last - it->first + 1;
 450                         start += skip;
 451                         numskips += skip;
 452                         ++it;
 453                 }
 454                 return numskips;
 455         }
 456
 457         void markMisspelledWords(pos_type const & first, pos_type const & last,
 458                                                          SpellChecker::Result result,
 459                                                          docstring const & word,
 460                                                          SkipPositions const & skips);
 461
 462         InsetCode ownerCode() const
 463         {
 464                 return inset_owner_ ? inset_owner_->lyxCode() : NO_CODE;
 465         }
 466
 467         /// Which Paragraph owns us?
 468         Paragraph * owner_;
 469
 470         /// In which Inset?
 471         Inset const * inset_owner_;
 472
 473         ///
 474         FontList fontlist_;
 475
 476         ///
 477         int id_;
 478
 479         ///
 480         ParagraphParameters params_;
 481
 482         /// for recording and looking up changes
 483         Changes changes_;
 484
 485         ///
 486         InsetList insetlist_;
 487
 488         /// end of label
 489         pos_type begin_of_body_;
 490
 491         typedef docstring TextContainer;
 492         ///
 493         TextContainer text_;
 494
 495         typedef set<docstring> Words;
 496         typedef map<string, Words> LangWordsMap;
 497         ///
 498         LangWordsMap words_;
 499         ///
 500         Layout const * layout_;
 501         ///
 502         SpellCheckerState speller_state_;
 503 };
 504
 505
 506 Paragraph::Private::Private(Paragraph * owner, Layout const & layout)
 507         : owner_(owner), inset_owner_(0), id_(-1), begin_of_body_(0), layout_(&layout)
 508 {
 509         text_.reserve(100);
 510 }
 511
 512
 513 //static
 514 int Paragraph::Private::make_id()
 515 {
 516         // The id is unique per session across buffers because it is used in
 517         // LFUN_PARAGRAPH_GOTO to switch to a different buffer, for instance in the
 518         // outliner.
 519         // (thread-safe)
 520         static atomic_uint next_id(0);
 521         return next_id++;
 522 }
 523
 524
 525 Paragraph::Private::Private(Private const & p, Paragraph * owner)
 526         : owner_(owner), inset_owner_(p.inset_owner_), fontlist_(p.fontlist_),
 527           id_(make_id()),
 528           params_(p.params_), changes_(p.changes_), insetlist_(p.insetlist_),
 529           begin_of_body_(p.begin_of_body_), text_(p.text_), words_(p.words_),
 530           layout_(p.layout_)
 531 {
 532         requestSpellCheck(p.text_.size());
 533 }
 534
 535
 536 Paragraph::Private::Private(Private const & p, Paragraph * owner,
 537         pos_type beg, pos_type end)
 538         : owner_(owner), inset_owner_(p.inset_owner_), id_(make_id()),
 539           params_(p.params_), changes_(p.changes_),
 540           insetlist_(p.insetlist_, beg, end),
 541           begin_of_body_(p.begin_of_body_), words_(p.words_),
 542           layout_(p.layout_)
 543 {
 544         if (beg >= pos_type(p.text_.size()))
 545                 return;
 546         text_ = p.text_.substr(beg, end - beg);
 547
 548         FontList::const_iterator fcit = fontlist_.begin();
 549         FontList::const_iterator fend = fontlist_.end();
 550         for (; fcit != fend; ++fcit) {
 551                 if (fcit->pos() < beg)
 552                         continue;
 553                 if (fcit->pos() >= end) {
 554                         // Add last entry in the fontlist_.
 555                         fontlist_.set(text_.size() - 1, fcit->font());
 556                         break;
 557                 }
 558                 // Add a new entry in the fontlist_.
 559                 fontlist_.set(fcit->pos() - beg, fcit->font());
 560         }
 561         requestSpellCheck(p.text_.size());
 562 }
 563
 564
 565 void Paragraph::addChangesToToc(DocIterator const & cdit,
 566         Buffer const & buf, bool output_active) const
 567 {
 568         d->changes_.addToToc(cdit, buf, output_active);
 569 }
 570
 571
 572 void Paragraph::addChangesToBuffer(Buffer const & buf) const
 573 {
 574         d->changes_.updateBuffer(buf);
 575 }
 576
 577
 578 bool Paragraph::isChangeUpdateRequired() const
 579 {
 580         return d->changes_.isUpdateRequired();
 581 }
 582
 583
 584 bool Paragraph::isDeleted(pos_type start, pos_type end) const
 585 {
 586         LASSERT(start >= 0 && start <= size(), return false);
 587         LASSERT(end > start && end <= size() + 1, return false);
 588
 589         return d->changes_.isDeleted(start, end);
 590 }
 591
 592
 593 bool Paragraph::isChanged(pos_type start, pos_type end) const
 594 {
 595         LASSERT(start >= 0 && start <= size(), return false);
 596         LASSERT(end > start && end <= size() + 1, return false);
 597
 598         return d->changes_.isChanged(start, end);
 599 }
 600
 601
 602 bool Paragraph::isMergedOnEndOfParDeletion(bool trackChanges) const
 603 {
 604         // keep the logic here in sync with the logic of eraseChars()
 605         if (!trackChanges)
 606                 return true;
 607
 608         Change const & change = d->changes_.lookup(size());
 609         return change.inserted() && change.currentAuthor();
 610 }
 611
 612
 613 void Paragraph::setChange(Change const & change)
 614 {
 615         // beware of the imaginary end-of-par character!
 616         d->changes_.set(change, 0, size() + 1);
 617
 618         /*
 619          * Propagate the change recursively - but not in case of DELETED!
 620          *
 621          * Imagine that your co-author makes changes in an existing inset. He
 622          * sends your document to you and you come to the conclusion that the
 623          * inset should go completely. If you erase it, LyX must not delete all
 624          * text within the inset. Otherwise, the change tracked insertions of
 625          * your co-author get lost and there is no way to restore them later.
 626          *
 627          * Conclusion: An inset's content should remain untouched if you delete it
 628          */
 629
 630         if (!change.deleted()) {
 631                 for (pos_type pos = 0; pos < size(); ++pos) {
 632                         if (Inset * inset = getInset(pos))
 633                                 inset->setChange(change);
 634                 }
 635         }
 636 }
 637
 638
 639 void Paragraph::setChange(pos_type pos, Change const & change)
 640 {
 641         LASSERT(pos >= 0 && pos <= size(), return);
 642         d->changes_.set(change, pos);
 643
 644         // see comment in setChange(Change const &) above
 645         if (!change.deleted() && pos < size())
 646                         if (Inset * inset = getInset(pos))
 647                                 inset->setChange(change);
 648 }
 649
 650
 651 Change const & Paragraph::lookupChange(pos_type pos) const
 652 {
 653         LBUFERR(pos >= 0 && pos <= size());
 654         return d->changes_.lookup(pos);
 655 }
 656
 657
 658 void Paragraph::acceptChanges(pos_type start, pos_type end)
 659 {
 660         LASSERT(start >= 0 && start <= size(), return);
 661         LASSERT(end > start && end <= size() + 1, return);
 662
 663         for (pos_type pos = start; pos < end; ++pos) {
 664                 switch (lookupChange(pos).type) {
 665                         case Change::UNCHANGED:
 666                                 // accept changes in nested inset
 667                                 if (Inset * inset = getInset(pos))
 668                                         inset->acceptChanges();
 669                                 break;
 670
 671                         case Change::INSERTED:
 672                                 d->changes_.set(Change(Change::UNCHANGED), pos);
 673                                 // also accept changes in nested inset
 674                                 if (Inset * inset = getInset(pos))
 675                                         inset->acceptChanges();
 676                                 break;
 677
 678                         case Change::DELETED:
 679                                 // Suppress access to non-existent
 680                                 // "end-of-paragraph char"
 681                                 if (pos < size()) {
 682                                         eraseChar(pos, false);
 683                                         --end;
 684                                         --pos;
 685                                 }
 686                                 break;
 687                 }
 688
 689         }
 690 }
 691
 692
 693 void Paragraph::rejectChanges(pos_type start, pos_type end)
 694 {
 695         LASSERT(start >= 0 && start <= size(), return);
 696         LASSERT(end > start && end <= size() + 1, return);
 697
 698         for (pos_type pos = start; pos < end; ++pos) {
 699                 switch (lookupChange(pos).type) {
 700                         case Change::UNCHANGED:
 701                                 // reject changes in nested inset
 702                                 if (Inset * inset = getInset(pos))
 703                                                 inset->rejectChanges();
 704                                 break;
 705
 706                         case Change::INSERTED:
 707                                 // Suppress access to non-existent
 708                                 // "end-of-paragraph char"
 709                                 if (pos < size()) {
 710                                         eraseChar(pos, false);
 711                                         --end;
 712                                         --pos;
 713                                 }
 714                                 break;
 715
 716                         case Change::DELETED:
 717                                 d->changes_.set(Change(Change::UNCHANGED), pos);
 718
 719                                 // Do NOT reject changes within a deleted inset!
 720                                 // There may be insertions of a co-author inside of it!
 721
 722                                 break;
 723                 }
 724         }
 725 }
 726
 727
 728 void Paragraph::Private::insertChar(pos_type pos, char_type c,
 729                 Change const & change)
 730 {
 731         LASSERT(pos >= 0 && pos <= int(text_.size()), return);
 732
 733         // track change
 734         changes_.insert(change, pos);
 735
 736         // This is actually very common when parsing buffers (and
 737         // maybe inserting ascii text)
 738         if (pos == pos_type(text_.size())) {
 739                 // when appending characters, no need to update tables
 740                 text_.push_back(c);
 741                 // but we want spell checking
 742                 requestSpellCheck(pos);
 743                 return;
 744         }
 745
 746         text_.insert(text_.begin() + pos, c);
 747
 748         // Update the font table.
 749         fontlist_.increasePosAfterPos(pos);
 750
 751         // Update the insets
 752         insetlist_.increasePosAfterPos(pos);
 753
 754         // Update list of misspelled positions
 755         speller_state_.increasePosAfterPos(pos);
 756 }
 757
 758
 759 bool Paragraph::insertInset(pos_type pos, Inset * inset,
 760                                    Font const & font, Change const & change)
 761 {
 762         LASSERT(inset, return false);
 763         LASSERT(pos >= 0 && pos <= size(), return false);
 764
 765         // Paragraph::insertInset() can be used in cut/copy/paste operation where
 766         // d->inset_owner_ is not set yet.
 767         if (d->inset_owner_ && !d->inset_owner_->insetAllowed(inset->lyxCode()))
 768                 return false;
 769
 770         d->insertChar(pos, META_INSET, change);
 771         LASSERT(d->text_[pos] == META_INSET, return false);
 772
 773         // Add a new entry in the insetlist_.
 774         d->insetlist_.insert(inset, pos);
 775
 776         // Some insets require run of spell checker
 777         requestSpellCheck(pos);
 778         setFont(pos, font);
 779         return true;
 780 }
 781
 782
 783 bool Paragraph::eraseChar(pos_type pos, bool trackChanges)
 784 {
 785         LASSERT(pos >= 0 && pos <= size(), return false);
 786
 787         // keep the logic here in sync with the logic of isMergedOnEndOfParDeletion()
 788
 789         if (trackChanges) {
 790                 Change change = d->changes_.lookup(pos);
 791
 792                 // set the character to DELETED if
 793                 //  a) it was previously unchanged or
 794                 //  b) it was inserted by a co-author
 795
 796                 if (!change.changed() ||
 797                       (change.inserted() && !change.currentAuthor())) {
 798                         setChange(pos, Change(Change::DELETED));
 799                         // request run of spell checker
 800                         requestSpellCheck(pos);
 801                         return false;
 802                 }
 803
 804                 if (change.deleted())
 805                         return false;
 806         }
 807
 808         // Don't physically access the imaginary end-of-paragraph character.
 809         // eraseChar() can only mark it as DELETED. A physical deletion of
 810         // end-of-par must be handled externally.
 811         if (pos == size()) {
 812                 return false;
 813         }
 814
 815         // track change
 816         d->changes_.erase(pos);
 817
 818         // if it is an inset, delete the inset entry
 819         if (d->text_[pos] == META_INSET)
 820                 d->insetlist_.erase(pos);
 821
 822         d->text_.erase(d->text_.begin() + pos);
 823
 824         // Update the fontlist_
 825         d->fontlist_.erase(pos);
 826
 827         // Update the insetlist_
 828         d->insetlist_.decreasePosAfterPos(pos);
 829
 830         // Update list of misspelled positions
 831         d->speller_state_.decreasePosAfterPos(pos);
 832         d->speller_state_.refreshLast(size());
 833
 834         return true;
 835 }
 836
 837
 838 int Paragraph::eraseChars(pos_type start, pos_type end, bool trackChanges)
 839 {
 840         LASSERT(start >= 0 && start <= size(), return 0);
 841         LASSERT(end >= start && end <= size() + 1, return 0);
 842
 843         pos_type i = start;
 844         for (pos_type count = end - start; count; --count) {
 845                 if (!eraseChar(i, trackChanges))
 846                         ++i;
 847         }
 848         return end - i;
 849 }
 850
 851
 852 int Paragraph::Private::latexSurrogatePair(otexstream & os, char_type c,
 853                 char_type next, OutputParams const & runparams)
 854 {
 855         // Writing next here may circumvent a possible font change between
 856         // c and next. Since next is only output if it forms a surrogate pair
 857         // with c we can ignore this:
 858         // A font change inside a surrogate pair does not make sense and is
 859         // hopefully impossible to input.
 860         // FIXME: change tracking
 861         // Is this correct WRT change tracking?
 862         Encoding const & encoding = *(runparams.encoding);
 863         docstring latex1 = encoding.latexChar(next).first;
 864         if (runparams.inIPA) {
 865                 string const tipashortcut = Encodings::TIPAShortcut(next);
 866                 if (!tipashortcut.empty()) {
 867                         latex1 = from_ascii(tipashortcut);
 868                 }
 869         }
 870         docstring const latex2 = encoding.latexChar(c).first;
 871         if (docstring(1, next) == latex1) {
 872                 // the encoding supports the combination
 873                 os << latex2 << latex1;
 874                 return latex1.length() + latex2.length();
 875         } else if (runparams.local_font &&
 876                    runparams.local_font->language()->lang() == "polutonikogreek") {
 877                 // polutonikogreek only works without the brackets
 878                 os << latex1 << latex2;
 879                 return latex1.length() + latex2.length();
 880         } else
 881                 os << latex1 << '{' << latex2 << '}';
 882         return latex1.length() + latex2.length() + 2;
 883 }
 884
 885
 886 bool Paragraph::Private::simpleTeXBlanks(OutputParams const & runparams,
 887                                        otexstream & os,
 888                                        pos_type i,
 889                                        unsigned int & column,
 890                                        Font const & font,
 891                                        Layout const & style)
 892 {
 893         if (style.pass_thru || runparams.pass_thru)
 894                 return false;
 895
 896         if (i + 1 < int(text_.size())) {
 897                 char_type next = text_[i + 1];
 898                 if (Encodings::isCombiningChar(next)) {
 899                         // This space has an accent, so we must always output it.
 900                         column += latexSurrogatePair(os, ' ', next, runparams) - 1;
 901                         return true;
 902                 }
 903         }
 904
 905         if (runparams.linelen > 0
 906             && column > runparams.linelen
 907             && i
 908             && text_[i - 1] != ' '
 909             && (i + 1 < int(text_.size()))
 910             // same in FreeSpacing mode
 911             && !owner_->isFreeSpacing()
 912             // In typewriter mode, we want to avoid
 913             // ! . ? : at the end of a line
 914             && !(font.fontInfo().family() == TYPEWRITER_FAMILY
 915                  && (text_[i - 1] == '.'
 916                      || text_[i - 1] == '?'
