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 "support/debug.h"
  57 #include "support/docstring_list.h"
  58 #include "support/ExceptionMessage.h"
  59 #include "support/gettext.h"
  60 #include "support/lassert.h"
  61 #include "support/lstrings.h"
  62 #include "support/textutils.h"
  63
  64 #include <atomic>
  65 #include <sstream>
  66 #include <vector>
  67
  68 using namespace std;
  69 using namespace lyx::support;
  70
  71 namespace lyx {
  72
  73 namespace {
  74
  75 /// Inset identifier (above 0x10ffff, for ucs-4)
  76 char_type const META_INSET = 0x200001;
  77
  78 }
  79
  80
  81 /////////////////////////////////////////////////////////////////////
  82 //
  83 // SpellResultRange
  84 //
  85 /////////////////////////////////////////////////////////////////////
  86
  87 class SpellResultRange {
  88 public:
  89         SpellResultRange(FontSpan range, SpellChecker::Result result)
  90         : range_(range), result_(result)
  91         {}
  92         ///
  93         FontSpan const & range() const { return range_; }
  94         ///
  95         void range(FontSpan const & r) { range_ = r; }
  96         ///
  97         SpellChecker::Result result() const { return result_; }
  98         ///
  99         void result(SpellChecker::Result r) { result_ = r; }
 100         ///
 101         bool contains(pos_type pos) const { return range_.contains(pos); }
 102         ///
 103         bool covered(FontSpan const & r) const
 104         {
 105                 // 1. first of new range inside current range or
 106                 // 2. last of new range inside current range or
 107                 // 3. first of current range inside new range or
 108                 // 4. last of current range inside new range
 109                 //FIXME: is this the same as !range_.intersect(r).empty() ?
 110                 return range_.contains(r.first) || range_.contains(r.last) ||
 111                         r.contains(range_.first) || r.contains(range_.last);
 112         }
 113         ///
 114         void shift(pos_type pos, int offset)
 115         {
 116                 if (range_.first > pos) {
 117                         range_.first += offset;
 118                         range_.last += offset;
 119                 } else if (range_.last >= pos) {
 120                         range_.last += offset;
 121                 }
 122         }
 123 private:
 124         FontSpan range_ ;
 125         SpellChecker::Result result_ ;
 126 };
 127
 128
 129 /////////////////////////////////////////////////////////////////////
 130 //
 131 // SpellCheckerState
 132 //
 133 /////////////////////////////////////////////////////////////////////
 134
 135 class SpellCheckerState {
 136 public:
 137         SpellCheckerState()
 138         {
 139                 needs_refresh_ = true;
 140                 current_change_number_ = 0;
 141         }
 142
 143         void setRange(FontSpan const & fp, SpellChecker::Result state)
 144         {
 145                 Ranges result;
 146                 RangesIterator et = ranges_.end();
 147                 RangesIterator it = ranges_.begin();
 148                 for (; it != et; ++it) {
 149                         if (!it->covered(fp))
 150                                 result.push_back(SpellResultRange(it->range(), it->result()));
 151                         else if (state == SpellChecker::WORD_OK) {
 152                                 // trim or split the current misspelled range
 153                                 // store misspelled ranges only
 154                                 FontSpan range = it->range();
 155                                 if (fp.first > range.first) {
 156                                         // misspelled area in front of WORD_OK
 157                                         range.last = fp.first - 1;
 158                                         result.push_back(SpellResultRange(range, it->result()));
 159                                         range = it->range();
 160                                 }
 161                                 if (fp.last < range.last) {
 162                                         // misspelled area after WORD_OK range
 163                                         range.first = fp.last + 1;
 164                                         result.push_back(SpellResultRange(range, it->result()));
 165                                 }
 166                         }
 167                 }
 168                 ranges_ = result;
 169                 if (state != SpellChecker::WORD_OK)
 170                         ranges_.push_back(SpellResultRange(fp, state));
 171         }
 172
 173         void increasePosAfterPos(pos_type pos)
 174         {
 175                 correctRangesAfterPos(pos, 1);
 176                 needsRefresh(pos);
 177         }
 178
 179         void decreasePosAfterPos(pos_type pos)
 180         {
 181                 correctRangesAfterPos(pos, -1);
 182                 needsRefresh(pos);
 183         }
 184
 185         void refreshLast(pos_type pos)
 186         {
 187                 if (pos < refresh_.last)
 188                         refresh_.last = pos;
 189         }
 190
 191         SpellChecker::Result getState(pos_type pos) const
 192         {
 193                 SpellChecker::Result result = SpellChecker::WORD_OK;
 194                 RangesIterator et = ranges_.end();
 195                 RangesIterator it = ranges_.begin();
 196                 for (; it != et; ++it) {
 197                         if(it->contains(pos)) {
 198                                 return it->result();
 199                         }
 200                 }
 201                 return result;
 202         }
 203
 204         FontSpan const & getRange(pos_type pos) const
 205         {
 206                 /// empty span to indicate mismatch
 207                 static FontSpan empty_;
 208                 RangesIterator et = ranges_.end();
 209                 RangesIterator it = ranges_.begin();
 210                 for (; it != et; ++it) {
 211                         if(it->contains(pos)) {
 212                                 return it->range();
 213                         }
 214                 }
 215                 return empty_;
 216         }
 217
 218         bool needsRefresh() const
 219         {
 220                 return needs_refresh_;
 221         }
 222
 223         SpellChecker::ChangeNumber currentChangeNumber() const
 224         {
 225                 return current_change_number_;
 226         }
 227
 228         void refreshRange(pos_type & first, pos_type & last) const
 229         {
 230                 first = refresh_.first;
 231                 last = refresh_.last;
 232         }
 233
 234         void needsRefresh(pos_type pos)
 235         {
 236                 if (needs_refresh_ && pos != -1) {
 237                         if (pos < refresh_.first)
 238                                 refresh_.first = pos;
 239                         if (pos > refresh_.last)
 240                                 refresh_.last = pos;
 241                 } else if (pos != -1) {
 242                         // init request check for neighbour positions too
 243                         refresh_.first = pos > 0 ? pos - 1 : 0;
 244                         // no need for special end of paragraph check
 245                         refresh_.last = pos + 1;
 246                 }
 247                 needs_refresh_ = pos != -1;
 248         }
 249
 250         void needsCompleteRefresh(SpellChecker::ChangeNumber change_number)
 251         {
 252                 needs_refresh_ = true;
 253                 refresh_.first = 0;
 254                 refresh_.last = -1;
 255                 current_change_number_ = change_number;
 256         }
 257 private:
 258         typedef vector<SpellResultRange> Ranges;
 259         typedef Ranges::const_iterator RangesIterator;
 260         Ranges ranges_;
 261         /// the area of the paragraph with pending spell check
 262         FontSpan refresh_;
 263         bool needs_refresh_;
 264         /// spell state cache version number
 265         SpellChecker::ChangeNumber current_change_number_;
 266
 267
 268         void correctRangesAfterPos(pos_type pos, int offset)
 269         {
 270                 RangesIterator et = ranges_.end();
 271                 Ranges::iterator it = ranges_.begin();
 272                 for (; it != et; ++it) {
 273                         it->shift(pos, offset);
 274                 }
 275         }
 276
 277 };
 278
 279 /////////////////////////////////////////////////////////////////////
 280 //
 281 // Paragraph::Private
 282 //
 283 /////////////////////////////////////////////////////////////////////
 284
 285 class Paragraph::Private
 286 {
 287         // Enforce our own "copy" constructor
 288         Private(Private const &) = delete;
 289         Private & operator=(Private const &) = delete;
 290         // Unique ID generator
 291         static int make_id();
 292 public:
 293         ///
 294         Private(Paragraph * owner, Layout const & layout);
 295         /// "Copy constructor"
 296         Private(Private const &, Paragraph * owner);
 297         /// Copy constructor from \p beg  to \p end
 298         Private(Private const &, Paragraph * owner, pos_type beg, pos_type end);
 299
 300         ///
 301         void insertChar(pos_type pos, char_type c, Change const & change);
 302
 303         /// Output the surrogate pair formed by \p c and \p next to \p os.
 304         /// \return the number of characters written.
 305         int latexSurrogatePair(otexstream & os, char_type c, char_type next,
 306                                OutputParams const &);
 307
 308         /// Output a space in appropriate formatting (or a surrogate pair
 309         /// if the next character is a combining character).
 310         /// \return whether a surrogate pair was output.
 311         bool simpleTeXBlanks(OutputParams const &,
 312                              otexstream &,
 313                              pos_type i,
 314                              unsigned int & column,
 315                              Font const & font,
 316                              Layout const & style);
 317
 318         /// Output consecutive unicode chars, belonging to the same script as
 319         /// specified by the latex macro \p ltx, to \p os starting from \p i.
 320         /// \return the number of characters written.
 321         int writeScriptChars(otexstream & os, docstring const & ltx,
 322                            Change const &, Encoding const &, pos_type & i);
 323
 324         /// This could go to ParagraphParameters if we want to.
 325         int startTeXParParams(BufferParams const &, otexstream &,
 326                               OutputParams const &) const;
 327
 328         /// This could go to ParagraphParameters if we want to.
 329         bool endTeXParParams(BufferParams const &, otexstream &,
 330                              OutputParams const &) const;
 331
 332         ///
 333         void latexInset(BufferParams const &,
 334                                    otexstream &,
 335                                    OutputParams &,
 336                                    Font & running_font,
 337                                    Font & basefont,
 338                                    Font const & outerfont,
 339                                    bool & open_font,
 340                                    Change & running_change,
 341                                    Layout const & style,
 342                                    pos_type & i,
 343                                    unsigned int & column);
 344
 345         ///
 346         void latexSpecialChar(
 347                                    otexstream & os,
 348                                    BufferParams const & bparams,
 349                                    OutputParams const & runparams,
 350                                    Font const & running_font,
 351                                    Change const & running_change,
 352                                    Layout const & style,
 353                                    pos_type & i,
 354                                    pos_type end_pos,
 355                                    unsigned int & column);
 356
 357         ///
 358         bool latexSpecialT1(
 359                 char_type const c,
 360                 otexstream & os,
 361                 pos_type i,
 362                 unsigned int & column);
 363         ///
 364         bool latexSpecialT3(
 365                 char_type const c,
 366                 otexstream & os,
 367                 pos_type i,
 368                 unsigned int & column);
 369
 370         ///
 371         void validate(LaTeXFeatures & features) const;
 372
 373         /// Checks if the paragraph contains only text and no inset or font change.
 374         bool onlyText(Buffer const & buf, Font const & outerfont,
 375                       pos_type initial) const;
 376
 377         /// a vector of speller skip positions
 378         typedef vector<FontSpan> SkipPositions;
 379         typedef SkipPositions::const_iterator SkipPositionsIterator;
 380
 381         void appendSkipPosition(SkipPositions & skips, pos_type const pos) const;
 382
 383         Language * getSpellLanguage(pos_type const from) const;
 384
 385         Language * locateSpellRange(pos_type & from, pos_type & to,
 386                                     SkipPositions & skips) const;
 387
 388         bool hasSpellerChange() const
 389         {
 390                 SpellChecker::ChangeNumber speller_change_number = 0;
 391                 if (theSpellChecker())
 392                         speller_change_number = theSpellChecker()->changeNumber();
 393                 return speller_change_number > speller_state_.currentChangeNumber();
 394         }
 395
 396         bool ignoreWord(docstring const & word) const ;
 397
 398         void setMisspelled(pos_type from, pos_type to, SpellChecker::Result state)
 399         {
 400                 pos_type textsize = owner_->size();
 401                 // check for sane arguments
 402                 if (to <= from || from >= textsize)
 403                         return;
 404                 FontSpan fp = FontSpan(from, to - 1);
 405                 speller_state_.setRange(fp, state);
 406         }
 407
 408         void requestSpellCheck(pos_type pos)
 409         {
 410                 if (pos == -1)
 411                         speller_state_.needsCompleteRefresh(speller_state_.currentChangeNumber());
 412                 else
 413                         speller_state_.needsRefresh(pos);
 414         }
 415
 416         void readySpellCheck()
 417         {
 418                 speller_state_.needsRefresh(-1);
 419         }
 420
 421         bool needsSpellCheck() const
 422         {
 423                 return speller_state_.needsRefresh();
 424         }
 425
 426         void rangeOfSpellCheck(pos_type & first, pos_type & last) const
 427         {
 428                 speller_state_.refreshRange(first, last);
 429                 if (last == -1) {
 430                         last = owner_->size();
 431                         return;
 432                 }
 433                 pos_type endpos = last;
 434                 owner_->locateWord(first, endpos, WHOLE_WORD);
 435                 if (endpos < last) {
 436                         endpos = last;
 437                         owner_->locateWord(last, endpos, WHOLE_WORD);
 438                 }
 439                 last = endpos;
 440         }
 441
 442         int countSkips(SkipPositionsIterator & it, SkipPositionsIterator const et,
 443                             int & start) const
 444         {
 445                 int numskips = 0;
 446                 while (it != et && it->first < start) {
 447                         int skip = it->last - it->first + 1;
 448                         start += skip;
 449                         numskips += skip;
 450                         ++it;
 451                 }
 452                 return numskips;
 453         }
 454
 455         void markMisspelledWords(pos_type const & first, pos_type const & last,
 456                                                          SpellChecker::Result result,
 457                                                          docstring const & word,
 458                                                          SkipPositions const & skips);
 459
 460         InsetCode ownerCode() const
 461         {
 462                 return inset_owner_ ? inset_owner_->lyxCode() : NO_CODE;
 463         }
 464
 465         /// Which Paragraph owns us?
 466         Paragraph * owner_;
 467
 468         /// In which Inset?
 469         Inset const * inset_owner_;
 470
 471         ///
 472         FontList fontlist_;
 473
 474         ///
 475         int id_;
 476
 477         ///
 478         ParagraphParameters params_;
 479
 480         /// for recording and looking up changes
 481         Changes changes_;
 482
 483         ///
 484         InsetList insetlist_;
 485
 486         /// end of label
 487         pos_type begin_of_body_;
 488
 489         typedef docstring TextContainer;
 490         ///
 491         TextContainer text_;
 492
 493         typedef set<docstring> Words;
 494         typedef map<string, Words> LangWordsMap;
 495         ///
 496         LangWordsMap words_;
 497         ///
 498         Layout const * layout_;
 499         ///
 500         SpellCheckerState speller_state_;
 501 };
 502
 503
 504 Paragraph::Private::Private(Paragraph * owner, Layout const & layout)
 505         : owner_(owner), inset_owner_(0), id_(-1), begin_of_body_(0), layout_(&layout)
 506 {
 507         text_.reserve(100);
 508 }
 509
 510
 511 //static
 512 int Paragraph::Private::make_id()
 513 {
 514         // The id is unique per session across buffers because it is used in
 515         // LFUN_PARAGRAPH_GOTO to switch to a different buffer, for instance in the
 516         // outliner.
 517         // (thread-safe)
 518         static atomic_uint next_id(0);
 519         return next_id++;
 520 }
 521
 522
 523 Paragraph::Private::Private(Private const & p, Paragraph * owner)
 524         : owner_(owner), inset_owner_(p.inset_owner_), fontlist_(p.fontlist_),
 525           id_(make_id()),
 526           params_(p.params_), changes_(p.changes_), insetlist_(p.insetlist_),
 527           begin_of_body_(p.begin_of_body_), text_(p.text_), words_(p.words_),
 528           layout_(p.layout_)
 529 {
 530         requestSpellCheck(p.text_.size());
 531 }
 532
 533
 534 Paragraph::Private::Private(Private const & p, Paragraph * owner,
 535         pos_type beg, pos_type end)
 536         : owner_(owner), inset_owner_(p.inset_owner_), id_(make_id()),
 537           params_(p.params_), changes_(p.changes_),
 538           insetlist_(p.insetlist_, beg, end),
 539           begin_of_body_(p.begin_of_body_), words_(p.words_),
 540           layout_(p.layout_)
 541 {
 542         if (beg >= pos_type(p.text_.size()))
 543                 return;
 544         text_ = p.text_.substr(beg, end - beg);
 545
 546         FontList::const_iterator fcit = fontlist_.begin();
 547         FontList::const_iterator fend = fontlist_.end();
 548         for (; fcit != fend; ++fcit) {
 549                 if (fcit->pos() < beg)
 550                         continue;
 551                 if (fcit->pos() >= end) {
 552                         // Add last entry in the fontlist_.
 553                         fontlist_.set(text_.size() - 1, fcit->font());
 554                         break;
 555                 }
 556                 // Add a new entry in the fontlist_.
