src/Paragraph.cpp

   1 /**
   2  * \file Paragraph.cpp
   3  * This file is part of LyX, the document processor.
   4  * Licence details can be found in the file COPYING.
   5  *
   6  * \author Asger Alstrup
   7  * \author Lars Gullik Bjønnes
   8  * \author Richard Heck (XHTML output)
   9  * \author Jean-Marc Lasgouttes
  10  * \author Angus Leeming
  11  * \author John Levon
  12  * \author André Pönitz
  13  * \author Dekel Tsur
  14  * \author Jürgen Vigna
  15  *
  16  * Full author contact details are available in file CREDITS.
  17  */
  18
  19 #include <config.h>
  20
  21 #include "Paragraph.h"
  22
  23 #include "LayoutFile.h"
  24 #include "Buffer.h"
  25 #include "BufferParams.h"
  26 #include "Changes.h"
  27 #include "Counters.h"
  28 #include "Encoding.h"
  29 #include "InsetList.h"
  30 #include "Language.h"
  31 #include "LaTeXFeatures.h"
  32 #include "Layout.h"
  33 #include "Length.h"
  34 #include "Font.h"
  35 #include "FontList.h"
  36 #include "LyXRC.h"
  37 #include "OutputParams.h"
  38 #include "output_latex.h"
  39 #include "output_xhtml.h"
  40 #include "ParagraphParameters.h"
  41 #include "SpellChecker.h"
  42 #include "sgml.h"
  43 #include "TextClass.h"
  44 #include "TexRow.h"
  45 #include "Text.h"
  46 #include "VSpace.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 <sstream>
  65 #include <vector>
  66
  67 using namespace std;
  68 using namespace lyx::support;
  69
  70 namespace lyx {
  71
  72 namespace {
  73 /// Inset identifier (above 0x10ffff, for ucs-4)
  74 char_type const META_INSET = 0x200001;
  75 };
  76
  77
  78 /////////////////////////////////////////////////////////////////////
  79 //
  80 // SpellResultRange
  81 //
  82 /////////////////////////////////////////////////////////////////////
  83
  84 class SpellResultRange {
  85 public:
  86         SpellResultRange(FontSpan range, SpellChecker::Result result)
  87         : range_(range), result_(result)
  88         {}
  89         ///
  90         FontSpan const & range() const { return range_; }
  91         ///
  92         void range(FontSpan const & r) { range_ = r; }
  93         ///
  94         SpellChecker::Result result() const { return result_; }
  95         ///
  96         void result(SpellChecker::Result r) { result_ = r; }
  97         ///
  98         bool inside(pos_type pos) const { return range_.inside(pos); }
  99         ///
 100         bool covered(FontSpan const & r) const
 101         {
 102                 // 1. first of new range inside current range or
 103                 // 2. last of new range inside current range or
 104                 // 3. first of current range inside new range or
 105                 // 4. last of current range inside new range
 106                 return range_.inside(r.first) || range_.inside(r.last) ||
 107                         r.inside(range_.first) || r.inside(range_.last);
 108         }
 109         ///
 110         void shift(pos_type pos, int offset)
 111         {
 112                 if (range_.first > pos) {
 113                         range_.first += offset;
 114                         range_.last += offset;
 115                 } else if (range_.last > pos) {
 116                         range_.last += offset;
 117                 }
 118         }
 119 private:
 120         FontSpan range_ ;
 121         SpellChecker::Result result_ ;
 122 };
 123
 124
 125 /////////////////////////////////////////////////////////////////////
 126 //
 127 // SpellCheckerState
 128 //
 129 /////////////////////////////////////////////////////////////////////
 130
 131 class SpellCheckerState {
 132 public:
 133         SpellCheckerState() {
 134                 needs_refresh_ = true;
 135                 current_change_number_ = 0;
 136         }
 137
 138         void setRange(FontSpan const fp, SpellChecker::Result state)
 139         {
 140                 eraseCoveredRanges(fp);
 141                 if (state != SpellChecker::WORD_OK)
 142                         ranges_.push_back(SpellResultRange(fp, state));
 143         }
 144
 145         void increasePosAfterPos(pos_type pos)
 146         {
 147                 correctRangesAfterPos(pos, 1);
 148                 needsRefresh(pos);
 149         }
 150
 151         void decreasePosAfterPos(pos_type pos)
 152         {
 153                 correctRangesAfterPos(pos, -1);
 154                 needsRefresh(pos);
 155         }
 156
 157         void refreshLast(pos_type pos)
 158         {
 159                 if (pos < refresh_.last)
 160                         refresh_.last = pos;
 161         }
 162
 163         SpellChecker::Result getState(pos_type pos) const
 164         {
 165                 SpellChecker::Result result = SpellChecker::WORD_OK;
 166                 RangesIterator et = ranges_.end();
 167                 RangesIterator it = ranges_.begin();
 168                 for (; it != et; ++it) {
 169                         if(it->inside(pos)) {
 170                                 return it->result();
 171                         }
 172                 }
 173                 return result;
 174         }
 175
 176         bool needsRefresh() const {
 177                 return needs_refresh_;
 178         }
 179
 180         SpellChecker::ChangeNumber currentChangeNumber() const {
 181                 return current_change_number_;
 182         }
 183
 184         void refreshRange(pos_type & first, pos_type & last) const {
 185                 first = refresh_.first;
 186                 last = refresh_.last;
 187         }
 188
 189         void needsRefresh(pos_type pos) {
 190                 if (needs_refresh_ && pos != -1) {
 191                         if (pos < refresh_.first)
 192                                 refresh_.first = pos;
 193                         if (pos > refresh_.last)
 194                                 refresh_.last = pos;
 195                 } else if (pos != -1) {
 196                         refresh_.first = pos;
 197                         refresh_.last = pos;
 198                 }
 199                 needs_refresh_ = pos != -1;
 200         }
 201
 202         void needsCompleteRefresh(SpellChecker::ChangeNumber change_number) {
 203                 needs_refresh_ = true;
 204                 refresh_.first = 0;
 205                 refresh_.last = -1;
 206                 current_change_number_ = change_number;
 207         }
 208
 209 private:
 210         typedef vector<SpellResultRange> Ranges;
 211         typedef Ranges::const_iterator RangesIterator;
 212         Ranges ranges_;
 213         /// the area of the paragraph with pending spell check
 214         FontSpan refresh_;
 215         bool needs_refresh_;
 216         /// spell state cache version number
 217         SpellChecker::ChangeNumber current_change_number_;
 218
 219
 220         void eraseCoveredRanges(FontSpan const fp)
 221         {
 222                 Ranges result;
 223                 RangesIterator et = ranges_.end();
 224                 RangesIterator it = ranges_.begin();
 225                 for (; it != et; ++it) {
 226                         if (!it->covered(fp))
 227                                 result.push_back(SpellResultRange(it->range(), it->result()));
 228                 }
 229                 ranges_ = result;
 230         }
 231
 232         void correctRangesAfterPos(pos_type pos, int offset)
 233         {
 234                 RangesIterator et = ranges_.end();
 235                 Ranges::iterator it = ranges_.begin();
 236                 for (; it != et; ++it) {
 237                         it->shift(pos, offset);
 238                 }
 239         }
 240
 241 };
 242
 243 /////////////////////////////////////////////////////////////////////
 244 //
 245 // Paragraph::Private
 246 //
 247 /////////////////////////////////////////////////////////////////////
 248
 249 class Paragraph::Private
 250 {
 251 public:
 252         ///
 253         Private(Paragraph * owner, Layout const & layout);
 254         /// "Copy constructor"
 255         Private(Private const &, Paragraph * owner);
 256         /// Copy constructor from \p beg  to \p end
 257         Private(Private const &, Paragraph * owner, pos_type beg, pos_type end);
 258
 259         ///
 260         void insertChar(pos_type pos, char_type c, Change const & change);
 261
 262         /// Output the surrogate pair formed by \p c and \p next to \p os.
 263         /// \return the number of characters written.
 264         int latexSurrogatePair(odocstream & os, char_type c, char_type next,
 265                                OutputParams const &);
 266
 267         /// Output a space in appropriate formatting (or a surrogate pair
 268         /// if the next character is a combining character).
 269         /// \return whether a surrogate pair was output.
 270         bool simpleTeXBlanks(OutputParams const &,
 271                              odocstream &, TexRow & texrow,
 272                              pos_type i,
 273                              unsigned int & column,
 274                              Font const & font,
 275                              Layout const & style);
 276
 277         /// Output consecutive unicode chars, belonging to the same script as
 278         /// specified by the latex macro \p ltx, to \p os starting from \p i.
 279         /// \return the number of characters written.
 280         int writeScriptChars(odocstream & os, docstring const & ltx,
 281                            Change const &, Encoding const &, pos_type & i);
 282
 283         /// This could go to ParagraphParameters if we want to.
 284         int startTeXParParams(BufferParams const &, odocstream &, TexRow &,
 285                               OutputParams const &) const;
 286
 287         /// This could go to ParagraphParameters if we want to.
 288         int endTeXParParams(BufferParams const &, odocstream &, TexRow &,
 289                             OutputParams const &) const;
 290
 291         ///
 292         void latexInset(BufferParams const &,
 293                                    odocstream &,
 294                                    TexRow & texrow, OutputParams &,
 295                                    Font & running_font,
 296                                    Font & basefont,
 297                                    Font const & outerfont,
 298                                    bool & open_font,
 299                                    Change & running_change,
 300                                    Layout const & style,
 301                                    pos_type & i,
 302                                    unsigned int & column);
 303
 304         ///
 305         void latexSpecialChar(
 306                                    odocstream & os,
 307                                    OutputParams const & runparams,
 308                                    Font const & running_font,
 309                                    Change const & running_change,
 310                                    Layout const & style,
 311                                    pos_type & i,
 312                                    unsigned int & column);
 313
 314         ///
 315         bool latexSpecialT1(
 316                 char_type const c,
 317                 odocstream & os,
 318                 pos_type i,
 319                 unsigned int & column);
 320         ///
 321         bool latexSpecialTypewriter(
 322                 char_type const c,
 323                 odocstream & os,
 324                 pos_type i,
 325                 unsigned int & column);
 326         ///
 327         bool latexSpecialPhrase(
 328                 odocstream & os,
 329                 pos_type & i,
 330                 unsigned int & column,
 331                 OutputParams const & runparams);
 332
 333         ///
 334         void validate(LaTeXFeatures & features) const;
 335
 336         /// Checks if the paragraph contains only text and no inset or font change.