 917                      || text_[i - 1] == ':'
 918                      || text_[i - 1] == '!'))) {
 919                 os << '\n';
 920                 os.texrow().start(owner_->id(), i + 1);
 921                 column = 0;
 922         } else if (style.free_spacing) {
 923                 os << '~';
 924         } else {
 925                 os << ' ';
 926         }
 927         return false;
 928 }
 929
 930
 931 int Paragraph::Private::writeScriptChars(otexstream & os,
 932                                          docstring const & ltx,
 933                                          Change const & runningChange,
 934                                          Encoding const & encoding,
 935                                          pos_type & i)
 936 {
 937         // FIXME: modifying i here is not very nice...
 938
 939         // We only arrive here when character text_[i] could not be translated
 940         // into the current latex encoding (or its latex translation has been forced,)
 941         // and it belongs to a known script.
 942         // TODO: We need \textcyr and \textgreek wrappers also for characters
 943         //       that can be encoded in the "LaTeX encoding" but not in the
 944         //       current *font encoding*.
 945         //       (See #9681 for details and test)
 946         // Parameter ltx contains the latex translation of text_[i] as specified
 947         // in the unicodesymbols file and is something like "\textXXX{<spec>}".
 948         // The latex macro name "textXXX" specifies the script to which text_[i]
 949         // belongs and we use it in order to check whether characters from the
 950         // same script immediately follow, such that we can collect them in a
 951         // single "\textXXX" macro. So, we have to retain "\textXXX{<spec>"
 952         // for the first char but only "<spec>" for all subsequent chars.
 953         docstring::size_type const brace1 = ltx.find_first_of(from_ascii("{"));
 954         docstring::size_type const brace2 = ltx.find_last_of(from_ascii("}"));
 955         string script = to_ascii(ltx.substr(1, brace1 - 1));
 956         int pos = 0;
 957         int length = brace2;
 958         bool closing_brace = true;
 959         if (script == "textgreek" && encoding.latexName() == "iso-8859-7") {
 960                 // Correct encoding is being used, so we can avoid \textgreek.
 961                 // TODO: wrong test: we need to check the *font encoding*
 962                 //       (i.e. the active language and its FontEncoding tag)
 963                 //       instead of the LaTeX *input encoding*!
 964                 //       See #9637 for details and test-cases.
 965                 pos = brace1 + 1;
 966                 length -= pos;
 967                 closing_brace = false;
 968         }
 969         os << ltx.substr(pos, length);
 970         int size = text_.size();
 971         while (i + 1 < size) {
 972                 char_type const next = text_[i + 1];
 973                 // Stop here if next character belongs to another script
 974                 // or there is a change in change tracking status.
 975                 if (!Encodings::isKnownScriptChar(next, script) ||
 976                     runningChange != owner_->lookupChange(i + 1))
 977                         break;
 978                 Font prev_font;
 979                 bool found = false;
 980                 FontList::const_iterator cit = fontlist_.begin();
 981                 FontList::const_iterator end = fontlist_.end();
 982                 for (; cit != end; ++cit) {
 983                         if (cit->pos() >= i && !found) {
 984                                 prev_font = cit->font();
 985                                 found = true;
 986                         }
 987                         if (cit->pos() >= i + 1)
 988                                 break;
 989                 }
 990                 // Stop here if there is a font attribute or encoding change.
 991                 if (found && cit != end && prev_font != cit->font())
 992                         break;
 993                 docstring const latex = encoding.latexChar(next).first;
 994                 docstring::size_type const b1 =
 995                                         latex.find_first_of(from_ascii("{"));
 996                 docstring::size_type const b2 =
 997                                         latex.find_last_of(from_ascii("}"));
 998                 int const len = b2 - b1 - 1;
 999                 os << latex.substr(b1 + 1, len);
1000                 length += len;
1001                 ++i;
1002         }
1003         if (closing_brace) {
1004                 os << '}';
1005                 ++length;
1006         }
1007         return length;
1008 }
1009
1010
1011 void Paragraph::Private::latexInset(BufferParams const & bparams,
1012                                     otexstream & os,
1013                                     OutputParams & runparams,
1014                                     Font & running_font,
1015                                     Font & basefont,
1016                                     Font const & outerfont,
1017                                     bool & open_font,
1018                                     Change & running_change,
1019                                     Layout const & style,
1020                                     pos_type & i,
1021                                     unsigned int & column)
1022 {
1023         Inset * inset = owner_->getInset(i);
1024         LBUFERR(inset);
1025
1026         if (style.pass_thru) {
1027                 odocstringstream ods;
1028                 inset->plaintext(ods, runparams);
1029                 os << ods.str();
1030                 return;
1031         }
1032
1033         // FIXME: move this to InsetNewline::latex
1034         if (inset->lyxCode() == NEWLINE_CODE || inset->lyxCode() == SEPARATOR_CODE) {
1035                 // newlines are handled differently here than
1036                 // the default in simpleTeXSpecialChars().
1037                 if (!style.newline_allowed) {
1038                         os << '\n';
1039                 } else {
1040                         if (open_font) {
1041                                 column += running_font.latexWriteEndChanges(
1042                                         os, bparams, runparams,
1043                                         basefont, basefont);
1044                                 open_font = false;
1045                         }
1046
1047                         if (running_font.fontInfo().family() == TYPEWRITER_FAMILY)
1048                                 os << '~';
1049
1050                         basefont = owner_->getLayoutFont(bparams, outerfont);
1051                         running_font = basefont;
1052
1053                         if (runparams.moving_arg)
1054                                 os << "\\protect ";
1055
1056                 }
1057                 os.texrow().start(owner_->id(), i + 1);
1058                 column = 0;
1059         }
1060
1061         if (owner_->isDeleted(i)) {
1062                 if( ++runparams.inDeletedInset == 1)
1063                         runparams.changeOfDeletedInset = owner_->lookupChange(i);
1064         }
1065
1066         if (inset->canTrackChanges()) {
1067                 column += Changes::latexMarkChange(os, bparams, running_change,
1068                         Change(Change::UNCHANGED), runparams);
1069                 running_change = Change(Change::UNCHANGED);
1070         }
1071
1072         bool close = false;
1073         odocstream::pos_type const len = os.os().tellp();
1074
1075         if (inset->forceLTR()
1076             && !runparams.use_polyglossia
1077             && running_font.isRightToLeft()
1078             // ERT is an exception, it should be output with no
1079             // decorations at all
1080             && inset->lyxCode() != ERT_CODE) {
1081                 if (running_font.language()->lang() == "farsi")
1082                         os << "\\beginL{}";
1083                 else
1084                         os << "\\L{";
1085                 close = true;
1086         }
1087
1088         // FIXME: Bug: we can have an empty font change here!
1089         // if there has just been a font change, we are going to close it
1090         // right now, which means stupid latex code like \textsf{}. AFAIK,
1091         // this does not harm dvi output. A minor bug, thus (JMarc)
1092
1093         // Some insets cannot be inside a font change command.
1094         // However, even such insets *can* be placed in \L or \R
1095         // or their equivalents (for RTL language switches), so we don't
1096         // close the language in those cases.
1097         // ArabTeX, though, cannot handle this special behavior, it seems.
1098         bool arabtex = basefont.language()->lang() == "arabic_arabtex"
1099                 || running_font.language()->lang() == "arabic_arabtex";
1100         if (open_font && !inset->inheritFont()) {
1101                 bool closeLanguage = arabtex
1102                         || basefont.isRightToLeft() == running_font.isRightToLeft();
1103                 unsigned int count = running_font.latexWriteEndChanges(os,
1104                         bparams, runparams, basefont, basefont, closeLanguage);
1105                 column += count;
1106                 // if any font properties were closed, update the running_font,
1107                 // making sure, however, to leave the language as it was
1108                 if (count > 0) {
1109                         // FIXME: probably a better way to keep track of the old
1110                         // language, than copying the entire font?
1111                         Font const copy_font(running_font);
1112                         basefont = owner_->getLayoutFont(bparams, outerfont);
1113                         running_font = basefont;
1114                         if (!closeLanguage)
1115                                 running_font.setLanguage(copy_font.language());
1116                         // leave font open if language is still open
1117                         open_font = (running_font.language() == basefont.language());
1118                         if (closeLanguage)
1119                                 runparams.local_font = &basefont;
1120                 }
1121         }
1122
1123         size_t const previous_row_count = os.texrow().rows();
1124
1125         try {
1126                 runparams.lastid = id_;
1127                 runparams.lastpos = i;
1128                 inset->latex(os, runparams);
1129         } catch (EncodingException & e) {
1130                 // add location information and throw again.
1131                 e.par_id = id_;
1132                 e.pos = i;
1133                 throw(e);
1134         }
1135
1136         if (close) {
1137                 if (running_font.language()->lang() == "farsi")
1138                                 os << "\\endL{}";
1139                         else
1140                                 os << '}';
1141         }
1142
1143         if (os.texrow().rows() > previous_row_count) {
1144                 os.texrow().start(owner_->id(), i + 1);
1145                 column = 0;
1146         } else {
1147                 column += (unsigned int)(os.os().tellp() - len);
1148         }
1149
1150         if (owner_->isDeleted(i))
1151                 --runparams.inDeletedInset;
1152 }
1153
1154
1155 void Paragraph::Private::latexSpecialChar(otexstream & os,
1156                                           BufferParams const & bparams,
1157                                           OutputParams const & runparams,
1158                                           Font const & running_font,
1159                                           Change const & running_change,
1160                                           Layout const & style,
1161                                           pos_type & i,
1162                                           pos_type end_pos,
1163                                           unsigned int & column)
1164 {
1165         // With polyglossia, brackets and stuff need not be reversed
1166         // in RTL scripts (see bug #8251)
1167         char_type const c = (runparams.use_polyglossia) ?
1168                 owner_->getUChar(bparams, i) : text_[i];
1169
1170         if (style.pass_thru || runparams.pass_thru
1171             || contains(style.pass_thru_chars, c)
1172             || contains(runparams.pass_thru_chars, c)) {
1173                 if (c != '\0') {
1174                         Encoding const * const enc = runparams.encoding;
1175                         if (enc && !enc->encodable(c))
1176                                 throw EncodingException(c);
1177                         os.put(c);
1178                 }
1179                 return;
1180         }
1181
1182         // TIPA uses its own T3 encoding
1183         if (runparams.inIPA && latexSpecialT3(c, os, i, column))
1184                 return;
1185         // If T1 font encoding is used, use the special
1186         // characters it provides.
1187         // NOTE: Some languages reset the font encoding internally to a
1188         //       non-standard font encoding. If we are using such a language,
1189         //       we do not output special T1 chars.
1190         if (!runparams.inIPA && !running_font.language()->internalFontEncoding()
1191             && bparams.font_encoding() == "T1" && latexSpecialT1(c, os, i, column))
1192                 return;
1193
1194         // Otherwise, we use what LaTeX provides us.
1195         switch (c) {
1196         case '\\':
1197                 os << "\\textbackslash{}";
1198                 column += 15;
1199                 break;
1200         case '<':
1201                 os << "\\textless{}";
1202                 column += 10;
1203                 break;
1204         case '>':
1205                 os << "\\textgreater{}";
1206                 column += 13;
1207                 break;
1208         case '|':
1209                 os << "\\textbar{}";
1210                 column += 9;
1211                 break;
1212         case '-':
1213                 os << '-';
1214                 if (i + 1 < static_cast<pos_type>(text_.size()) &&
1215                     (end_pos == -1 || i + 1 < end_pos) &&
1216                     text_[i+1] == '-') {
1217                         // Prevent "--" becoming an endash and "---" becoming
1218                         // an emdash.
1219                         // Within \ttfamily, "--" is merged to "-" (no endash)
1220                         // so we avoid this rather irritating ligature as well
1221                         os << "{}";
1222                         column += 2;
1223                 }
1224                 break;
1225         case '\"':
1226                 os << "\\char`\\\"{}";
1227                 column += 9;
1228                 break;
1229
1230         case '$': case '&':
1231         case '%': case '#': case '{':
1232         case '}': case '_':
1233                 os << '\\';
1234                 os.put(c);
1235                 column += 1;
1236                 break;
1237
1238         case '~':
1239                 os << "\\textasciitilde{}";
1240                 column += 16;
1241                 break;
1242
1243         case '^':
1244                 os << "\\textasciicircum{}";
1245                 column += 17;
1246                 break;
1247
1248         case '*':
1249         case '[':
1250         case ']':
1251                 // avoid being mistaken for optional arguments
1252                 os << '{';
1253                 os.put(c);
1254                 os << '}';
1255                 column += 2;
1256                 break;
1257
1258         case ' ':
1259                 // Blanks are printed before font switching.
1260                 // Sure? I am not! (try nice-latex)
1261                 // I am sure it's correct. LyX might be smarter
1262                 // in the future, but for now, nothing wrong is
1263                 // written. (Asger)
1264                 break;
1265
1266         default:
1267                 if (c == '\0')
1268                         return;
1269
1270                 Encoding const & encoding = *(runparams.encoding);
1271                 char_type next = '\0';
1272                 if (i + 1 < int(text_.size())) {
1273                         next = text_[i + 1];
1274                         if (Encodings::isCombiningChar(next)) {
1275                                 column += latexSurrogatePair(os, c, next, runparams) - 1;
1276                                 ++i;
1277                                 break;
1278                         }
1279                 }
1280                 string script;
1281                 pair<docstring, bool> latex = encoding.latexChar(c);
1282                 docstring nextlatex;
1283                 bool nexttipas = false;
1284                 string nexttipashortcut;
1285                 if (next != '\0' && next != META_INSET && encoding.encodable(next)) {
1286                         nextlatex = encoding.latexChar(next).first;
1287                         if (runparams.inIPA) {
1288                                 nexttipashortcut = Encodings::TIPAShortcut(next);
1289                                 nexttipas = !nexttipashortcut.empty();
1290                         }
1291                 }
1292                 bool tipas = false;
1293                 if (runparams.inIPA) {
1294                         string const tipashortcut = Encodings::TIPAShortcut(c);
1295                         if (!tipashortcut.empty()) {
1296                                 latex.first = from_ascii(tipashortcut);
1297                                 latex.second = false;
1298                                 tipas = true;
1299                         }
1300                 }
1301                 if (Encodings::isKnownScriptChar(c, script)
1302                     && prefixIs(latex.first, from_ascii("\\" + script)))
1303                         column += writeScriptChars(os, latex.first,
1304                                         running_change, encoding, i) - 1;
1305                 else if (latex.second
1306                          && ((!prefixIs(nextlatex, '\\')
1307                                && !prefixIs(nextlatex, '{')
1308                                && !prefixIs(nextlatex, '}'))
1309                              || (nexttipas
1310                                  && !prefixIs(from_ascii(nexttipashortcut), '\\')))
1311                          && !tipas) {
1312                         // Prevent eating of a following
1313                         // space or command corruption by
1314                         // following characters
1315                         if (next == ' ' || next == '\0') {
1316                                 column += latex.first.length() + 1;
1317                                 os << latex.first << "{}";
1318                         } else {
1319                                 column += latex.first.length();
1320                                 os << latex.first << " ";
1321                         }
1322                 } else {
1323                         column += latex.first.length() - 1;
1324                         os << latex.first;
1325                 }
1326                 break;
1327         }
1328 }
1329
1330
1331 bool Paragraph::Private::latexSpecialT1(char_type const c, otexstream & os,
1332         pos_type i, unsigned int & column)
1333 {
1334         switch (c) {
1335         case '>':
1336         case '<':
1337                 os.put(c);
1338                 // In T1 encoding, these characters exist
1339                 // but we should avoid ligatures
1340                 if (i + 1 >= int(text_.size()) || text_[i + 1] != c)
1341                         return true;
1342                 os << "\\textcompwordmark{}";
1343                 column += 19;
1344                 return true;
1345         case '|':
1346                 os.put(c);
1347                 return true;
1348         case '\"':
1349                 // soul.sty breaks with \char`\"
1350                 os << "\\textquotedbl{}";
1351                 column += 14;
1352                 return true;
1353         default:
1354                 return false;
1355         }
1356 }
1357
1358
1359 bool Paragraph::Private::latexSpecialT3(char_type const c, otexstream & os,
1360         pos_type /*i*/, unsigned int & column)
1361 {
1362         switch (c) {
1363         case '*':
1364         case '[':
1365         case ']':
1366         case '\"':
1367                 os.put(c);
1368                 return true;
1369         case '|':
1370                 os << "\\textvertline{}";
1371                 column += 14;
1372                 return true;
1373         default:
1374                 return false;
1375         }
1376 }
1377
1378
1379 void Paragraph::Private::validate(LaTeXFeatures & features) const
1380 {
1381         if (layout_->inpreamble && inset_owner_) {
1382                 bool const is_command = layout_->latextype == LATEX_COMMAND;
1383                 Buffer const & buf = inset_owner_->buffer();
1384                 BufferParams const & bp = features.runparams().is_child
1385                         ? buf.masterParams() : buf.params();
1386                 Font f;
1387                 // Using a string stream here circumvents the encoding
1388                 // switching machinery of odocstream. Therefore the
1389                 // output is wrong if this paragraph contains content
1390                 // that needs to switch encoding.