 557                 fontlist_.set(fcit->pos() - beg, fcit->font());
 558         }
 559         requestSpellCheck(p.text_.size());
 560 }
 561
 562
 563 void Paragraph::addChangesToToc(DocIterator const & cdit,
 564         Buffer const & buf, bool output_active) const
 565 {
 566         d->changes_.addToToc(cdit, buf, output_active);
 567 }
 568
 569
 570 void Paragraph::addChangesToBuffer(Buffer const & buf) const
 571 {
 572         d->changes_.updateBuffer(buf);
 573 }
 574
 575
 576 bool Paragraph::isChangeUpdateRequired() const
 577 {
 578         return d->changes_.isUpdateRequired();
 579 }
 580
 581
 582 bool Paragraph::isDeleted(pos_type start, pos_type end) const
 583 {
 584         LASSERT(start >= 0 && start <= size(), return false);
 585         LASSERT(end > start && end <= size() + 1, return false);
 586
 587         return d->changes_.isDeleted(start, end);
 588 }
 589
 590
 591 bool Paragraph::isChanged(pos_type start, pos_type end) const
 592 {
 593         LASSERT(start >= 0 && start <= size(), return false);
 594         LASSERT(end > start && end <= size() + 1, return false);
 595
 596         return d->changes_.isChanged(start, end);
 597 }
 598
 599
 600 bool Paragraph::isMergedOnEndOfParDeletion(bool trackChanges) const
 601 {
 602         // keep the logic here in sync with the logic of eraseChars()
 603         if (!trackChanges)
 604                 return true;
 605
 606         Change const & change = d->changes_.lookup(size());
 607         return change.inserted() && change.currentAuthor();
 608 }
 609
 610
 611 void Paragraph::setChange(Change const & change)
 612 {
 613         // beware of the imaginary end-of-par character!
 614         d->changes_.set(change, 0, size() + 1);
 615
 616         /*
 617          * Propagate the change recursively - but not in case of DELETED!
 618          *
 619          * Imagine that your co-author makes changes in an existing inset. He
 620          * sends your document to you and you come to the conclusion that the
 621          * inset should go completely. If you erase it, LyX must not delete all
 622          * text within the inset. Otherwise, the change tracked insertions of
 623          * your co-author get lost and there is no way to restore them later.
 624          *
 625          * Conclusion: An inset's content should remain untouched if you delete it
 626          */
 627
 628         if (!change.deleted()) {
 629                 for (pos_type pos = 0; pos < size(); ++pos) {
 630                         if (Inset * inset = getInset(pos))
 631                                 inset->setChange(change);
 632                 }
 633         }
 634 }
 635
 636
 637 void Paragraph::setChange(pos_type pos, Change const & change)
 638 {
 639         LASSERT(pos >= 0 && pos <= size(), return);
 640         d->changes_.set(change, pos);
 641
 642         // see comment in setChange(Change const &) above
 643         if (!change.deleted() && pos < size())
 644                         if (Inset * inset = getInset(pos))
 645                                 inset->setChange(change);
 646 }
 647
 648
 649 Change const & Paragraph::lookupChange(pos_type pos) const
 650 {
 651         LBUFERR(pos >= 0 && pos <= size());
 652         return d->changes_.lookup(pos);
 653 }
 654
 655
 656 void Paragraph::acceptChanges(pos_type start, pos_type end)
 657 {
 658         LASSERT(start >= 0 && start <= size(), return);
 659         LASSERT(end > start && end <= size() + 1, return);
 660
 661         for (pos_type pos = start; pos < end; ++pos) {
 662                 switch (lookupChange(pos).type) {
 663                         case Change::UNCHANGED:
 664                                 // accept changes in nested inset
 665                                 if (Inset * inset = getInset(pos))
 666                                         inset->acceptChanges();
 667                                 break;
 668
 669                         case Change::INSERTED:
 670                                 d->changes_.set(Change(Change::UNCHANGED), pos);
 671                                 // also accept changes in nested inset
 672                                 if (Inset * inset = getInset(pos))
 673                                         inset->acceptChanges();
 674                                 break;
 675
 676                         case Change::DELETED:
 677                                 // Suppress access to non-existent
 678                                 // "end-of-paragraph char"
 679                                 if (pos < size()) {
 680                                         eraseChar(pos, false);
 681                                         --end;
 682                                         --pos;
 683                                 }
 684                                 break;
 685                 }
 686
 687         }
 688 }
 689
 690
 691 void Paragraph::rejectChanges(pos_type start, pos_type end)
 692 {
 693         LASSERT(start >= 0 && start <= size(), return);
 694         LASSERT(end > start && end <= size() + 1, return);
 695
 696         for (pos_type pos = start; pos < end; ++pos) {
 697                 switch (lookupChange(pos).type) {
 698                         case Change::UNCHANGED:
 699                                 // reject changes in nested inset
 700                                 if (Inset * inset = getInset(pos))
 701                                                 inset->rejectChanges();
 702                                 break;
 703
 704                         case Change::INSERTED:
 705                                 // Suppress access to non-existent
 706                                 // "end-of-paragraph char"
 707                                 if (pos < size()) {
 708                                         eraseChar(pos, false);
 709                                         --end;
 710                                         --pos;
 711                                 }
 712                                 break;
 713
 714                         case Change::DELETED:
 715                                 d->changes_.set(Change(Change::UNCHANGED), pos);
 716
 717                                 // Do NOT reject changes within a deleted inset!
 718                                 // There may be insertions of a co-author inside of it!
 719
 720                                 break;
 721                 }
 722         }
 723 }
 724
 725
 726 void Paragraph::Private::insertChar(pos_type pos, char_type c,
 727                 Change const & change)
 728 {
 729         LASSERT(pos >= 0 && pos <= int(text_.size()), return);
 730
 731         // track change
 732         changes_.insert(change, pos);
 733
 734         // This is actually very common when parsing buffers (and
 735         // maybe inserting ascii text)
 736         if (pos == pos_type(text_.size())) {
 737                 // when appending characters, no need to update tables
 738                 text_.push_back(c);
 739                 // but we want spell checking
 740                 requestSpellCheck(pos);
 741                 return;
 742         }
 743
 744         text_.insert(text_.begin() + pos, c);
 745
 746         // Update the font table.
 747         fontlist_.increasePosAfterPos(pos);
 748
 749         // Update the insets
 750         insetlist_.increasePosAfterPos(pos);
 751
 752         // Update list of misspelled positions
 753         speller_state_.increasePosAfterPos(pos);
 754 }
 755
 756
 757 bool Paragraph::insertInset(pos_type pos, Inset * inset,
 758                                    Font const & font, Change const & change)
 759 {
 760         LASSERT(inset, return false);
 761         LASSERT(pos >= 0 && pos <= size(), return false);
 762
 763         // Paragraph::insertInset() can be used in cut/copy/paste operation where
 764         // d->inset_owner_ is not set yet.
 765         if (d->inset_owner_ && !d->inset_owner_->insetAllowed(inset->lyxCode()))
 766                 return false;
 767
 768         d->insertChar(pos, META_INSET, change);
 769         LASSERT(d->text_[pos] == META_INSET, return false);
 770
 771         // Add a new entry in the insetlist_.
 772         d->insetlist_.insert(inset, pos);
 773
 774         // Some insets require run of spell checker
 775         requestSpellCheck(pos);
 776         setFont(pos, font);
 777         return true;
 778 }
 779
 780
 781 bool Paragraph::eraseChar(pos_type pos, bool trackChanges)
 782 {
 783         LASSERT(pos >= 0 && pos <= size(), return false);
 784
 785         // keep the logic here in sync with the logic of isMergedOnEndOfParDeletion()
 786
 787         if (trackChanges) {
 788                 Change change = d->changes_.lookup(pos);
 789
 790                 // set the character to DELETED if
 791                 //  a) it was previously unchanged or
 792                 //  b) it was inserted by a co-author
 793
 794                 if (!change.changed() ||
 795                       (change.inserted() && !change.currentAuthor())) {
 796                         setChange(pos, Change(Change::DELETED));
 797                         // request run of spell checker
 798                         requestSpellCheck(pos);
 799                         return false;
 800                 }
 801
 802                 if (change.deleted())
 803                         return false;
 804         }
 805
 806         // Don't physically access the imaginary end-of-paragraph character.
 807         // eraseChar() can only mark it as DELETED. A physical deletion of
 808         // end-of-par must be handled externally.
 809         if (pos == size()) {
 810                 return false;
 811         }
 812
 813         // track change
 814         d->changes_.erase(pos);
 815
 816         // if it is an inset, delete the inset entry
 817         if (d->text_[pos] == META_INSET)
 818                 d->insetlist_.erase(pos);
 819
 820         d->text_.erase(d->text_.begin() + pos);
 821
 822         // Update the fontlist_
 823         d->fontlist_.erase(pos);
 824
 825         // Update the insetlist_
 826         d->insetlist_.decreasePosAfterPos(pos);
 827
 828         // Update list of misspelled positions
 829         d->speller_state_.decreasePosAfterPos(pos);
 830         d->speller_state_.refreshLast(size());
 831
 832         return true;
 833 }
 834
 835
 836 int Paragraph::eraseChars(pos_type start, pos_type end, bool trackChanges)
 837 {
 838         LASSERT(start >= 0 && start <= size(), return 0);
 839         LASSERT(end >= start && end <= size() + 1, return 0);
 840
 841         pos_type i = start;
 842         for (pos_type count = end - start; count; --count) {
 843                 if (!eraseChar(i, trackChanges))
 844                         ++i;
 845         }
 846         return end - i;
 847 }
 848
 849
 850 int Paragraph::Private::latexSurrogatePair(otexstream & os, char_type c,
 851                 char_type next, OutputParams const & runparams)
 852 {
 853         // Writing next here may circumvent a possible font change between
 854         // c and next. Since next is only output if it forms a surrogate pair
 855         // with c we can ignore this:
 856         // A font change inside a surrogate pair does not make sense and is
 857         // hopefully impossible to input.
 858         // FIXME: change tracking
 859         // Is this correct WRT change tracking?
 860         Encoding const & encoding = *(runparams.encoding);
 861         docstring latex1 = encoding.latexChar(next).first;
 862         if (runparams.inIPA) {
 863                 string const tipashortcut = Encodings::TIPAShortcut(next);
 864                 if (!tipashortcut.empty()) {
 865                         latex1 = from_ascii(tipashortcut);
 866                 }
 867         }
 868         docstring const latex2 = encoding.latexChar(c).first;
 869         if (docstring(1, next) == latex1) {
 870                 // the encoding supports the combination
 871                 os << latex2 << latex1;
 872                 return latex1.length() + latex2.length();
 873         } else if (runparams.local_font &&
 874                    runparams.local_font->language()->lang() == "polutonikogreek") {
 875                 // polutonikogreek only works without the brackets
 876                 os << latex1 << latex2;
 877                 return latex1.length() + latex2.length();
 878         } else
 879                 os << latex1 << '{' << latex2 << '}';
 880         return latex1.length() + latex2.length() + 2;
 881 }
 882
 883
 884 bool Paragraph::Private::simpleTeXBlanks(OutputParams const & runparams,
 885                                        otexstream & os,
 886                                        pos_type i,
 887                                        unsigned int & column,
 888                                        Font const & font,
 889                                        Layout const & style)
 890 {
 891         if (style.pass_thru || runparams.pass_thru)
 892                 return false;
 893
 894         if (i + 1 < int(text_.size())) {
 895                 char_type next = text_[i + 1];
 896                 if (Encodings::isCombiningChar(next)) {
 897                         // This space has an accent, so we must always output it.
 898                         column += latexSurrogatePair(os, ' ', next, runparams) - 1;
 899                         return true;
 900                 }
 901         }
 902
 903         if (runparams.linelen > 0
 904             && column > runparams.linelen
 905             && i
 906             && text_[i - 1] != ' '
 907             && (i + 1 < int(text_.size()))
 908             // same in FreeSpacing mode
 909             && !owner_->isFreeSpacing()
 910             // In typewriter mode, we want to avoid
 911             // ! . ? : at the end of a line
 912             && !(font.fontInfo().family() == TYPEWRITER_FAMILY
 913                  && (text_[i - 1] == '.'
 914                      || text_[i - 1] == '?'
 915                      || text_[i - 1] == ':'
 916                      || text_[i - 1] == '!'))) {
 917                 os << '\n';
 918                 os.texrow().start(owner_->id(), i + 1);
 919                 column = 0;
 920         } else if (style.free_spacing) {
 921                 os << '~';
 922         } else {
 923                 os << ' ';
 924         }
 925         return false;
 926 }
 927
 928
 929 int Paragraph::Private::writeScriptChars(otexstream & os,
 930                                          docstring const & ltx,
 931                                          Change const & runningChange,
 932                                          Encoding const & encoding,
 933                                          pos_type & i)
 934 {
 935         // FIXME: modifying i here is not very nice...
 936
 937         // We only arrive here when character text_[i] could not be translated
 938         // into the current latex encoding (or its latex translation has been forced,)
 939         // and it belongs to a known script.
 940         // TODO: We need \textcyr and \textgreek wrappers also for characters
 941         //       that can be encoded in the "LaTeX encoding" but not in the
 942         //       current *font encoding*.
 943         //       (See #9681 for details and test)
 944         // Parameter ltx contains the latex translation of text_[i] as specified
 945         // in the unicodesymbols file and is something like "\textXXX{<spec>}".
 946         // The latex macro name "textXXX" specifies the script to which text_[i]
 947         // belongs and we use it in order to check whether characters from the
 948         // same script immediately follow, such that we can collect them in a
 949         // single "\textXXX" macro. So, we have to retain "\textXXX{<spec>"
 950         // for the first char but only "<spec>" for all subsequent chars.
 951         docstring::size_type const brace1 = ltx.find_first_of(from_ascii("{"));
 952         docstring::size_type const brace2 = ltx.find_last_of(from_ascii("}"));
 953         string script = to_ascii(ltx.substr(1, brace1 - 1));
 954         int pos = 0;
 955         int length = brace2;
 956         bool closing_brace = true;
 957         if (script == "textgreek" && encoding.latexName() == "iso-8859-7") {
 958                 // Correct encoding is being used, so we can avoid \textgreek.
 959                 // TODO: wrong test: we need to check the *font encoding*
 960                 //       (i.e. the active language and its FontEncoding tag)
 961                 //       instead of the LaTeX *input encoding*!
 962                 //       See #9637 for details and test-cases.
 963                 pos = brace1 + 1;
 964                 length -= pos;
 965                 closing_brace = false;
 966         }
 967         os << ltx.substr(pos, length);
 968         int size = text_.size();
 969         while (i + 1 < size) {
 970                 char_type const next = text_[i + 1];
 971                 // Stop here if next character belongs to another script
 972                 // or there is a change in change tracking status.
 973                 if (!Encodings::isKnownScriptChar(next, script) ||
 974                     runningChange != owner_->lookupChange(i + 1))
 975                         break;
 976                 Font prev_font;
 977                 bool found = false;
 978                 FontList::const_iterator cit = fontlist_.begin();
 979                 FontList::const_iterator end = fontlist_.end();
 980                 for (; cit != end; ++cit) {
 981                         if (cit->pos() >= i && !found) {
 982                                 prev_font = cit->font();
 983                                 found = true;
 984                         }
 985                         if (cit->pos() >= i + 1)
 986                                 break;
 987                 }
 988                 // Stop here if there is a font attribute or encoding change.
 989                 if (found && cit != end && prev_font != cit->font())
 990                         break;
 991                 docstring const latex = encoding.latexChar(next).first;
 992                 docstring::size_type const b1 =
 993                                         latex.find_first_of(from_ascii("{"));
 994                 docstring::size_type const b2 =
 995                                         latex.find_last_of(from_ascii("}"));
 996                 int const len = b2 - b1 - 1;
 997                 os << latex.substr(b1 + 1, len);
 998                 length += len;
 999                 ++i;
1000         }
1001         if (closing_brace) {
1002                 os << '}';
1003                 ++length;
1004         }
1005         return length;
1006 }
1007
1008
1009 void Paragraph::Private::latexInset(BufferParams const & bparams,
1010                                     otexstream & os,
1011                                     OutputParams & runparams,
1012                                     Font & running_font,
1013                                     Font & basefont,
1014                                     Font const & outerfont,
1015                                     bool & open_font,
1016                                     Change & running_change,
1017                                     Layout const & style,
1018                                     pos_type & i,
1019                                     unsigned int & column)
1020 {
1021         Inset * inset = owner_->getInset(i);
1022         LBUFERR(inset);
1023
1024         if (style.pass_thru) {
1025                 odocstringstream ods;
1026                 inset->plaintext(ods, runparams);
1027                 os << ods.str();
1028                 return;
1029         }
1030
1031         // FIXME: move this to InsetNewline::latex
1032         if (inset->lyxCode() == NEWLINE_CODE || inset->lyxCode() == SEPARATOR_CODE) {
1033                 // newlines are handled differently here than
1034                 // the default in simpleTeXSpecialChars().