 337         bool onlyText(Buffer const & buf, Font const & outerfont,
 338                       pos_type initial) const;
 339
 340         /// match a string against a particular point in the paragraph
 341         bool isTextAt(string const & str, pos_type pos) const;
 342
 343         /// a vector of speller skip positions
 344         typedef vector<FontSpan> SkipPositions;
 345         typedef SkipPositions::const_iterator SkipPositionsIterator;
 346
 347         void appendSkipPosition(SkipPositions & skips, pos_type const pos) const;
 348
 349         Language * getSpellLanguage(pos_type const from) const;
 350
 351         Language * locateSpellRange(pos_type & from, pos_type & to,
 352                                                                 SkipPositions & skips) const;
 353
 354         bool hasSpellerChange() const {
 355                 SpellChecker::ChangeNumber speller_change_number = 0;
 356                 if (theSpellChecker())
 357                         speller_change_number = theSpellChecker()->changeNumber();
 358                 return speller_change_number > speller_state_.currentChangeNumber();
 359         }
 360
 361         void setMisspelled(pos_type from, pos_type to, SpellChecker::Result state)
 362         {
 363                 pos_type textsize = owner_->size();
 364                 // check for sane arguments
 365                 if (to < from || from >= textsize)
 366                         return;
 367                 FontSpan fp = FontSpan(from, to);
 368                 // don't mark end of paragraph
 369                 if (fp.last >= textsize)
 370                         fp.last = textsize - 1;
 371                 speller_state_.setRange(fp, state);
 372         }
 373
 374         void requestSpellCheck(pos_type pos) {
 375                 speller_state_.needsRefresh(pos);
 376         }
 377
 378         void readySpellCheck() {
 379                 speller_state_.needsRefresh(-1);
 380         }
 381
 382         bool needsSpellCheck() const
 383         {
 384                 return speller_state_.needsRefresh();
 385         }
 386
 387         void rangeOfSpellCheck(pos_type & first, pos_type & last) const
 388         {
 389                 speller_state_.refreshRange(first, last);
 390                 if (last == -1) {
 391                         last = owner_->size();
 392                         return;
 393                 }
 394                 pos_type endpos = last;
 395                 owner_->locateWord(first, endpos, WHOLE_WORD);
 396                 if (endpos < last) {
 397                         endpos = last;
 398                         owner_->locateWord(last, endpos, WHOLE_WORD);
 399                 }
 400                 last = endpos;
 401         }
 402
 403         int countSkips(SkipPositionsIterator & it, SkipPositionsIterator const et,
 404                             int & start) const
 405         {
 406                 int numskips = 0;
 407                 while (it != et && it->first < start) {
 408                         int skip = it->last - it->first + 1;
 409                         start += skip;
 410                         numskips += skip;
 411                         ++it;
 412                 }
 413                 return numskips;
 414         }
 415
 416         void markMisspelledWords(pos_type const & first, pos_type const & last,
 417                                                          SpellChecker::Result result,
 418                                                          docstring const & word,
 419                                                          SkipPositions const & skips);
 420
 421         InsetCode ownerCode() const
 422         {
 423                 return inset_owner_ ? inset_owner_->lyxCode() : NO_CODE;
 424         }
 425
 426         /// Which Paragraph owns us?
 427         Paragraph * owner_;
 428
 429         /// In which Inset?
 430         Inset const * inset_owner_;
 431
 432         ///
 433         FontList fontlist_;
 434
 435         ///
 436         int id_;
 437
 438         ///
 439         ParagraphParameters params_;
 440
 441         /// for recording and looking up changes
 442         Changes changes_;
 443
 444         ///
 445         InsetList insetlist_;
 446
 447         /// end of label
 448         pos_type begin_of_body_;
 449
 450         typedef docstring TextContainer;
 451         ///
 452         TextContainer text_;
 453
 454         typedef set<docstring> Words;
 455         typedef map<Language, Words> LangWordsMap;
 456         ///
 457         LangWordsMap words_;
 458         ///
 459         Layout const * layout_;
 460         ///
 461         SpellCheckerState speller_state_;
 462 };
 463
 464
 465 namespace {
 466
 467 struct special_phrase {
 468         string phrase;
 469         docstring macro;
 470         bool builtin;
 471 };
 472
 473 special_phrase const special_phrases[] = {
 474         { "LyX", from_ascii("\\LyX{}"), false },
 475         { "TeX", from_ascii("\\TeX{}"), true },
 476         { "LaTeX2e", from_ascii("\\LaTeXe{}"), true },
 477         { "LaTeX", from_ascii("\\LaTeX{}"), true },
 478 };
 479
 480 size_t const phrases_nr = sizeof(special_phrases)/sizeof(special_phrase);
 481
 482 } // namespace anon
 483
 484
 485 Paragraph::Private::Private(Paragraph * owner, Layout const & layout)
 486         : owner_(owner), inset_owner_(0), id_(-1), begin_of_body_(0), layout_(&layout)
 487 {
 488         text_.reserve(100);
 489 }
 490
 491
 492 // Initialization of the counter for the paragraph id's,
 493 //
 494 // FIXME: There should be a more intelligent way to generate and use the
 495 // paragraph ids per buffer instead a global static counter for all InsetText
 496 // in the running program.
 497 static int paragraph_id = -1;
 498
 499 Paragraph::Private::Private(Private const & p, Paragraph * owner)
 500         : owner_(owner), inset_owner_(p.inset_owner_), fontlist_(p.fontlist_),
 501           params_(p.params_), changes_(p.changes_), insetlist_(p.insetlist_),
 502           begin_of_body_(p.begin_of_body_), text_(p.text_), words_(p.words_),
 503           layout_(p.layout_)
 504 {
 505         id_ = ++paragraph_id;
 506         requestSpellCheck(p.text_.size());
 507 }
 508
 509
 510 Paragraph::Private::Private(Private const & p, Paragraph * owner,
 511         pos_type beg, pos_type end)
 512         : owner_(owner), inset_owner_(p.inset_owner_),
 513           params_(p.params_), changes_(p.changes_),
 514           insetlist_(p.insetlist_, beg, end),
 515           begin_of_body_(p.begin_of_body_), words_(p.words_),
 516           layout_(p.layout_)
 517 {
 518         id_ = ++paragraph_id;
 519         if (beg >= pos_type(p.text_.size()))
 520                 return;
 521         text_ = p.text_.substr(beg, end - beg);
 522
 523         FontList::const_iterator fcit = fontlist_.begin();
 524         FontList::const_iterator fend = fontlist_.end();
 525         for (; fcit != fend; ++fcit) {
 526                 if (fcit->pos() < beg)
 527                         continue;
 528                 if (fcit->pos() >= end) {
 529                         // Add last entry in the fontlist_.
 530                         fontlist_.set(text_.size() - 1, fcit->font());
 531                         break;
 532                 }
 533                 // Add a new entry in the fontlist_.
 534                 fontlist_.set(fcit->pos() - beg, fcit->font());
 535         }
 536         requestSpellCheck(p.text_.size());
 537 }
 538
 539
 540 void Paragraph::addChangesToToc(DocIterator const & cdit,
 541         Buffer const & buf) const
 542 {
 543         d->changes_.addToToc(cdit, buf);
 544 }
 545
 546
 547 bool Paragraph::isDeleted(pos_type start, pos_type end) const
 548 {
 549         LASSERT(start >= 0 && start <= size(), /**/);
 550         LASSERT(end > start && end <= size() + 1, /**/);
 551
 552         return d->changes_.isDeleted(start, end);
 553 }
 554
 555
 556 bool Paragraph::isChanged(pos_type start, pos_type end) const
 557 {
 558         LASSERT(start >= 0 && start <= size(), /**/);
 559         LASSERT(end > start && end <= size() + 1, /**/);
 560
 561         return d->changes_.isChanged(start, end);
 562 }
 563
 564
 565 bool Paragraph::isMergedOnEndOfParDeletion(bool trackChanges) const
 566 {
 567         // keep the logic here in sync with the logic of eraseChars()
 568         if (!trackChanges)
 569                 return true;
 570
 571         Change const change = d->changes_.lookup(size());
 572         return change.inserted() && change.currentAuthor();
 573 }
 574
 575
 576 void Paragraph::setChange(Change const & change)
 577 {
 578         // beware of the imaginary end-of-par character!
 579         d->changes_.set(change, 0, size() + 1);
 580
 581         /*
 582          * Propagate the change recursively - but not in case of DELETED!
 583          *
 584          * Imagine that your co-author makes changes in an existing inset. He
 585          * sends your document to you and you come to the conclusion that the
 586          * inset should go completely. If you erase it, LyX must not delete all
 587          * text within the inset. Otherwise, the change tracked insertions of
 588          * your co-author get lost and there is no way to restore them later.
 589          *
 590          * Conclusion: An inset's content should remain untouched if you delete it
 591          */
 592
 593         if (!change.deleted()) {
 594                 for (pos_type pos = 0; pos < size(); ++pos) {
 595                         if (Inset * inset = getInset(pos))
 596                                 inset->setChange(change);
 597                 }
 598         }
 599 }
 600
 601
 602 void Paragraph::setChange(pos_type pos, Change const & change)
 603 {
 604         LASSERT(pos >= 0 && pos <= size(), /**/);
 605         d->changes_.set(change, pos);
 606
 607         // see comment in setChange(Change const &) above
 608         if (!change.deleted() && pos < size())
 609                         if (Inset * inset = getInset(pos))
 610                                 inset->setChange(change);
 611 }
 612
 613
 614 Change const & Paragraph::lookupChange(pos_type pos) const
 615 {
 616         LASSERT(pos >= 0 && pos <= size(), /**/);
 617         return d->changes_.lookup(pos);
 618 }
 619
 620
 621 void Paragraph::acceptChanges(pos_type start, pos_type end)
 622 {
 623         LASSERT(start >= 0 && start <= size(), /**/);
 624         LASSERT(end > start && end <= size() + 1, /**/);
 625
 626         for (pos_type pos = start; pos < end; ++pos) {
 627                 switch (lookupChange(pos).type) {
 628                         case Change::UNCHANGED:
 629                                 // accept changes in nested inset
 630                                 if (Inset * inset = getInset(pos))
 631                                         inset->acceptChanges();
 632                                 break;
 633
 634                         case Change::INSERTED:
 635                                 d->changes_.set(Change(Change::UNCHANGED), pos);
 636                                 // also accept changes in nested inset
 637                                 if (Inset * inset = getInset(pos))
 638                                         inset->acceptChanges();
 639                                 break;
 640
 641                         case Change::DELETED:
 642                                 // Suppress access to non-existent
 643                                 // "end-of-paragraph char"
 644                                 if (pos < size()) {
 645                                         eraseChar(pos, false);
 646                                         --end;
 647                                         --pos;
 648                                 }
 649                                 break;
 650                 }
 651
 652         }
 653 }
 654
 655
 656 void Paragraph::rejectChanges(pos_type start, pos_type end)
 657 {
 658         LASSERT(start >= 0 && start <= size(), /**/);
 659         LASSERT(end > start && end <= size() + 1, /**/);
 660
 661         for (pos_type pos = start; pos < end; ++pos) {
 662                 switch (lookupChange(pos).type) {
 663                         case Change::UNCHANGED:
 664                                 // reject changes in nested inset
 665                                 if (Inset * inset = getInset(pos))
 666                                                 inset->rejectChanges();
 667                                 break;
 668
 669                         case Change::INSERTED:
 670                                 // Suppress access to non-existent
 671                                 // "end-of-paragraph char"
 672                                 if (pos < size()) {
 673                                         eraseChar(pos, false);
 674                                         --end;
 675                                         --pos;
 676                                 }
 677                                 break;
 678
 679                         case Change::DELETED:
 680                                 d->changes_.set(Change(Change::UNCHANGED), pos);
 681
 682                                 // Do NOT reject changes within a deleted inset!