1391                 otexstringstream os;
1392                 if (is_command) {
1393                         os << '\\' << from_ascii(layout_->latexname());
1394                         // we have to provide all the optional arguments here, even though
1395                         // the last one is the only one we care about.
1396                         // Separate handling of optional argument inset.
1397                         if (!layout_->latexargs().empty()) {
1398                                 OutputParams rp = features.runparams();
1399                                 rp.local_font = &owner_->getFirstFontSettings(bp);
1400                                 latexArgInsets(*owner_, os, rp, layout_->latexargs());
1401                         }
1402                         os << from_ascii(layout_->latexparam());
1403                 }
1404                 size_t const length = os.length();
1405                 // this will output "{" at the beginning, but not at the end
1406                 owner_->latex(bp, f, os, features.runparams(), 0, -1, true);
1407                 if (os.length() > length) {
1408                         if (is_command) {
1409                                 os << '}';
1410                                 if (!layout_->postcommandargs().empty()) {
1411                                         OutputParams rp = features.runparams();
1412                                         rp.local_font = &owner_->getFirstFontSettings(bp);
1413                                         latexArgInsets(*owner_, os, rp, layout_->postcommandargs(), "post:");
1414                                 }
1415                         }
1416                         features.addPreambleSnippet(os.release(), true);
1417                 }
1418         }
1419
1420         if (features.runparams().flavor == OutputParams::HTML
1421             && layout_->htmltitle()) {
1422                 features.setHTMLTitle(owner_->asString(AS_STR_INSETS | AS_STR_SKIPDELETE));
1423         }
1424
1425         // check the params.
1426         if (!params_.spacing().isDefault())
1427                 features.require("setspace");
1428
1429         // then the layouts
1430         features.useLayout(layout_->name());
1431
1432         // then the fonts
1433         fontlist_.validate(features);
1434
1435         // then the indentation
1436         if (!params_.leftIndent().zero())
1437                 features.require("ParagraphLeftIndent");
1438
1439         // then the insets
1440         InsetList::const_iterator icit = insetlist_.begin();
1441         InsetList::const_iterator iend = insetlist_.end();
1442         for (; icit != iend; ++icit) {
1443                 if (icit->inset) {
1444                         features.inDeletedInset(owner_->isDeleted(icit->pos));
1445                         icit->inset->validate(features);
1446                         features.inDeletedInset(false);
1447                         if (layout_->needprotect &&
1448                             icit->inset->lyxCode() == FOOT_CODE)
1449                                 features.require("NeedLyXFootnoteCode");
1450                 }
1451         }
1452
1453         // then the contents
1454         for (pos_type i = 0; i < int(text_.size()) ; ++i) {
1455                 BufferEncodings::validate(text_[i], features);
1456         }
1457 }
1458
1459 /////////////////////////////////////////////////////////////////////
1460 //
1461 // Paragraph
1462 //
1463 /////////////////////////////////////////////////////////////////////
1464
1465 namespace {
1466         Layout const emptyParagraphLayout;
1467 }
1468
1469 Paragraph::Paragraph()
1470         : d(new Paragraph::Private(this, emptyParagraphLayout))
1471 {
1472         itemdepth = 0;
1473         d->params_.clear();
1474 }
1475
1476
1477 Paragraph::Paragraph(Paragraph const & par)
1478         : itemdepth(par.itemdepth),
1479         d(new Paragraph::Private(*par.d, this))
1480 {
1481         registerWords();
1482 }
1483
1484
1485 Paragraph::Paragraph(Paragraph const & par, pos_type beg, pos_type end)
1486         : itemdepth(par.itemdepth),
1487         d(new Paragraph::Private(*par.d, this, beg, end))
1488 {
1489         registerWords();
1490 }
1491
1492
1493 Paragraph & Paragraph::operator=(Paragraph const & par)
1494 {
1495         // needed as we will destroy the private part before copying it
1496         if (&par != this) {
1497                 itemdepth = par.itemdepth;
1498
1499                 deregisterWords();
1500                 delete d;
1501                 d = new Private(*par.d, this);
1502                 registerWords();
1503         }
1504         return *this;
1505 }
1506
1507
1508 Paragraph::~Paragraph()
1509 {
1510         deregisterWords();
1511         delete d;
1512 }
1513
1514
1515 namespace {
1516
1517 // this shall be called just before every "os << ..." action.
1518 void flushString(ostream & os, docstring & s)
1519 {
1520         os << to_utf8(s);
1521         s.erase();
1522 }
1523
1524 }
1525
1526
1527 void Paragraph::write(ostream & os, BufferParams const & bparams,
1528         depth_type & dth) const
1529 {
1530         // The beginning or end of a deeper (i.e. nested) area?
1531         if (dth != d->params_.depth()) {
1532                 if (d->params_.depth() > dth) {
1533                         while (d->params_.depth() > dth) {
1534                                 os << "\n\\begin_deeper";
1535                                 ++dth;
1536                         }
1537                 } else {
1538                         while (d->params_.depth() < dth) {
1539                                 os << "\n\\end_deeper";
1540                                 --dth;
1541                         }
1542                 }
1543         }
1544
1545         // First write the layout
1546         os << "\n\\begin_layout " << to_utf8(d->layout_->name()) << '\n';
1547
1548         d->params_.write(os);
1549
1550         Font font1(inherit_font, bparams.language);
1551
1552         Change running_change = Change(Change::UNCHANGED);
1553
1554         // this string is used as a buffer to avoid repetitive calls
1555         // to to_utf8(), which turn out to be expensive (JMarc)
1556         docstring write_buffer;
1557
1558         int column = 0;
1559         for (pos_type i = 0; i <= size(); ++i) {
1560
1561                 Change const & change = lookupChange(i);
1562                 if (change != running_change)
1563                         flushString(os, write_buffer);
1564                 Changes::lyxMarkChange(os, bparams, column, running_change, change);
1565                 running_change = change;
1566
1567                 if (i == size())
1568                         break;
1569
1570                 // Write font changes
1571                 Font font2 = getFontSettings(bparams, i);
1572                 if (font2 != font1) {
1573                         flushString(os, write_buffer);
1574                         font2.lyxWriteChanges(font1, os);
1575                         column = 0;
1576                         font1 = font2;
1577                 }
1578
1579                 char_type const c = d->text_[i];
1580                 switch (c) {
1581                 case META_INSET:
1582                         if (Inset const * inset = getInset(i)) {
1583                                 flushString(os, write_buffer);
1584                                 if (inset->directWrite()) {
1585                                         // international char, let it write
1586                                         // code directly so it's shorter in
1587                                         // the file
1588                                         inset->write(os);
1589                                 } else {
1590                                         if (i)
1591                                                 os << '\n';
1592                                         os << "\\begin_inset ";
1593                                         inset->write(os);
1594                                         os << "\n\\end_inset\n\n";
1595                                         column = 0;
1596                                 }
1597                                 // FIXME This can be removed again once the mystery
1598                                 // crash has been resolved.
1599                                 os << flush;
1600                         }
1601                         break;
1602                 case '\\':
1603                         flushString(os, write_buffer);
1604                         os << "\n\\backslash\n";
1605                         column = 0;
1606                         break;
1607                 case '.':
1608                         flushString(os, write_buffer);
1609                         if (i + 1 < size() && d->text_[i + 1] == ' ') {
1610                                 os << ".\n";
1611                                 column = 0;
1612                         } else
1613                                 os << '.';
1614                         break;
1615                 default:
1616                         if ((column > 70 && c == ' ')
1617                             || column > 79) {
1618                                 flushString(os, write_buffer);
1619                                 os << '\n';
1620                                 column = 0;
1621                         }
1622                         // this check is to amend a bug. LyX sometimes
1623                         // inserts '\0' this could cause problems.
1624                         if (c != '\0')
1625                                 write_buffer.push_back(c);
1626                         else
1627                                 LYXERR0("NUL char in structure.");
1628                         ++column;
1629                         break;
1630                 }
1631         }
1632
1633         flushString(os, write_buffer);
1634         os << "\n\\end_layout\n";
1635         // FIXME This can be removed again once the mystery
1636         // crash has been resolved.
1637         os << flush;
1638 }
1639
1640
1641 void Paragraph::validate(LaTeXFeatures & features) const
1642 {
1643         d->validate(features);
1644 }
1645
1646
1647 void Paragraph::insert(pos_type start, docstring const & str,
1648                        Font const & font, Change const & change)
1649 {
1650         for (size_t i = 0, n = str.size(); i != n ; ++i)
1651                 insertChar(start + i, str[i], font, change);
1652 }
1653
1654
1655 void Paragraph::appendChar(char_type c, Font const & font,
1656                 Change const & change)
1657 {
1658         // track change
1659         d->changes_.insert(change, d->text_.size());
1660         // when appending characters, no need to update tables
1661         d->text_.push_back(c);
1662         setFont(d->text_.size() - 1, font);
1663         d->requestSpellCheck(d->text_.size() - 1);
1664 }
1665
1666
1667 void Paragraph::appendString(docstring const & s, Font const & font,
1668                 Change const & change)
1669 {
1670         pos_type end = s.size();
1671         size_t oldsize = d->text_.size();
1672         size_t newsize = oldsize + end;
1673         size_t capacity = d->text_.capacity();
1674         if (newsize >= capacity)
1675                 d->text_.reserve(max(capacity + 100, newsize));
1676
1677         // when appending characters, no need to update tables
1678         d->text_.append(s);
1679
1680         // FIXME: Optimize this!
1681         for (size_t i = oldsize; i != newsize; ++i) {
1682                 // track change
1683                 d->changes_.insert(change, i);
1684                 d->requestSpellCheck(i);
1685         }
1686         d->fontlist_.set(oldsize, font);
1687         d->fontlist_.set(newsize - 1, font);
1688 }
1689
1690
1691 void Paragraph::insertChar(pos_type pos, char_type c,
1692                            bool trackChanges)
1693 {
1694         d->insertChar(pos, c, Change(trackChanges ?
1695                            Change::INSERTED : Change::UNCHANGED));
1696 }
1697
1698
1699 void Paragraph::insertChar(pos_type pos, char_type c,
1700                            Font const & font, bool trackChanges)
1701 {
1702         d->insertChar(pos, c, Change(trackChanges ?
1703                            Change::INSERTED : Change::UNCHANGED));
1704         setFont(pos, font);
1705 }
1706
1707
1708 void Paragraph::insertChar(pos_type pos, char_type c,
1709                            Font const & font, Change const & change)
1710 {
1711         d->insertChar(pos, c, change);
1712         setFont(pos, font);
1713 }
1714
1715
1716 void Paragraph::resetFonts(Font const & font)
1717 {
1718         d->fontlist_.clear();
1719         d->fontlist_.set(0, font);
1720         d->fontlist_.set(d->text_.size() - 1, font);
1721 }
1722
1723 // Gets uninstantiated font setting at position.
1724 Font const & Paragraph::getFontSettings(BufferParams const & bparams,
1725                                          pos_type pos) const
1726 {
1727         if (pos > size()) {
1728                 LYXERR0("pos: " << pos << " size: " << size());
1729                 LBUFERR(false);
1730         }
1731
1732         FontList::const_iterator cit = d->fontlist_.fontIterator(pos);
1733         if (cit != d->fontlist_.end())
1734                 return cit->font();
1735
1736         if (pos == size() && !empty())
1737                 return getFontSettings(bparams, pos - 1);
1738
1739         // Optimisation: avoid a full font instantiation if there is no
1740         // language change from previous call.
1741         static Font previous_font;
1742         static Language const * previous_lang = 0;
1743         Language const * lang = getParLanguage(bparams);
1744         if (lang != previous_lang) {
1745                 previous_lang = lang;
1746                 previous_font = Font(inherit_font, lang);
1747         }
1748         return previous_font;
1749 }
1750
1751
1752 FontSpan Paragraph::fontSpan(pos_type pos) const
1753 {
1754         LBUFERR(pos <= size());
1755
1756         if (pos == size())
1757                 return FontSpan(pos, pos);
1758
1759         pos_type start = 0;
1760         FontList::const_iterator cit = d->fontlist_.begin();
1761         FontList::const_iterator end = d->fontlist_.end();
1762         for (; cit != end; ++cit) {
1763                 if (cit->pos() >= pos) {
1764                         if (pos >= beginOfBody())
1765                                 return FontSpan(max(start, beginOfBody()),
1766                                                 cit->pos());
1767                         else
1768                                 return FontSpan(start,
1769                                                 min(beginOfBody() - 1,
1770                                                          cit->pos()));
1771                 }
1772                 start = cit->pos() + 1;
1773         }
1774
1775         // This should not happen, but if so, we take no chances.
1776         LYXERR0("Paragraph::fontSpan: position not found in fontinfo table!");
1777         LASSERT(false, return FontSpan(pos, pos));
1778 }
1779
1780
1781 // Gets uninstantiated font setting at position 0
1782 Font const & Paragraph::getFirstFontSettings(BufferParams const & bparams) const
1783 {
1784         if (!empty() && !d->fontlist_.empty())
1785                 return d->fontlist_.begin()->font();
1786
1787         // Optimisation: avoid a full font instantiation if there is no
1788         // language change from previous call.
1789         static Font previous_font;
1790         static Language const * previous_lang = 0;
1791         if (bparams.language != previous_lang) {
1792                 previous_lang = bparams.language;
1793                 previous_font = Font(inherit_font, bparams.language);
1794         }
1795
1796         return previous_font;
1797 }
1798
1799
1800 // Gets the fully instantiated font at a given position in a paragraph
1801 // This is basically the same function as Text::GetFont() in text2.cpp.