1035                 if (!style.newline_allowed) {
1036                         os << '\n';
1037                 } else {
1038                         if (open_font) {
1039                                 column += running_font.latexWriteEndChanges(
1040                                         os, bparams, runparams,
1041                                         basefont, basefont);
1042                                 open_font = false;
1043                         }
1044
1045                         if (running_font.fontInfo().family() == TYPEWRITER_FAMILY)
1046                                 os << '~';
1047
1048                         basefont = owner_->getLayoutFont(bparams, outerfont);
1049                         running_font = basefont;
1050
1051                         if (runparams.moving_arg)
1052                                 os << "\\protect ";
1053
1054                 }
1055                 os.texrow().start(owner_->id(), i + 1);
1056                 column = 0;
1057         }
1058
1059         if (owner_->isDeleted(i)) {
1060                 if( ++runparams.inDeletedInset == 1)
1061                         runparams.changeOfDeletedInset = owner_->lookupChange(i);
1062         }
1063
1064         if (inset->canTrackChanges()) {
1065                 column += Changes::latexMarkChange(os, bparams, running_change,
1066                         Change(Change::UNCHANGED), runparams);
1067                 running_change = Change(Change::UNCHANGED);
1068         }
1069
1070         bool close = false;
1071         odocstream::pos_type const len = os.os().tellp();
1072
1073         if (inset->forceLTR()
1074             && !runparams.use_polyglossia
1075             && running_font.isRightToLeft()
1076             // ERT is an exception, it should be output with no
1077             // decorations at all
1078             && inset->lyxCode() != ERT_CODE) {
1079                 if (running_font.language()->lang() == "farsi")
1080                         os << "\\beginL{}";
1081                 else
1082                         os << "\\L{";
1083                 close = true;
1084         }
1085
1086         // FIXME: Bug: we can have an empty font change here!
1087         // if there has just been a font change, we are going to close it
1088         // right now, which means stupid latex code like \textsf{}. AFAIK,
1089         // this does not harm dvi output. A minor bug, thus (JMarc)
1090
1091         // Some insets cannot be inside a font change command.
1092         // However, even such insets *can* be placed in \L or \R
1093         // or their equivalents (for RTL language switches), so we don't
1094         // close the language in those cases.
1095         // ArabTeX, though, cannot handle this special behavior, it seems.
1096         bool arabtex = basefont.language()->lang() == "arabic_arabtex"
1097                 || running_font.language()->lang() == "arabic_arabtex";
1098         if (open_font && !inset->inheritFont()) {
1099                 bool closeLanguage = arabtex
1100                         || basefont.isRightToLeft() == running_font.isRightToLeft();
1101                 unsigned int count = running_font.latexWriteEndChanges(os,
1102                         bparams, runparams, basefont, basefont, closeLanguage);
1103                 column += count;
1104                 // if any font properties were closed, update the running_font,
1105                 // making sure, however, to leave the language as it was
1106                 if (count > 0) {
1107                         // FIXME: probably a better way to keep track of the old
1108                         // language, than copying the entire font?
1109                         Font const copy_font(running_font);
1110                         basefont = owner_->getLayoutFont(bparams, outerfont);
1111                         running_font = basefont;
1112                         if (!closeLanguage)
1113                                 running_font.setLanguage(copy_font.language());
1114                         // leave font open if language is still open
1115                         open_font = (running_font.language() == basefont.language());
1116                         if (closeLanguage)
1117                                 runparams.local_font = &basefont;
1118                 }
1119         }
1120
1121         size_t const previous_row_count = os.texrow().rows();
1122
1123         try {
1124                 runparams.lastid = id_;
1125                 runparams.lastpos = i;
1126                 inset->latex(os, runparams);
1127         } catch (EncodingException & e) {
1128                 // add location information and throw again.
1129                 e.par_id = id_;
1130                 e.pos = i;
1131                 throw(e);
1132         }
1133
1134         if (close) {
1135                 if (running_font.language()->lang() == "farsi")
1136                                 os << "\\endL{}";
1137                         else
1138                                 os << '}';
1139         }
1140
1141         if (os.texrow().rows() > previous_row_count) {
1142                 os.texrow().start(owner_->id(), i + 1);
1143                 column = 0;
1144         } else {
1145                 column += (unsigned int)(os.os().tellp() - len);
1146         }
1147
1148         if (owner_->isDeleted(i))
1149                 --runparams.inDeletedInset;
1150 }
1151
1152
1153 void Paragraph::Private::latexSpecialChar(otexstream & os,
1154                                           BufferParams const & bparams,
1155                                           OutputParams const & runparams,
1156                                           Font const & running_font,
1157                                           Change const & running_change,
1158                                           Layout const & style,
1159                                           pos_type & i,
1160                                           pos_type end_pos,
1161                                           unsigned int & column)
1162 {
1163         // With polyglossia, brackets and stuff need not be reversed
1164         // in RTL scripts (see bug #8251)
1165         char_type const c = (runparams.use_polyglossia) ?
1166                 owner_->getUChar(bparams, i) : text_[i];
1167
1168         if (style.pass_thru || runparams.pass_thru
1169             || contains(style.pass_thru_chars, c)
1170             || contains(runparams.pass_thru_chars, c)) {
1171                 if (c != '\0') {
1172                         Encoding const * const enc = runparams.encoding;
1173                         if (enc && !enc->encodable(c))
1174                                 throw EncodingException(c);
1175                         os.put(c);
1176                 }
1177                 return;
1178         }
1179
1180         // TIPA uses its own T3 encoding
1181         if (runparams.inIPA && latexSpecialT3(c, os, i, column))
1182                 return;
1183         // If T1 font encoding is used, use the special
1184         // characters it provides.
1185         // NOTE: Some languages reset the font encoding internally to a
1186         //       non-standard font encoding. If we are using such a language,
1187         //       we do not output special T1 chars.
1188         if (!runparams.inIPA && !running_font.language()->internalFontEncoding()
1189             && bparams.font_encoding() == "T1" && latexSpecialT1(c, os, i, column))
1190                 return;
1191
1192         // Otherwise, we use what LaTeX provides us.
1193         switch (c) {
1194         case '\\':
1195                 os << "\\textbackslash{}";
1196                 column += 15;
1197                 break;
1198         case '<':
1199                 os << "\\textless{}";
1200                 column += 10;
1201                 break;
1202         case '>':
1203                 os << "\\textgreater{}";
1204                 column += 13;
1205                 break;
1206         case '|':
1207                 os << "\\textbar{}";
1208                 column += 9;
1209                 break;
1210         case '-':
1211                 os << '-';
1212                 if (i + 1 < static_cast<pos_type>(text_.size()) &&
1213                     (end_pos == -1 || i + 1 < end_pos) &&
1214                     text_[i+1] == '-') {
1215                         // Prevent "--" becoming an endash and "---" becoming
1216                         // an emdash.
1217                         // Within \ttfamily, "--" is merged to "-" (no endash)
1218                         // so we avoid this rather irritating ligature as well
1219                         os << "{}";
1220                         column += 2;
1221                 }
1222                 break;
1223         case '\"':
1224                 os << "\\char`\\\"{}";
1225                 column += 9;
1226                 break;
1227
1228         case '$': case '&':
1229         case '%': case '#': case '{':
1230         case '}': case '_':
1231                 os << '\\';
1232                 os.put(c);
1233                 column += 1;
1234                 break;
1235
1236         case '~':
1237                 os << "\\textasciitilde{}";
1238                 column += 16;
1239                 break;
1240
1241         case '^':
1242                 os << "\\textasciicircum{}";
1243                 column += 17;
1244                 break;
1245
1246         case '*':
1247         case '[':
1248         case ']':
1249                 // avoid being mistaken for optional arguments
1250                 os << '{';
1251                 os.put(c);
1252                 os << '}';
1253                 column += 2;
1254                 break;
1255
1256         case ' ':
1257                 // Blanks are printed before font switching.
1258                 // Sure? I am not! (try nice-latex)
1259                 // I am sure it's correct. LyX might be smarter
1260                 // in the future, but for now, nothing wrong is
1261                 // written. (Asger)
1262                 break;
1263
1264         default:
1265                 if (c == '\0')
1266                         return;
1267
1268                 Encoding const & encoding = *(runparams.encoding);
1269                 char_type next = '\0';
1270                 if (i + 1 < int(text_.size())) {
1271                         next = text_[i + 1];
1272                         if (Encodings::isCombiningChar(next)) {
1273                                 column += latexSurrogatePair(os, c, next, runparams) - 1;
1274                                 ++i;
1275                                 break;
1276                         }
1277                 }
1278                 string script;
1279                 pair<docstring, bool> latex = encoding.latexChar(c);
1280                 docstring nextlatex;
1281                 bool nexttipas = false;
1282                 string nexttipashortcut;
1283                 if (next != '\0' && next != META_INSET && encoding.encodable(next)) {
1284                         nextlatex = encoding.latexChar(next).first;
1285                         if (runparams.inIPA) {
1286                                 nexttipashortcut = Encodings::TIPAShortcut(next);
1287                                 nexttipas = !nexttipashortcut.empty();
1288                         }
1289                 }
1290                 bool tipas = false;
1291                 if (runparams.inIPA) {
1292                         string const tipashortcut = Encodings::TIPAShortcut(c);
1293                         if (!tipashortcut.empty()) {
1294                                 latex.first = from_ascii(tipashortcut);
1295                                 latex.second = false;
1296                                 tipas = true;
1297                         }
1298                 }
1299                 if (Encodings::isKnownScriptChar(c, script)
1300                     && prefixIs(latex.first, from_ascii("\\" + script)))
1301                         column += writeScriptChars(os, latex.first,
1302                                         running_change, encoding, i) - 1;
1303                 else if (latex.second
1304                          && ((!prefixIs(nextlatex, '\\')
1305                                && !prefixIs(nextlatex, '{')
1306                                && !prefixIs(nextlatex, '}'))
1307                              || (nexttipas
1308                                  && !prefixIs(from_ascii(nexttipashortcut), '\\')))
1309                          && !tipas) {
1310                         // Prevent eating of a following
1311                         // space or command corruption by
1312                         // following characters
1313                         if (next == ' ' || next == '\0') {
1314                                 column += latex.first.length() + 1;
1315                                 os << latex.first << "{}";
1316                         } else {
1317                                 column += latex.first.length();
1318                                 os << latex.first << " ";
1319                         }
1320                 } else {
1321                         column += latex.first.length() - 1;
1322                         os << latex.first;
1323                 }
1324                 break;
1325         }
1326 }
1327
1328
1329 bool Paragraph::Private::latexSpecialT1(char_type const c, otexstream & os,
1330         pos_type i, unsigned int & column)
1331 {
1332         switch (c) {
1333         case '>':
1334         case '<':
1335                 os.put(c);
1336                 // In T1 encoding, these characters exist
1337                 // but we should avoid ligatures
1338                 if (i + 1 >= int(text_.size()) || text_[i + 1] != c)
1339                         return true;
1340                 os << "\\textcompwordmark{}";
1341                 column += 19;
1342                 return true;
1343         case '|':
1344                 os.put(c);
1345                 return true;
1346         case '\"':
1347                 // soul.sty breaks with \char`\"
1348                 os << "\\textquotedbl{}";
1349                 column += 14;
1350                 return true;
1351         default:
1352                 return false;
1353         }
1354 }
1355
1356
1357 bool Paragraph::Private::latexSpecialT3(char_type const c, otexstream & os,
1358         pos_type /*i*/, unsigned int & column)
1359 {
1360         switch (c) {
1361         case '*':
1362         case '[':
1363         case ']':
1364         case '\"':
1365                 os.put(c);
1366                 return true;
1367         case '|':
1368                 os << "\\textvertline{}";
1369                 column += 14;
1370                 return true;
1371         default:
1372                 return false;
1373         }
1374 }
1375
1376
1377 void Paragraph::Private::validate(LaTeXFeatures & features) const
1378 {
1379         if (layout_->inpreamble && inset_owner_) {
1380                 bool const is_command = layout_->latextype == LATEX_COMMAND;
1381                 Buffer const & buf = inset_owner_->buffer();
1382                 BufferParams const & bp = features.runparams().is_child
1383                         ? buf.masterParams() : buf.params();
1384                 Font f;
1385                 // Using a string stream here circumvents the encoding
1386                 // switching machinery of odocstream. Therefore the
1387                 // output is wrong if this paragraph contains content
1388                 // that needs to switch encoding.
1389                 odocstringstream ods;
1390                 otexstream os(ods);
1391                 if (is_command) {
1392                         os << '\\' << from_ascii(layout_->latexname());
1393                         // we have to provide all the optional arguments here, even though
1394                         // the last one is the only one we care about.
1395                         // Separate handling of optional argument inset.
1396                         if (!layout_->latexargs().empty()) {
1397                                 OutputParams rp = features.runparams();
1398                                 rp.local_font = &owner_->getFirstFontSettings(bp);
1399                                 latexArgInsets(*owner_, os, rp, layout_->latexargs());
1400                         }
1401                         os << from_ascii(layout_->latexparam());
1402                 }
1403                 docstring::size_type const length = ods.str().length();
1404                 // this will output "{" at the beginning, but not at the end
1405                 owner_->latex(bp, f, os, features.runparams(), 0, -1, true);
1406                 if (ods.str().length() > length) {
1407                         if (is_command) {
1408                                 ods << '}';
1409                                 if (!layout_->postcommandargs().empty()) {
1410                                         OutputParams rp = features.runparams();
1411                                         rp.local_font = &owner_->getFirstFontSettings(bp);
1412                                         latexArgInsets(*owner_, os, rp, layout_->postcommandargs(), "post:");
1413                                 }
1414                         }
1415                         string const snippet = to_utf8(ods.str());
1416                         features.addPreambleSnippet(snippet, 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                         icit->inset->validate(features);
1445                         if (layout_->needprotect &&
1446                             icit->inset->lyxCode() == FOOT_CODE)
1447                                 features.require("NeedLyXFootnoteCode");
1448                 }
1449         }
1450
1451         // then the contents
1452         for (pos_type i = 0; i < int(text_.size()) ; ++i) {
1453                 BufferEncodings::validate(text_[i], features);
1454         }
1455 }
1456
1457 /////////////////////////////////////////////////////////////////////
1458 //
1459 // Paragraph
1460 //
1461 /////////////////////////////////////////////////////////////////////
1462
1463 namespace {
1464         Layout const emptyParagraphLayout;
1465 }
1466
1467 Paragraph::Paragraph()
1468         : d(new Paragraph::Private(this, emptyParagraphLayout))
1469 {
1470         itemdepth = 0;
1471         d->params_.clear();
1472 }
1473
1474
1475 Paragraph::Paragraph(Paragraph const & par)
1476         : itemdepth(par.itemdepth),
1477         d(new Paragraph::Private(*par.d, this))
1478 {
1479         registerWords();
1480 }
1481
1482
1483 Paragraph::Paragraph(Paragraph const & par, pos_type beg, pos_type end)
1484         : itemdepth(par.itemdepth),
1485         d(new Paragraph::Private(*par.d, this, beg, end))
1486 {
1487         registerWords();
1488 }
1489
1490
1491 Paragraph & Paragraph::operator=(Paragraph const & par)
1492 {
1493         // needed as we will destroy the private part before copying it
1494         if (&par != this) {
1495                 itemdepth = par.itemdepth;
1496
1497                 deregisterWords();
1498                 delete d;
1499                 d = new Private(*par.d, this);
1500                 registerWords();
1501         }
1502         return *this;
1503 }
1504
1505
1506 Paragraph::~Paragraph()
1507 {
1508         deregisterWords();
1509         delete d;
1510 }
1511
1512
1513 namespace {
1514
1515 // this shall be called just before every "os << ..." action.
1516 void flushString(ostream & os, docstring & s)
1517 {
1518         os << to_utf8(s);
1519         s.erase();
1520 }
1521
1522 }
1523
1524
1525 void Paragraph::write(ostream & os, BufferParams const & bparams,
1526         depth_type & dth) const
1527 {
1528         // The beginning or end of a deeper (i.e. nested) area?
1529         if (dth != d->params_.depth()) {
1530                 if (d->params_.depth() > dth) {
1531                         while (d->params_.depth() > dth) {
1532                                 os << "\n\\begin_deeper";
1533                                 ++dth;
1534                         }
1535                 } else {
1536                         while (d->params_.depth() < dth) {
1537                                 os << "\n\\end_deeper";
1538                                 --dth;
1539                         }
1540                 }
1541         }
1542
1543         // First write the layout
1544         os << "\n\\begin_layout " << to_utf8(d->layout_->name()) << '\n';
1545
1546         d->params_.write(os);
1547
1548         Font font1(inherit_font, bparams.language);
1549
1550         Change running_change = Change(Change::UNCHANGED);
1551
1552         // this string is used as a buffer to avoid repetitive calls
1553         // to to_utf8(), which turn out to be expensive (JMarc)
1554         docstring write_buffer;
1555
1556         int column = 0;
1557         for (pos_type i = 0; i <= size(); ++i) {
1558
1559                 Change const & change = lookupChange(i);
1560                 if (change != running_change)
1561                         flushString(os, write_buffer);
1562                 Changes::lyxMarkChange(os, bparams, column, running_change, change);
1563                 running_change = change;
1564
1565                 if (i == size())
1566                         break;
1567
1568                 // Write font changes
1569                 Font font2 = getFontSettings(bparams, i);
1570                 if (font2 != font1) {
1571                         flushString(os, write_buffer);
1572                         font2.lyxWriteChanges(font1, os);
1573                         column = 0;
1574                         font1 = font2;
1575                 }
1576
1577                 char_type const c = d->text_[i];
1578                 switch (c) {
1579                 case META_INSET:
1580                         if (Inset const * inset = getInset(i)) {
1581                                 flushString(os, write_buffer);
1582                                 if (inset->directWrite()) {
1583                                         // international char, let it write
1584                                         // code directly so it's shorter in
1585                                         // the file
1586                                         inset->write(os);
1587                                 } else {
1588                                         if (i)
1589                                                 os << '\n';
1590                                         os << "\\begin_inset ";
1591                                         inset->write(os);
1592                                         os << "\n\\end_inset\n\n";
1593                                         column = 0;
1594                                 }
1595                                 // FIXME This can be removed again once the mystery
1596                                 // crash has been resolved.