 683                                 // There may be insertions of a co-author inside of it!
 684
 685                                 break;
 686                 }
 687         }
 688 }
 689
 690
 691 void Paragraph::Private::insertChar(pos_type pos, char_type c,
 692                 Change const & change)
 693 {
 694         LASSERT(pos >= 0 && pos <= int(text_.size()), /**/);
 695
 696         // track change
 697         changes_.insert(change, pos);
 698
 699         // This is actually very common when parsing buffers (and
 700         // maybe inserting ascii text)
 701         if (pos == pos_type(text_.size())) {
 702                 // when appending characters, no need to update tables
 703                 text_.push_back(c);
 704                 // but we want spell checking
 705                 requestSpellCheck(pos);
 706                 return;
 707         }
 708
 709         text_.insert(text_.begin() + pos, c);
 710
 711         // Update the font table.
 712         fontlist_.increasePosAfterPos(pos);
 713
 714         // Update the insets
 715         insetlist_.increasePosAfterPos(pos);
 716
 717         // Update list of misspelled positions
 718         speller_state_.increasePosAfterPos(pos);
 719 }
 720
 721
 722 bool Paragraph::insertInset(pos_type pos, Inset * inset,
 723                                    Change const & change)
 724 {
 725         LASSERT(inset, /**/);
 726         LASSERT(pos >= 0 && pos <= size(), /**/);
 727
 728         // Paragraph::insertInset() can be used in cut/copy/paste operation where
 729         // d->inset_owner_ is not set yet.
 730         if (d->inset_owner_ && !d->inset_owner_->insetAllowed(inset->lyxCode()))
 731                 return false;
 732
 733         d->insertChar(pos, META_INSET, change);
 734         LASSERT(d->text_[pos] == META_INSET, /**/);
 735
 736         // Add a new entry in the insetlist_.
 737         d->insetlist_.insert(inset, pos);
 738
 739         // Some insets require run of spell checker
 740         requestSpellCheck(pos);
 741         return true;
 742 }
 743
 744
 745 bool Paragraph::eraseChar(pos_type pos, bool trackChanges)
 746 {
 747         LASSERT(pos >= 0 && pos <= size(), return false);
 748
 749         // keep the logic here in sync with the logic of isMergedOnEndOfParDeletion()
 750
 751         if (trackChanges) {
 752                 Change change = d->changes_.lookup(pos);
 753
 754                 // set the character to DELETED if
 755                 //  a) it was previously unchanged or
 756                 //  b) it was inserted by a co-author
 757
 758                 if (!change.changed() ||
 759                       (change.inserted() && !change.currentAuthor())) {
 760                         setChange(pos, Change(Change::DELETED));
 761                         // request run of spell checker
 762                         requestSpellCheck(pos);
 763                         return false;
 764                 }
 765
 766                 if (change.deleted())
 767                         return false;
 768         }
 769
 770         // Don't physically access the imaginary end-of-paragraph character.
 771         // eraseChar() can only mark it as DELETED. A physical deletion of
 772         // end-of-par must be handled externally.
 773         if (pos == size()) {
 774                 return false;
 775         }
 776
 777         // track change
 778         d->changes_.erase(pos);
 779
 780         // if it is an inset, delete the inset entry
 781         if (d->text_[pos] == META_INSET)
 782                 d->insetlist_.erase(pos);
 783
 784         d->text_.erase(d->text_.begin() + pos);
 785
 786         // Update the fontlist_
 787         d->fontlist_.erase(pos);
 788
 789         // Update the insetlist_
 790         d->insetlist_.decreasePosAfterPos(pos);
 791
 792         // Update list of misspelled positions
 793         d->speller_state_.decreasePosAfterPos(pos);
 794         d->speller_state_.refreshLast(size());
 795
 796         return true;
 797 }
 798
 799
 800 int Paragraph::eraseChars(pos_type start, pos_type end, bool trackChanges)
 801 {
 802         LASSERT(start >= 0 && start <= size(), /**/);
 803         LASSERT(end >= start && end <= size() + 1, /**/);
 804
 805         pos_type i = start;
 806         for (pos_type count = end - start; count; --count) {
 807                 if (!eraseChar(i, trackChanges))
 808                         ++i;
 809         }
 810         return end - i;
 811 }
 812
 813
 814 int Paragraph::Private::latexSurrogatePair(odocstream & os, char_type c,
 815                 char_type next, OutputParams const & runparams)
 816 {
 817         // Writing next here may circumvent a possible font change between
 818         // c and next. Since next is only output if it forms a surrogate pair
 819         // with c we can ignore this:
 820         // A font change inside a surrogate pair does not make sense and is
 821         // hopefully impossible to input.
 822         // FIXME: change tracking
 823         // Is this correct WRT change tracking?
 824         Encoding const & encoding = *(runparams.encoding);
 825         docstring const latex1 = encoding.latexChar(next);
 826         docstring const latex2 = encoding.latexChar(c);
 827         if (docstring(1, next) == latex1) {
 828                 // the encoding supports the combination
 829                 os << latex2 << latex1;
 830                 return latex1.length() + latex2.length();
 831         } else if (runparams.local_font &&
 832                    runparams.local_font->language()->lang() == "polutonikogreek") {
 833                 // polutonikogreek only works without the brackets
 834                 os << latex1 << latex2;
 835                 return latex1.length() + latex2.length();
 836         } else
 837                 os << latex1 << '{' << latex2 << '}';
 838         return latex1.length() + latex2.length() + 2;
 839 }
 840
 841
 842 bool Paragraph::Private::simpleTeXBlanks(OutputParams const & runparams,
 843                                        odocstream & os, TexRow & texrow,
 844                                        pos_type i,
 845                                        unsigned int & column,
 846                                        Font const & font,
 847                                        Layout const & style)
 848 {
 849         if (style.pass_thru || runparams.pass_thru)
 850                 return false;
 851
 852         if (i + 1 < int(text_.size())) {
 853                 char_type next = text_[i + 1];
 854                 if (Encodings::isCombiningChar(next)) {
 855                         // This space has an accent, so we must always output it.
 856                         column += latexSurrogatePair(os, ' ', next, runparams) - 1;
 857                         return true;
 858                 }
 859         }
 860
 861         if (runparams.linelen > 0
 862             && column > runparams.linelen
 863             && i
 864             && text_[i - 1] != ' '
 865             && (i + 1 < int(text_.size()))
 866             // same in FreeSpacing mode
 867             && !owner_->isFreeSpacing()
 868             // In typewriter mode, we want to avoid
 869             // ! . ? : at the end of a line
 870             && !(font.fontInfo().family() == TYPEWRITER_FAMILY
 871                  && (text_[i - 1] == '.'
 872                      || text_[i - 1] == '?'
 873                      || text_[i - 1] == ':'
 874                      || text_[i - 1] == '!'))) {
 875                 os << '\n';
 876                 texrow.newline();
 877                 texrow.start(owner_->id(), i + 1);
 878                 column = 0;
 879         } else if (style.free_spacing) {
 880                 os << '~';
 881         } else {
 882                 os << ' ';
 883         }
 884         return false;
 885 }
 886
 887
 888 int Paragraph::Private::writeScriptChars(odocstream & os,
 889                                          docstring const & ltx,
 890                                          Change const & runningChange,
 891                                          Encoding const & encoding,
 892                                          pos_type & i)
 893 {
 894         // FIXME: modifying i here is not very nice...
 895
 896         // We only arrive here when a proper language for character text_[i] has
 897         // not been specified (i.e., it could not be translated in the current
 898         // latex encoding) or its latex translation has been forced, and it
 899         // belongs to a known script.
 900         // Parameter ltx contains the latex translation of text_[i] as specified
 901         // in the unicodesymbols file and is something like "\textXXX{<spec>}".
 902         // The latex macro name "textXXX" specifies the script to which text_[i]
 903         // belongs and we use it in order to check whether characters from the
 904         // same script immediately follow, such that we can collect them in a
 905         // single "\textXXX" macro. So, we have to retain "\textXXX{<spec>"
 906         // for the first char but only "<spec>" for all subsequent chars.
 907         docstring::size_type const brace1 = ltx.find_first_of(from_ascii("{"));
 908         docstring::size_type const brace2 = ltx.find_last_of(from_ascii("}"));
 909         string script = to_ascii(ltx.substr(1, brace1 - 1));
 910         int pos = 0;
 911         int length = brace2;
 912         bool closing_brace = true;
 913         if (script == "textgreek" && encoding.latexName() == "iso-8859-7") {
 914                 // Correct encoding is being used, so we can avoid \textgreek.
 915                 pos = brace1 + 1;
 916                 length -= pos;
 917                 closing_brace = false;
 918         }
 919         os << ltx.substr(pos, length);
 920         int size = text_.size();
 921         while (i + 1 < size) {
 922                 char_type const next = text_[i + 1];
 923                 // Stop here if next character belongs to another script
 924                 // or there is a change in change tracking status.