1802 // The difference is that this one is used for generating the LaTeX file,
1803 // and thus cosmetic "improvements" are disallowed: This has to deliver
1804 // the true picture of the buffer. (Asger)
1805 Font const Paragraph::getFont(BufferParams const & bparams, pos_type pos,
1806                                  Font const & outerfont) const
1807 {
1808         LBUFERR(pos >= 0);
1809
1810         Font font = getFontSettings(bparams, pos);
1811
1812         pos_type const body_pos = beginOfBody();
1813         FontInfo & fi = font.fontInfo();
1814         if (pos < body_pos)
1815                 fi.realize(d->layout_->labelfont);
1816         else
1817                 fi.realize(d->layout_->font);
1818
1819         fi.realize(outerfont.fontInfo());
1820         fi.realize(bparams.getFont().fontInfo());
1821
1822         return font;
1823 }
1824
1825
1826 Font const Paragraph::getLabelFont
1827         (BufferParams const & bparams, Font const & outerfont) const
1828 {
1829         FontInfo tmpfont = d->layout_->labelfont;
1830         tmpfont.realize(outerfont.fontInfo());
1831         tmpfont.realize(bparams.getFont().fontInfo());
1832         return Font(tmpfont, getParLanguage(bparams));
1833 }
1834
1835
1836 Font const Paragraph::getLayoutFont
1837         (BufferParams const & bparams, Font const & outerfont) const
1838 {
1839         FontInfo tmpfont = d->layout_->font;
1840         tmpfont.realize(outerfont.fontInfo());
1841         tmpfont.realize(bparams.getFont().fontInfo());
1842         return Font(tmpfont, getParLanguage(bparams));
1843 }
1844
1845
1846 char_type Paragraph::getUChar(BufferParams const & bparams, pos_type pos) const
1847 {
1848         char_type c = d->text_[pos];
1849         if (!getFontSettings(bparams, pos).isRightToLeft())
1850                 return c;
1851
1852         // FIXME: The arabic special casing is due to the difference of arabic
1853         // round brackets input introduced in r18599. Check if this should be
1854         // unified with Hebrew or at least if all bracket types should be
1855         // handled the same (file format change in either case).
1856         string const & lang = getFontSettings(bparams, pos).language()->lang();
1857         bool const arabic = lang == "arabic_arabtex" || lang == "arabic_arabi"
1858                 || lang == "farsi";
1859         char_type uc = c;
1860         switch (c) {
1861         case '(':
1862                 uc = arabic ? c : ')';
1863                 break;
1864         case ')':
1865                 uc = arabic ? c : '(';
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         case '<':
1880                 uc = '>';
1881                 break;
1882         case '>':
1883                 uc = '<';
1884                 break;
1885         }
1886
1887         return uc;
1888 }
1889
1890
1891 void Paragraph::setFont(pos_type pos, Font const & font)
1892 {
1893         LASSERT(pos <= size(), return);
1894
1895         // First, reduce font against layout/label font
1896         // Update: The setCharFont() routine in text2.cpp already
1897         // reduces font, so we don't need to do that here. (Asger)
1898
1899         d->fontlist_.set(pos, font);
1900 }
1901
1902
1903 void Paragraph::makeSameLayout(Paragraph const & par)
1904 {
1905         d->layout_ = par.d->layout_;
1906         d->params_ = par.d->params_;
1907 }
1908
1909
1910 bool Paragraph::stripLeadingSpaces(bool trackChanges)
1911 {
1912         if (isFreeSpacing())
1913                 return false;
1914
1915         int pos = 0;
1916         int count = 0;
1917
1918         while (pos < size() && (isNewline(pos) || isLineSeparator(pos))) {
1919                 if (eraseChar(pos, trackChanges))
1920                         ++count;
1921                 else
1922                         ++pos;
1923         }
1924
1925         return count > 0 || pos > 0;
1926 }
1927
1928
1929 bool Paragraph::hasSameLayout(Paragraph const & par) const
1930 {
1931         return par.d->layout_ == d->layout_
1932                 && d->params_.sameLayout(par.d->params_);
1933 }
1934
1935
1936 depth_type Paragraph::getDepth() const
1937 {
1938         return d->params_.depth();
1939 }
1940
1941
1942 depth_type Paragraph::getMaxDepthAfter() const
1943 {
1944         if (d->layout_->isEnvironment())
1945                 return d->params_.depth() + 1;
1946         else
1947                 return d->params_.depth();
1948 }
1949
1950
1951 char Paragraph::getAlign() const
1952 {
1953         if (d->params_.align() == LYX_ALIGN_LAYOUT)
1954                 return d->layout_->align;
1955         else
1956                 return d->params_.align();
1957 }
1958
1959
1960 docstring const & Paragraph::labelString() const
1961 {
1962         return d->params_.labelString();
1963 }
1964
1965
1966 // the next two functions are for the manual labels
1967 docstring const Paragraph::getLabelWidthString() const
1968 {
1969         if (d->layout_->margintype == MARGIN_MANUAL
1970             || d->layout_->latextype == LATEX_BIB_ENVIRONMENT)
1971                 return d->params_.labelWidthString();
1972         else
1973                 return _("Senseless with this layout!");
1974 }
1975
1976
1977 void Paragraph::setLabelWidthString(docstring const & s)
1978 {
1979         d->params_.labelWidthString(s);
1980 }
1981
1982
1983 docstring Paragraph::expandLabel(Layout const & layout,
1984                 BufferParams const & bparams) const
1985 {
1986         return expandParagraphLabel(layout, bparams, true);
1987 }
1988
1989
1990 docstring Paragraph::expandDocBookLabel(Layout const & layout,
1991                 BufferParams const & bparams) const
1992 {
1993         return expandParagraphLabel(layout, bparams, false);
1994 }
1995
1996
1997 docstring Paragraph::expandParagraphLabel(Layout const & layout,
1998                 BufferParams const & bparams, bool process_appendix) const
1999 {
2000         DocumentClass const & tclass = bparams.documentClass();
2001         string const & lang = getParLanguage(bparams)->code();
2002         bool const in_appendix = process_appendix && d->params_.appendix();
2003         docstring fmt = translateIfPossible(layout.labelstring(in_appendix), lang);
2004
2005         if (fmt.empty() && !layout.counter.empty())
2006                 return tclass.counters().theCounter(layout.counter, lang);
2007
2008         // handle 'inherited level parts' in 'fmt',
2009         // i.e. the stuff between '@' in   '@Section@.\arabic{subsection}'
2010         size_t const i = fmt.find('@', 0);
2011         if (i != docstring::npos) {
2012                 size_t const j = fmt.find('@', i + 1);
2013                 if (j != docstring::npos) {
2014                         docstring parent(fmt, i + 1, j - i - 1);
2015                         docstring label = from_ascii("??");
2016                         if (tclass.hasLayout(parent))
2017                                 label = expandParagraphLabel(tclass[parent], bparams,
2018                                                       process_appendix);
2019                         fmt = docstring(fmt, 0, i) + label
2020                                 + docstring(fmt, j + 1, docstring::npos);
2021                 }
2022         }
2023
2024         return tclass.counters().counterLabel(fmt, lang);
2025 }
2026
2027
2028 void Paragraph::applyLayout(Layout const & new_layout)
2029 {
2030         d->layout_ = &new_layout;
2031         LyXAlignment const oldAlign = d->params_.align();
2032
2033         if (!(oldAlign & d->layout_->alignpossible)) {
2034                 frontend::Alert::warning(_("Alignment not permitted"),
2035                         _("The new layout does not permit the alignment previously used.\nSetting to default."));
2036                 d->params_.align(LYX_ALIGN_LAYOUT);
2037         }
2038 }
2039
2040
2041 pos_type Paragraph::beginOfBody() const
2042 {
2043         return d->begin_of_body_;
2044 }
2045
2046
2047 void Paragraph::setBeginOfBody()
2048 {
2049         if (d->layout_->labeltype != LABEL_MANUAL) {
2050                 d->begin_of_body_ = 0;
2051                 return;
2052         }
2053
2054         // Unroll the first two cycles of the loop
2055         // and remember the previous character to
2056         // remove unnecessary getChar() calls
2057         pos_type i = 0;
2058         pos_type end = size();
2059         if (i < end && !(isNewline(i) || isEnvSeparator(i))) {
2060                 ++i;
2061                 if (i < end) {
2062                         char_type previous_char = d->text_[i];
2063                         if (!(isNewline(i) || isEnvSeparator(i))) {
2064                                 ++i;
2065                                 while (i < end && previous_char != ' ') {
2066                                         char_type temp = d->text_[i];
2067                                         if (isNewline(i) || isEnvSeparator(i))
2068                                                 break;
2069                                         ++i;
2070                                         previous_char = temp;
2071                                 }
2072                         }
2073                 }
2074         }
2075
2076         d->begin_of_body_ = i;
2077 }
2078
2079
2080 bool Paragraph::allowParagraphCustomization() const
2081 {
2082         return inInset().allowParagraphCustomization();
2083 }
2084
2085
2086 bool Paragraph::usePlainLayout() const
2087 {
2088         return inInset().usePlainLayout();
2089 }
2090
2091
2092 bool Paragraph::isPassThru() const
2093 {
2094         return inInset().isPassThru() || d->layout_->pass_thru;
2095 }
2096
2097 namespace {
2098
2099 // paragraphs inside floats need different alignment tags to avoid
2100 // unwanted space
2101
2102 bool noTrivlistCentering(InsetCode code)
2103 {
2104         return code == FLOAT_CODE
2105                || code == WRAP_CODE
2106                || code == CELL_CODE;
2107 }
2108
2109
2110 string correction(string const & orig)
2111 {
2112         if (orig == "flushleft")
2113                 return "raggedright";
2114         if (orig == "flushright")
2115                 return "raggedleft";
2116         if (orig == "center")
2117                 return "centering";
2118         return orig;
2119 }
2120
2121
2122 bool corrected_env(otexstream & os, string const & suffix, string const & env,
2123         InsetCode code, bool const lastpar, int & col)
2124 {
2125         string macro = suffix + "{";
2126         if (noTrivlistCentering(code)) {
2127                 if (lastpar) {
2128                         // the last paragraph in non-trivlist-aligned
2129                         // context is special (to avoid unwanted whitespace)
2130                         if (suffix == "\\begin") {
2131                                 macro = "\\" + correction(env) + "{}";
2132                                 os << from_ascii(macro);
2133                                 col += macro.size();
2134                                 return true;
2135                         }
2136                         return false;
2137                 }
2138                 macro += correction(env);
2139         } else
2140                 macro += env;
2141         macro += "}";
2142         if (suffix == "\\par\\end") {
2143                 os << breakln;
2144                 col = 0;
2145         }
2146         os << from_ascii(macro);
2147         col += macro.size();
2148         if (suffix == "\\begin") {
2149                 os << breakln;
2150                 col = 0;
2151         }
2152         return true;
2153 }
2154
2155 } // namespace anon
2156
2157
2158 int Paragraph::Private::startTeXParParams(BufferParams const & bparams,
2159                         otexstream & os, OutputParams const & runparams) const
2160 {
2161         int column = 0;
2162
2163         bool canindent =
2164                 (bparams.paragraph_separation == BufferParams::ParagraphIndentSeparation) ?
2165                         (layout_->toggle_indent != ITOGGLE_NEVER) :
2166                         (layout_->toggle_indent == ITOGGLE_ALWAYS);
2167
2168         if (canindent && params_.noindent() && !layout_->pass_thru) {
2169                 os << "\\noindent ";
2170                 column += 10;
2171         }
2172
2173         LyXAlignment const curAlign = params_.align();
2174
2175         if (curAlign == layout_->align)
2176                 return column;
2177
2178         switch (curAlign) {
2179         case LYX_ALIGN_NONE:
2180         case LYX_ALIGN_BLOCK:
2181         case LYX_ALIGN_LAYOUT:
2182         case LYX_ALIGN_SPECIAL:
2183         case LYX_ALIGN_DECIMAL:
2184                 break;
2185         case LYX_ALIGN_LEFT:
2186         case LYX_ALIGN_RIGHT:
2187         case LYX_ALIGN_CENTER:
2188                 if (runparams.moving_arg) {
2189                         os << "\\protect";
2190                         column += 8;
2191                 }
2192                 break;
2193         }
2194
2195         string const begin_tag = "\\begin";
2196         InsetCode code = ownerCode();
2197         bool const lastpar = runparams.isLastPar;
2198
2199         switch (curAlign) {
2200         case LYX_ALIGN_NONE:
2201         case LYX_ALIGN_BLOCK:
2202         case LYX_ALIGN_LAYOUT:
2203         case LYX_ALIGN_SPECIAL:
2204         case LYX_ALIGN_DECIMAL:
2205                 break;
2206         case LYX_ALIGN_LEFT: {
2207                 if (owner_->getParLanguage(bparams)->babel() != "hebrew")
2208                         corrected_env(os, begin_tag, "flushleft", code, lastpar, column);
2209                 else
2210                         corrected_env(os, begin_tag, "flushright", code, lastpar, column);
2211                 break;
2212         } case LYX_ALIGN_RIGHT: {
2213                 if (owner_->getParLanguage(bparams)->babel() != "hebrew")
2214                         corrected_env(os, begin_tag, "flushright", code, lastpar, column);
2215                 else
2216                         corrected_env(os, begin_tag, "flushleft", code, lastpar, column);
2217                 break;
2218         } case LYX_ALIGN_CENTER: {
2219                 corrected_env(os, begin_tag, "center", code, lastpar, column);
2220                 break;
2221         }
2222         }
2223
2224         return column;
2225 }
2226
2227
2228 bool Paragraph::Private::endTeXParParams(BufferParams const & bparams,
2229                         otexstream & os, OutputParams const & runparams) const
2230 {
2231         LyXAlignment const curAlign = params_.align();
2232
2233         if (curAlign == layout_->align)
2234                 return false;
2235
2236         switch (curAlign) {
2237         case LYX_ALIGN_NONE:
2238         case LYX_ALIGN_BLOCK:
2239         case LYX_ALIGN_LAYOUT:
2240         case LYX_ALIGN_SPECIAL:
2241         case LYX_ALIGN_DECIMAL:
2242                 break;
2243         case LYX_ALIGN_LEFT:
2244         case LYX_ALIGN_RIGHT:
2245         case LYX_ALIGN_CENTER:
2246                 if (runparams.moving_arg)
2247                         os << "\\protect";
2248                 break;
2249         }
2250
2251         bool output = false;
2252         int col = 0;
2253         string const end_tag = "\\par\\end";
2254         InsetCode code = ownerCode();
2255         bool const lastpar = runparams.isLastPar;
2256
2257         switch (curAlign) {
2258         case LYX_ALIGN_NONE:
2259         case LYX_ALIGN_BLOCK:
2260         case LYX_ALIGN_LAYOUT:
2261         case LYX_ALIGN_SPECIAL:
2262         case LYX_ALIGN_DECIMAL:
2263                 break;
2264         case LYX_ALIGN_LEFT: {
2265                 if (owner_->getParLanguage(bparams)->babel() != "hebrew")
2266                         output = corrected_env(os, end_tag, "flushleft", code, lastpar, col);
2267                 else
2268                         output = corrected_env(os, end_tag, "flushright", code, lastpar, col);
2269                 break;
2270         } case LYX_ALIGN_RIGHT: {
2271                 if (owner_->getParLanguage(bparams)->babel() != "hebrew")
2272                         output = corrected_env(os, end_tag, "flushright", code, lastpar, col);
2273                 else
2274                         output = corrected_env(os, end_tag, "flushleft", code, lastpar, col);
2275                 break;
2276         } case LYX_ALIGN_CENTER: {
2277                 corrected_env(os, end_tag, "center", code, lastpar, col);
2278                 break;
2279         }
2280         }
2281
2282         return output || lastpar;
2283 }
2284
2285
2286 // This one spits out the text of the paragraph
2287 void Paragraph::latex(BufferParams const & bparams,
2288         Font const & outerfont,
2289         otexstream & os,
2290         OutputParams const & runparams,
2291         int start_pos, int end_pos, bool force) const
2292 {
2293         LYXERR(Debug::LATEX, "Paragraph::latex...     " << this);
2294
2295         // FIXME This check should not be needed. Perhaps issue an
2296         // error if it triggers.
2297         Layout const & style = inInset().forcePlainLayout() ?
2298                 bparams.documentClass().plainLayout() : *d->layout_;
2299
2300         if (!force && style.inpreamble)
2301                 return;
2302
2303         bool const allowcust = allowParagraphCustomization();
2304
2305         // Current base font for all inherited font changes, without any
2306         // change caused by an individual character, except for the language:
2307         // It is set to the language of the first character.