1597                                 os << flush;
1598                         }
1599                         break;
1600                 case '\\':
1601                         flushString(os, write_buffer);
1602                         os << "\n\\backslash\n";
1603                         column = 0;
1604                         break;
1605                 case '.':
1606                         flushString(os, write_buffer);
1607                         if (i + 1 < size() && d->text_[i + 1] == ' ') {
1608                                 os << ".\n";
1609                                 column = 0;
1610                         } else
1611                                 os << '.';
1612                         break;
1613                 default:
1614                         if ((column > 70 && c == ' ')
1615                             || column > 79) {
1616                                 flushString(os, write_buffer);
1617                                 os << '\n';
1618                                 column = 0;
1619                         }
1620                         // this check is to amend a bug. LyX sometimes
1621                         // inserts '\0' this could cause problems.
1622                         if (c != '\0')
1623                                 write_buffer.push_back(c);
1624                         else
1625                                 LYXERR0("NUL char in structure.");
1626                         ++column;
1627                         break;
1628                 }
1629         }
1630
1631         flushString(os, write_buffer);
1632         os << "\n\\end_layout\n";
1633         // FIXME This can be removed again once the mystery
1634         // crash has been resolved.
1635         os << flush;
1636 }
1637
1638
1639 void Paragraph::validate(LaTeXFeatures & features) const
1640 {
1641         d->validate(features);
1642 }
1643
1644
1645 void Paragraph::insert(pos_type start, docstring const & str,
1646                        Font const & font, Change const & change)
1647 {
1648         for (size_t i = 0, n = str.size(); i != n ; ++i)
1649                 insertChar(start + i, str[i], font, change);
1650 }
1651
1652
1653 void Paragraph::appendChar(char_type c, Font const & font,
1654                 Change const & change)
1655 {
1656         // track change
1657         d->changes_.insert(change, d->text_.size());
1658         // when appending characters, no need to update tables
1659         d->text_.push_back(c);
1660         setFont(d->text_.size() - 1, font);
1661         d->requestSpellCheck(d->text_.size() - 1);
1662 }
1663
1664
1665 void Paragraph::appendString(docstring const & s, Font const & font,
1666                 Change const & change)
1667 {
1668         pos_type end = s.size();
1669         size_t oldsize = d->text_.size();
1670         size_t newsize = oldsize + end;
1671         size_t capacity = d->text_.capacity();
1672         if (newsize >= capacity)
1673                 d->text_.reserve(max(capacity + 100, newsize));
1674
1675         // when appending characters, no need to update tables
1676         d->text_.append(s);
1677
1678         // FIXME: Optimize this!
1679         for (size_t i = oldsize; i != newsize; ++i) {
1680                 // track change
1681                 d->changes_.insert(change, i);
1682                 d->requestSpellCheck(i);
1683         }
1684         d->fontlist_.set(oldsize, font);
1685         d->fontlist_.set(newsize - 1, font);
1686 }
1687
1688
1689 void Paragraph::insertChar(pos_type pos, char_type c,
1690                            bool trackChanges)
1691 {
1692         d->insertChar(pos, c, Change(trackChanges ?
1693                            Change::INSERTED : Change::UNCHANGED));
1694 }
1695
1696
1697 void Paragraph::insertChar(pos_type pos, char_type c,
1698                            Font const & font, bool trackChanges)
1699 {
1700         d->insertChar(pos, c, Change(trackChanges ?
1701                            Change::INSERTED : Change::UNCHANGED));
1702         setFont(pos, font);
1703 }
1704
1705
1706 void Paragraph::insertChar(pos_type pos, char_type c,
1707                            Font const & font, Change const & change)
1708 {
1709         d->insertChar(pos, c, change);
1710         setFont(pos, font);
1711 }
1712
1713
1714 void Paragraph::resetFonts(Font const & font)
1715 {
1716         d->fontlist_.clear();
1717         d->fontlist_.set(0, font);
1718         d->fontlist_.set(d->text_.size() - 1, font);
1719 }
1720
1721 // Gets uninstantiated font setting at position.
1722 Font const & Paragraph::getFontSettings(BufferParams const & bparams,
1723                                          pos_type pos) const
1724 {
1725         if (pos > size()) {
1726                 LYXERR0("pos: " << pos << " size: " << size());
1727                 LBUFERR(false);
1728         }
1729
1730         FontList::const_iterator cit = d->fontlist_.fontIterator(pos);
1731         if (cit != d->fontlist_.end())
1732                 return cit->font();
1733
1734         if (pos == size() && !empty())
1735                 return getFontSettings(bparams, pos - 1);
1736
1737         // Optimisation: avoid a full font instantiation if there is no
1738         // language change from previous call.
1739         static Font previous_font;
1740         static Language const * previous_lang = 0;
1741         Language const * lang = getParLanguage(bparams);
1742         if (lang != previous_lang) {
1743                 previous_lang = lang;
1744                 previous_font = Font(inherit_font, lang);
1745         }
1746         return previous_font;
1747 }
1748
1749
1750 FontSpan Paragraph::fontSpan(pos_type pos) const
1751 {
1752         LBUFERR(pos <= size());
1753
1754         if (pos == size())
1755                 return FontSpan(pos, pos);
1756
1757         pos_type start = 0;
1758         FontList::const_iterator cit = d->fontlist_.begin();
1759         FontList::const_iterator end = d->fontlist_.end();
1760         for (; cit != end; ++cit) {
1761                 if (cit->pos() >= pos) {
1762                         if (pos >= beginOfBody())
1763                                 return FontSpan(max(start, beginOfBody()),
1764                                                 cit->pos());
1765                         else
1766                                 return FontSpan(start,
1767                                                 min(beginOfBody() - 1,
1768                                                          cit->pos()));
1769                 }
1770                 start = cit->pos() + 1;
1771         }
1772
1773         // This should not happen, but if so, we take no chances.
1774         LYXERR0("Paragraph::fontSpan: position not found in fontinfo table!");
1775         LASSERT(false, return FontSpan(pos, pos));
1776 }
1777
1778
1779 // Gets uninstantiated font setting at position 0
1780 Font const & Paragraph::getFirstFontSettings(BufferParams const & bparams) const
1781 {
1782         if (!empty() && !d->fontlist_.empty())
1783                 return d->fontlist_.begin()->font();
1784
1785         // Optimisation: avoid a full font instantiation if there is no
1786         // language change from previous call.
1787         static Font previous_font;
1788         static Language const * previous_lang = 0;
1789         if (bparams.language != previous_lang) {
1790                 previous_lang = bparams.language;
1791                 previous_font = Font(inherit_font, bparams.language);
1792         }
1793
1794         return previous_font;
1795 }
1796
1797
1798 // Gets the fully instantiated font at a given position in a paragraph
1799 // This is basically the same function as Text::GetFont() in text2.cpp.
1800 // The difference is that this one is used for generating the LaTeX file,
1801 // and thus cosmetic "improvements" are disallowed: This has to deliver
1802 // the true picture of the buffer. (Asger)
1803 Font const Paragraph::getFont(BufferParams const & bparams, pos_type pos,
1804                                  Font const & outerfont) const
1805 {
1806         LBUFERR(pos >= 0);
1807
1808         Font font = getFontSettings(bparams, pos);
1809
1810         pos_type const body_pos = beginOfBody();
1811         FontInfo & fi = font.fontInfo();
1812         if (pos < body_pos)
1813                 fi.realize(d->layout_->labelfont);
1814         else
1815                 fi.realize(d->layout_->font);
1816
1817         fi.realize(outerfont.fontInfo());
1818         fi.realize(bparams.getFont().fontInfo());
1819
1820         return font;
1821 }
1822
1823
1824 Font const Paragraph::getLabelFont
1825         (BufferParams const & bparams, Font const & outerfont) const
1826 {
1827         FontInfo tmpfont = d->layout_->labelfont;
1828         tmpfont.realize(outerfont.fontInfo());
1829         tmpfont.realize(bparams.getFont().fontInfo());
1830         return Font(tmpfont, getParLanguage(bparams));
1831 }
1832
1833
1834 Font const Paragraph::getLayoutFont
1835         (BufferParams const & bparams, Font const & outerfont) const
1836 {
1837         FontInfo tmpfont = d->layout_->font;
1838         tmpfont.realize(outerfont.fontInfo());
1839         tmpfont.realize(bparams.getFont().fontInfo());
1840         return Font(tmpfont, getParLanguage(bparams));
1841 }
1842
1843
1844 char_type Paragraph::getUChar(BufferParams const & bparams, pos_type pos) const
1845 {
1846         char_type c = d->text_[pos];
1847         if (!getFontSettings(bparams, pos).isRightToLeft())
1848                 return c;
1849
1850         // FIXME: The arabic special casing is due to the difference of arabic
1851         // round brackets input introduced in r18599. Check if this should be
1852         // unified with Hebrew or at least if all bracket types should be
1853         // handled the same (file format change in either case).
1854         string const & lang = getFontSettings(bparams, pos).language()->lang();
1855         bool const arabic = lang == "arabic_arabtex" || lang == "arabic_arabi"
1856                 || lang == "farsi";
1857         char_type uc = c;
1858         switch (c) {
1859         case '(':
1860                 uc = arabic ? c : ')';
1861                 break;
1862         case ')':
1863                 uc = arabic ? c : '(';
1864                 break;
1865         case '[':
1866                 uc = ']';
1867                 break;
1868         case ']':
1869                 uc = '[';
1870                 break;
1871         case '{':
1872                 uc = '}';
1873                 break;
1874         case '}':
1875                 uc = '{';
1876                 break;
1877         case '<':
1878                 uc = '>';
1879                 break;
1880         case '>':
1881                 uc = '<';
1882                 break;
1883         }
1884
1885         return uc;
1886 }
1887
1888
1889 void Paragraph::setFont(pos_type pos, Font const & font)
1890 {
1891         LASSERT(pos <= size(), return);
1892
1893         // First, reduce font against layout/label font
1894         // Update: The setCharFont() routine in text2.cpp already
1895         // reduces font, so we don't need to do that here. (Asger)
1896
1897         d->fontlist_.set(pos, font);
1898 }
1899
1900
1901 void Paragraph::makeSameLayout(Paragraph const & par)
1902 {
1903         d->layout_ = par.d->layout_;
1904         d->params_ = par.d->params_;
1905 }
1906
1907
1908 bool Paragraph::stripLeadingSpaces(bool trackChanges)
1909 {
1910         if (isFreeSpacing())
1911                 return false;
1912
1913         int pos = 0;
1914         int count = 0;
1915
1916         while (pos < size() && (isNewline(pos) || isLineSeparator(pos))) {
1917                 if (eraseChar(pos, trackChanges))
1918                         ++count;
1919                 else
1920                         ++pos;
1921         }
1922
1923         return count > 0 || pos > 0;
1924 }
1925
1926
1927 bool Paragraph::hasSameLayout(Paragraph const & par) const
1928 {
1929         return par.d->layout_ == d->layout_
1930                 && d->params_.sameLayout(par.d->params_);
1931 }
1932
1933
1934 depth_type Paragraph::getDepth() const
1935 {
1936         return d->params_.depth();
1937 }
1938
1939
1940 depth_type Paragraph::getMaxDepthAfter() const
1941 {
1942         if (d->layout_->isEnvironment())
1943                 return d->params_.depth() + 1;
1944         else
1945                 return d->params_.depth();
1946 }
1947
1948
1949 char Paragraph::getAlign() const
1950 {
1951         if (d->params_.align() == LYX_ALIGN_LAYOUT)
1952                 return d->layout_->align;
1953         else
1954                 return d->params_.align();
1955 }
1956
1957
1958 docstring const & Paragraph::labelString() const
1959 {
1960         return d->params_.labelString();
1961 }
1962
1963
1964 // the next two functions are for the manual labels
1965 docstring const Paragraph::getLabelWidthString() const
1966 {
1967         if (d->layout_->margintype == MARGIN_MANUAL
1968             || d->layout_->latextype == LATEX_BIB_ENVIRONMENT)
1969                 return d->params_.labelWidthString();
1970         else
1971                 return _("Senseless with this layout!");
1972 }
1973
1974
1975 void Paragraph::setLabelWidthString(docstring const & s)
1976 {
1977         d->params_.labelWidthString(s);
1978 }
1979
1980
1981 docstring Paragraph::expandLabel(Layout const & layout,
1982                 BufferParams const & bparams) const
1983 {
1984         return expandParagraphLabel(layout, bparams, true);
1985 }
1986
1987
1988 docstring Paragraph::expandDocBookLabel(Layout const & layout,
1989                 BufferParams const & bparams) const
1990 {
1991         return expandParagraphLabel(layout, bparams, false);
1992 }
1993
1994
1995 docstring Paragraph::expandParagraphLabel(Layout const & layout,
1996                 BufferParams const & bparams, bool process_appendix) const
1997 {
1998         DocumentClass const & tclass = bparams.documentClass();
1999         string const & lang = getParLanguage(bparams)->code();
2000         bool const in_appendix = process_appendix && d->params_.appendix();
2001         docstring fmt = translateIfPossible(layout.labelstring(in_appendix), lang);
2002
2003         if (fmt.empty() && !layout.counter.empty())
2004                 return tclass.counters().theCounter(layout.counter, lang);
2005
2006         // handle 'inherited level parts' in 'fmt',
2007         // i.e. the stuff between '@' in   '@Section@.\arabic{subsection}'
2008         size_t const i = fmt.find('@', 0);
2009         if (i != docstring::npos) {
2010                 size_t const j = fmt.find('@', i + 1);
2011                 if (j != docstring::npos) {
2012                         docstring parent(fmt, i + 1, j - i - 1);
2013                         docstring label = from_ascii("??");
2014                         if (tclass.hasLayout(parent))
2015                                 label = expandParagraphLabel(tclass[parent], bparams,
2016                                                       process_appendix);
2017                         fmt = docstring(fmt, 0, i) + label
2018                                 + docstring(fmt, j + 1, docstring::npos);
2019                 }
2020         }
2021
2022         return tclass.counters().counterLabel(fmt, lang);
2023 }
2024
2025
2026 void Paragraph::applyLayout(Layout const & new_layout)
2027 {
2028         d->layout_ = &new_layout;
2029         LyXAlignment const oldAlign = d->params_.align();
2030
2031         if (!(oldAlign & d->layout_->alignpossible)) {
2032                 frontend::Alert::warning(_("Alignment not permitted"),
2033                         _("The new layout does not permit the alignment previously used.\nSetting to default."));
2034                 d->params_.align(LYX_ALIGN_LAYOUT);
2035         }
2036 }
2037
2038
2039 pos_type Paragraph::beginOfBody() const
2040 {
2041         return d->begin_of_body_;
2042 }
2043
2044
2045 void Paragraph::setBeginOfBody()
2046 {
2047         if (d->layout_->labeltype != LABEL_MANUAL) {
2048                 d->begin_of_body_ = 0;
2049                 return;
2050         }
2051
2052         // Unroll the first two cycles of the loop
2053         // and remember the previous character to
2054         // remove unnecessary getChar() calls
2055         pos_type i = 0;
2056         pos_type end = size();
2057         if (i < end && !(isNewline(i) || isEnvSeparator(i))) {
2058                 ++i;
2059                 if (i < end) {
2060                         char_type previous_char = d->text_[i];
2061                         if (!(isNewline(i) || isEnvSeparator(i))) {
2062                                 ++i;
2063                                 while (i < end && previous_char != ' ') {
2064                                         char_type temp = d->text_[i];
2065                                         if (isNewline(i) || isEnvSeparator(i))
2066                                                 break;
2067                                         ++i;
2068                                         previous_char = temp;
2069                                 }
2070                         }
2071                 }
2072         }
2073
2074         d->begin_of_body_ = i;
2075 }
2076
2077
2078 bool Paragraph::allowParagraphCustomization() const
2079 {
2080         return inInset().allowParagraphCustomization();
2081 }
2082
2083
2084 bool Paragraph::usePlainLayout() const
2085 {
2086         return inInset().usePlainLayout();
2087 }
2088
2089
2090 bool Paragraph::isPassThru() const
2091 {
2092         return inInset().isPassThru() || d->layout_->pass_thru;
2093 }
2094
2095 namespace {
2096
2097 // paragraphs inside floats need different alignment tags to avoid
2098 // unwanted space
2099
2100 bool noTrivlistCentering(InsetCode code)
2101 {
2102         return code == FLOAT_CODE
2103                || code == WRAP_CODE
2104                || code == CELL_CODE;
2105 }
2106
2107
2108 string correction(string const & orig)
2109 {
2110         if (orig == "flushleft")
2111                 return "raggedright";
2112         if (orig == "flushright")
2113                 return "raggedleft";
2114         if (orig == "center")
2115                 return "centering";
2116         return orig;
2117 }
2118
2119
2120 bool corrected_env(otexstream & os, string const & suffix, string const & env,
2121         InsetCode code, bool const lastpar, int & col)
2122 {
2123         string macro = suffix + "{";
2124         if (noTrivlistCentering(code)) {
2125                 if (lastpar) {
2126                         // the last paragraph in non-trivlist-aligned
2127                         // context is special (to avoid unwanted whitespace)
2128                         if (suffix == "\\begin") {
2129                                 macro = "\\" + correction(env) + "{}";
2130                                 os << from_ascii(macro);
2131                                 col += macro.size();
2132                                 return true;
2133                         }
2134                         return false;
2135                 }
2136                 macro += correction(env);
2137         } else
2138                 macro += env;
2139         macro += "}";
2140         if (suffix == "\\par\\end") {
2141                 os << breakln;
2142                 col = 0;
2143         }
2144         os << from_ascii(macro);
2145         col += macro.size();
2146         if (suffix == "\\begin") {
2147                 os << breakln;
2148                 col = 0;
2149         }
2150         return true;
2151 }
2152
2153 } // namespace anon
2154
2155
2156 int Paragraph::Private::startTeXParParams(BufferParams const & bparams,
2157                         otexstream & os, OutputParams const & runparams) const
2158 {
2159         int column = 0;
2160
2161         bool canindent =
2162                 (bparams.paragraph_separation == BufferParams::ParagraphIndentSeparation) ?