 925                 if (!Encodings::isKnownScriptChar(next, script) ||
 926                     runningChange != owner_->lookupChange(i + 1))
 927                         break;
 928                 Font prev_font;
 929                 bool found = false;
 930                 FontList::const_iterator cit = fontlist_.begin();
 931                 FontList::const_iterator end = fontlist_.end();
 932                 for (; cit != end; ++cit) {
 933                         if (cit->pos() >= i && !found) {
 934                                 prev_font = cit->font();
 935                                 found = true;
 936                         }
 937                         if (cit->pos() >= i + 1)
 938                                 break;
 939                 }
 940                 // Stop here if there is a font attribute or encoding change.
 941                 if (found && cit != end && prev_font != cit->font())
 942                         break;
 943                 docstring const latex = encoding.latexChar(next);
 944                 docstring::size_type const b1 =
 945                                         latex.find_first_of(from_ascii("{"));
 946                 docstring::size_type const b2 =
 947                                         latex.find_last_of(from_ascii("}"));
 948                 int const len = b2 - b1 - 1;
 949                 os << latex.substr(b1 + 1, len);
 950                 length += len;
 951                 ++i;
 952         }
 953         if (closing_brace) {
 954                 os << '}';
 955                 ++length;
 956         }
 957         return length;
 958 }
 959
 960
 961 bool Paragraph::Private::isTextAt(string const & str, pos_type pos) const
 962 {
 963         pos_type const len = str.length();
 964
 965         // is the paragraph large enough?
 966         if (pos + len > int(text_.size()))
 967                 return false;
 968
 969         // does the wanted text start at point?
 970         for (string::size_type i = 0; i < str.length(); ++i) {
 971                 // Caution: direct comparison of characters works only
 972                 // because str is pure ASCII.
 973                 if (str[i] != text_[pos + i])
 974                         return false;
 975         }
 976
 977         return fontlist_.hasChangeInRange(pos, len);
 978 }
 979
 980
 981 void Paragraph::Private::latexInset(BufferParams const & bparams,
 982                                     odocstream & os,
 983                                     TexRow & texrow,
 984                                     OutputParams & runparams,
 985                                     Font & running_font,
 986                                     Font & basefont,
 987                                     Font const & outerfont,
 988                                     bool & open_font,
 989                                     Change & running_change,
 990                                     Layout const & style,
 991                                     pos_type & i,
 992                                     unsigned int & column)
 993 {
 994         Inset * inset = owner_->getInset(i);
 995         LASSERT(inset, /**/);
 996
 997         if (style.pass_thru) {
 998                 inset->plaintext(os, runparams);
 999                 return;
1000         }
1001
1002         // FIXME: move this to InsetNewline::latex
1003         if (inset->lyxCode() == NEWLINE_CODE) {
1004                 // newlines are handled differently here than
1005                 // the default in simpleTeXSpecialChars().
1006                 if (!style.newline_allowed) {
1007                         os << '\n';
1008                 } else {
1009                         if (open_font) {
1010                                 column += running_font.latexWriteEndChanges(
1011                                         os, bparams, runparams,
1012                                         basefont, basefont);
1013                                 open_font = false;
1014                         }
1015
1016                         if (running_font.fontInfo().family() == TYPEWRITER_FAMILY)
1017                                 os << '~';
1018
1019                         basefont = owner_->getLayoutFont(bparams, outerfont);
1020                         running_font = basefont;
1021
1022                         if (runparams.moving_arg)
1023                                 os << "\\protect ";
1024
1025                 }
1026                 texrow.newline();
1027                 texrow.start(owner_->id(), i + 1);
1028                 column = 0;
1029         }
1030
1031         if (owner_->isDeleted(i)) {
1032                 if( ++runparams.inDeletedInset == 1)
1033                         runparams.changeOfDeletedInset = owner_->lookupChange(i);
1034         }
1035
1036         if (inset->canTrackChanges()) {
1037                 column += Changes::latexMarkChange(os, bparams, running_change,
1038                         Change(Change::UNCHANGED), runparams);
1039                 running_change = Change(Change::UNCHANGED);
1040         }
1041
1042         bool close = false;
1043         odocstream::pos_type const len = os.tellp();
1044
1045         if (inset->forceLTR()
1046             && running_font.isRightToLeft()
1047             // ERT is an exception, it should be output with no
1048             // decorations at all
1049             && inset->lyxCode() != ERT_CODE) {
1050                 if (running_font.language()->lang() == "farsi")
1051                         os << "\\beginL{}";
1052                 else
1053                         os << "\\L{";
1054                 close = true;
1055         }
1056
1057         // FIXME: Bug: we can have an empty font change here!
1058         // if there has just been a font change, we are going to close it
1059         // right now, which means stupid latex code like \textsf{}. AFAIK,
1060         // this does not harm dvi output. A minor bug, thus (JMarc)
1061
1062         // Some insets cannot be inside a font change command.
1063         // However, even such insets *can* be placed in \L or \R
1064         // or their equivalents (for RTL language switches), so we don't
1065         // close the language in those cases.
1066         // ArabTeX, though, cannot handle this special behavior, it seems.
1067         bool arabtex = basefont.language()->lang() == "arabic_arabtex"
1068                 || running_font.language()->lang() == "arabic_arabtex";
1069         if (open_font && inset->noFontChange()) {
1070                 bool closeLanguage = arabtex
1071                         || basefont.isRightToLeft() == running_font.isRightToLeft();
1072                 unsigned int count = running_font.latexWriteEndChanges(os,
1073                         bparams, runparams, basefont, basefont, closeLanguage);
1074                 column += count;
1075                 // if any font properties were closed, update the running_font,
1076                 // making sure, however, to leave the language as it was
1077                 if (count > 0) {
1078                         // FIXME: probably a better way to keep track of the old
1079                         // language, than copying the entire font?
1080                         Font const copy_font(running_font);
1081                         basefont = owner_->getLayoutFont(bparams, outerfont);
1082                         running_font = basefont;
1083                         if (!closeLanguage)
1084                                 running_font.setLanguage(copy_font.language());
1085                         // leave font open if language is still open
1086                         open_font = (running_font.language() == basefont.language());
1087                         if (closeLanguage)
1088                                 runparams.local_font = &basefont;
1089                 }
1090         }
1091
1092         int tmp;
1093
1094         try {
1095                 tmp = inset->latex(os, runparams);
1096         } catch (EncodingException & e) {
1097                 // add location information and throw again.
1098                 e.par_id = id_;
1099                 e.pos = i;
1100                 throw(e);
1101         }
1102
1103         if (close) {
1104                 if (running_font.language()->lang() == "farsi")
1105                                 os << "\\endL{}";
1106                         else
1107                                 os << '}';
1108         }
1109
1110         if (tmp) {
1111                 texrow.newlines(tmp);
1112                 texrow.start(owner_->id(), i + 1);
1113                 column = 0;
1114         } else {
1115                 column += (unsigned int)(os.tellp() - len);
1116         }
1117
1118         if (owner_->isDeleted(i))
1119                 --runparams.inDeletedInset;
1120 }
1121
1122
1123 void Paragraph::Private::latexSpecialChar(
1124                                              odocstream & os,
1125                                              OutputParams const & runparams,
1126                                              Font const & running_font,
1127                                              Change const & running_change,
1128                                              Layout const & style,
1129                                              pos_type & i,
1130                                              unsigned int & column)
1131 {
1132         char_type const c = text_[i];
1133
1134         if (style.pass_thru || runparams.pass_thru) {
1135                 if (c != '\0')
1136                         // FIXME UNICODE: This can fail if c cannot
1137                         // be encoded in the current encoding.
1138                         os.put(c);
1139                 return;
1140         }
1141
1142         // If T1 font encoding is used, use the special
1143         // characters it provides.
1144         // NOTE: some languages reset the font encoding
1145         // internally
1146         if (!running_font.language()->internalFontEncoding()
1147             && lyxrc.fontenc == "T1" && latexSpecialT1(c, os, i, column))
1148                 return;
1149
1150         // \tt font needs special treatment
1151         if (running_font.fontInfo().family() == TYPEWRITER_FAMILY
1152                 && latexSpecialTypewriter(c, os, i, column))
1153                 return;
1154
1155         // Otherwise, we use what LaTeX provides us.
1156         switch (c) {
1157         case '\\':
1158                 os << "\\textbackslash{}";
1159                 column += 15;
1160                 break;
1161         case '<':
1162                 os << "\\textless{}";
1163                 column += 10;
1164                 break;
1165         case '>':
1166                 os << "\\textgreater{}";
1167                 column += 13;
1168                 break;
1169         case '|':
1170                 os << "\\textbar{}";
1171                 column += 9;
1172                 break;
1173         case '-':
1174                 os << '-';
1175                 break;
1176         case '\"':
1177                 os << "\\char`\\\"{}";
1178                 column += 9;
1179                 break;
1180
1181         case '$': case '&':
1182         case '%': case '#': case '{':
1183         case '}': case '_':
1184                 os << '\\';
1185                 os.put(c);
1186                 column += 1;
1187                 break;
1188
1189         case '~':
1190                 os << "\\textasciitilde{}";
1191                 column += 16;
1192                 break;
1193
1194         case '^':
1195                 os << "\\textasciicircum{}";
1196                 column += 17;
1197                 break;
1198
1199         case '*':
1200         case '[':
1201         case ']':
1202                 // avoid being mistaken for optional arguments
1203                 os << '{';
1204                 os.put(c);
1205                 os << '}';
1206                 column += 2;
1207                 break;
1208
1209         case ' ':
1210                 // Blanks are printed before font switching.
1211                 // Sure? I am not! (try nice-latex)
1212                 // I am sure it's correct. LyX might be smarter
1213                 // in the future, but for now, nothing wrong is
1214                 // written. (Asger)
1215                 break;
1216
1217         default:
1218                 // LyX, LaTeX etc.