2308         // As long as we are in the label, this font is the base font of the
2309         // label. Before the first body character it is set to the base font
2310         // of the body.
2311         Font basefont;
2312
2313         // Maybe we have to create a optional argument.
2314         pos_type body_pos = beginOfBody();
2315         unsigned int column = 0;
2316
2317         if (body_pos > 0) {
2318                 // the optional argument is kept in curly brackets in
2319                 // case it contains a ']'
2320                 // This is not strictly needed, but if this is changed it
2321                 // would be a file format change, and tex2lyx would need
2322                 // to be adjusted, since it unconditionally removes the
2323                 // braces when it parses \item.
2324                 os << "[{";
2325                 column += 2;
2326                 basefont = getLabelFont(bparams, outerfont);
2327         } else {
2328                 basefont = getLayoutFont(bparams, outerfont);
2329         }
2330
2331         // Which font is currently active?
2332         Font running_font(basefont);
2333         // Do we have an open font change?
2334         bool open_font = false;
2335
2336         Change runningChange = Change(Change::UNCHANGED);
2337
2338         Encoding const * const prev_encoding = runparams.encoding;
2339
2340         os.texrow().start(id(), 0);
2341
2342         // if the paragraph is empty, the loop will not be entered at all
2343         if (empty()) {
2344                 if (style.isCommand()) {
2345                         os << '{';
2346                         ++column;
2347                 }
2348                 if (!style.leftdelim().empty()) {
2349                         os << style.leftdelim();
2350                         column += style.leftdelim().size();
2351                 }
2352                 if (allowcust)
2353                         column += d->startTeXParParams(bparams, os, runparams);
2354         }
2355
2356         for (pos_type i = 0; i < size(); ++i) {
2357                 // First char in paragraph or after label?
2358                 if (i == body_pos) {
2359                         if (body_pos > 0) {
2360                                 if (open_font) {
2361                                         column += running_font.latexWriteEndChanges(
2362                                                 os, bparams, runparams,
2363                                                 basefont, basefont);
2364                                         open_font = false;
2365                                 }
2366                                 basefont = getLayoutFont(bparams, outerfont);
2367                                 running_font = basefont;
2368
2369                                 column += Changes::latexMarkChange(os, bparams,
2370                                                 runningChange, Change(Change::UNCHANGED),
2371                                                 runparams);
2372                                 runningChange = Change(Change::UNCHANGED);
2373
2374                                 os << "}] ";
2375                                 column +=3;
2376                         }
2377                         if (style.isCommand()) {
2378                                 os << '{';
2379                                 ++column;
2380                         }
2381
2382                         if (!style.leftdelim().empty()) {
2383                                 os << style.leftdelim();
2384                                 column += style.leftdelim().size();
2385                         }
2386
2387                         if (allowcust)
2388                                 column += d->startTeXParParams(bparams, os,
2389                                                             runparams);
2390                 }
2391
2392                 Change const & change = runparams.inDeletedInset
2393                         ? runparams.changeOfDeletedInset : lookupChange(i);
2394
2395                 if (bparams.output_changes && runningChange != change) {
2396                         if (open_font) {
2397                                 column += running_font.latexWriteEndChanges(
2398                                                 os, bparams, runparams, basefont, basefont);
2399                                 open_font = false;
2400                         }
2401                         basefont = getLayoutFont(bparams, outerfont);
2402                         running_font = basefont;
2403
2404                         // Check whether a display math inset follows
2405                         if (d->text_[i] == META_INSET
2406                             && i >= start_pos && (end_pos == -1 || i < end_pos)) {
2407                                 InsetMath const * im = getInset(i)->asInsetMath();
2408                                 if (im && im->asHullInset()) {
2409                                         switch (im->asHullInset()->getType()) {
2410                                         case hullEquation:
2411                                         case hullEqnArray:
2412                                         case hullAlign:
2413                                         case hullFlAlign:
2414                                         case hullGather:
2415                                         case hullMultline:
2416                                                 runparams.inDisplayMath = true;
2417                                                 break;
2418                                         }
2419                                 }
2420                         }
2421
2422                         column += Changes::latexMarkChange(os, bparams, runningChange,
2423                                                            change, runparams);
2424                         runningChange = change;
2425                 }
2426
2427                 // do not output text which is marked deleted
2428                 // if change tracking output is disabled
2429                 if (!bparams.output_changes && change.deleted()) {
2430                         continue;
2431                 }
2432
2433                 ++column;
2434
2435                 // Fully instantiated font
2436                 Font const current_font = getFont(bparams, i, outerfont);
2437
2438                 Font const last_font = running_font;
2439
2440                 // Do we need to close the previous font?
2441                 if (open_font &&
2442                     (current_font != running_font ||
2443                      current_font.language() != running_font.language()))
2444                 {
2445                         column += running_font.latexWriteEndChanges(
2446                                         os, bparams, runparams, basefont,
2447                                         (i == body_pos-1) ? basefont : current_font);
2448                         running_font = basefont;
2449                         open_font = false;
2450                 }
2451
2452                 string const running_lang = runparams.use_polyglossia ?
2453                         running_font.language()->polyglossia() : running_font.language()->babel();
2454                 // close babel's font environment before opening CJK.
2455                 string const lang_end_command = runparams.use_polyglossia ?
2456                         "\\end{$$lang}" : lyxrc.language_command_end;
2457                 if (!running_lang.empty() &&
2458                     current_font.language()->encoding()->package() == Encoding::CJK) {
2459                                 string end_tag = subst(lang_end_command,
2460                                                         "$$lang",
2461                                                         running_lang);
2462                                 os << from_ascii(end_tag);
2463                                 column += end_tag.length();
2464                                 if (runparams.use_polyglossia)
2465                                         popPolyglossiaLang();
2466                 }
2467
2468                 // Switch file encoding if necessary (and allowed)
2469                 if (!runparams.pass_thru && !style.pass_thru &&
2470                     runparams.encoding->package() != Encoding::none &&
2471                     current_font.language()->encoding()->package() != Encoding::none) {
2472                         pair<bool, int> const enc_switch =
2473                                 switchEncoding(os.os(), bparams, runparams,
2474                                         *(current_font.language()->encoding()));
2475                         if (enc_switch.first) {
2476                                 column += enc_switch.second;
2477                                 runparams.encoding = current_font.language()->encoding();
2478                         }
2479                 }
2480
2481                 char_type const c = d->text_[i];
2482
2483                 // Do we need to change font?
2484                 if ((current_font != running_font ||
2485                      current_font.language() != running_font.language()) &&
2486                         i != body_pos - 1)
2487                 {
2488                         odocstringstream ods;
2489                         column += current_font.latexWriteStartChanges(ods, bparams,
2490                                                               runparams, basefont,
2491                                                               last_font);
2492                         running_font = current_font;
2493                         open_font = true;
2494                         docstring fontchange = ods.str();
2495                         // check whether the fontchange ends with a \\textcolor
2496                         // modifier and the text starts with a space (bug 4473)
2497                         docstring const last_modifier = rsplit(fontchange, '\\');
2498                         if (prefixIs(last_modifier, from_ascii("textcolor")) && c == ' ')
2499                                 os << fontchange << from_ascii("{}");
2500                         // check if the fontchange ends with a trailing blank
2501                         // (like "\small " (see bug 3382)
2502                         else if (suffixIs(fontchange, ' ') && c == ' ')
2503                                 os << fontchange.substr(0, fontchange.size() - 1)
2504                                    << from_ascii("{}");
2505                         else
2506                                 os << fontchange;
2507                 }
2508
2509                 // FIXME: think about end_pos implementation...
2510                 if (c == ' ' && i >= start_pos && (end_pos == -1 || i < end_pos)) {
2511                         // FIXME: integrate this case in latexSpecialChar
2512                         // Do not print the separation of the optional argument
2513                         // if style.pass_thru is false. This works because
2514                         // latexSpecialChar ignores spaces if
2515                         // style.pass_thru is false.
2516                         if (i != body_pos - 1) {
2517                                 if (d->simpleTeXBlanks(runparams, os,
2518                                                 i, column, current_font, style)) {
2519                                         // A surrogate pair was output. We
2520                                         // must not call latexSpecialChar
2521                                         // in this iteration, since it would output
2522                                         // the combining character again.
2523                                         ++i;
2524                                         continue;
2525                                 }
2526                         }
2527                 }
2528
2529                 OutputParams rp = runparams;
2530                 rp.free_spacing = style.free_spacing;
2531                 rp.local_font = &current_font;
2532                 rp.intitle = style.intitle;
2533
2534                 // Two major modes:  LaTeX or plain
2535                 // Handle here those cases common to both modes
2536                 // and then split to handle the two modes separately.
2537                 if (c == META_INSET) {
2538                         if (i >= start_pos && (end_pos == -1 || i < end_pos)) {
2539                                 d->latexInset(bparams, os, rp, running_font,
2540                                                 basefont, outerfont, open_font,
2541                                                 runningChange, style, i, column);
2542                         }
2543                 } else {
2544                         if (i >= start_pos && (end_pos == -1 || i < end_pos)) {
2545                                 try {
2546                                         d->latexSpecialChar(os, bparams, rp, running_font, runningChange,
2547                                                             style, i, end_pos, column);
2548                                 } catch (EncodingException & e) {
2549                                 if (runparams.dryrun) {
2550                                         os << "<" << _("LyX Warning: ")
2551                                            << _("uncodable character") << " '";
2552                                         os.put(c);
2553                                         os << "'>";
2554                                 } else {
2555                                         // add location information and throw again.
2556                                         e.par_id = id();
2557                                         e.pos = i;
2558                                         throw(e);
2559                                 }
2560                         }
2561                 }
2562                 }
2563
2564                 // Set the encoding to that returned from latexSpecialChar (see
2565                 // comment for encoding member in OutputParams.h)
2566                 runparams.encoding = rp.encoding;
2567         }
2568
2569         // If we have an open font definition, we have to close it
2570         if (open_font) {
2571 #ifdef FIXED_LANGUAGE_END_DETECTION
2572                 if (next_) {
2573                         running_font.latexWriteEndChanges(os, bparams,
2574                                         runparams, basefont,
2575                                         next_->getFont(bparams, 0, outerfont));
2576                 } else {
2577                         running_font.latexWriteEndChanges(os, bparams,
2578                                         runparams, basefont, basefont);
2579                 }
2580 #else
2581 //FIXME: For now we ALWAYS have to close the foreign font settings if they are
2582 //FIXME: there as we start another \selectlanguage with the next paragraph if
2583 //FIXME: we are in need of this. This should be fixed sometime (Jug)
2584                 running_font.latexWriteEndChanges(os, bparams, runparams,
2585                                 basefont, basefont);
2586 #endif
2587         }
2588
2589         column += Changes::latexMarkChange(os, bparams, runningChange,
2590                                            Change(Change::UNCHANGED), runparams);
2591
2592         // Needed if there is an optional argument but no contents.
2593         if (body_pos > 0 && body_pos == size()) {
2594                 os << "}]~";
2595         }
2596
2597         if (!style.rightdelim().empty()) {
2598                 os << style.rightdelim();
2599                 column += style.rightdelim().size();
2600         }
2601
2602         if (allowcust && d->endTeXParParams(bparams, os, runparams)
2603             && runparams.encoding != prev_encoding) {
2604                 runparams.encoding = prev_encoding;
2605                 os << setEncoding(prev_encoding->iconvName());
2606         }
2607
2608         LYXERR(Debug::LATEX, "Paragraph::latex... done " << this);
2609 }
2610
2611
2612 bool Paragraph::emptyTag() const
2613 {
2614         for (pos_type i = 0; i < size(); ++i) {
2615                 if (Inset const * inset = getInset(i)) {
2616                         InsetCode lyx_code = inset->lyxCode();
2617                         // FIXME testing like that is wrong. What is
2618                         // the intent?