2163                         (layout_->toggle_indent != ITOGGLE_NEVER) :
2164                         (layout_->toggle_indent == ITOGGLE_ALWAYS);
2165
2166         if (canindent && params_.noindent() && !layout_->pass_thru) {
2167                 os << "\\noindent ";
2168                 column += 10;
2169         }
2170
2171         LyXAlignment const curAlign = params_.align();
2172
2173         if (curAlign == layout_->align)
2174                 return column;
2175
2176         switch (curAlign) {
2177         case LYX_ALIGN_NONE:
2178         case LYX_ALIGN_BLOCK:
2179         case LYX_ALIGN_LAYOUT:
2180         case LYX_ALIGN_SPECIAL:
2181         case LYX_ALIGN_DECIMAL:
2182                 break;
2183         case LYX_ALIGN_LEFT:
2184         case LYX_ALIGN_RIGHT:
2185         case LYX_ALIGN_CENTER:
2186                 if (runparams.moving_arg) {
2187                         os << "\\protect";
2188                         column += 8;
2189                 }
2190                 break;
2191         }
2192
2193         string const begin_tag = "\\begin";
2194         InsetCode code = ownerCode();
2195         bool const lastpar = runparams.isLastPar;
2196
2197         switch (curAlign) {
2198         case LYX_ALIGN_NONE:
2199         case LYX_ALIGN_BLOCK:
2200         case LYX_ALIGN_LAYOUT:
2201         case LYX_ALIGN_SPECIAL:
2202         case LYX_ALIGN_DECIMAL:
2203                 break;
2204         case LYX_ALIGN_LEFT: {
2205                 if (owner_->getParLanguage(bparams)->babel() != "hebrew")
2206                         corrected_env(os, begin_tag, "flushleft", code, lastpar, column);
2207                 else
2208                         corrected_env(os, begin_tag, "flushright", code, lastpar, column);
2209                 break;
2210         } case LYX_ALIGN_RIGHT: {
2211                 if (owner_->getParLanguage(bparams)->babel() != "hebrew")
2212                         corrected_env(os, begin_tag, "flushright", code, lastpar, column);
2213                 else
2214                         corrected_env(os, begin_tag, "flushleft", code, lastpar, column);
2215                 break;
2216         } case LYX_ALIGN_CENTER: {
2217                 corrected_env(os, begin_tag, "center", code, lastpar, column);
2218                 break;
2219         }
2220         }
2221
2222         return column;
2223 }
2224
2225
2226 bool Paragraph::Private::endTeXParParams(BufferParams const & bparams,
2227                         otexstream & os, OutputParams const & runparams) const
2228 {
2229         LyXAlignment const curAlign = params_.align();
2230
2231         if (curAlign == layout_->align)
2232                 return false;
2233
2234         switch (curAlign) {
2235         case LYX_ALIGN_NONE:
2236         case LYX_ALIGN_BLOCK:
2237         case LYX_ALIGN_LAYOUT:
2238         case LYX_ALIGN_SPECIAL:
2239         case LYX_ALIGN_DECIMAL:
2240                 break;
2241         case LYX_ALIGN_LEFT:
2242         case LYX_ALIGN_RIGHT:
2243         case LYX_ALIGN_CENTER:
2244                 if (runparams.moving_arg)
2245                         os << "\\protect";
2246                 break;
2247         }
2248
2249         bool output = false;
2250         int col = 0;
2251         string const end_tag = "\\par\\end";
2252         InsetCode code = ownerCode();
2253         bool const lastpar = runparams.isLastPar;
2254
2255         switch (curAlign) {
2256         case LYX_ALIGN_NONE:
2257         case LYX_ALIGN_BLOCK:
2258         case LYX_ALIGN_LAYOUT:
2259         case LYX_ALIGN_SPECIAL:
2260         case LYX_ALIGN_DECIMAL:
2261                 break;
2262         case LYX_ALIGN_LEFT: {
2263                 if (owner_->getParLanguage(bparams)->babel() != "hebrew")
2264                         output = corrected_env(os, end_tag, "flushleft", code, lastpar, col);
2265                 else
2266                         output = corrected_env(os, end_tag, "flushright", code, lastpar, col);
2267                 break;
2268         } case LYX_ALIGN_RIGHT: {
2269                 if (owner_->getParLanguage(bparams)->babel() != "hebrew")
2270                         output = corrected_env(os, end_tag, "flushright", code, lastpar, col);
2271                 else
2272                         output = corrected_env(os, end_tag, "flushleft", code, lastpar, col);
2273                 break;
2274         } case LYX_ALIGN_CENTER: {
2275                 corrected_env(os, end_tag, "center", code, lastpar, col);
2276                 break;
2277         }
2278         }
2279
2280         return output || lastpar;
2281 }
2282
2283
2284 // This one spits out the text of the paragraph
2285 void Paragraph::latex(BufferParams const & bparams,
2286         Font const & outerfont,
2287         otexstream & os,
2288         OutputParams const & runparams,
2289         int start_pos, int end_pos, bool force) const
2290 {
2291         LYXERR(Debug::LATEX, "Paragraph::latex...     " << this);
2292
2293         // FIXME This check should not be needed. Perhaps issue an
2294         // error if it triggers.
2295         Layout const & style = inInset().forcePlainLayout() ?
2296                 bparams.documentClass().plainLayout() : *d->layout_;
2297
2298         if (!force && style.inpreamble)
2299                 return;
2300
2301         bool const allowcust = allowParagraphCustomization();
2302
2303         // Current base font for all inherited font changes, without any
2304         // change caused by an individual character, except for the language:
2305         // It is set to the language of the first character.
2306         // As long as we are in the label, this font is the base font of the
2307         // label. Before the first body character it is set to the base font
2308         // of the body.
2309         Font basefont;
2310
2311         // Maybe we have to create a optional argument.
2312         pos_type body_pos = beginOfBody();
2313         unsigned int column = 0;
2314
2315         if (body_pos > 0) {
2316                 // the optional argument is kept in curly brackets in
2317                 // case it contains a ']'
2318                 // This is not strictly needed, but if this is changed it
2319                 // would be a file format change, and tex2lyx would need
2320                 // to be adjusted, since it unconditionally removes the
2321                 // braces when it parses \item.
2322                 os << "[{";
2323                 column += 2;
2324                 basefont = getLabelFont(bparams, outerfont);
2325         } else {
2326                 basefont = getLayoutFont(bparams, outerfont);
2327         }
2328
2329         // Which font is currently active?
2330         Font running_font(basefont);
2331         // Do we have an open font change?
2332         bool open_font = false;
2333
2334         Change runningChange = Change(Change::UNCHANGED);
2335
2336         Encoding const * const prev_encoding = runparams.encoding;
2337
2338         os.texrow().start(id(), 0);
2339
2340         // if the paragraph is empty, the loop will not be entered at all
2341         if (empty()) {
2342                 if (style.isCommand()) {
2343                         os << '{';
2344                         ++column;
2345                 }
2346                 if (!style.leftdelim().empty()) {
2347                         os << style.leftdelim();
2348                         column += style.leftdelim().size();
2349                 }
2350                 if (allowcust)
2351                         column += d->startTeXParParams(bparams, os, runparams);
2352         }
2353
2354         for (pos_type i = 0; i < size(); ++i) {
2355                 // First char in paragraph or after label?
2356                 if (i == body_pos) {
2357                         if (body_pos > 0) {
2358                                 if (open_font) {
2359                                         column += running_font.latexWriteEndChanges(
2360                                                 os, bparams, runparams,
2361                                                 basefont, basefont);
2362                                         open_font = false;
2363                                 }
2364                                 basefont = getLayoutFont(bparams, outerfont);
2365                                 running_font = basefont;
2366
2367                                 column += Changes::latexMarkChange(os, bparams,
2368                                                 runningChange, Change(Change::UNCHANGED),
2369                                                 runparams);
2370                                 runningChange = Change(Change::UNCHANGED);
2371
2372                                 os << "}] ";
2373                                 column +=3;
2374                         }
2375                         if (style.isCommand()) {
2376                                 os << '{';
2377                                 ++column;
2378                         }
2379
2380                         if (!style.leftdelim().empty()) {
2381                                 os << style.leftdelim();
2382                                 column += style.leftdelim().size();
2383                         }
2384
2385                         if (allowcust)
2386                                 column += d->startTeXParParams(bparams, os,
2387                                                             runparams);
2388                 }
2389
2390                 Change const & change = runparams.inDeletedInset
2391                         ? runparams.changeOfDeletedInset : lookupChange(i);
2392
2393                 if (bparams.output_changes && runningChange != change) {
2394                         if (open_font) {
2395                                 column += running_font.latexWriteEndChanges(
2396                                                 os, bparams, runparams, basefont, basefont);
2397                                 open_font = false;
2398                         }
2399                         basefont = getLayoutFont(bparams, outerfont);
2400                         running_font = basefont;
2401
2402                         column += Changes::latexMarkChange(os, bparams, runningChange,
2403                                                            change, runparams);
2404                         runningChange = change;
2405                 }
2406
2407                 // do not output text which is marked deleted
2408                 // if change tracking output is disabled
2409                 if (!bparams.output_changes && change.deleted()) {
2410                         continue;
2411                 }
2412
2413                 ++column;
2414
2415                 // Fully instantiated font
2416                 Font const current_font = getFont(bparams, i, outerfont);
2417
2418                 Font const last_font = running_font;
2419
2420                 // Do we need to close the previous font?
2421                 if (open_font &&
2422                     (current_font != running_font ||
2423                      current_font.language() != running_font.language()))
2424                 {
2425                         column += running_font.latexWriteEndChanges(
2426                                         os, bparams, runparams, basefont,
2427                                         (i == body_pos-1) ? basefont : current_font);
2428                         running_font = basefont;
2429                         open_font = false;
2430                 }
2431
2432                 string const running_lang = runparams.use_polyglossia ?
2433                         running_font.language()->polyglossia() : running_font.language()->babel();
2434                 // close babel's font environment before opening CJK.
2435                 string const lang_end_command = runparams.use_polyglossia ?
2436                         "\\end{$$lang}" : lyxrc.language_command_end;
2437                 if (!running_lang.empty() &&
2438                     current_font.language()->encoding()->package() == Encoding::CJK) {
2439                                 string end_tag = subst(lang_end_command,
2440                                                         "$$lang",
2441                                                         running_lang);
2442                                 os << from_ascii(end_tag);
2443                                 column += end_tag.length();
2444                                 if (runparams.use_polyglossia)
2445                                         popPolyglossiaLang();
2446                 }
2447
2448                 // Switch file encoding if necessary (and allowed)
2449                 if (!runparams.pass_thru && !style.pass_thru &&
2450                     runparams.encoding->package() != Encoding::none &&
2451                     current_font.language()->encoding()->package() != Encoding::none) {
2452                         pair<bool, int> const enc_switch =
2453                                 switchEncoding(os.os(), bparams, runparams,
2454                                         *(current_font.language()->encoding()));
2455                         if (enc_switch.first) {
2456                                 column += enc_switch.second;
2457                                 runparams.encoding = current_font.language()->encoding();
2458                         }
2459                 }
2460
2461                 char_type const c = d->text_[i];
2462
2463                 // Do we need to change font?
2464                 if ((current_font != running_font ||
2465                      current_font.language() != running_font.language()) &&
2466                         i != body_pos - 1)
2467                 {
2468                         odocstringstream ods;
2469                         column += current_font.latexWriteStartChanges(ods, bparams,
2470                                                               runparams, basefont,
2471                                                               last_font);
2472                         running_font = current_font;
2473                         open_font = true;
2474                         docstring fontchange = ods.str();
2475                         // check whether the fontchange ends with a \\textcolor
2476                         // modifier and the text starts with a space (bug 4473)
2477                         docstring const last_modifier = rsplit(fontchange, '\\');
2478                         if (prefixIs(last_modifier, from_ascii("textcolor")) && c == ' ')
2479                                 os << fontchange << from_ascii("{}");
2480                         // check if the fontchange ends with a trailing blank
2481                         // (like "\small " (see bug 3382)
2482                         else if (suffixIs(fontchange, ' ') && c == ' ')
2483                                 os << fontchange.substr(0, fontchange.size() - 1)
2484                                    << from_ascii("{}");
2485                         else
2486                                 os << fontchange;
2487                 }
2488
2489                 // FIXME: think about end_pos implementation...
2490                 if (c == ' ' && i >= start_pos && (end_pos == -1 || i < end_pos)) {
2491                         // FIXME: integrate this case in latexSpecialChar
2492                         // Do not print the separation of the optional argument
2493                         // if style.pass_thru is false. This works because
2494                         // latexSpecialChar ignores spaces if
2495                         // style.pass_thru is false.
2496                         if (i != body_pos - 1) {
2497                                 if (d->simpleTeXBlanks(runparams, os,
2498                                                 i, column, current_font, style)) {
2499                                         // A surrogate pair was output. We
2500                                         // must not call latexSpecialChar
2501                                         // in this iteration, since it would output
2502                                         // the combining character again.
2503                                         ++i;
2504                                         continue;
2505                                 }
2506                         }
2507                 }
2508
2509                 OutputParams rp = runparams;
2510                 rp.free_spacing = style.free_spacing;
2511                 rp.local_font = &current_font;
2512                 rp.intitle = style.intitle;
2513
2514                 // Two major modes:  LaTeX or plain
2515                 // Handle here those cases common to both modes
2516                 // and then split to handle the two modes separately.
2517                 if (c == META_INSET) {
2518                         if (i >= start_pos && (end_pos == -1 || i < end_pos)) {
2519                                 d->latexInset(bparams, os, rp, running_font,
2520                                                 basefont, outerfont, open_font,
2521                                                 runningChange, style, i, column);
2522                         }
2523                 } else {
2524                         if (i >= start_pos && (end_pos == -1 || i < end_pos)) {
2525                                 try {
2526                                         d->latexSpecialChar(os, bparams, rp, running_font, runningChange,
2527                                                             style, i, end_pos, column);
2528                                 } catch (EncodingException & e) {
2529                                 if (runparams.dryrun) {
2530                                         os << "<" << _("LyX Warning: ")
2531                                            << _("uncodable character") << " '";
2532                                         os.put(c);
2533                                         os << "'>";
2534                                 } else {
2535                                         // add location information and throw again.
2536                                         e.par_id = id();
2537                                         e.pos = i;
2538                                         throw(e);
2539                                 }
2540                         }
2541                 }
2542                 }
2543
2544                 // Set the encoding to that returned from latexSpecialChar (see
2545                 // comment for encoding member in OutputParams.h)
2546                 runparams.encoding = rp.encoding;
2547         }
2548
2549         // If we have an open font definition, we have to close it
2550         if (open_font) {
2551 #ifdef FIXED_LANGUAGE_END_DETECTION
2552                 if (next_) {
2553                         running_font.latexWriteEndChanges(os, bparams,
2554                                         runparams, basefont,
2555                                         next_->getFont(bparams, 0, outerfont));
2556                 } else {
2557                         running_font.latexWriteEndChanges(os, bparams,
2558                                         runparams, basefont, basefont);
2559                 }
2560 #else
2561 //FIXME: For now we ALWAYS have to close the foreign font settings if they are
2562 //FIXME: there as we start another \selectlanguage with the next paragraph if
2563 //FIXME: we are in need of this. This should be fixed sometime (Jug)
2564                 running_font.latexWriteEndChanges(os, bparams, runparams,
2565                                 basefont, basefont);
2566 #endif
2567         }
2568
2569         column += Changes::latexMarkChange(os, bparams, runningChange,
2570                                            Change(Change::UNCHANGED), runparams);
2571
2572         // Needed if there is an optional argument but no contents.
2573         if (body_pos > 0 && body_pos == size()) {
2574                 os << "}]~";
2575         }
2576
2577         if (!style.rightdelim().empty()) {
2578                 os << style.rightdelim();
2579                 column += style.rightdelim().size();
2580         }
2581
2582         if (allowcust && d->endTeXParParams(bparams, os, runparams)
2583             && runparams.encoding != prev_encoding) {
2584                 runparams.encoding = prev_encoding;
2585                 os << setEncoding(prev_encoding->iconvName());
2586         }
2587
2588         LYXERR(Debug::LATEX, "Paragraph::latex... done " << this);
2589 }
2590
2591
2592 bool Paragraph::emptyTag() const
2593 {
2594         for (pos_type i = 0; i < size(); ++i) {
2595                 if (Inset const * inset = getInset(i)) {
2596                         InsetCode lyx_code = inset->lyxCode();
2597                         // FIXME testing like that is wrong. What is
2598                         // the intent?