1219                 if (latexSpecialPhrase(os, i, column, runparams))
1220                         return;
1221
1222                 if (c == '\0')
1223                         return;
1224
1225                 Encoding const & encoding = *(runparams.encoding);
1226                 if (i + 1 < int(text_.size())) {
1227                         char_type next = text_[i + 1];
1228                         if (Encodings::isCombiningChar(next)) {
1229                                 column += latexSurrogatePair(os, c, next, runparams) - 1;
1230                                 ++i;
1231                                 break;
1232                         }
1233                 }
1234                 string script;
1235                 docstring const latex = encoding.latexChar(c);
1236                 if (Encodings::isKnownScriptChar(c, script)
1237                     && prefixIs(latex, from_ascii("\\" + script)))
1238                         column += writeScriptChars(os, latex,
1239                                         running_change, encoding, i) - 1;
1240                 else if (latex.length() > 1 && latex[latex.length() - 1] != '}') {
1241                         // Prevent eating of a following
1242                         // space or command corruption by
1243                         // following characters
1244                         column += latex.length() + 1;
1245                         os << latex << "{}";
1246                 } else {
1247                         column += latex.length() - 1;
1248                         os << latex;
1249                 }
1250                 break;
1251         }
1252 }
1253
1254
1255 bool Paragraph::Private::latexSpecialT1(char_type const c, odocstream & os,
1256         pos_type i, unsigned int & column)
1257 {
1258         switch (c) {
1259         case '>':
1260         case '<':
1261                 os.put(c);
1262                 // In T1 encoding, these characters exist
1263                 // but we should avoid ligatures
1264                 if (i + 1 >= int(text_.size()) || text_[i + 1] != c)
1265                         return true;
1266                 os << "\\textcompwordmark{}";
1267                 column += 19;
1268                 return true;
1269         case '|':
1270                 os.put(c);
1271                 return true;
1272         case '\"':
1273                 // soul.sty breaks with \char`\"
1274                 os << "\\textquotedbl{}";
1275                 column += 14;
1276                 return true;
1277         default:
1278                 return false;
1279         }
1280 }
1281
1282
1283 bool Paragraph::Private::latexSpecialTypewriter(char_type const c, odocstream & os,
1284         pos_type i, unsigned int & column)
1285 {
1286         switch (c) {
1287         case '-':
1288                 // within \ttfamily, "--" is merged to "-" (no endash)
1289                 // so we avoid this rather irritating ligature
1290                 if (i + 1 < int(text_.size()) && text_[i + 1] == '-') {
1291                         os << "-{}";
1292                         column += 2;
1293                 } else
1294                         os << '-';
1295                 return true;
1296
1297         // everything else has to be checked separately
1298         // (depending on the encoding)
1299         default:
1300                 return false;
1301         }
1302 }
1303
1304
1305 bool Paragraph::Private::latexSpecialPhrase(odocstream & os, pos_type & i,
1306         unsigned int & column, OutputParams const & runparams)
1307 {
1308         // FIXME: if we have "LaTeX" with a font
1309         // change in the middle (before the 'T', then
1310         // the "TeX" part is still special cased.
1311         // Really we should only operate this on
1312         // "words" for some definition of word
1313
1314         for (size_t pnr = 0; pnr < phrases_nr; ++pnr) {
1315                 if (!isTextAt(special_phrases[pnr].phrase, i))
1316                         continue;
1317                 if (runparams.moving_arg)
1318                         os << "\\protect";
1319                 os << special_phrases[pnr].macro;
1320                 i += special_phrases[pnr].phrase.length() - 1;
1321                 column += special_phrases[pnr].macro.length() - 1;
1322                 return true;
1323         }
1324         return false;
1325 }
1326
1327
1328 void Paragraph::Private::validate(LaTeXFeatures & features) const
1329 {
1330         if (layout_->inpreamble && inset_owner_) {
1331                 Buffer const & buf = inset_owner_->buffer();
1332                 BufferParams const & bp = buf.params();
1333                 Font f;
1334                 TexRow tr;
1335                 // Using a string stream here circumvents the encoding
1336                 // switching machinery of odocstream. Therefore the
1337                 // output is wrong if this paragraph contains content
1338                 // that needs to switch encoding.
1339                 odocstringstream ods;
1340                 ods << '\\' << from_ascii(layout_->latexname());
1341                 // we have to provide all the optional arguments here, even though
1342                 // the last one is the only one we care about.
1343                 // Separate handling of optional argument inset.
1344                 if (layout_->optargs != 0 || layout_->reqargs != 0)
1345                         latexArgInsets(*owner_, ods, features.runparams(),
1346                                        layout_->reqargs, layout_->optargs);
1347                 else
1348                         ods << from_ascii(layout_->latexparam());
1349                 docstring::size_type const length = ods.str().length();
1350                 // this will output "{" at the beginning, but not at the end
1351                 owner_->latex(bp, f, ods, tr, features.runparams(), 0, -1, true);
1352                 if (ods.str().length() > length) {
1353                         ods << '}';
1354                         string const snippet = to_utf8(ods.str());
1355                         features.addPreambleSnippet(snippet);
1356                 }
1357         }
1358
1359         if (features.runparams().flavor == OutputParams::HTML
1360             && layout_->htmltitle()) {
1361                 features.setHTMLTitle(owner_->asString(AS_STR_INSETS));
1362         }
1363
1364         // check the params.
1365         if (!params_.spacing().isDefault())
1366                 features.require("setspace");
1367
1368         // then the layouts
1369         features.useLayout(layout_->name());
1370
1371         // then the fonts
1372         fontlist_.validate(features);
1373
1374         // then the indentation
1375         if (!params_.leftIndent().zero())
1376                 features.require("ParagraphLeftIndent");
1377
1378         // then the insets
1379         InsetList::const_iterator icit = insetlist_.begin();
1380         InsetList::const_iterator iend = insetlist_.end();
1381         for (; icit != iend; ++icit) {
1382                 if (icit->inset) {
1383                         icit->inset->validate(features);
1384                         if (layout_->needprotect &&
1385                             icit->inset->lyxCode() == FOOT_CODE)
1386                                 features.require("NeedLyXFootnoteCode");
1387                 }
1388         }
1389
1390         // then the contents
1391         for (pos_type i = 0; i < int(text_.size()) ; ++i) {
1392                 for (size_t pnr = 0; pnr < phrases_nr; ++pnr) {
1393                         if (!special_phrases[pnr].builtin
1394                             && isTextAt(special_phrases[pnr].phrase, i)) {
1395                                 features.require(special_phrases[pnr].phrase);
1396                                 break;
1397                         }
1398                 }
1399                 Encodings::validate(text_[i], features);
1400         }
1401 }
1402
1403 /////////////////////////////////////////////////////////////////////
1404 //
1405 // Paragraph
1406 //
1407 /////////////////////////////////////////////////////////////////////
1408
1409 namespace {
1410         Layout const emptyParagraphLayout;
1411 }
1412
1413 Paragraph::Paragraph()
1414         : d(new Paragraph::Private(this, emptyParagraphLayout))
1415 {
1416         itemdepth = 0;
1417         d->params_.clear();
1418 }
1419
1420
1421 Paragraph::Paragraph(Paragraph const & par)
1422         : itemdepth(par.itemdepth),
1423         d(new Paragraph::Private(*par.d, this))
1424 {
1425         registerWords();
1426 }
1427
1428
1429 Paragraph::Paragraph(Paragraph const & par, pos_type beg, pos_type end)
1430         : itemdepth(par.itemdepth),
1431         d(new Paragraph::Private(*par.d, this, beg, end))
1432 {
1433         registerWords();
1434 }
1435
1436
1437 Paragraph & Paragraph::operator=(Paragraph const & par)
1438 {
1439         // needed as we will destroy the private part before copying it
1440         if (&par != this) {
1441                 itemdepth = par.itemdepth;
1442
1443                 deregisterWords();
1444                 delete d;
1445                 d = new Private(*par.d, this);
1446                 registerWords();
1447         }
1448         return *this;
1449 }
1450
1451
1452 Paragraph::~Paragraph()
1453 {
1454         deregisterWords();
1455         delete d;
1456 }
1457
1458
1459 namespace {
1460
1461 // this shall be called just before every "os << ..." action.
1462 void flushString(ostream & os, docstring & s)
1463 {
1464         os << to_utf8(s);
1465         s.erase();
1466 }
1467
1468 }
1469
1470
1471 void Paragraph::write(ostream & os, BufferParams const & bparams,
1472         depth_type & dth) const
1473 {
1474         // The beginning or end of a deeper (i.e. nested) area?
1475         if (dth != d->params_.depth()) {
1476                 if (d->params_.depth() > dth) {
1477                         while (d->params_.depth() > dth) {
1478                                 os << "\n\\begin_deeper";
1479                                 ++dth;
1480                         }
1481                 } else {
1482                         while (d->params_.depth() < dth) {
1483                                 os << "\n\\end_deeper";
1484                                 --dth;
1485                         }
1486                 }
1487         }
1488
1489         // First write the layout
1490         os << "\n\\begin_layout " << to_utf8(d->layout_->name()) << '\n';
1491
1492         d->params_.write(os);
1493
1494         Font font1(inherit_font, bparams.language);
1495
1496         Change running_change = Change(Change::UNCHANGED);
1497
1498         // this string is used as a buffer to avoid repetitive calls
1499         // to to_utf8(), which turn out to be expensive (JMarc)
1500         docstring write_buffer;
1501
1502         int column = 0;
1503         for (pos_type i = 0; i <= size(); ++i) {
1504
1505                 Change const change = lookupChange(i);
1506                 if (change != running_change)
1507                         flushString(os, write_buffer);
1508                 Changes::lyxMarkChange(os, bparams, column, running_change, change);
1509                 running_change = change;
1510
1511                 if (i == size())
1512                         break;
1513
1514                 // Write font changes
1515                 Font font2 = getFontSettings(bparams, i);
1516                 if (font2 != font1) {
1517                         flushString(os, write_buffer);
1518                         font2.lyxWriteChanges(font1, os);
1519                         column = 0;
1520                         font1 = font2;
1521                 }
1522
1523                 char_type const c = d->text_[i];
1524                 switch (c) {
1525                 case META_INSET:
1526                         if (Inset const * inset = getInset(i)) {
1527                                 flushString(os, write_buffer);
1528                                 if (inset->directWrite()) {
1529                                         // international char, let it write
1530                                         // code directly so it's shorter in
1531                                         // the file
1532                                         inset->write(os);
1533                                 } else {
1534                                         if (i)
1535                                                 os << '\n';
1536                                         os << "\\begin_inset ";
1537                                         inset->write(os);
1538                                         os << "\n\\end_inset\n\n";
1539                                         column = 0;
1540                                 }
1541                         }
1542                         break;
1543                 case '\\':
1544                         flushString(os, write_buffer);
1545                         os << "\n\\backslash\n";
1546                         column = 0;
1547                         break;
1548                 case '.':
1549                         flushString(os, write_buffer);
1550                         if (i + 1 < size() && d->text_[i + 1] == ' ') {
1551                                 os << ".\n";
1552                                 column = 0;
1553                         } else
1554                                 os << '.';
1555                         break;
1556                 default:
1557                         if ((column > 70 && c == ' ')
1558                             || column > 79) {
1559                                 flushString(os, write_buffer);
1560                                 os << '\n';
1561                                 column = 0;
1562                         }
1563                         // this check is to amend a bug. LyX sometimes
1564                         // inserts '\0' this could cause problems.