2619                         if (lyx_code != TOC_CODE &&
2620                             lyx_code != INCLUDE_CODE &&
2621                             lyx_code != GRAPHICS_CODE &&
2622                             lyx_code != ERT_CODE &&
2623                             lyx_code != LISTINGS_CODE &&
2624                             lyx_code != FLOAT_CODE &&
2625                             lyx_code != TABULAR_CODE) {
2626                                 return false;
2627                         }
2628                 } else {
2629                         char_type c = d->text_[i];
2630                         if (c != ' ' && c != '\t')
2631                                 return false;
2632                 }
2633         }
2634         return true;
2635 }
2636
2637
2638 string Paragraph::getID(Buffer const & buf, OutputParams const & runparams)
2639         const
2640 {
2641         for (pos_type i = 0; i < size(); ++i) {
2642                 if (Inset const * inset = getInset(i)) {
2643                         InsetCode lyx_code = inset->lyxCode();
2644                         if (lyx_code == LABEL_CODE) {
2645                                 InsetLabel const * const il = static_cast<InsetLabel const *>(inset);
2646                                 docstring const & id = il->getParam("name");
2647                                 return "id='" + to_utf8(sgml::cleanID(buf, runparams, id)) + "'";
2648                         }
2649                 }
2650         }
2651         return string();
2652 }
2653
2654
2655 pos_type Paragraph::firstWordDocBook(odocstream & os, OutputParams const & runparams)
2656         const
2657 {
2658         pos_type i;
2659         for (i = 0; i < size(); ++i) {
2660                 if (Inset const * inset = getInset(i)) {
2661                         inset->docbook(os, runparams);
2662                 } else {
2663                         char_type c = d->text_[i];
2664                         if (c == ' ')
2665                                 break;
2666                         os << sgml::escapeChar(c);
2667                 }
2668         }
2669         return i;
2670 }
2671
2672
2673 pos_type Paragraph::firstWordLyXHTML(XHTMLStream & xs, OutputParams const & runparams)
2674         const
2675 {
2676         pos_type i;
2677         for (i = 0; i < size(); ++i) {
2678                 if (Inset const * inset = getInset(i)) {
2679                         inset->xhtml(xs, runparams);
2680                 } else {
2681                         char_type c = d->text_[i];
2682                         if (c == ' ')
2683                                 break;
2684                         xs << c;
2685                 }
2686         }
2687         return i;
2688 }
2689
2690
2691 bool Paragraph::Private::onlyText(Buffer const & buf, Font const & outerfont, pos_type initial) const
2692 {
2693         Font font_old;
2694         pos_type size = text_.size();
2695         for (pos_type i = initial; i < size; ++i) {
2696                 Font font = owner_->getFont(buf.params(), i, outerfont);
2697                 if (text_[i] == META_INSET)
2698                         return false;
2699                 if (i != initial && font != font_old)
2700                         return false;
2701                 font_old = font;
2702         }
2703
2704         return true;
2705 }
2706
2707
2708 void Paragraph::simpleDocBookOnePar(Buffer const & buf,
2709                                     odocstream & os,
2710                                     OutputParams const & runparams,
2711                                     Font const & outerfont,
2712                                     pos_type initial) const
2713 {
2714         bool emph_flag = false;
2715
2716         Layout const & style = *d->layout_;
2717         FontInfo font_old =
2718                 style.labeltype == LABEL_MANUAL ? style.labelfont : style.font;
2719
2720         if (style.pass_thru && !d->onlyText(buf, outerfont, initial))
2721                 os << "]]>";
2722
2723         // parsing main loop
2724         for (pos_type i = initial; i < size(); ++i) {
2725                 Font font = getFont(buf.params(), i, outerfont);
2726
2727                 // handle <emphasis> tag
2728                 if (font_old.emph() != font.fontInfo().emph()) {
2729                         if (font.fontInfo().emph() == FONT_ON) {
2730                                 os << "<emphasis>";
2731                                 emph_flag = true;
2732                         } else if (i != initial) {
2733                                 os << "</emphasis>";
2734                                 emph_flag = false;
2735                         }
2736                 }
2737
2738                 if (Inset const * inset = getInset(i)) {
2739                         inset->docbook(os, runparams);
2740                 } else {
2741                         char_type c = d->text_[i];
2742
2743                         if (style.pass_thru)
2744                                 os.put(c);
2745                         else
2746                                 os << sgml::escapeChar(c);
2747                 }
2748                 font_old = font.fontInfo();
2749         }
2750
2751         if (emph_flag) {
2752                 os << "</emphasis>";
2753         }
2754
2755         if (style.free_spacing)
2756                 os << '\n';
2757         if (style.pass_thru && !d->onlyText(buf, outerfont, initial))
2758                 os << "<![CDATA[";
2759 }
2760
2761
2762 namespace {
2763 void doFontSwitch(vector<html::FontTag> & tagsToOpen,
2764                   vector<html::EndFontTag> & tagsToClose,
2765                   bool & flag, FontState curstate, html::FontTypes type)
2766 {
2767         if (curstate == FONT_ON) {
2768                 tagsToOpen.push_back(html::FontTag(type));
2769                 flag = true;
2770         } else if (flag) {
2771                 tagsToClose.push_back(html::EndFontTag(type));
2772                 flag = false;
2773         }
2774 }
2775 }
2776
2777
2778 docstring Paragraph::simpleLyXHTMLOnePar(Buffer const & buf,
2779                                     XHTMLStream & xs,
2780                                     OutputParams const & runparams,
2781                                     Font const & outerfont,
2782                                     bool start_paragraph, bool close_paragraph,
2783                                     pos_type initial) const
2784 {
2785         docstring retval;
2786
2787         // track whether we have opened these tags
2788         bool emph_flag = false;
2789         bool bold_flag = false;
2790         bool noun_flag = false;
2791         bool ubar_flag = false;
2792         bool dbar_flag = false;
2793         bool sout_flag = false;
2794         bool wave_flag = false;
2795         // shape tags
2796         bool shap_flag = false;
2797         // family tags
2798         bool faml_flag = false;
2799         // size tags
2800         bool size_flag = false;
2801
2802         Layout const & style = *d->layout_;
2803
2804         if (start_paragraph)
2805                 xs.startDivision(allowEmpty());
2806
2807         FontInfo font_old =
2808                 style.labeltype == LABEL_MANUAL ? style.labelfont : style.font;
2809
2810         FontShape  curr_fs   = INHERIT_SHAPE;
2811         FontFamily curr_fam  = INHERIT_FAMILY;
2812         FontSize   curr_size = FONT_SIZE_INHERIT;
2813
2814         string const default_family =
2815                 buf.masterBuffer()->params().fonts_default_family;
2816
2817         vector<html::FontTag> tagsToOpen;
2818         vector<html::EndFontTag> tagsToClose;
2819
2820         // parsing main loop
2821         for (pos_type i = initial; i < size(); ++i) {
2822                 // let's not show deleted material in the output
2823                 if (isDeleted(i))
2824                         continue;
2825
2826                 Font const font = getFont(buf.masterBuffer()->params(), i, outerfont);
2827
2828                 // emphasis
2829                 FontState curstate = font.fontInfo().emph();
2830                 if (font_old.emph() != curstate)
2831                         doFontSwitch(tagsToOpen, tagsToClose, emph_flag, curstate, html::FT_EMPH);
2832
2833                 // noun
2834                 curstate = font.fontInfo().noun();
2835                 if (font_old.noun() != curstate)
2836                         doFontSwitch(tagsToOpen, tagsToClose, noun_flag, curstate, html::FT_NOUN);
2837
2838                 // underbar
2839                 curstate = font.fontInfo().underbar();
2840                 if (font_old.underbar() != curstate)
2841                         doFontSwitch(tagsToOpen, tagsToClose, ubar_flag, curstate, html::FT_UBAR);
2842
2843                 // strikeout
2844                 curstate = font.fontInfo().strikeout();
2845                 if (font_old.strikeout() != curstate)
2846                         doFontSwitch(tagsToOpen, tagsToClose, sout_flag, curstate, html::FT_SOUT);
2847
2848                 // double underbar
2849                 curstate = font.fontInfo().uuline();
2850                 if (font_old.uuline() != curstate)
2851                         doFontSwitch(tagsToOpen, tagsToClose, dbar_flag, curstate, html::FT_DBAR);
2852
2853                 // wavy line
2854                 curstate = font.fontInfo().uwave();
2855                 if (font_old.uwave() != curstate)
2856                         doFontSwitch(tagsToOpen, tagsToClose, wave_flag, curstate, html::FT_WAVE);
2857
2858                 // bold
2859                 // a little hackish, but allows us to reuse what we have.
2860                 curstate = (font.fontInfo().series() == BOLD_SERIES ? FONT_ON : FONT_OFF);
2861                 if (font_old.series() != font.fontInfo().series())
2862                         doFontSwitch(tagsToOpen, tagsToClose, bold_flag, curstate, html::FT_BOLD);
2863
2864                 // Font shape
2865                 curr_fs = font.fontInfo().shape();
2866                 FontShape old_fs = font_old.shape();
2867                 if (old_fs != curr_fs) {
2868                         if (shap_flag) {
2869                                 switch (old_fs) {
2870                                 case ITALIC_SHAPE:
2871                                         tagsToClose.push_back(html::EndFontTag(html::FT_ITALIC));
2872                                         break;
2873                                 case SLANTED_SHAPE:
2874                                         tagsToClose.push_back(html::EndFontTag(html::FT_SLANTED));
2875                                         break;
2876                                 case SMALLCAPS_SHAPE:
2877                                         tagsToClose.push_back(html::EndFontTag(html::FT_SMALLCAPS));
2878                                         break;
2879                                 case UP_SHAPE:
2880                                 case INHERIT_SHAPE:
2881                                         break;
2882                                 default:
2883                                         // the other tags are for internal use
2884                                         LATTEST(false);
2885                                         break;
2886                                 }
2887                                 shap_flag = false;
2888                         }
2889                         switch (curr_fs) {
2890                         case ITALIC_SHAPE:
2891                                 tagsToOpen.push_back(html::FontTag(html::FT_ITALIC));
2892                                 shap_flag = true;
2893                                 break;
2894                         case SLANTED_SHAPE:
2895                                 tagsToOpen.push_back(html::FontTag(html::FT_SLANTED));
2896                                 shap_flag = true;
2897                                 break;
2898                         case SMALLCAPS_SHAPE:
2899                                 tagsToOpen.push_back(html::FontTag(html::FT_SMALLCAPS));
2900                                 shap_flag = true;
2901                                 break;
2902                         case UP_SHAPE:
2903                         case INHERIT_SHAPE:
2904                                 break;
2905                         default:
2906                                 // the other tags are for internal use
2907                                 LATTEST(false);
2908                                 break;
2909                         }
2910                 }
2911
2912                 // Font family
2913                 curr_fam = font.fontInfo().family();
2914                 FontFamily old_fam = font_old.family();
2915                 if (old_fam != curr_fam) {
2916                         if (faml_flag) {
2917                                 switch (old_fam) {
2918                                 case ROMAN_FAMILY:
2919                                         tagsToClose.push_back(html::EndFontTag(html::FT_ROMAN));
2920                                         break;
2921                                 case SANS_FAMILY:
2922                                         tagsToClose.push_back(html::EndFontTag(html::FT_SANS));
2923                                         break;
2924                                 case TYPEWRITER_FAMILY:
2925                                         tagsToClose.push_back(html::EndFontTag(html::FT_TYPE));
2926                                         break;
2927                                 case INHERIT_FAMILY:
2928                                         break;
2929                                 default:
2930                                         // the other tags are for internal use
2931                                         LATTEST(false);
2932                                         break;
2933                                 }
2934                                 faml_flag = false;
2935                         }
2936                         switch (curr_fam) {
2937                         case ROMAN_FAMILY:
2938                                 // we will treat a "default" font family as roman, since we have
2939                                 // no other idea what to do.
2940                                 if (default_family != "rmdefault" && default_family != "default") {
2941                                         tagsToOpen.push_back(html::FontTag(html::FT_ROMAN));
2942                                         faml_flag = true;
2943                                 }
2944                                 break;
2945                         case SANS_FAMILY:
2946                                 if (default_family != "sfdefault") {
2947                                         tagsToOpen.push_back(html::FontTag(html::FT_SANS));
2948                                         faml_flag = true;
2949                                 }
2950                                 break;
2951                         case TYPEWRITER_FAMILY:
2952                                 if (default_family != "ttdefault") {
2953                                         tagsToOpen.push_back(html::FontTag(html::FT_TYPE));
2954                                         faml_flag = true;
2955                                 }
2956                                 break;
2957                         case INHERIT_FAMILY:
2958                                 break;
2959                         default:
2960                                 // the other tags are for internal use
2961                                 LATTEST(false);
2962                                 break;
2963                         }
2964                 }
2965
2966                 // Font size
2967                 curr_size = font.fontInfo().size();
2968                 FontSize old_size = font_old.size();
2969                 if (old_size != curr_size) {
2970                         if (size_flag) {
2971                                 switch (old_size) {
2972                                 case FONT_SIZE_TINY:
2973                                         tagsToClose.push_back(html::EndFontTag(html::FT_SIZE_TINY));
2974                                         break;
2975                                 case FONT_SIZE_SCRIPT:
2976                                         tagsToClose.push_back(html::EndFontTag(html::FT_SIZE_SCRIPT));
2977                                         break;
2978                                 case FONT_SIZE_FOOTNOTE:
2979                                         tagsToClose.push_back(html::EndFontTag(html::FT_SIZE_FOOTNOTE));
2980                                         break;
2981                                 case FONT_SIZE_SMALL:
2982                                         tagsToClose.push_back(html::EndFontTag(html::FT_SIZE_SMALL));
2983                                         break;
2984                                 case FONT_SIZE_LARGE:
2985                                         tagsToClose.push_back(html::EndFontTag(html::FT_SIZE_LARGE));
2986                                         break;
2987                                 case FONT_SIZE_LARGER:
2988                                         tagsToClose.push_back(html::EndFontTag(html::FT_SIZE_LARGER));
2989                                         break;
2990                                 case FONT_SIZE_LARGEST:
2991                                         tagsToClose.push_back(html::EndFontTag(html::FT_SIZE_LARGEST));
2992                                         break;
2993                                 case FONT_SIZE_HUGE:
2994                                         tagsToClose.push_back(html::EndFontTag(html::FT_SIZE_HUGE));
2995                                         break;
2996                                 case FONT_SIZE_HUGER:
2997                                         tagsToClose.push_back(html::EndFontTag(html::FT_SIZE_HUGER));
2998                                         break;
2999                                 case FONT_SIZE_INCREASE:
3000                                         tagsToClose.push_back(html::EndFontTag(html::FT_SIZE_INCREASE));
3001                                         break;
3002                                 case FONT_SIZE_DECREASE:
3003                                         tagsToClose.push_back(html::EndFontTag(html::FT_SIZE_DECREASE));
3004                                         break;
3005                                 case FONT_SIZE_INHERIT:
3006                                 case FONT_SIZE_NORMAL:
3007                                         break;
3008                                 default:
3009                                         // the other tags are for internal use
3010                                         LATTEST(false);
3011                                         break;
3012                                 }
3013                                 size_flag = false;
3014                         }
3015                         switch (curr_size) {
3016                         case FONT_SIZE_TINY:
3017                                 tagsToOpen.push_back(html::FontTag(html::FT_SIZE_TINY));
3018                                 size_flag = true;
3019                                 break;
3020                         case FONT_SIZE_SCRIPT:
3021                                 tagsToOpen.push_back(html::FontTag(html::FT_SIZE_SCRIPT));
3022                                 size_flag = true;
3023                                 break;
3024                         case FONT_SIZE_FOOTNOTE:
3025                                 tagsToOpen.push_back(html::FontTag(html::FT_SIZE_FOOTNOTE));
3026                                 size_flag = true;
3027                                 break;
3028                         case FONT_SIZE_SMALL:
3029                                 tagsToOpen.push_back(html::FontTag(html::FT_SIZE_SMALL));
3030                                 size_flag = true;
3031                                 break;
3032                         case FONT_SIZE_LARGE:
3033                                 tagsToOpen.push_back(html::FontTag(html::FT_SIZE_LARGE));
3034                                 size_flag = true;
3035                                 break;
3036                         case FONT_SIZE_LARGER:
3037                                 tagsToOpen.push_back(html::FontTag(html::FT_SIZE_LARGER));
3038                                 size_flag = true;
3039                                 break;
3040                         case FONT_SIZE_LARGEST:
3041                                 tagsToOpen.push_back(html::FontTag(html::FT_SIZE_LARGEST));
3042                                 size_flag = true;
3043                                 break;
3044                         case FONT_SIZE_HUGE:
3045                                 tagsToOpen.push_back(html::FontTag(html::FT_SIZE_HUGE));
3046                                 size_flag = true;
3047                                 break;
3048                         case FONT_SIZE_HUGER:
3049                                 tagsToOpen.push_back(html::FontTag(html::FT_SIZE_HUGER));
3050                                 size_flag = true;
3051                                 break;
3052                         case FONT_SIZE_INCREASE:
3053                                 tagsToOpen.push_back(html::FontTag(html::FT_SIZE_INCREASE));
3054                                 size_flag = true;
3055                                 break;
3056                         case FONT_SIZE_DECREASE:
3057                                 tagsToOpen.push_back(html::FontTag(html::FT_SIZE_DECREASE));
3058                                 size_flag = true;
3059                                 break;
3060                         case FONT_SIZE_NORMAL:
3061                         case FONT_SIZE_INHERIT:
3062                                 break;
3063                         default:
3064                                 // the other tags are for internal use
3065                                 LATTEST(false);
3066                                 break;
3067                         }
3068                 }
3069
3070                 // FIXME XHTML
3071                 // Other such tags? What about the other text ranges?
3072
3073                 vector<html::EndFontTag>::const_iterator cit = tagsToClose.begin();
3074                 vector<html::EndFontTag>::const_iterator cen = tagsToClose.end();
3075                 for (; cit != cen; ++cit)
3076                         xs << *cit;
3077
3078                 vector<html::FontTag>::const_iterator sit = tagsToOpen.begin();
3079                 vector<html::FontTag>::const_iterator sen = tagsToOpen.end();
3080                 for (; sit != sen; ++sit)
3081                         xs << *sit;
3082
3083                 tagsToClose.clear();
3084                 tagsToOpen.clear();
3085
3086                 Inset const * inset = getInset(i);
3087                 if (inset) {
3088                         if (!runparams.for_toc || inset->isInToc()) {
3089                                 OutputParams np = runparams;
3090                                 np.local_font = &font;
3091                                 // If the paragraph has size 1, then we are in the "special
3092                                 // case" where we do not output the containing paragraph info
3093                                 if (!inset->getLayout().htmlisblock() && size() != 1)
3094                                         np.html_in_par = true;
3095                                 retval += inset->xhtml(xs, np);
3096                         }
3097                 } else {
3098                         char_type c = getUChar(buf.masterBuffer()->params(), i);
3099                         xs << c;
3100                 }
3101                 font_old = font.fontInfo();
3102         }
3103
3104         // FIXME XHTML
3105         // I'm worried about what happens if a branch, say, is itself
3106         // wrapped in some font stuff. I think that will not work.