2599                         if (lyx_code != TOC_CODE &&
2600                             lyx_code != INCLUDE_CODE &&
2601                             lyx_code != GRAPHICS_CODE &&
2602                             lyx_code != ERT_CODE &&
2603                             lyx_code != LISTINGS_CODE &&
2604                             lyx_code != FLOAT_CODE &&
2605                             lyx_code != TABULAR_CODE) {
2606                                 return false;
2607                         }
2608                 } else {
2609                         char_type c = d->text_[i];
2610                         if (c != ' ' && c != '\t')
2611                                 return false;
2612                 }
2613         }
2614         return true;
2615 }
2616
2617
2618 string Paragraph::getID(Buffer const & buf, OutputParams const & runparams)
2619         const
2620 {
2621         for (pos_type i = 0; i < size(); ++i) {
2622                 if (Inset const * inset = getInset(i)) {
2623                         InsetCode lyx_code = inset->lyxCode();
2624                         if (lyx_code == LABEL_CODE) {
2625                                 InsetLabel const * const il = static_cast<InsetLabel const *>(inset);
2626                                 docstring const & id = il->getParam("name");
2627                                 return "id='" + to_utf8(sgml::cleanID(buf, runparams, id)) + "'";
2628                         }
2629                 }
2630         }
2631         return string();
2632 }
2633
2634
2635 pos_type Paragraph::firstWordDocBook(odocstream & os, OutputParams const & runparams)
2636         const
2637 {
2638         pos_type i;
2639         for (i = 0; i < size(); ++i) {
2640                 if (Inset const * inset = getInset(i)) {
2641                         inset->docbook(os, runparams);
2642                 } else {
2643                         char_type c = d->text_[i];
2644                         if (c == ' ')
2645                                 break;
2646                         os << sgml::escapeChar(c);
2647                 }
2648         }
2649         return i;
2650 }
2651
2652
2653 pos_type Paragraph::firstWordLyXHTML(XHTMLStream & xs, OutputParams const & runparams)
2654         const
2655 {
2656         pos_type i;
2657         for (i = 0; i < size(); ++i) {
2658                 if (Inset const * inset = getInset(i)) {
2659                         inset->xhtml(xs, runparams);
2660                 } else {
2661                         char_type c = d->text_[i];
2662                         if (c == ' ')
2663                                 break;
2664                         xs << c;
2665                 }
2666         }
2667         return i;
2668 }
2669
2670
2671 bool Paragraph::Private::onlyText(Buffer const & buf, Font const & outerfont, pos_type initial) const
2672 {
2673         Font font_old;
2674         pos_type size = text_.size();
2675         for (pos_type i = initial; i < size; ++i) {
2676                 Font font = owner_->getFont(buf.params(), i, outerfont);
2677                 if (text_[i] == META_INSET)
2678                         return false;
2679                 if (i != initial && font != font_old)
2680                         return false;
2681                 font_old = font;
2682         }
2683
2684         return true;
2685 }
2686
2687
2688 void Paragraph::simpleDocBookOnePar(Buffer const & buf,
2689                                     odocstream & os,
2690                                     OutputParams const & runparams,
2691                                     Font const & outerfont,
2692                                     pos_type initial) const
2693 {
2694         bool emph_flag = false;
2695
2696         Layout const & style = *d->layout_;
2697         FontInfo font_old =
2698                 style.labeltype == LABEL_MANUAL ? style.labelfont : style.font;
2699
2700         if (style.pass_thru && !d->onlyText(buf, outerfont, initial))
2701                 os << "]]>";
2702
2703         // parsing main loop
2704         for (pos_type i = initial; i < size(); ++i) {
2705                 Font font = getFont(buf.params(), i, outerfont);
2706
2707                 // handle <emphasis> tag
2708                 if (font_old.emph() != font.fontInfo().emph()) {
2709                         if (font.fontInfo().emph() == FONT_ON) {
2710                                 os << "<emphasis>";
2711                                 emph_flag = true;
2712                         } else if (i != initial) {
2713                                 os << "</emphasis>";
2714                                 emph_flag = false;
2715                         }
2716                 }
2717
2718                 if (Inset const * inset = getInset(i)) {
2719                         inset->docbook(os, runparams);
2720                 } else {
2721                         char_type c = d->text_[i];
2722
2723                         if (style.pass_thru)
2724                                 os.put(c);
2725                         else
2726                                 os << sgml::escapeChar(c);
2727                 }
2728                 font_old = font.fontInfo();
2729         }
2730
2731         if (emph_flag) {
2732                 os << "</emphasis>";
2733         }
2734
2735         if (style.free_spacing)
2736                 os << '\n';
2737         if (style.pass_thru && !d->onlyText(buf, outerfont, initial))
2738                 os << "<![CDATA[";
2739 }
2740
2741
2742 namespace {
2743 void doFontSwitch(vector<html::FontTag> & tagsToOpen,
2744                   vector<html::EndFontTag> & tagsToClose,
2745                   bool & flag, FontState curstate, html::FontTypes type)
2746 {
2747         if (curstate == FONT_ON) {
2748                 tagsToOpen.push_back(html::FontTag(type));
2749                 flag = true;
2750         } else if (flag) {
2751                 tagsToClose.push_back(html::EndFontTag(type));
2752                 flag = false;
2753         }
2754 }
2755 }
2756
2757
2758 docstring Paragraph::simpleLyXHTMLOnePar(Buffer const & buf,
2759                                     XHTMLStream & xs,
2760                                     OutputParams const & runparams,
2761                                     Font const & outerfont,
2762                                     bool start_paragraph, bool close_paragraph,
2763                                     pos_type initial) const
2764 {
2765         docstring retval;
2766
2767         // track whether we have opened these tags
2768         bool emph_flag = false;
2769         bool bold_flag = false;
2770         bool noun_flag = false;
2771         bool ubar_flag = false;
2772         bool dbar_flag = false;
2773         bool sout_flag = false;
2774         bool wave_flag = false;
2775         // shape tags
2776         bool shap_flag = false;
2777         // family tags
2778         bool faml_flag = false;
2779         // size tags
2780         bool size_flag = false;
2781
2782         Layout const & style = *d->layout_;
2783
2784         if (start_paragraph)
2785                 xs.startDivision(allowEmpty());
2786
2787         FontInfo font_old =
2788                 style.labeltype == LABEL_MANUAL ? style.labelfont : style.font;
2789
2790         FontShape  curr_fs   = INHERIT_SHAPE;
2791         FontFamily curr_fam  = INHERIT_FAMILY;
2792         FontSize   curr_size = FONT_SIZE_INHERIT;
2793
2794         string const default_family =
2795                 buf.masterBuffer()->params().fonts_default_family;
2796
2797         vector<html::FontTag> tagsToOpen;
2798         vector<html::EndFontTag> tagsToClose;
2799
2800         // parsing main loop
2801         for (pos_type i = initial; i < size(); ++i) {
2802                 // let's not show deleted material in the output
2803                 if (isDeleted(i))
2804                         continue;
2805
2806                 Font const font = getFont(buf.masterBuffer()->params(), i, outerfont);
2807
2808                 // emphasis
2809                 FontState curstate = font.fontInfo().emph();
2810                 if (font_old.emph() != curstate)
2811                         doFontSwitch(tagsToOpen, tagsToClose, emph_flag, curstate, html::FT_EMPH);
2812
2813                 // noun
2814                 curstate = font.fontInfo().noun();
2815                 if (font_old.noun() != curstate)
2816                         doFontSwitch(tagsToOpen, tagsToClose, noun_flag, curstate, html::FT_NOUN);
2817
2818                 // underbar
2819                 curstate = font.fontInfo().underbar();
2820                 if (font_old.underbar() != curstate)
2821                         doFontSwitch(tagsToOpen, tagsToClose, ubar_flag, curstate, html::FT_UBAR);
2822
2823                 // strikeout
2824                 curstate = font.fontInfo().strikeout();
2825                 if (font_old.strikeout() != curstate)
2826                         doFontSwitch(tagsToOpen, tagsToClose, sout_flag, curstate, html::FT_SOUT);
2827
2828                 // double underbar
2829                 curstate = font.fontInfo().uuline();
2830                 if (font_old.uuline() != curstate)
2831                         doFontSwitch(tagsToOpen, tagsToClose, dbar_flag, curstate, html::FT_DBAR);
2832
2833                 // wavy line
2834                 curstate = font.fontInfo().uwave();
2835                 if (font_old.uwave() != curstate)
2836                         doFontSwitch(tagsToOpen, tagsToClose, wave_flag, curstate, html::FT_WAVE);
2837
2838                 // bold
2839                 // a little hackish, but allows us to reuse what we have.
2840                 curstate = (font.fontInfo().series() == BOLD_SERIES ? FONT_ON : FONT_OFF);
2841                 if (font_old.series() != font.fontInfo().series())
2842                         doFontSwitch(tagsToOpen, tagsToClose, bold_flag, curstate, html::FT_BOLD);
2843
2844                 // Font shape
2845                 curr_fs = font.fontInfo().shape();
2846                 FontShape old_fs = font_old.shape();
2847                 if (old_fs != curr_fs) {
2848                         if (shap_flag) {
2849                                 switch (old_fs) {
2850                                 case ITALIC_SHAPE:
2851                                         tagsToClose.push_back(html::EndFontTag(html::FT_ITALIC));
2852                                         break;
2853                                 case SLANTED_SHAPE:
2854                                         tagsToClose.push_back(html::EndFontTag(html::FT_SLANTED));
2855                                         break;
2856                                 case SMALLCAPS_SHAPE:
2857                                         tagsToClose.push_back(html::EndFontTag(html::FT_SMALLCAPS));
2858                                         break;
2859                                 case UP_SHAPE:
2860                                 case INHERIT_SHAPE:
2861                                         break;
2862                                 default:
2863                                         // the other tags are for internal use
2864                                         LATTEST(false);
2865                                         break;
2866                                 }
2867                                 shap_flag = false;
2868                         }
2869                         switch (curr_fs) {
2870                         case ITALIC_SHAPE:
2871                                 tagsToOpen.push_back(html::FontTag(html::FT_ITALIC));
2872                                 shap_flag = true;
2873                                 break;
2874                         case SLANTED_SHAPE:
2875                                 tagsToOpen.push_back(html::FontTag(html::FT_SLANTED));
2876                                 shap_flag = true;
2877                                 break;
2878                         case SMALLCAPS_SHAPE:
2879                                 tagsToOpen.push_back(html::FontTag(html::FT_SMALLCAPS));
2880                                 shap_flag = true;
2881                                 break;
2882                         case UP_SHAPE:
2883                         case INHERIT_SHAPE:
2884                                 break;
2885                         default:
2886                                 // the other tags are for internal use
2887                                 LATTEST(false);
2888                                 break;
2889                         }
2890                 }
2891
2892                 // Font family
2893                 curr_fam = font.fontInfo().family();
2894                 FontFamily old_fam = font_old.family();
2895                 if (old_fam != curr_fam) {
2896                         if (faml_flag) {
2897                                 switch (old_fam) {
2898                                 case ROMAN_FAMILY:
2899                                         tagsToClose.push_back(html::EndFontTag(html::FT_ROMAN));
2900                                         break;
2901                                 case SANS_FAMILY:
2902                                         tagsToClose.push_back(html::EndFontTag(html::FT_SANS));
2903                                         break;
2904                                 case TYPEWRITER_FAMILY:
2905                                         tagsToClose.push_back(html::EndFontTag(html::FT_TYPE));
2906                                         break;
2907                                 case INHERIT_FAMILY:
2908                                         break;
2909                                 default:
2910                                         // the other tags are for internal use
2911                                         LATTEST(false);
2912                                         break;
2913                                 }
2914                                 faml_flag = false;
2915                         }
2916                         switch (curr_fam) {
2917                         case ROMAN_FAMILY:
2918                                 // we will treat a "default" font family as roman, since we have
2919                                 // no other idea what to do.
2920                                 if (default_family != "rmdefault" && default_family != "default") {
2921                                         tagsToOpen.push_back(html::FontTag(html::FT_ROMAN));
2922                                         faml_flag = true;
2923                                 }
2924                                 break;
2925                         case SANS_FAMILY:
2926                                 if (default_family != "sfdefault") {
2927                                         tagsToOpen.push_back(html::FontTag(html::FT_SANS));
2928                                         faml_flag = true;
2929                                 }
2930                                 break;
2931                         case TYPEWRITER_FAMILY:
2932                                 if (default_family != "ttdefault") {
2933                                         tagsToOpen.push_back(html::FontTag(html::FT_TYPE));
2934                                         faml_flag = true;
2935                                 }
2936                                 break;
2937                         case INHERIT_FAMILY:
2938                                 break;
2939                         default:
2940                                 // the other tags are for internal use
2941                                 LATTEST(false);
2942                                 break;
2943                         }
2944                 }
2945
2946                 // Font size
2947                 curr_size = font.fontInfo().size();
2948                 FontSize old_size = font_old.size();
2949                 if (old_size != curr_size) {
2950                         if (size_flag) {
2951                                 switch (old_size) {
2952                                 case FONT_SIZE_TINY:
2953                                         tagsToClose.push_back(html::EndFontTag(html::FT_SIZE_TINY));
2954                                         break;
2955                                 case FONT_SIZE_SCRIPT:
2956                                         tagsToClose.push_back(html::EndFontTag(html::FT_SIZE_SCRIPT));
2957                                         break;
2958                                 case FONT_SIZE_FOOTNOTE:
2959                                         tagsToClose.push_back(html::EndFontTag(html::FT_SIZE_FOOTNOTE));
2960                                         break;
2961                                 case FONT_SIZE_SMALL:
2962                                         tagsToClose.push_back(html::EndFontTag(html::FT_SIZE_SMALL));
2963                                         break;
2964                                 case FONT_SIZE_LARGE:
2965                                         tagsToClose.push_back(html::EndFontTag(html::FT_SIZE_LARGE));
2966                                         break;
2967                                 case FONT_SIZE_LARGER:
2968                                         tagsToClose.push_back(html::EndFontTag(html::FT_SIZE_LARGER));
2969                                         break;
2970                                 case FONT_SIZE_LARGEST:
2971                                         tagsToClose.push_back(html::EndFontTag(html::FT_SIZE_LARGEST));
2972                                         break;
2973                                 case FONT_SIZE_HUGE:
2974                                         tagsToClose.push_back(html::EndFontTag(html::FT_SIZE_HUGE));
2975                                         break;
2976                                 case FONT_SIZE_HUGER:
2977                                         tagsToClose.push_back(html::EndFontTag(html::FT_SIZE_HUGER));
2978                                         break;
2979                                 case FONT_SIZE_INCREASE:
2980                                         tagsToClose.push_back(html::EndFontTag(html::FT_SIZE_INCREASE));
2981                                         break;
2982                                 case FONT_SIZE_DECREASE:
2983                                         tagsToClose.push_back(html::EndFontTag(html::FT_SIZE_DECREASE));
2984                                         break;
2985                                 case FONT_SIZE_INHERIT:
2986                                 case FONT_SIZE_NORMAL:
2987                                         break;
2988                                 default:
2989                                         // the other tags are for internal use
2990                                         LATTEST(false);
2991                                         break;
2992                                 }
2993                                 size_flag = false;
2994                         }
2995                         switch (curr_size) {
2996                         case FONT_SIZE_TINY:
2997                                 tagsToOpen.push_back(html::FontTag(html::FT_SIZE_TINY));
2998                                 size_flag = true;
2999                                 break;
3000                         case FONT_SIZE_SCRIPT:
3001                                 tagsToOpen.push_back(html::FontTag(html::FT_SIZE_SCRIPT));
3002                                 size_flag = true;
3003                                 break;
3004                         case FONT_SIZE_FOOTNOTE:
3005                                 tagsToOpen.push_back(html::FontTag(html::FT_SIZE_FOOTNOTE));
3006                                 size_flag = true;
3007                                 break;
3008                         case FONT_SIZE_SMALL:
3009                                 tagsToOpen.push_back(html::FontTag(html::FT_SIZE_SMALL));
3010                                 size_flag = true;
3011                                 break;
3012                         case FONT_SIZE_LARGE:
3013                                 tagsToOpen.push_back(html::FontTag(html::FT_SIZE_LARGE));
3014                                 size_flag = true;
3015                                 break;
3016                         case FONT_SIZE_LARGER:
3017                                 tagsToOpen.push_back(html::FontTag(html::FT_SIZE_LARGER));
3018                                 size_flag = true;
3019                                 break;
3020                         case FONT_SIZE_LARGEST:
3021                                 tagsToOpen.push_back(html::FontTag(html::FT_SIZE_LARGEST));
3022                                 size_flag = true;
3023                                 break;
3024                         case FONT_SIZE_HUGE:
3025                                 tagsToOpen.push_back(html::FontTag(html::FT_SIZE_HUGE));
3026                                 size_flag = true;
3027                                 break;
3028                         case FONT_SIZE_HUGER:
3029                                 tagsToOpen.push_back(html::FontTag(html::FT_SIZE_HUGER));
3030                                 size_flag = true;
3031                                 break;
3032                         case FONT_SIZE_INCREASE:
3033                                 tagsToOpen.push_back(html::FontTag(html::FT_SIZE_INCREASE));
3034                                 size_flag = true;
3035                                 break;
3036                         case FONT_SIZE_DECREASE:
3037                                 tagsToOpen.push_back(html::FontTag(html::FT_SIZE_DECREASE));
3038                                 size_flag = true;
3039                                 break;
3040                         case FONT_SIZE_NORMAL:
3041                         case FONT_SIZE_INHERIT:
3042                                 break;
3043                         default:
3044                                 // the other tags are for internal use
3045                                 LATTEST(false);
3046                                 break;
3047                         }
3048                 }
3049
3050                 // FIXME XHTML
3051                 // Other such tags? What about the other text ranges?