1565                         if (c != '\0')
1566                                 write_buffer.push_back(c);
1567                         else
1568                                 LYXERR0("NUL char in structure.");
1569                         ++column;
1570                         break;
1571                 }
1572         }
1573
1574         flushString(os, write_buffer);
1575         os << "\n\\end_layout\n";
1576 }
1577
1578
1579 void Paragraph::validate(LaTeXFeatures & features) const
1580 {
1581         d->validate(features);
1582 }
1583
1584
1585 void Paragraph::insert(pos_type start, docstring const & str,
1586                        Font const & font, Change const & change)
1587 {
1588         for (size_t i = 0, n = str.size(); i != n ; ++i)
1589                 insertChar(start + i, str[i], font, change);
1590 }
1591
1592
1593 void Paragraph::appendChar(char_type c, Font const & font,
1594                 Change const & change)
1595 {
1596         // track change
1597         d->changes_.insert(change, d->text_.size());
1598         // when appending characters, no need to update tables
1599         d->text_.push_back(c);
1600         setFont(d->text_.size() - 1, font);
1601 }
1602
1603
1604 void Paragraph::appendString(docstring const & s, Font const & font,
1605                 Change const & change)
1606 {
1607         pos_type end = s.size();
1608         size_t oldsize = d->text_.size();
1609         size_t newsize = oldsize + end;
1610         size_t capacity = d->text_.capacity();
1611         if (newsize >= capacity)
1612                 d->text_.reserve(max(capacity + 100, newsize));
1613
1614         // when appending characters, no need to update tables
1615         d->text_.append(s);
1616
1617         // FIXME: Optimize this!
1618         for (size_t i = oldsize; i != newsize; ++i) {
1619                 // track change
1620                 d->changes_.insert(change, i);
1621         }
1622         d->fontlist_.set(oldsize, font);
1623         d->fontlist_.set(newsize - 1, font);
1624 }
1625
1626
1627 void Paragraph::insertChar(pos_type pos, char_type c,
1628                            bool trackChanges)
1629 {
1630         d->insertChar(pos, c, Change(trackChanges ?
1631                            Change::INSERTED : Change::UNCHANGED));
1632 }
1633
1634
1635 void Paragraph::insertChar(pos_type pos, char_type c,
1636                            Font const & font, bool trackChanges)
1637 {
1638         d->insertChar(pos, c, Change(trackChanges ?
1639                            Change::INSERTED : Change::UNCHANGED));
1640         setFont(pos, font);
1641 }
1642
1643
1644 void Paragraph::insertChar(pos_type pos, char_type c,
1645                            Font const & font, Change const & change)
1646 {
1647         d->insertChar(pos, c, change);
1648         setFont(pos, font);
1649 }
1650
1651
1652 bool Paragraph::insertInset(pos_type pos, Inset * inset,
1653                             Font const & font, Change const & change)
1654 {
1655         bool const success = insertInset(pos, inset, change);
1656         // Set the font/language of the inset...
1657         setFont(pos, font);
1658         return success;
1659 }
1660
1661
1662 void Paragraph::resetFonts(Font const & font)
1663 {
1664         d->fontlist_.clear();
1665         d->fontlist_.set(0, font);
1666         d->fontlist_.set(d->text_.size() - 1, font);
1667 }
1668
1669 // Gets uninstantiated font setting at position.
1670 Font const & Paragraph::getFontSettings(BufferParams const & bparams,
1671                                          pos_type pos) const
1672 {
1673         if (pos > size()) {
1674                 LYXERR0("pos: " << pos << " size: " << size());
1675                 LASSERT(pos <= size(), /**/);
1676         }
1677
1678         FontList::const_iterator cit = d->fontlist_.fontIterator(pos);
1679         if (cit != d->fontlist_.end())
1680                 return cit->font();
1681
1682         if (pos == size() && !empty())
1683                 return getFontSettings(bparams, pos - 1);
1684
1685         // Optimisation: avoid a full font instantiation if there is no
1686         // language change from previous call.
1687         static Font previous_font;
1688         static Language const * previous_lang = 0;
1689         Language const * lang = getParLanguage(bparams);
1690         if (lang != previous_lang) {
1691                 previous_lang = lang;
1692                 previous_font = Font(inherit_font, lang);
1693         }
1694         return previous_font;
1695 }
1696
1697
1698 FontSpan Paragraph::fontSpan(pos_type pos) const
1699 {
1700         LASSERT(pos <= size(), /**/);
1701         pos_type start = 0;
1702
1703         FontList::const_iterator cit = d->fontlist_.begin();
1704         FontList::const_iterator end = d->fontlist_.end();
1705         for (; cit != end; ++cit) {
1706                 if (cit->pos() >= pos) {
1707                         if (pos >= beginOfBody())
1708                                 return FontSpan(max(start, beginOfBody()),
1709                                                 cit->pos());
1710                         else
1711                                 return FontSpan(start,
1712                                                 min(beginOfBody() - 1,
1713                                                          cit->pos()));
1714                 }
1715                 start = cit->pos() + 1;
1716         }
1717
1718         // This should not happen, but if so, we take no chances.
1719         // LYXERR0("Paragraph::getEndPosOfFontSpan: This should not happen!");
1720         return FontSpan(pos, pos);
1721 }
1722
1723
1724 // Gets uninstantiated font setting at position 0
1725 Font const & Paragraph::getFirstFontSettings(BufferParams const & bparams) const
1726 {
1727         if (!empty() && !d->fontlist_.empty())
1728                 return d->fontlist_.begin()->font();
1729
1730         // Optimisation: avoid a full font instantiation if there is no
1731         // language change from previous call.
1732         static Font previous_font;
1733         static Language const * previous_lang = 0;
1734         if (bparams.language != previous_lang) {
1735                 previous_lang = bparams.language;
1736                 previous_font = Font(inherit_font, bparams.language);
1737         }
1738
1739         return previous_font;
1740 }
1741
1742
1743 // Gets the fully instantiated font at a given position in a paragraph
1744 // This is basically the same function as Text::GetFont() in text2.cpp.
1745 // The difference is that this one is used for generating the LaTeX file,
1746 // and thus cosmetic "improvements" are disallowed: This has to deliver
1747 // the true picture of the buffer. (Asger)
1748 Font const Paragraph::getFont(BufferParams const & bparams, pos_type pos,
1749                                  Font const & outerfont) const
1750 {
1751         LASSERT(pos >= 0, /**/);
1752
1753         Font font = getFontSettings(bparams, pos);
1754
1755         pos_type const body_pos = beginOfBody();
1756         FontInfo & fi = font.fontInfo();
1757         if (pos < body_pos)
1758                 fi.realize(d->layout_->labelfont);
1759         else
1760                 fi.realize(d->layout_->font);
1761
1762         fi.realize(outerfont.fontInfo());
1763         fi.realize(bparams.getFont().fontInfo());
1764
1765         return font;
1766 }
1767
1768
1769 Font const Paragraph::getLabelFont
1770         (BufferParams const & bparams, Font const & outerfont) const
1771 {
1772         FontInfo tmpfont = d->layout_->labelfont;
1773         tmpfont.realize(outerfont.fontInfo());
1774         tmpfont.realize(bparams.getFont().fontInfo());
1775         return Font(tmpfont, getParLanguage(bparams));
1776 }
1777
1778
1779 Font const Paragraph::getLayoutFont
1780         (BufferParams const & bparams, Font const & outerfont) const
1781 {
1782         FontInfo tmpfont = d->layout_->font;
1783         tmpfont.realize(outerfont.fontInfo());
1784         tmpfont.realize(bparams.getFont().fontInfo());
1785         return Font(tmpfont, getParLanguage(bparams));
1786 }
1787
1788
1789 /// Returns the height of the highest font in range
1790 FontSize Paragraph::highestFontInRange
1791         (pos_type startpos, pos_type endpos, FontSize def_size) const
1792 {
1793         return d->fontlist_.highestInRange(startpos, endpos, def_size);
1794 }
1795
1796
1797 char_type Paragraph::getUChar(BufferParams const & bparams, pos_type pos) const
1798 {
1799         char_type c = d->text_[pos];
1800         if (!lyxrc.rtl_support)
1801                 return c;
1802
1803         char_type uc = c;
1804         switch (c) {
1805         case '(':
1806                 uc = ')';
1807                 break;
1808         case ')':
1809                 uc = '(';
1810                 break;
1811         case '[':
1812                 uc = ']';
1813                 break;
1814         case ']':
1815                 uc = '[';
1816                 break;
1817         case '{':
1818                 uc = '}';
1819                 break;
1820         case '}':
1821                 uc = '{';
1822                 break;
1823         case '<':
1824                 uc = '>';
1825                 break;
1826         case '>':
1827                 uc = '<';
1828                 break;
1829         }
1830         if (uc != c && getFontSettings(bparams, pos).isRightToLeft())
1831                 return uc;
1832         return c;
1833 }
1834
1835
1836 void Paragraph::setFont(pos_type pos, Font const & font)
1837 {
1838         LASSERT(pos <= size(), /**/);
1839
1840         // First, reduce font against layout/label font
1841         // Update: The setCharFont() routine in text2.cpp already
1842         // reduces font, so we don't need to do that here. (Asger)
1843
1844         d->fontlist_.set(pos, font);
1845 }
1846
1847
1848 void Paragraph::makeSameLayout(Paragraph const & par)
1849 {
1850         d->layout_ = par.d->layout_;
1851         d->params_ = par.d->params_;
1852 }
1853
1854
1855 bool Paragraph::stripLeadingSpaces(bool trackChanges)
1856 {
1857         if (isFreeSpacing())
1858                 return false;
1859
1860         int pos = 0;
1861         int count = 0;
1862
1863         while (pos < size() && (isNewline(pos) || isLineSeparator(pos))) {
1864                 if (eraseChar(pos, trackChanges))
1865                         ++count;
1866                 else
1867                         ++pos;
1868         }
1869
1870         return count > 0 || pos > 0;
1871 }
1872
1873
1874 bool Paragraph::hasSameLayout(Paragraph const & par) const
1875 {
1876         return par.d->layout_ == d->layout_
1877                 && d->params_.sameLayout(par.d->params_);
1878 }
1879
1880
1881 depth_type Paragraph::getDepth() const
1882 {
1883         return d->params_.depth();
1884 }
1885
1886
1887 depth_type Paragraph::getMaxDepthAfter() const
1888 {
1889         if (d->layout_->isEnvironment())
1890                 return d->params_.depth() + 1;
1891         else
1892                 return d->params_.depth();
1893 }
1894
1895
1896 char Paragraph::getAlign() const
1897 {
1898         if (d->params_.align() == LYX_ALIGN_LAYOUT)
1899                 return d->layout_->align;
1900         else
1901                 return d->params_.align();
1902 }
1903
1904
1905 docstring const & Paragraph::labelString() const
1906 {
1907         return d->params_.labelString();
1908 }
1909
1910
1911 // the next two functions are for the manual labels
1912 docstring const Paragraph::getLabelWidthString() const
1913 {
1914         if (d->layout_->margintype == MARGIN_MANUAL
1915             || d->layout_->latextype == LATEX_BIB_ENVIRONMENT)
1916                 return d->params_.labelWidthString();
1917         else
1918                 return _("Senseless with this layout!");
1919 }
1920
1921
1922 void Paragraph::setLabelWidthString(docstring const & s)
1923 {
1924         d->params_.labelWidthString(s);
1925 }
1926
1927
1928 docstring Paragraph::expandLabel(Layout const & layout,
1929                 BufferParams const & bparams) const