3107         xs.closeFontTags();
3108         if (close_paragraph)
3109                 xs.endDivision();
3110
3111         return retval;
3112 }
3113
3114
3115 bool Paragraph::isHfill(pos_type pos) const
3116 {
3117         Inset const * inset = getInset(pos);
3118         return inset && inset->isHfill();
3119 }
3120
3121
3122 bool Paragraph::isNewline(pos_type pos) const
3123 {
3124         Inset const * inset = getInset(pos);
3125         return inset && inset->lyxCode() == NEWLINE_CODE;
3126 }
3127
3128
3129 bool Paragraph::isEnvSeparator(pos_type pos) const
3130 {
3131         Inset const * inset = getInset(pos);
3132         return inset && inset->lyxCode() == SEPARATOR_CODE;
3133 }
3134
3135
3136 bool Paragraph::isLineSeparator(pos_type pos) const
3137 {
3138         char_type const c = d->text_[pos];
3139         if (isLineSeparatorChar(c))
3140                 return true;
3141         Inset const * inset = getInset(pos);
3142         return inset && inset->isLineSeparator();
3143 }
3144
3145
3146 bool Paragraph::isWordSeparator(pos_type pos) const
3147 {
3148         if (pos == size())
3149                 return true;
3150         if (Inset const * inset = getInset(pos))
3151                 return !inset->isLetter();
3152         // if we have a hard hyphen (no en- or emdash) or apostrophe
3153         // we pass this to the spell checker
3154         // FIXME: this method is subject to change, visit
3155         // https://bugzilla.mozilla.org/show_bug.cgi?id=355178
3156         // to get an impression how complex this is.
3157         if (isHardHyphenOrApostrophe(pos))
3158                 return false;
3159         char_type const c = d->text_[pos];
3160         // We want to pass the escape chars to the spellchecker
3161         docstring const escape_chars = from_utf8(lyxrc.spellchecker_esc_chars);
3162         return !isLetterChar(c) && !isDigitASCII(c) && !contains(escape_chars, c);
3163 }
3164
3165
3166 bool Paragraph::isHardHyphenOrApostrophe(pos_type pos) const
3167 {
3168         pos_type const psize = size();
3169         if (pos >= psize)
3170                 return false;
3171         char_type const c = d->text_[pos];
3172         if (c != '-' && c != '\'')
3173                 return false;
3174         int nextpos = pos + 1;
3175         int prevpos = pos > 0 ? pos - 1 : 0;
3176         if ((nextpos == psize || isSpace(nextpos))
3177                 && (pos == 0 || isSpace(prevpos)))
3178                 return false;
3179         return true;
3180 }
3181
3182
3183 FontSpan const & Paragraph::getSpellRange(pos_type pos) const
3184 {
3185         return d->speller_state_.getRange(pos);
3186 }
3187
3188
3189 bool Paragraph::isChar(pos_type pos) const
3190 {
3191         if (Inset const * inset = getInset(pos))
3192                 return inset->isChar();
3193         char_type const c = d->text_[pos];
3194         return !isLetterChar(c) && !isDigitASCII(c) && !lyx::isSpace(c);
3195 }
3196
3197
3198 bool Paragraph::isSpace(pos_type pos) const
3199 {
3200         if (Inset const * inset = getInset(pos))
3201                 return inset->isSpace();
3202         char_type const c = d->text_[pos];
3203         return lyx::isSpace(c);
3204 }
3205
3206
3207 Language const *
3208 Paragraph::getParLanguage(BufferParams const & bparams) const
3209 {
3210         if (!empty())
3211                 return getFirstFontSettings(bparams).language();
3212         // FIXME: we should check the prev par as well (Lgb)
3213         return bparams.language;
3214 }
3215
3216
3217 bool Paragraph::isRTL(BufferParams const & bparams) const
3218 {
3219         return getParLanguage(bparams)->rightToLeft()
3220                 && !inInset().getLayout().forceLTR();
3221 }
3222
3223
3224 void Paragraph::changeLanguage(BufferParams const & bparams,
3225                                Language const * from, Language const * to)
3226 {
3227         // change language including dummy font change at the end
3228         for (pos_type i = 0; i <= size(); ++i) {
3229                 Font font = getFontSettings(bparams, i);
3230                 if (font.language() == from) {
3231                         font.setLanguage(to);
3232                         setFont(i, font);
3233                         d->requestSpellCheck(i);
3234                 }
3235         }
3236 }
3237
3238
3239 bool Paragraph::isMultiLingual(BufferParams const & bparams) const
3240 {
3241         Language const * doc_language = bparams.language;
3242         FontList::const_iterator cit = d->fontlist_.begin();
3243         FontList::const_iterator end = d->fontlist_.end();
3244
3245         for (; cit != end; ++cit)
3246                 if (cit->font().language() != ignore_language &&
3247                     cit->font().language() != latex_language &&
3248                     cit->font().language() != doc_language)
3249                         return true;
3250         return false;
3251 }
3252
3253
3254 void Paragraph::getLanguages(std::set<Language const *> & languages) const
3255 {
3256         FontList::const_iterator cit = d->fontlist_.begin();
3257         FontList::const_iterator end = d->fontlist_.end();
3258
3259         for (; cit != end; ++cit) {
3260                 Language const * lang = cit->font().language();
3261                 if (lang != ignore_language &&
3262                     lang != latex_language)
3263                         languages.insert(lang);
3264         }
3265 }
3266
3267
3268 docstring Paragraph::asString(int options) const
3269 {
3270         return asString(0, size(), options);
3271 }
3272
3273
3274 docstring Paragraph::asString(pos_type beg, pos_type end, int options, const OutputParams *runparams) const
3275 {
3276         odocstringstream os;
3277
3278         if (beg == 0
3279             && options & AS_STR_LABEL
3280             && !d->params_.labelString().empty())
3281                 os << d->params_.labelString() << ' ';
3282
3283         for (pos_type i = beg; i < end; ++i) {
3284                 if ((options & AS_STR_SKIPDELETE) && isDeleted(i))
3285                         continue;
3286                 char_type const c = d->text_[i];
3287                 if (isPrintable(c) || c == '\t'
3288                     || (c == '\n' && (options & AS_STR_NEWLINES)))
3289                         os.put(c);
3290                 else if (c == META_INSET && (options & AS_STR_INSETS)) {
3291                         if (c == META_INSET && (options & AS_STR_PLAINTEXT)) {
3292                                 LASSERT(runparams != 0, return docstring());
3293                                 getInset(i)->plaintext(os, *runparams);
3294                         } else {
3295                                 getInset(i)->toString(os);
3296                         }
3297                 }
3298         }
3299
3300         return os.str();
3301 }
3302
3303
3304 void Paragraph::forOutliner(docstring & os, size_t const maxlen,
3305                                                         bool const shorten) const
3306 {
3307         size_t tmplen = shorten ? maxlen + 1 : maxlen;
3308         if (!d->params_.labelString().empty())
3309                 os += d->params_.labelString() + ' ';
3310         for (pos_type i = 0; i < size() && os.length() < tmplen; ++i) {
3311                 if (isDeleted(i))
3312                         continue;
3313                 char_type const c = d->text_[i];
3314                 if (isPrintable(c))
3315                         os += c;
3316                 else if (c == META_INSET)
3317                         getInset(i)->forOutliner(os, tmplen, false);
3318         }
3319         if (shorten)
3320                 Text::shortenForOutliner(os, maxlen);
3321 }
3322
3323
3324 void Paragraph::setInsetOwner(Inset const * inset)
3325 {
3326         d->inset_owner_ = inset;
3327 }
3328
3329
3330 int Paragraph::id() const
3331 {
3332         return d->id_;
3333 }
3334
3335
3336 void Paragraph::setId(int id)
3337 {
3338         d->id_ = id;
3339 }
3340
3341
3342 Layout const & Paragraph::layout() const
3343 {
3344         return *d->layout_;
3345 }
3346
3347
3348 void Paragraph::setLayout(Layout const & layout)
3349 {
3350         d->layout_ = &layout;
3351 }
3352
3353
3354 void Paragraph::setDefaultLayout(DocumentClass const & tc)
3355 {
3356         setLayout(tc.defaultLayout());
3357 }
3358
3359
3360 void Paragraph::setPlainLayout(DocumentClass const & tc)
3361 {
3362         setLayout(tc.plainLayout());
3363 }
3364
3365
3366 void Paragraph::setPlainOrDefaultLayout(DocumentClass const & tclass)
3367 {
3368         if (usePlainLayout())
3369                 setPlainLayout(tclass);
3370         else
3371                 setDefaultLayout(tclass);
3372 }
3373
3374
3375 Inset const & Paragraph::inInset() const
3376 {
3377         LBUFERR(d->inset_owner_);
3378         return *d->inset_owner_;
3379 }
3380
3381
3382 ParagraphParameters & Paragraph::params()
3383 {
3384         return d->params_;
3385 }
3386
3387
3388 ParagraphParameters const & Paragraph::params() const
3389 {
3390         return d->params_;
3391 }
3392
3393
3394 bool Paragraph::isFreeSpacing() const
3395 {
3396         if (d->layout_->free_spacing)
3397                 return true;
3398         return d->inset_owner_ && d->inset_owner_->isFreeSpacing();
3399 }
3400
3401
3402 bool Paragraph::allowEmpty() const
3403 {
3404         if (d->layout_->keepempty)
3405                 return true;
3406         return d->inset_owner_ && d->inset_owner_->allowEmpty();
3407 }
3408
3409
3410 bool Paragraph::brokenBiblio() const
3411 {
3412         // there is a problem if there is no bibitem at position 0 or
3413         // if there is another bibitem in the paragraph.
3414         return d->layout_->labeltype == LABEL_BIBLIO
3415                 && (d->insetlist_.find(BIBITEM_CODE) != 0
3416                     || d->insetlist_.find(BIBITEM_CODE, 1) > 0);
3417 }
3418
3419
3420 int Paragraph::fixBiblio(Buffer const & buffer)
3421 {
3422         // FIXME: What about the case where paragraph is not BIBLIO
3423         // but there is an InsetBibitem?
3424         // FIXME: when there was already an inset at 0, the return value is 1,
3425         // which does not tell whether another inset has been remove; the
3426         // cursor cannot be correctly updated.
3427
3428         if (d->layout_->labeltype != LABEL_BIBLIO)
3429                 return 0;
3430
3431         bool const track_changes = buffer.params().track_changes;
3432         int bibitem_pos = d->insetlist_.find(BIBITEM_CODE);
3433         bool const hasbibitem0 = bibitem_pos == 0;
3434
3435         if (hasbibitem0) {
3436                 bibitem_pos = d->insetlist_.find(BIBITEM_CODE, 1);
3437                 // There was an InsetBibitem at pos 0, and no other one => OK
3438                 if (bibitem_pos == -1)
3439                         return 0;
3440                 // there is a bibitem at the 0 position, but since
3441                 // there is a second one, we copy the second on the
3442                 // first. We're assuming there are at most two of
3443                 // these, which there should be.
3444                 // FIXME: why does it make sense to do that rather
3445                 // than keep the first? (JMarc)
3446                 Inset * inset = releaseInset(bibitem_pos);
3447                 d->insetlist_.begin()->inset = inset;
3448                 return -bibitem_pos;
3449         }
3450
3451         // We need to create an inset at the beginning
3452         Inset * inset = 0;
3453         if (bibitem_pos > 0) {
3454                 // there was one somewhere in the paragraph, let's move it
3455                 inset = d->insetlist_.release(bibitem_pos);
3456                 eraseChar(bibitem_pos, track_changes);
3457         } else
3458                 // make a fresh one
3459                 inset = new InsetBibitem(const_cast<Buffer *>(&buffer),
3460                                          InsetCommandParams(BIBITEM_CODE));
3461
3462         Font font(inherit_font, buffer.params().language);
3463         insertInset(0, inset, font, Change(track_changes ? Change::INSERTED
3464                                                    : Change::UNCHANGED));
3465
3466         return 1;
3467 }
3468
3469
3470 void Paragraph::checkAuthors(AuthorList const & authorList)
3471 {
3472         d->changes_.checkAuthors(authorList);
3473 }
3474
3475
3476 bool Paragraph::isChanged(pos_type pos) const
3477 {
3478         return lookupChange(pos).changed();
3479 }
3480
3481
3482 bool Paragraph::isInserted(pos_type pos) const
3483 {
3484         return lookupChange(pos).inserted();
3485 }
3486
3487
3488 bool Paragraph::isDeleted(pos_type pos) const
3489 {
3490         return lookupChange(pos).deleted();
3491 }
3492
3493
3494 InsetList const & Paragraph::insetList() const
3495 {
3496         return d->insetlist_;
3497 }
3498
3499
3500 void Paragraph::setBuffer(Buffer & b)
3501 {
3502         d->insetlist_.setBuffer(b);
3503 }
3504
3505
3506 Inset * Paragraph::releaseInset(pos_type pos)
3507 {
3508         Inset * inset = d->insetlist_.release(pos);
3509         /// does not honour change tracking!
3510         eraseChar(pos, false);
3511         return inset;
3512 }
3513
3514
3515 Inset * Paragraph::getInset(pos_type pos)
3516 {
3517         return (pos < pos_type(d->text_.size()) && d->text_[pos] == META_INSET)
3518                  ? d->insetlist_.get(pos) : 0;
3519 }
3520
3521
3522 Inset const * Paragraph::getInset(pos_type pos) const
3523 {
3524         return (pos < pos_type(d->text_.size()) && d->text_[pos] == META_INSET)
3525                  ? d->insetlist_.get(pos) : 0;
3526 }
3527
3528
3529 void Paragraph::changeCase(BufferParams const & bparams, pos_type pos,
3530                 pos_type & right, TextCase action)
3531 {
3532         // process sequences of modified characters; in change
3533         // tracking mode, this approach results in much better
3534         // usability than changing case on a char-by-char basis
3535         // We also need to track the current font, since font
3536         // changes within sequences can occur.
3537         vector<pair<char_type, Font> > changes;
3538
3539         bool const trackChanges = bparams.track_changes;
3540
3541         bool capitalize = true;
3542
3543         for (; pos < right; ++pos) {
3544                 char_type oldChar = d->text_[pos];
3545                 char_type newChar = oldChar;
3546
3547                 // ignore insets and don't play with deleted text!
3548                 if (oldChar != META_INSET && !isDeleted(pos)) {
3549                         switch (action) {
3550                                 case text_lowercase:
3551                                         newChar = lowercase(oldChar);
3552                                         break;
3553                                 case text_capitalization:
3554                                         if (capitalize) {
3555                                                 newChar = uppercase(oldChar);
3556                                                 capitalize = false;
3557                                         }
3558                                         break;
3559                                 case text_uppercase:
3560                                         newChar = uppercase(oldChar);
3561                                         break;
3562                         }
3563                 }
3564
3565                 if (isWordSeparator(pos) || isDeleted(pos)) {
3566                         // permit capitalization again
3567                         capitalize = true;
3568                 }
3569
3570                 if (oldChar != newChar) {
3571                         changes.push_back(make_pair(newChar, getFontSettings(bparams, pos)));
3572                         if (pos != right - 1)
3573                                 continue;
3574                         // step behind the changing area
3575                         pos++;
3576                 }
3577
3578                 int erasePos = pos - changes.size();
3579                 for (size_t i = 0; i < changes.size(); i++) {
3580                         insertChar(pos, changes[i].first,
3581                                    changes[i].second,
3582                                    trackChanges);
3583                         if (!eraseChar(erasePos, trackChanges)) {
3584                                 ++erasePos;
3585                                 ++pos; // advance
3586                                 ++right; // expand selection
3587                         }
3588                 }
3589                 changes.clear();
3590         }
3591 }
3592
3593
3594 int Paragraph::find(docstring const & str, bool cs, bool mw,
3595                 pos_type start_pos, bool del) const
3596 {
3597         pos_type pos = start_pos;
3598         int const strsize = str.length();
3599         int i = 0;
3600         pos_type const parsize = d->text_.size();
3601         for (i = 0; i < strsize && pos < parsize; ++i, ++pos) {
3602                 // Ignore "invisible" letters such as ligature breaks
3603                 // and hyphenation chars while searching
3604                 while (pos < parsize - 1 && isInset(pos)) {
3605                         odocstringstream os;
3606                         getInset(pos)->toString(os);
3607                         if (!getInset(pos)->isLetter() || !os.str().empty())
3608                                 break;
3609                         pos++;
3610                 }
3611                 if (cs && str[i] != d->text_[pos])
3612                         break;
3613                 if (!cs && uppercase(str[i]) != uppercase(d->text_[pos]))
3614                         break;
3615                 if (!del && isDeleted(pos))
3616                         break;
3617         }
3618
3619         if (i != strsize)
3620                 return 0;
3621
3622         // if necessary, check whether string matches word
3623         if (mw) {
3624                 if (start_pos > 0 && !isWordSeparator(start_pos - 1))
3625                         return 0;
3626                 if (pos < parsize
3627                         && !isWordSeparator(pos))
3628                         return 0;
3629         }
3630
3631         return pos - start_pos;
3632 }
3633
3634
3635 char_type Paragraph::getChar(pos_type pos) const
3636 {
3637         return d->text_[pos];
3638 }
3639
3640
3641 pos_type Paragraph::size() const
3642 {
3643         return d->text_.size();
3644 }
3645
3646
3647 bool Paragraph::empty() const
3648 {
3649         return d->text_.empty();
3650 }
3651
3652
3653 bool Paragraph::isInset(pos_type pos) const
3654 {
3655         return d->text_[pos] == META_INSET;
3656 }
3657
3658
3659 bool Paragraph::isSeparator(pos_type pos) const
3660 {
3661         //FIXME: Are we sure this can be the only separator?