3052
3053                 vector<html::EndFontTag>::const_iterator cit = tagsToClose.begin();
3054                 vector<html::EndFontTag>::const_iterator cen = tagsToClose.end();
3055                 for (; cit != cen; ++cit)
3056                         xs << *cit;
3057
3058                 vector<html::FontTag>::const_iterator sit = tagsToOpen.begin();
3059                 vector<html::FontTag>::const_iterator sen = tagsToOpen.end();
3060                 for (; sit != sen; ++sit)
3061                         xs << *sit;
3062
3063                 tagsToClose.clear();
3064                 tagsToOpen.clear();
3065
3066                 Inset const * inset = getInset(i);
3067                 if (inset) {
3068                         if (!runparams.for_toc || inset->isInToc()) {
3069                                 OutputParams np = runparams;
3070                                 np.local_font = &font;
3071                                 // If the paragraph has size 1, then we are in the "special
3072                                 // case" where we do not output the containing paragraph info
3073                                 if (!inset->getLayout().htmlisblock() && size() != 1)
3074                                         np.html_in_par = true;
3075                                 retval += inset->xhtml(xs, np);
3076                         }
3077                 } else {
3078                         char_type c = getUChar(buf.masterBuffer()->params(), i);
3079                         xs << c;
3080                 }
3081                 font_old = font.fontInfo();
3082         }
3083
3084         // FIXME XHTML
3085         // I'm worried about what happens if a branch, say, is itself
3086         // wrapped in some font stuff. I think that will not work.
3087         xs.closeFontTags();
3088         if (close_paragraph)
3089                 xs.endDivision();
3090
3091         return retval;
3092 }
3093
3094
3095 bool Paragraph::isHfill(pos_type pos) const
3096 {
3097         Inset const * inset = getInset(pos);
3098         return inset && inset->isHfill();
3099 }
3100
3101
3102 bool Paragraph::isNewline(pos_type pos) const
3103 {
3104         Inset const * inset = getInset(pos);
3105         return inset && inset->lyxCode() == NEWLINE_CODE;
3106 }
3107
3108
3109 bool Paragraph::isEnvSeparator(pos_type pos) const
3110 {
3111         Inset const * inset = getInset(pos);
3112         return inset && inset->lyxCode() == SEPARATOR_CODE;
3113 }
3114
3115
3116 bool Paragraph::isLineSeparator(pos_type pos) const
3117 {
3118         char_type const c = d->text_[pos];
3119         if (isLineSeparatorChar(c))
3120                 return true;
3121         Inset const * inset = getInset(pos);
3122         return inset && inset->isLineSeparator();
3123 }
3124
3125
3126 bool Paragraph::isWordSeparator(pos_type pos) const
3127 {
3128         if (pos == size())
3129                 return true;
3130         if (Inset const * inset = getInset(pos))
3131                 return !inset->isLetter();
3132         // if we have a hard hyphen (no en- or emdash) or apostrophe
3133         // we pass this to the spell checker
3134         // FIXME: this method is subject to change, visit
3135         // https://bugzilla.mozilla.org/show_bug.cgi?id=355178
3136         // to get an impression how complex this is.
3137         if (isHardHyphenOrApostrophe(pos))
3138                 return false;
3139         char_type const c = d->text_[pos];
3140         // We want to pass the escape chars to the spellchecker
3141         docstring const escape_chars = from_utf8(lyxrc.spellchecker_esc_chars);
3142         return !isLetterChar(c) && !isDigitASCII(c) && !contains(escape_chars, c);
3143 }
3144
3145
3146 bool Paragraph::isHardHyphenOrApostrophe(pos_type pos) const
3147 {
3148         pos_type const psize = size();
3149         if (pos >= psize)
3150                 return false;
3151         char_type const c = d->text_[pos];
3152         if (c != '-' && c != '\'')
3153                 return false;
3154         int nextpos = pos + 1;
3155         int prevpos = pos > 0 ? pos - 1 : 0;
3156         if ((nextpos == psize || isSpace(nextpos))
3157                 && (pos == 0 || isSpace(prevpos)))
3158                 return false;
3159         return true;
3160 }
3161
3162
3163 FontSpan const & Paragraph::getSpellRange(pos_type pos) const
3164 {
3165         return d->speller_state_.getRange(pos);
3166 }
3167
3168
3169 bool Paragraph::isChar(pos_type pos) const
3170 {
3171         if (Inset const * inset = getInset(pos))
3172                 return inset->isChar();
3173         char_type const c = d->text_[pos];
3174         return !isLetterChar(c) && !isDigitASCII(c) && !lyx::isSpace(c);
3175 }
3176
3177
3178 bool Paragraph::isSpace(pos_type pos) const
3179 {
3180         if (Inset const * inset = getInset(pos))
3181                 return inset->isSpace();
3182         char_type const c = d->text_[pos];
3183         return lyx::isSpace(c);
3184 }
3185
3186
3187 Language const *
3188 Paragraph::getParLanguage(BufferParams const & bparams) const
3189 {
3190         if (!empty())
3191                 return getFirstFontSettings(bparams).language();
3192         // FIXME: we should check the prev par as well (Lgb)
3193         return bparams.language;
3194 }
3195
3196
3197 bool Paragraph::isRTL(BufferParams const & bparams) const
3198 {
3199         return getParLanguage(bparams)->rightToLeft()
3200                 && !inInset().getLayout().forceLTR();
3201 }
3202
3203
3204 void Paragraph::changeLanguage(BufferParams const & bparams,
3205                                Language const * from, Language const * to)
3206 {
3207         // change language including dummy font change at the end
3208         for (pos_type i = 0; i <= size(); ++i) {
3209                 Font font = getFontSettings(bparams, i);
3210                 if (font.language() == from) {
3211                         font.setLanguage(to);
3212                         setFont(i, font);
3213                         d->requestSpellCheck(i);
3214                 }
3215         }
3216 }
3217
3218
3219 bool Paragraph::isMultiLingual(BufferParams const & bparams) const
3220 {
3221         Language const * doc_language = bparams.language;
3222         FontList::const_iterator cit = d->fontlist_.begin();
3223         FontList::const_iterator end = d->fontlist_.end();
3224
3225         for (; cit != end; ++cit)
3226                 if (cit->font().language() != ignore_language &&
3227                     cit->font().language() != latex_language &&
3228                     cit->font().language() != doc_language)
3229                         return true;
3230         return false;
3231 }
3232
3233
3234 void Paragraph::getLanguages(std::set<Language const *> & languages) const
3235 {
3236         FontList::const_iterator cit = d->fontlist_.begin();
3237         FontList::const_iterator end = d->fontlist_.end();
3238
3239         for (; cit != end; ++cit) {
3240                 Language const * lang = cit->font().language();
3241                 if (lang != ignore_language &&
3242                     lang != latex_language)
3243                         languages.insert(lang);
3244         }
3245 }
3246
3247
3248 docstring Paragraph::asString(int options) const
3249 {
3250         return asString(0, size(), options);
3251 }
3252
3253
3254 docstring Paragraph::asString(pos_type beg, pos_type end, int options, const OutputParams *runparams) const
3255 {
3256         odocstringstream os;
3257
3258         if (beg == 0
3259             && options & AS_STR_LABEL
3260             && !d->params_.labelString().empty())
3261                 os << d->params_.labelString() << ' ';
3262
3263         for (pos_type i = beg; i < end; ++i) {
3264                 if ((options & AS_STR_SKIPDELETE) && isDeleted(i))
3265                         continue;
3266                 char_type const c = d->text_[i];
3267                 if (isPrintable(c) || c == '\t'
3268                     || (c == '\n' && (options & AS_STR_NEWLINES)))
3269                         os.put(c);
3270                 else if (c == META_INSET && (options & AS_STR_INSETS)) {
3271                         if (c == META_INSET && (options & AS_STR_PLAINTEXT)) {
3272                                 LASSERT(runparams != 0, return docstring());
3273                                 getInset(i)->plaintext(os, *runparams);
3274                         } else {
3275                                 getInset(i)->toString(os);
3276                         }
3277                 }
3278         }
3279
3280         return os.str();
3281 }
3282
3283
3284 void Paragraph::forOutliner(docstring & os, size_t const maxlen,
3285                                                         bool const shorten) const
3286 {
3287         size_t tmplen = shorten ? maxlen + 1 : maxlen;
3288         if (!d->params_.labelString().empty())
3289                 os += d->params_.labelString() + ' ';
3290         for (pos_type i = 0; i < size() && os.length() < tmplen; ++i) {
3291                 if (isDeleted(i))
3292                         continue;
3293                 char_type const c = d->text_[i];
3294                 if (isPrintable(c))
3295                         os += c;
3296                 else if (c == META_INSET)
3297                         getInset(i)->forOutliner(os, tmplen, false);
3298         }
3299         if (shorten)
3300                 Text::shortenForOutliner(os, maxlen);
3301 }
3302
3303
3304 void Paragraph::setInsetOwner(Inset const * inset)
3305 {
3306         d->inset_owner_ = inset;
3307 }
3308
3309
3310 int Paragraph::id() const
3311 {
3312         return d->id_;
3313 }
3314
3315
3316 void Paragraph::setId(int id)
3317 {
3318         d->id_ = id;
3319 }
3320
3321
3322 Layout const & Paragraph::layout() const
3323 {
3324         return *d->layout_;
3325 }
3326
3327
3328 void Paragraph::setLayout(Layout const & layout)
3329 {
3330         d->layout_ = &layout;
3331 }
3332
3333
3334 void Paragraph::setDefaultLayout(DocumentClass const & tc)
3335 {
3336         setLayout(tc.defaultLayout());
3337 }
3338
3339
3340 void Paragraph::setPlainLayout(DocumentClass const & tc)
3341 {
3342         setLayout(tc.plainLayout());
3343 }
3344
3345
3346 void Paragraph::setPlainOrDefaultLayout(DocumentClass const & tclass)
3347 {
3348         if (usePlainLayout())
3349                 setPlainLayout(tclass);
3350         else
3351                 setDefaultLayout(tclass);
3352 }
3353
3354
3355 Inset const & Paragraph::inInset() const
3356 {
3357         LBUFERR(d->inset_owner_);
3358         return *d->inset_owner_;
3359 }
3360
3361
3362 ParagraphParameters & Paragraph::params()
3363 {
3364         return d->params_;
3365 }
3366
3367
3368 ParagraphParameters const & Paragraph::params() const
3369 {
3370         return d->params_;
3371 }
3372
3373
3374 bool Paragraph::isFreeSpacing() const
3375 {
3376         if (d->layout_->free_spacing)
3377                 return true;
3378         return d->inset_owner_ && d->inset_owner_->isFreeSpacing();
3379 }
3380
3381
3382 bool Paragraph::allowEmpty() const
3383 {
3384         if (d->layout_->keepempty)
3385                 return true;
3386         return d->inset_owner_ && d->inset_owner_->allowEmpty();
3387 }
3388
3389
3390 bool Paragraph::brokenBiblio() const
3391 {
3392         // there is a problem if there is no bibitem at position 0 or
3393         // if there is another bibitem in the paragraph.
3394         return d->layout_->labeltype == LABEL_BIBLIO
3395                 && (d->insetlist_.find(BIBITEM_CODE) != 0
3396                     || d->insetlist_.find(BIBITEM_CODE, 1) > 0);
3397 }
3398
3399
3400 int Paragraph::fixBiblio(Buffer const & buffer)
3401 {
3402         // FIXME: What about the case where paragraph is not BIBLIO
3403         // but there is an InsetBibitem?
3404         // FIXME: when there was already an inset at 0, the return value is 1,
3405         // which does not tell whether another inset has been remove; the
3406         // cursor cannot be correctly updated.
3407
3408         if (d->layout_->labeltype != LABEL_BIBLIO)
3409                 return 0;
3410
3411         bool const track_changes = buffer.params().track_changes;
3412         int bibitem_pos = d->insetlist_.find(BIBITEM_CODE);
3413         bool const hasbibitem0 = bibitem_pos == 0;
3414
3415         if (hasbibitem0) {
3416                 bibitem_pos = d->insetlist_.find(BIBITEM_CODE, 1);
3417                 // There was an InsetBibitem at pos 0, and no other one => OK
3418                 if (bibitem_pos == -1)
3419                         return 0;
3420                 // there is a bibitem at the 0 position, but since
3421                 // there is a second one, we copy the second on the
3422                 // first. We're assuming there are at most two of
3423                 // these, which there should be.
3424                 // FIXME: why does it make sense to do that rather
3425                 // than keep the first? (JMarc)
3426                 Inset * inset = releaseInset(bibitem_pos);
3427                 d->insetlist_.begin()->inset = inset;
3428                 return -bibitem_pos;
3429         }
3430
3431         // We need to create an inset at the beginning
3432         Inset * inset = 0;
3433         if (bibitem_pos > 0) {
3434                 // there was one somewhere in the paragraph, let's move it
3435                 inset = d->insetlist_.release(bibitem_pos);
3436                 eraseChar(bibitem_pos, track_changes);
3437         } else
3438                 // make a fresh one
3439                 inset = new InsetBibitem(const_cast<Buffer *>(&buffer),
3440                                          InsetCommandParams(BIBITEM_CODE));
3441
3442         Font font(inherit_font, buffer.params().language);
3443         insertInset(0, inset, font, Change(track_changes ? Change::INSERTED
3444                                                    : Change::UNCHANGED));
3445
3446         return 1;
3447 }
3448
3449
3450 void Paragraph::checkAuthors(AuthorList const & authorList)
3451 {
3452         d->changes_.checkAuthors(authorList);
3453 }
3454
3455
3456 bool Paragraph::isChanged(pos_type pos) const
3457 {
3458         return lookupChange(pos).changed();
3459 }
3460
3461
3462 bool Paragraph::isInserted(pos_type pos) const
3463 {
3464         return lookupChange(pos).inserted();
3465 }
3466
3467
3468 bool Paragraph::isDeleted(pos_type pos) const
3469 {
3470         return lookupChange(pos).deleted();
3471 }
3472
3473
3474 InsetList const & Paragraph::insetList() const
3475 {
3476         return d->insetlist_;
3477 }
3478
3479
3480 void Paragraph::setBuffer(Buffer & b)
3481 {
3482         d->insetlist_.setBuffer(b);
3483 }
3484
3485
3486 Inset * Paragraph::releaseInset(pos_type pos)
3487 {
3488         Inset * inset = d->insetlist_.release(pos);
3489         /// does not honour change tracking!
3490         eraseChar(pos, false);
3491         return inset;
3492 }
3493
3494
3495 Inset * Paragraph::getInset(pos_type pos)
3496 {
3497         return (pos < pos_type(d->text_.size()) && d->text_[pos] == META_INSET)
3498                  ? d->insetlist_.get(pos) : 0;
3499 }
3500
3501
3502 Inset const * Paragraph::getInset(pos_type pos) const
3503 {
3504         return (pos < pos_type(d->text_.size()) && d->text_[pos] == META_INSET)
3505                  ? d->insetlist_.get(pos) : 0;
3506 }
3507
3508
3509 void Paragraph::changeCase(BufferParams const & bparams, pos_type pos,
3510                 pos_type & right, TextCase action)
3511 {
3512         // process sequences of modified characters; in change
3513         // tracking mode, this approach results in much better
3514         // usability than changing case on a char-by-char basis
3515         // We also need to track the current font, since font
3516         // changes within sequences can occur.
3517         vector<pair<char_type, Font> > changes;
3518
3519         bool const trackChanges = bparams.track_changes;
3520
3521         bool capitalize = true;
3522
3523         for (; pos < right; ++pos) {
3524                 char_type oldChar = d->text_[pos];
3525                 char_type newChar = oldChar;
3526
3527                 // ignore insets and don't play with deleted text!