1930 {
1931         return expandParagraphLabel(layout, bparams, true);
1932 }
1933
1934
1935 docstring Paragraph::expandDocBookLabel(Layout const & layout,
1936                 BufferParams const & bparams) const
1937 {
1938         return expandParagraphLabel(layout, bparams, false);
1939 }
1940
1941
1942 docstring Paragraph::expandParagraphLabel(Layout const & layout,
1943                 BufferParams const & bparams, bool process_appendix) const
1944 {
1945         DocumentClass const & tclass = bparams.documentClass();
1946         string const & lang = getParLanguage(bparams)->code();
1947         bool const in_appendix = process_appendix && d->params_.appendix();
1948         docstring fmt = translateIfPossible(layout.labelstring(in_appendix), lang);
1949
1950         if (fmt.empty() && layout.labeltype == LABEL_COUNTER
1951             && !layout.counter.empty())
1952                 return tclass.counters().theCounter(layout.counter, lang);
1953
1954         // handle 'inherited level parts' in 'fmt',
1955         // i.e. the stuff between '@' in   '@Section@.\arabic{subsection}'
1956         size_t const i = fmt.find('@', 0);
1957         if (i != docstring::npos) {
1958                 size_t const j = fmt.find('@', i + 1);
1959                 if (j != docstring::npos) {
1960                         docstring parent(fmt, i + 1, j - i - 1);
1961                         docstring label = from_ascii("??");
1962                         if (tclass.hasLayout(parent))
1963                                 docstring label = expandParagraphLabel(tclass[parent], bparams,
1964                                                       process_appendix);
1965                         fmt = docstring(fmt, 0, i) + label
1966                                 + docstring(fmt, j + 1, docstring::npos);
1967                 }
1968         }
1969
1970         return tclass.counters().counterLabel(fmt, lang);
1971 }
1972
1973
1974 void Paragraph::applyLayout(Layout const & new_layout)
1975 {
1976         d->layout_ = &new_layout;
1977         LyXAlignment const oldAlign = d->params_.align();
1978
1979         if (!(oldAlign & d->layout_->alignpossible)) {
1980                 frontend::Alert::warning(_("Alignment not permitted"),
1981                         _("The new layout does not permit the alignment previously used.\nSetting to default."));
1982                 d->params_.align(LYX_ALIGN_LAYOUT);
1983         }
1984 }
1985
1986
1987 pos_type Paragraph::beginOfBody() const
1988 {
1989         return d->begin_of_body_;
1990 }
1991
1992
1993 void Paragraph::setBeginOfBody()
1994 {
1995         if (d->layout_->labeltype != LABEL_MANUAL) {
1996                 d->begin_of_body_ = 0;
1997                 return;
1998         }
1999
2000         // Unroll the first two cycles of the loop
2001         // and remember the previous character to
2002         // remove unnecessary getChar() calls
2003         pos_type i = 0;
2004         pos_type end = size();
2005         if (i < end && !isNewline(i)) {
2006                 ++i;
2007                 char_type previous_char = 0;
2008                 char_type temp = 0;
2009                 if (i < end) {
2010                         previous_char = d->text_[i];
2011                         if (!isNewline(i)) {
2012                                 ++i;
2013                                 while (i < end && previous_char != ' ') {
2014                                         temp = d->text_[i];
2015                                         if (isNewline(i))
2016                                                 break;
2017                                         ++i;
2018                                         previous_char = temp;
2019                                 }
2020                         }
2021                 }
2022         }
2023
2024         d->begin_of_body_ = i;
2025 }
2026
2027
2028 bool Paragraph::allowParagraphCustomization() const
2029 {
2030         return inInset().allowParagraphCustomization();
2031 }
2032
2033
2034 bool Paragraph::usePlainLayout() const
2035 {
2036         return inInset().usePlainLayout();
2037 }
2038
2039
2040 namespace {
2041
2042 // paragraphs inside floats need different alignment tags to avoid
2043 // unwanted space
2044
2045 bool noTrivlistCentering(InsetCode code)
2046 {
2047         return code == FLOAT_CODE
2048                || code == WRAP_CODE
2049                || code == CELL_CODE;
2050 }
2051
2052
2053 string correction(string const & orig)
2054 {
2055         if (orig == "flushleft")
2056                 return "raggedright";
2057         if (orig == "flushright")
2058                 return "raggedleft";
2059         if (orig == "center")
2060                 return "centering";
2061         return orig;
2062 }
2063
2064
2065 string const corrected_env(string const & suffix, string const & env,
2066         InsetCode code, bool const lastpar)
2067 {
2068         string output = suffix + "{";
2069         if (noTrivlistCentering(code)) {
2070                 if (lastpar) {
2071                         // the last paragraph in non-trivlist-aligned
2072                         // context is special (to avoid unwanted whitespace)
2073                         if (suffix == "\\begin")
2074                                 return "\\" + correction(env) + "{}";
2075                         return string();
2076                 }
2077                 output += correction(env);
2078         } else
2079                 output += env;
2080         output += "}";
2081         if (suffix == "\\begin")
2082                 output += "\n";
2083         return output;
2084 }
2085
2086
2087 void adjust_row_column(string const & str, TexRow & texrow, int & column)
2088 {
2089         if (!contains(str, "\n"))
2090                 column += str.size();
2091         else {
2092                 string tmp;
2093                 texrow.newline();
2094                 column = rsplit(str, tmp, '\n').size();
2095         }
2096 }
2097
2098 } // namespace anon
2099
2100
2101 int Paragraph::Private::startTeXParParams(BufferParams const & bparams,
2102                                  odocstream & os, TexRow & texrow,
2103                                  OutputParams const & runparams) const
2104 {
2105         int column = 0;
2106
2107         if (params_.noindent()) {
2108                 os << "\\noindent ";
2109                 column += 10;
2110         }
2111
2112         LyXAlignment const curAlign = params_.align();
2113
2114         if (curAlign == layout_->align)
2115                 return column;
2116
2117         switch (curAlign) {
2118         case LYX_ALIGN_NONE:
2119         case LYX_ALIGN_BLOCK:
2120         case LYX_ALIGN_LAYOUT:
2121         case LYX_ALIGN_SPECIAL:
2122         case LYX_ALIGN_DECIMAL:
2123                 break;
2124         case LYX_ALIGN_LEFT:
2125         case LYX_ALIGN_RIGHT:
2126         case LYX_ALIGN_CENTER:
2127                 if (runparams.moving_arg) {
2128                         os << "\\protect";
2129                         column += 8;
2130                 }
2131                 break;
2132         }
2133
2134         string const begin_tag = "\\begin";
2135         InsetCode code = ownerCode();
2136         bool const lastpar = runparams.isLastPar;
2137
2138         switch (curAlign) {
2139         case LYX_ALIGN_NONE:
2140         case LYX_ALIGN_BLOCK:
2141         case LYX_ALIGN_LAYOUT:
2142         case LYX_ALIGN_SPECIAL:
2143         case LYX_ALIGN_DECIMAL:
2144                 break;
2145         case LYX_ALIGN_LEFT: {
2146                 string output;
2147                 if (owner_->getParLanguage(bparams)->babel() != "hebrew")
2148                         output = corrected_env(begin_tag, "flushleft", code, lastpar);
2149                 else
2150                         output = corrected_env(begin_tag, "flushright", code, lastpar);
2151                 os << from_ascii(output);
2152                 adjust_row_column(output, texrow, column);
2153                 break;
2154         } case LYX_ALIGN_RIGHT: {
2155                 string output;
2156                 if (owner_->getParLanguage(bparams)->babel() != "hebrew")
2157                         output = corrected_env(begin_tag, "flushright", code, lastpar);
2158                 else
2159                         output = corrected_env(begin_tag, "flushleft", code, lastpar);
2160                 os << from_ascii(output);
2161                 adjust_row_column(output, texrow, column);
2162                 break;
2163         } case LYX_ALIGN_CENTER: {
2164                 string output;
2165                 output = corrected_env(begin_tag, "center", code, lastpar);
2166                 os << from_ascii(output);
2167                 adjust_row_column(output, texrow, column);
2168                 break;
2169         }
2170         }
2171
2172         return column;
2173 }
2174
2175
2176 int Paragraph::Private::endTeXParParams(BufferParams const & bparams,
2177                                odocstream & os, TexRow & texrow,
2178                                OutputParams const & runparams) const
2179 {
2180         int column = 0;
2181
2182         LyXAlignment const curAlign = params_.align();
2183
2184         if (curAlign == layout_->align)
2185                 return column;
2186
2187         switch (curAlign) {
2188         case LYX_ALIGN_NONE:
2189         case LYX_ALIGN_BLOCK:
2190         case LYX_ALIGN_LAYOUT:
2191         case LYX_ALIGN_SPECIAL:
2192         case LYX_ALIGN_DECIMAL:
2193                 break;
2194         case LYX_ALIGN_LEFT:
2195         case LYX_ALIGN_RIGHT:
2196         case LYX_ALIGN_CENTER:
2197                 if (runparams.moving_arg) {
2198                         os << "\\protect";
2199                         column = 8;
2200                 }
2201                 break;
2202         }
2203
2204         string const end_tag = "\n\\par\\end";
2205         InsetCode code = ownerCode();
2206         bool const lastpar = runparams.isLastPar;
2207
2208         switch (curAlign) {
2209         case LYX_ALIGN_NONE:
2210         case LYX_ALIGN_BLOCK:
2211         case LYX_ALIGN_LAYOUT:
2212         case LYX_ALIGN_SPECIAL:
2213         case LYX_ALIGN_DECIMAL:
2214                 break;
2215         case LYX_ALIGN_LEFT: {
2216                 string output;
2217                 if (owner_->getParLanguage(bparams)->babel() != "hebrew")
2218                         output = corrected_env(end_tag, "flushleft", code, lastpar);
2219                 else
2220                         output = corrected_env(end_tag, "flushright", code, lastpar);
2221                 os << from_ascii(output);
2222                 adjust_row_column(output, texrow, column);
2223                 break;
2224         } case LYX_ALIGN_RIGHT: {
2225                 string output;
2226                 if (owner_->getParLanguage(bparams)->babel() != "hebrew")
2227                         output = corrected_env(end_tag, "flushright", code, lastpar);
2228                 else
2229                         output = corrected_env(end_tag, "flushleft", code, lastpar);
2230                 os << from_ascii(output);
2231                 adjust_row_column(output, texrow, column);
2232                 break;
2233         } case LYX_ALIGN_CENTER: {
2234                 string output;
2235                 output = corrected_env(end_tag, "center", code, lastpar);
2236                 os << from_ascii(output);
2237                 adjust_row_column(output, texrow, column);
2238                 break;
2239         }
2240         }
2241
2242         return column;
2243 }
2244
2245
2246 // This one spits out the text of the paragraph
2247 void Paragraph::latex(BufferParams const & bparams,
2248         Font const & outerfont,
2249         odocstream & os, TexRow & texrow,
2250         OutputParams const & runparams,
2251         int start_pos, int end_pos, bool force) const
2252 {
2253         LYXERR(Debug::LATEX, "Paragraph::latex...     " << this);
2254
2255         // FIXME This check should not be needed. Perhaps issue an
2256         // error if it triggers.