3662         return d->text_[pos] == ' ';
3663 }
3664
3665
3666 void Paragraph::deregisterWords()
3667 {
3668         Private::LangWordsMap::const_iterator itl = d->words_.begin();
3669         Private::LangWordsMap::const_iterator ite = d->words_.end();
3670         for (; itl != ite; ++itl) {
3671                 WordList * wl = theWordList(itl->first);
3672                 Private::Words::const_iterator it = (itl->second).begin();
3673                 Private::Words::const_iterator et = (itl->second).end();
3674                 for (; it != et; ++it)
3675                         wl->remove(*it);
3676         }
3677         d->words_.clear();
3678 }
3679
3680
3681 void Paragraph::locateWord(pos_type & from, pos_type & to,
3682         word_location const loc) const
3683 {
3684         switch (loc) {
3685         case WHOLE_WORD_STRICT:
3686                 if (from == 0 || from == size()
3687                     || isWordSeparator(from)
3688                     || isWordSeparator(from - 1)) {
3689                         to = from;
3690                         return;
3691                 }
3692                 // no break here, we go to the next
3693
3694         case WHOLE_WORD:
3695                 // If we are already at the beginning of a word, do nothing
3696                 if (!from || isWordSeparator(from - 1))
3697                         break;
3698                 // no break here, we go to the next
3699
3700         case PREVIOUS_WORD:
3701                 // always move the cursor to the beginning of previous word
3702                 while (from && !isWordSeparator(from - 1))
3703                         --from;
3704                 break;
3705         case NEXT_WORD:
3706                 LYXERR0("Paragraph::locateWord: NEXT_WORD not implemented yet");
3707                 break;
3708         case PARTIAL_WORD:
3709                 // no need to move the 'from' cursor
3710                 break;
3711         }
3712         to = from;
3713         while (to < size() && !isWordSeparator(to))
3714                 ++to;
3715 }
3716
3717
3718 void Paragraph::collectWords()
3719 {
3720         for (pos_type pos = 0; pos < size(); ++pos) {
3721                 if (isWordSeparator(pos))
3722                         continue;
3723                 pos_type from = pos;
3724                 locateWord(from, pos, WHOLE_WORD);
3725                 // Work around MSVC warning: The statement
3726                 // if (pos < from + lyxrc.completion_minlength)
3727                 // triggers a signed vs. unsigned warning.
3728                 // I don't know why this happens, it could be a MSVC bug, or
3729                 // related to LLP64 (windows) vs. LP64 (unix) programming
3730                 // model, or the C++ standard might be ambigous in the section
3731                 // defining the "usual arithmetic conversions". However, using
3732                 // a temporary variable is safe and works on all compilers.
3733                 pos_type const endpos = from + lyxrc.completion_minlength;
3734                 if (pos < endpos)
3735                         continue;
3736                 FontList::const_iterator cit = d->fontlist_.fontIterator(from);
3737                 if (cit == d->fontlist_.end())
3738                         return;
3739                 Language const * lang = cit->font().language();
3740                 docstring const word = asString(from, pos, AS_STR_NONE);
3741                 d->words_[lang->lang()].insert(word);
3742         }
3743 }
3744
3745
3746 void Paragraph::registerWords()
3747 {
3748         Private::LangWordsMap::const_iterator itl = d->words_.begin();
3749         Private::LangWordsMap::const_iterator ite = d->words_.end();
3750         for (; itl != ite; ++itl) {
3751                 WordList * wl = theWordList(itl->first);
3752                 Private::Words::const_iterator it = (itl->second).begin();
3753                 Private::Words::const_iterator et = (itl->second).end();
3754                 for (; it != et; ++it)
3755                         wl->insert(*it);
3756         }
3757 }
3758
3759
3760 void Paragraph::updateWords()
3761 {
3762         deregisterWords();
3763         collectWords();
3764         registerWords();
3765 }
3766
3767
3768 void Paragraph::Private::appendSkipPosition(SkipPositions & skips, pos_type const pos) const
3769 {
3770         SkipPositionsIterator begin = skips.begin();
3771         SkipPositions::iterator end = skips.end();
3772         if (pos > 0 && begin < end) {
3773                 --end;
3774                 if (end->last == pos - 1) {
3775                         end->last = pos;
3776                         return;
3777                 }
3778         }
3779         skips.insert(end, FontSpan(pos, pos));
3780 }
3781
3782
3783 Language * Paragraph::Private::locateSpellRange(
3784         pos_type & from, pos_type & to,
3785         SkipPositions & skips) const
3786 {
3787         // skip leading white space
3788         while (from < to && owner_->isWordSeparator(from))
3789                 ++from;
3790         // don't check empty range
3791         if (from >= to)
3792                 return 0;
3793         // get current language
3794         Language * lang = getSpellLanguage(from);
3795         pos_type last = from;
3796         bool samelang = true;
3797         bool sameinset = true;
3798         while (last < to && samelang && sameinset) {
3799                 // hop to end of word
3800                 while (last < to && !owner_->isWordSeparator(last)) {
3801                         if (owner_->getInset(last)) {
3802                                 appendSkipPosition(skips, last);
3803                         } else if (owner_->isDeleted(last)) {
3804                                 appendSkipPosition(skips, last);
3805                         }
3806                         ++last;
3807                 }
3808                 // hop to next word while checking for insets
3809                 while (sameinset && last < to && owner_->isWordSeparator(last)) {
3810                         if (Inset const * inset = owner_->getInset(last))
3811                                 sameinset = inset->isChar() && inset->isLetter();
3812                         if (sameinset && owner_->isDeleted(last)) {
3813                                 appendSkipPosition(skips, last);
3814                         }
3815                         if (sameinset)
3816                                 last++;
3817                 }
3818                 if (sameinset && last < to) {
3819                         // now check for language change
3820                         samelang = lang == getSpellLanguage(last);
3821                 }
3822         }
3823         // if language change detected backstep is needed
3824         if (!samelang)
3825                 --last;
3826         to = last;
3827         return lang;
3828 }
3829
3830
3831 Language * Paragraph::Private::getSpellLanguage(pos_type const from) const
3832 {
3833         Language * lang =
3834                 const_cast<Language *>(owner_->getFontSettings(
3835                         inset_owner_->buffer().params(), from).language());
3836         if (lang == inset_owner_->buffer().params().language
3837                 && !lyxrc.spellchecker_alt_lang.empty()) {
3838                 string lang_code;
3839                 string const lang_variety =
3840                         split(lyxrc.spellchecker_alt_lang, lang_code, '-');
3841                 lang->setCode(lang_code);
3842                 lang->setVariety(lang_variety);
3843         }
3844         return lang;
3845 }
3846
3847
3848 void Paragraph::requestSpellCheck(pos_type pos)
3849 {
3850         d->requestSpellCheck(pos);
3851 }
3852
3853
3854 bool Paragraph::needsSpellCheck() const
3855 {
3856         SpellChecker::ChangeNumber speller_change_number = 0;
3857         if (theSpellChecker())
3858                 speller_change_number = theSpellChecker()->changeNumber();
3859         if (speller_change_number > d->speller_state_.currentChangeNumber()) {
3860                 d->speller_state_.needsCompleteRefresh(speller_change_number);
3861         }
3862         return d->needsSpellCheck();
3863 }
3864
3865
3866 bool Paragraph::Private::ignoreWord(docstring const & word) const
3867 {
3868         // Ignore words with digits
3869         // FIXME: make this customizable
3870         // (note that some checkers ignore words with digits by default)
3871         docstring::const_iterator cit = word.begin();
3872         docstring::const_iterator const end = word.end();
3873         for (; cit != end; ++cit) {
3874                 if (isNumber((*cit)))
3875                         return true;
3876         }
3877         return false;
3878 }
3879
3880
3881 SpellChecker::Result Paragraph::spellCheck(pos_type & from, pos_type & to,
3882         WordLangTuple & wl, docstring_list & suggestions,
3883         bool do_suggestion, bool check_learned) const
3884 {
3885         SpellChecker::Result result = SpellChecker::WORD_OK;
3886         SpellChecker * speller = theSpellChecker();
3887         if (!speller)
3888                 return result;
3889
3890         if (!d->layout_->spellcheck || !inInset().allowSpellCheck())
3891                 return result;
3892
3893         locateWord(from, to, WHOLE_WORD);
3894         if (from == to || from >= size())
3895                 return result;
3896
3897         docstring word = asString(from, to, AS_STR_INSETS | AS_STR_SKIPDELETE);
3898         Language * lang = d->getSpellLanguage(from);
3899
3900         wl = WordLangTuple(word, lang);
3901
3902         if (word.empty())
3903                 return result;
3904
3905         if (needsSpellCheck() || check_learned) {
3906                 pos_type end = to;
3907                 if (!d->ignoreWord(word)) {
3908                         bool const trailing_dot = to < size() && d->text_[to] == '.';
3909                         result = speller->check(wl);
3910                         if (SpellChecker::misspelled(result) && trailing_dot) {
3911                                 wl = WordLangTuple(word.append(from_ascii(".")), lang);
3912                                 result = speller->check(wl);
3913                                 if (!SpellChecker::misspelled(result)) {
3914                                         LYXERR(Debug::GUI, "misspelled word is correct with dot: \"" <<
3915                                            word << "\" [" <<
3916                                            from << ".." << to << "]");
3917                                 } else {
3918                                         // spell check with dot appended failed too
3919                                         // restore original word/lang value
3920                                         word = asString(from, to, AS_STR_INSETS | AS_STR_SKIPDELETE);
3921                                         wl = WordLangTuple(word, lang);
3922                                 }
3923                         }
3924                 }
3925                 if (!SpellChecker::misspelled(result)) {
3926                         // area up to the begin of the next word is not misspelled
3927                         while (end < size() && isWordSeparator(end))
3928                                 ++end;
3929                 }
3930                 d->setMisspelled(from, end, result);
3931         } else {
3932                 result = d->speller_state_.getState(from);
3933         }
3934
3935         if (do_suggestion)
3936                 suggestions.clear();
3937
3938         if (SpellChecker::misspelled(result)) {
3939                 LYXERR(Debug::GUI, "misspelled word: \"" <<
3940                            word << "\" [" <<
3941                            from << ".." << to << "]");
3942                 if (do_suggestion)
3943                         speller->suggest(wl, suggestions);
3944         }
3945         return result;
3946 }
3947
3948
3949 void Paragraph::Private::markMisspelledWords(
3950         pos_type const & first, pos_type const & last,
3951         SpellChecker::Result result,
3952         docstring const & word,
3953         SkipPositions const & skips)
3954 {
3955         if (!SpellChecker::misspelled(result)) {
3956                 setMisspelled(first, last, SpellChecker::WORD_OK);
3957                 return;
3958         }
3959         int snext = first;
3960         SpellChecker * speller = theSpellChecker();
3961         // locate and enumerate the error positions
3962         int nerrors = speller->numMisspelledWords();
3963         int numskipped = 0;
3964         SkipPositionsIterator it = skips.begin();
3965         SkipPositionsIterator et = skips.end();
3966         for (int index = 0; index < nerrors; ++index) {
3967                 int wstart;
3968                 int wlen = 0;
3969                 speller->misspelledWord(index, wstart, wlen);
3970                 /// should not happen if speller supports range checks
3971                 if (!wlen) continue;
3972                 docstring const misspelled = word.substr(wstart, wlen);
3973                 wstart += first + numskipped;
3974                 if (snext < wstart) {
3975                         /// mark the range of correct spelling
3976                         numskipped += countSkips(it, et, wstart);
3977                         setMisspelled(snext,
3978                                 wstart - 1, SpellChecker::WORD_OK);
3979                 }
3980                 snext = wstart + wlen;
3981                 numskipped += countSkips(it, et, snext);
3982                 /// mark the range of misspelling
3983                 setMisspelled(wstart, snext, result);
3984                 LYXERR(Debug::GUI, "misspelled word: \"" <<
3985                            misspelled << "\" [" <<
3986                            wstart << ".." << (snext-1) << "]");
3987                 ++snext;
3988         }
3989         if (snext <= last) {
3990                 /// mark the range of correct spelling at end
3991                 setMisspelled(snext, last, SpellChecker::WORD_OK);
3992         }
3993 }
3994
3995
3996 void Paragraph::spellCheck() const
3997 {
3998         SpellChecker * speller = theSpellChecker();
3999         if (!speller || empty() ||!needsSpellCheck())
4000                 return;
4001         pos_type start;
4002         pos_type endpos;
4003         d->rangeOfSpellCheck(start, endpos);
4004         if (speller->canCheckParagraph()) {
4005                 // loop until we leave the range
4006                 for (pos_type first = start; first < endpos; ) {
4007                         pos_type last = endpos;
4008                         Private::SkipPositions skips;
4009                         Language * lang = d->locateSpellRange(first, last, skips);
4010                         if (first >= endpos)
4011                                 break;
4012                         // start the spell checker on the unit of meaning
4013                         docstring word = asString(first, last, AS_STR_INSETS + AS_STR_SKIPDELETE);
4014                         WordLangTuple wl = WordLangTuple(word, lang);
4015                         SpellChecker::Result result = word.size() ?
4016                                 speller->check(wl) : SpellChecker::WORD_OK;
4017                         d->markMisspelledWords(first, last, result, word, skips);
4018                         first = ++last;
4019                 }
4020         } else {
4021                 static docstring_list suggestions;
4022                 pos_type to = endpos;
4023                 while (start < endpos) {
4024                         WordLangTuple wl;
4025                         spellCheck(start, to, wl, suggestions, false);
4026                         start = to + 1;
4027                 }
4028         }
4029         d->readySpellCheck();
4030 }
4031
4032
4033 bool Paragraph::isMisspelled(pos_type pos, bool check_boundary) const
4034 {
4035         bool result = SpellChecker::misspelled(d->speller_state_.getState(pos));
4036         if (result || pos <= 0 || pos > size())
4037                 return result;
4038         if (check_boundary && (pos == size() || isWordSeparator(pos)))
4039                 result = SpellChecker::misspelled(d->speller_state_.getState(pos - 1));
4040         return result;
4041 }
4042
4043
4044 string Paragraph::magicLabel() const
4045 {
4046         stringstream ss;
4047         ss << "magicparlabel-" << id();
4048         return ss.str();
4049 }
4050
4051
4052 } // namespace lyx