3528                 if (oldChar != META_INSET && !isDeleted(pos)) {
3529                         switch (action) {
3530                                 case text_lowercase:
3531                                         newChar = lowercase(oldChar);
3532                                         break;
3533                                 case text_capitalization:
3534                                         if (capitalize) {
3535                                                 newChar = uppercase(oldChar);
3536                                                 capitalize = false;
3537                                         }
3538                                         break;
3539                                 case text_uppercase:
3540                                         newChar = uppercase(oldChar);
3541                                         break;
3542                         }
3543                 }
3544
3545                 if (isWordSeparator(pos) || isDeleted(pos)) {
3546                         // permit capitalization again
3547                         capitalize = true;
3548                 }
3549
3550                 if (oldChar != newChar) {
3551                         changes.push_back(make_pair(newChar, getFontSettings(bparams, pos)));
3552                         if (pos != right - 1)
3553                                 continue;
3554                         // step behind the changing area
3555                         pos++;
3556                 }
3557
3558                 int erasePos = pos - changes.size();
3559                 for (size_t i = 0; i < changes.size(); i++) {
3560                         insertChar(pos, changes[i].first,
3561                                    changes[i].second,
3562                                    trackChanges);
3563                         if (!eraseChar(erasePos, trackChanges)) {
3564                                 ++erasePos;
3565                                 ++pos; // advance
3566                                 ++right; // expand selection
3567                         }
3568                 }
3569                 changes.clear();
3570         }
3571 }
3572
3573
3574 int Paragraph::find(docstring const & str, bool cs, bool mw,
3575                 pos_type start_pos, bool del) const
3576 {
3577         pos_type pos = start_pos;
3578         int const strsize = str.length();
3579         int i = 0;
3580         pos_type const parsize = d->text_.size();
3581         for (i = 0; i < strsize && pos < parsize; ++i, ++pos) {
3582                 // Ignore "invisible" letters such as ligature breaks
3583                 // and hyphenation chars while searching
3584                 while (pos < parsize - 1 && isInset(pos)) {
3585                         odocstringstream os;
3586                         getInset(pos)->toString(os);
3587                         if (!getInset(pos)->isLetter() || !os.str().empty())
3588                                 break;
3589                         pos++;
3590                 }
3591                 if (cs && str[i] != d->text_[pos])
3592                         break;
3593                 if (!cs && uppercase(str[i]) != uppercase(d->text_[pos]))
3594                         break;
3595                 if (!del && isDeleted(pos))
3596                         break;
3597         }
3598
3599         if (i != strsize)
3600                 return 0;
3601
3602         // if necessary, check whether string matches word
3603         if (mw) {
3604                 if (start_pos > 0 && !isWordSeparator(start_pos - 1))
3605                         return 0;
3606                 if (pos < parsize
3607                         && !isWordSeparator(pos))
3608                         return 0;
3609         }
3610
3611         return pos - start_pos;
3612 }
3613
3614
3615 char_type Paragraph::getChar(pos_type pos) const
3616 {
3617         return d->text_[pos];
3618 }
3619
3620
3621 pos_type Paragraph::size() const
3622 {
3623         return d->text_.size();
3624 }
3625
3626
3627 bool Paragraph::empty() const
3628 {
3629         return d->text_.empty();
3630 }
3631
3632
3633 bool Paragraph::isInset(pos_type pos) const
3634 {
3635         return d->text_[pos] == META_INSET;
3636 }
3637
3638
3639 bool Paragraph::isSeparator(pos_type pos) const
3640 {
3641         //FIXME: Are we sure this can be the only separator?
3642         return d->text_[pos] == ' ';
3643 }
3644
3645
3646 void Paragraph::deregisterWords()
3647 {
3648         Private::LangWordsMap::const_iterator itl = d->words_.begin();
3649         Private::LangWordsMap::const_iterator ite = d->words_.end();
3650         for (; itl != ite; ++itl) {
3651                 WordList * wl = theWordList(itl->first);
3652                 Private::Words::const_iterator it = (itl->second).begin();
3653                 Private::Words::const_iterator et = (itl->second).end();
3654                 for (; it != et; ++it)
3655                         wl->remove(*it);
3656         }
3657         d->words_.clear();
3658 }
3659
3660
3661 void Paragraph::locateWord(pos_type & from, pos_type & to,
3662         word_location const loc) const
3663 {
3664         switch (loc) {
3665         case WHOLE_WORD_STRICT:
3666                 if (from == 0 || from == size()
3667                     || isWordSeparator(from)
3668                     || isWordSeparator(from - 1)) {
3669                         to = from;
3670                         return;
3671                 }
3672                 // no break here, we go to the next
3673
3674         case WHOLE_WORD:
3675                 // If we are already at the beginning of a word, do nothing
3676                 if (!from || isWordSeparator(from - 1))
3677                         break;
3678                 // no break here, we go to the next
3679
3680         case PREVIOUS_WORD:
3681                 // always move the cursor to the beginning of previous word
3682                 while (from && !isWordSeparator(from - 1))
3683                         --from;
3684                 break;
3685         case NEXT_WORD:
3686                 LYXERR0("Paragraph::locateWord: NEXT_WORD not implemented yet");
3687                 break;
3688         case PARTIAL_WORD:
3689                 // no need to move the 'from' cursor
3690                 break;
3691         }
3692         to = from;
3693         while (to < size() && !isWordSeparator(to))
3694                 ++to;
3695 }
3696
3697
3698 void Paragraph::collectWords()
3699 {
3700         for (pos_type pos = 0; pos < size(); ++pos) {
3701                 if (isWordSeparator(pos))
3702                         continue;
3703                 pos_type from = pos;
3704                 locateWord(from, pos, WHOLE_WORD);
3705                 // Work around MSVC warning: The statement
3706                 // if (pos < from + lyxrc.completion_minlength)
3707                 // triggers a signed vs. unsigned warning.
3708                 // I don't know why this happens, it could be a MSVC bug, or
3709                 // related to LLP64 (windows) vs. LP64 (unix) programming
3710                 // model, or the C++ standard might be ambigous in the section
3711                 // defining the "usual arithmetic conversions". However, using
3712                 // a temporary variable is safe and works on all compilers.
3713                 pos_type const endpos = from + lyxrc.completion_minlength;
3714                 if (pos < endpos)
3715                         continue;
3716                 FontList::const_iterator cit = d->fontlist_.fontIterator(from);
3717                 if (cit == d->fontlist_.end())
3718                         return;
3719                 Language const * lang = cit->font().language();
3720                 docstring const word = asString(from, pos, AS_STR_NONE);
3721                 d->words_[lang->lang()].insert(word);
3722         }
3723 }
3724
3725
3726 void Paragraph::registerWords()
3727 {
3728         Private::LangWordsMap::const_iterator itl = d->words_.begin();
3729         Private::LangWordsMap::const_iterator ite = d->words_.end();
3730         for (; itl != ite; ++itl) {
3731                 WordList * wl = theWordList(itl->first);
3732                 Private::Words::const_iterator it = (itl->second).begin();
3733                 Private::Words::const_iterator et = (itl->second).end();
3734                 for (; it != et; ++it)
3735                         wl->insert(*it);
3736         }
3737 }
3738
3739
3740 void Paragraph::updateWords()
3741 {
3742         deregisterWords();
3743         collectWords();
3744         registerWords();
3745 }
3746
3747
3748 void Paragraph::Private::appendSkipPosition(SkipPositions & skips, pos_type const pos) const
3749 {
3750         SkipPositionsIterator begin = skips.begin();
3751         SkipPositions::iterator end = skips.end();
3752         if (pos > 0 && begin < end) {
3753                 --end;
3754                 if (end->last == pos - 1) {
3755                         end->last = pos;
3756                         return;
3757                 }
3758         }
3759         skips.insert(end, FontSpan(pos, pos));
3760 }
3761
3762
3763 Language * Paragraph::Private::locateSpellRange(
3764         pos_type & from, pos_type & to,
3765         SkipPositions & skips) const
3766 {
3767         // skip leading white space
3768         while (from < to && owner_->isWordSeparator(from))
3769                 ++from;
3770         // don't check empty range
3771         if (from >= to)
3772                 return 0;
3773         // get current language
3774         Language * lang = getSpellLanguage(from);
3775         pos_type last = from;
3776         bool samelang = true;
3777         bool sameinset = true;
3778         while (last < to && samelang && sameinset) {
3779                 // hop to end of word
3780                 while (last < to && !owner_->isWordSeparator(last)) {
3781                         if (owner_->getInset(last)) {
3782                                 appendSkipPosition(skips, last);
3783                         } else if (owner_->isDeleted(last)) {
3784                                 appendSkipPosition(skips, last);
3785                         }
3786                         ++last;
3787                 }
3788                 // hop to next word while checking for insets
3789                 while (sameinset && last < to && owner_->isWordSeparator(last)) {
3790                         if (Inset const * inset = owner_->getInset(last))
3791                                 sameinset = inset->isChar() && inset->isLetter();
3792                         if (sameinset && owner_->isDeleted(last)) {
3793                                 appendSkipPosition(skips, last);
3794                         }
3795                         if (sameinset)
3796                                 last++;
3797                 }
3798                 if (sameinset && last < to) {
3799                         // now check for language change
3800                         samelang = lang == getSpellLanguage(last);
3801                 }
3802         }
3803         // if language change detected backstep is needed
3804         if (!samelang)
3805                 --last;
3806         to = last;
3807         return lang;
3808 }
3809
3810
3811 Language * Paragraph::Private::getSpellLanguage(pos_type const from) const
3812 {
3813         Language * lang =
3814                 const_cast<Language *>(owner_->getFontSettings(
3815                         inset_owner_->buffer().params(), from).language());
3816         if (lang == inset_owner_->buffer().params().language
3817                 && !lyxrc.spellchecker_alt_lang.empty()) {
3818                 string lang_code;
3819                 string const lang_variety =
3820                         split(lyxrc.spellchecker_alt_lang, lang_code, '-');
3821                 lang->setCode(lang_code);
3822                 lang->setVariety(lang_variety);
3823         }
3824         return lang;
3825 }
3826
3827
3828 void Paragraph::requestSpellCheck(pos_type pos)
3829 {
3830         d->requestSpellCheck(pos);
3831 }
3832
3833
3834 bool Paragraph::needsSpellCheck() const
3835 {
3836         SpellChecker::ChangeNumber speller_change_number = 0;
3837         if (theSpellChecker())
3838                 speller_change_number = theSpellChecker()->changeNumber();
3839         if (speller_change_number > d->speller_state_.currentChangeNumber()) {
3840                 d->speller_state_.needsCompleteRefresh(speller_change_number);
3841         }
3842         return d->needsSpellCheck();
3843 }
3844
3845
3846 bool Paragraph::Private::ignoreWord(docstring const & word) const
3847 {
3848         // Ignore words with digits
3849         // FIXME: make this customizable
3850         // (note that some checkers ignore words with digits by default)
3851         docstring::const_iterator cit = word.begin();
3852         docstring::const_iterator const end = word.end();
3853         for (; cit != end; ++cit) {
3854                 if (isNumber((*cit)))
3855                         return true;
3856         }
3857         return false;
3858 }
3859
3860
3861 SpellChecker::Result Paragraph::spellCheck(pos_type & from, pos_type & to,
3862         WordLangTuple & wl, docstring_list & suggestions,
3863         bool do_suggestion, bool check_learned) const
3864 {
3865         SpellChecker::Result result = SpellChecker::WORD_OK;
3866         SpellChecker * speller = theSpellChecker();
3867         if (!speller)
3868                 return result;
3869
3870         if (!d->layout_->spellcheck || !inInset().allowSpellCheck())
3871                 return result;
3872
3873         locateWord(from, to, WHOLE_WORD);
3874         if (from == to || from >= size())
3875                 return result;
3876
3877         docstring word = asString(from, to, AS_STR_INSETS | AS_STR_SKIPDELETE);
3878         Language * lang = d->getSpellLanguage(from);
3879
3880         wl = WordLangTuple(word, lang);
3881
3882         if (word.empty())
3883                 return result;
3884
3885         if (needsSpellCheck() || check_learned) {
3886                 pos_type end = to;
3887                 if (!d->ignoreWord(word)) {
3888                         bool const trailing_dot = to < size() && d->text_[to] == '.';
3889                         result = speller->check(wl);
3890                         if (SpellChecker::misspelled(result) && trailing_dot) {
3891                                 wl = WordLangTuple(word.append(from_ascii(".")), lang);
3892                                 result = speller->check(wl);
3893                                 if (!SpellChecker::misspelled(result)) {
3894                                         LYXERR(Debug::GUI, "misspelled word is correct with dot: \"" <<
3895                                            word << "\" [" <<
3896                                            from << ".." << to << "]");
3897                                 } else {
3898                                         // spell check with dot appended failed too
3899                                         // restore original word/lang value
3900                                         word = asString(from, to, AS_STR_INSETS | AS_STR_SKIPDELETE);
3901                                         wl = WordLangTuple(word, lang);
3902                                 }
3903                         }
3904                 }
3905                 if (!SpellChecker::misspelled(result)) {
3906                         // area up to the begin of the next word is not misspelled
3907                         while (end < size() && isWordSeparator(end))
3908                                 ++end;
3909                 }
3910                 d->setMisspelled(from, end, result);
3911         } else {
3912                 result = d->speller_state_.getState(from);
3913         }
3914
3915         if (do_suggestion)
3916                 suggestions.clear();
3917
3918         if (SpellChecker::misspelled(result)) {
3919                 LYXERR(Debug::GUI, "misspelled word: \"" <<
3920                            word << "\" [" <<
3921                            from << ".." << to << "]");
3922                 if (do_suggestion)
3923                         speller->suggest(wl, suggestions);
3924         }
3925         return result;
3926 }
3927
3928
3929 void Paragraph::Private::markMisspelledWords(
3930         pos_type const & first, pos_type const & last,
3931         SpellChecker::Result result,
3932         docstring const & word,
3933         SkipPositions const & skips)
3934 {
3935         if (!SpellChecker::misspelled(result)) {
3936                 setMisspelled(first, last, SpellChecker::WORD_OK);
3937                 return;
3938         }
3939         int snext = first;
3940         SpellChecker * speller = theSpellChecker();
3941         // locate and enumerate the error positions
3942         int nerrors = speller->numMisspelledWords();
3943         int numskipped = 0;
3944         SkipPositionsIterator it = skips.begin();
3945         SkipPositionsIterator et = skips.end();
3946         for (int index = 0; index < nerrors; ++index) {
3947                 int wstart;
3948                 int wlen = 0;
3949                 speller->misspelledWord(index, wstart, wlen);
3950                 /// should not happen if speller supports range checks
3951                 if (!wlen) continue;
3952                 docstring const misspelled = word.substr(wstart, wlen);
3953                 wstart += first + numskipped;
3954                 if (snext < wstart) {
3955                         /// mark the range of correct spelling
3956                         numskipped += countSkips(it, et, wstart);
3957                         setMisspelled(snext,
3958                                 wstart - 1, SpellChecker::WORD_OK);
3959                 }
3960                 snext = wstart + wlen;
3961                 numskipped += countSkips(it, et, snext);
3962                 /// mark the range of misspelling
3963                 setMisspelled(wstart, snext, result);
3964                 LYXERR(Debug::GUI, "misspelled word: \"" <<
3965                            misspelled << "\" [" <<
3966                            wstart << ".." << (snext-1) << "]");
3967                 ++snext;
3968         }
3969         if (snext <= last) {
3970                 /// mark the range of correct spelling at end
3971                 setMisspelled(snext, last, SpellChecker::WORD_OK);
3972         }
3973 }
3974
3975
3976 void Paragraph::spellCheck() const
3977 {
3978         SpellChecker * speller = theSpellChecker();
3979         if (!speller || empty() ||!needsSpellCheck())
3980                 return;
3981         pos_type start;
3982         pos_type endpos;
3983         d->rangeOfSpellCheck(start, endpos);
3984         if (speller->canCheckParagraph()) {
3985                 // loop until we leave the range
3986                 for (pos_type first = start; first < endpos; ) {
3987                         pos_type last = endpos;
3988                         Private::SkipPositions skips;
3989                         Language * lang = d->locateSpellRange(first, last, skips);
3990                         if (first >= endpos)
3991                                 break;
3992                         // start the spell checker on the unit of meaning
3993                         docstring word = asString(first, last, AS_STR_INSETS + AS_STR_SKIPDELETE);
3994                         WordLangTuple wl = WordLangTuple(word, lang);
3995                         SpellChecker::Result result = word.size() ?
3996                                 speller->check(wl) : SpellChecker::WORD_OK;
3997                         d->markMisspelledWords(first, last, result, word, skips);
3998                         first = ++last;
3999                 }
4000         } else {
4001                 static docstring_list suggestions;
4002                 pos_type to = endpos;
4003                 while (start < endpos) {
4004                         WordLangTuple wl;
4005                         spellCheck(start, to, wl, suggestions, false);
4006                         start = to + 1;
4007                 }
4008         }
4009         d->readySpellCheck();
4010 }
4011
4012
4013 bool Paragraph::isMisspelled(pos_type pos, bool check_boundary) const
4014 {
4015         bool result = SpellChecker::misspelled(d->speller_state_.getState(pos));
4016         if (result || pos <= 0 || pos > size())
4017                 return result;
4018         if (check_boundary && (pos == size() || isWordSeparator(pos)))
4019                 result = SpellChecker::misspelled(d->speller_state_.getState(pos - 1));
4020         return result;
4021 }
4022
4023
4024 string Paragraph::magicLabel() const
4025 {
4026         stringstream ss;
4027         ss << "magicparlabel-" << id();
4028         return ss.str();
4029 }
4030
4031
4032 } // namespace lyx