2257         Layout const & style = inInset().forcePlainLayout() ?
2258                 bparams.documentClass().plainLayout() : *d->layout_;
2259
2260         if (!force && style.inpreamble)
2261                 return;
2262
2263         bool const allowcust = allowParagraphCustomization();
2264
2265         // Current base font for all inherited font changes, without any
2266         // change caused by an individual character, except for the language:
2267         // It is set to the language of the first character.
2268         // As long as we are in the label, this font is the base font of the
2269         // label. Before the first body character it is set to the base font
2270         // of the body.
2271         Font basefont;
2272
2273         // Maybe we have to create a optional argument.
2274         pos_type body_pos = beginOfBody();
2275         unsigned int column = 0;
2276
2277         if (body_pos > 0) {
2278                 // the optional argument is kept in curly brackets in
2279                 // case it contains a ']'
2280                 os << "[{";
2281                 column += 2;
2282                 basefont = getLabelFont(bparams, outerfont);
2283         } else {
2284                 basefont = getLayoutFont(bparams, outerfont);
2285         }
2286
2287         // Which font is currently active?
2288         Font running_font(basefont);
2289         // Do we have an open font change?
2290         bool open_font = false;
2291
2292         Change runningChange = Change(Change::UNCHANGED);
2293
2294         Encoding const * const prev_encoding = runparams.encoding;
2295
2296         texrow.start(id(), 0);
2297
2298         // if the paragraph is empty, the loop will not be entered at all
2299         if (empty()) {
2300                 if (style.isCommand()) {
2301                         os << '{';
2302                         ++column;
2303                 }
2304                 if (allowcust)
2305                         column += d->startTeXParParams(bparams, os, texrow,
2306                                                     runparams);
2307         }
2308
2309         for (pos_type i = 0; i < size(); ++i) {
2310                 // First char in paragraph or after label?
2311                 if (i == body_pos) {
2312                         if (body_pos > 0) {
2313                                 if (open_font) {
2314                                         column += running_font.latexWriteEndChanges(
2315                                                 os, bparams, runparams,
2316                                                 basefont, basefont);
2317                                         open_font = false;
2318                                 }
2319                                 basefont = getLayoutFont(bparams, outerfont);
2320                                 running_font = basefont;
2321
2322                                 column += Changes::latexMarkChange(os, bparams,
2323                                                 runningChange, Change(Change::UNCHANGED),
2324                                                 runparams);
2325                                 runningChange = Change(Change::UNCHANGED);
2326
2327                                 os << "}] ";
2328                                 column +=3;
2329                         }
2330                         if (style.isCommand()) {
2331                                 os << '{';
2332                                 ++column;
2333                         }
2334
2335                         if (allowcust)
2336                                 column += d->startTeXParParams(bparams, os,
2337                                                             texrow,
2338                                                             runparams);
2339                 }
2340
2341                 Change const & change = runparams.inDeletedInset ? runparams.changeOfDeletedInset
2342                                                                  : lookupChange(i);
2343
2344                 if (bparams.outputChanges && runningChange != change) {
2345                         if (open_font) {
2346                                 column += running_font.latexWriteEndChanges(
2347                                                 os, bparams, runparams, basefont, basefont);
2348                                 open_font = false;
2349                         }
2350                         basefont = getLayoutFont(bparams, outerfont);
2351                         running_font = basefont;
2352
2353                         column += Changes::latexMarkChange(os, bparams, runningChange,
2354                                                            change, runparams);
2355                         runningChange = change;
2356                 }
2357
2358                 // do not output text which is marked deleted
2359                 // if change tracking output is disabled
2360                 if (!bparams.outputChanges && change.deleted()) {
2361                         continue;
2362                 }
2363
2364                 ++column;
2365
2366                 // Fully instantiated font
2367                 Font const font = getFont(bparams, i, outerfont);
2368
2369                 Font const last_font = running_font;
2370
2371                 // Do we need to close the previous font?
2372                 if (open_font &&
2373                     (font != running_font ||
2374                      font.language() != running_font.language()))
2375                 {
2376                         column += running_font.latexWriteEndChanges(
2377                                         os, bparams, runparams, basefont,
2378                                         (i == body_pos-1) ? basefont : font);
2379                         running_font = basefont;
2380                         open_font = false;
2381                 }
2382
2383                 string const running_lang = runparams.use_polyglossia ?
2384                         running_font.language()->polyglossia() : running_font.language()->babel();
2385                 // close babel's font environment before opening CJK.
2386                 string const lang_end_command = runparams.use_polyglossia ?
2387                         "\\end{$$lang}" : lyxrc.language_command_end;
2388                 if (!running_lang.empty() &&
2389                     font.language()->encoding()->package() == Encoding::CJK) {
2390                                 string end_tag = subst(lang_end_command,
2391                                                         "$$lang",
2392                                                         running_lang);
2393                                 os << from_ascii(end_tag);
2394                                 column += end_tag.length();
2395                 }
2396
2397                 // Switch file encoding if necessary (and allowed)
2398                 if (!runparams.pass_thru && !style.pass_thru &&
2399                     runparams.encoding->package() != Encoding::none &&
2400                     font.language()->encoding()->package() != Encoding::none) {
2401                         pair<bool, int> const enc_switch = switchEncoding(os, bparams,
2402                                         runparams, *(font.language()->encoding()));
2403                         if (enc_switch.first) {
2404                                 column += enc_switch.second;
2405                                 runparams.encoding = font.language()->encoding();
2406                         }
2407                 }
2408
2409                 char_type const c = d->text_[i];
2410
2411                 // Do we need to change font?
2412                 if ((font != running_font ||
2413                      font.language() != running_font.language()) &&
2414                         i != body_pos - 1)
2415                 {
2416                         odocstringstream ods;
2417                         column += font.latexWriteStartChanges(ods, bparams,
2418                                                               runparams, basefont,
2419                                                               last_font);
2420                         running_font = font;
2421                         open_font = true;
2422                         docstring fontchange = ods.str();
2423                         // check whether the fontchange ends with a \\textcolor
2424                         // modifier and the text starts with a space (bug 4473)
2425                         docstring const last_modifier = rsplit(fontchange, '\\');
2426                         if (prefixIs(last_modifier, from_ascii("textcolor")) && c == ' ')
2427                                 os << fontchange << from_ascii("{}");
2428                         // check if the fontchange ends with a trailing blank
2429                         // (like "\small " (see bug 3382)
2430                         else if (suffixIs(fontchange, ' ') && c == ' ')
2431                                 os << fontchange.substr(0, fontchange.size() - 1)
2432                                    << from_ascii("{}");
2433                         else
2434                                 os << fontchange;
2435                 }
2436
2437                 // FIXME: think about end_pos implementation...
2438                 if (c == ' ' && i >= start_pos && (end_pos == -1 || i < end_pos)) {
2439                         // FIXME: integrate this case in latexSpecialChar
2440                         // Do not print the separation of the optional argument
2441                         // if style.pass_thru is false. This works because
2442                         // latexSpecialChar ignores spaces if
2443                         // style.pass_thru is false.
2444                         if (i != body_pos - 1) {
2445                                 if (d->simpleTeXBlanks(
2446                                                 runparams, os, texrow,
2447                                                 i, column, font, style)) {
2448                                         // A surrogate pair was output. We
2449                                         // must not call latexSpecialChar
2450                                         // in this iteration, since it would output
2451                                         // the combining character again.
2452                                         ++i;
2453                                         continue;
2454                                 }
2455                         }
2456                 }
2457
2458                 OutputParams rp = runparams;
2459                 rp.free_spacing = style.free_spacing;
2460                 rp.local_font = &font;
2461                 rp.intitle = style.intitle;
2462
2463                 // Two major modes:  LaTeX or plain
2464                 // Handle here those cases common to both modes
2465                 // and then split to handle the two modes separately.
2466                 if (c == META_INSET) {
2467                         if (i >= start_pos && (end_pos == -1 || i < end_pos)) {
2468                                 d->latexInset(bparams, os,
2469                                                 texrow, rp, running_font,
2470                                                 basefont, outerfont, open_font,
2471                                                 runningChange, style, i, column);
2472                         }
2473                 } else {
2474                         if (i >= start_pos && (end_pos == -1 || i < end_pos)) {
2475                                 try {
2476                                         d->latexSpecialChar(os, rp, running_font, runningChange,
2477                                                 style, i, column);
2478                                 } catch (EncodingException & e) {
2479                                 if (runparams.dryrun) {
2480                                         os << "<" << _("LyX Warning: ")
2481                                            << _("uncodable character") << " '";
2482                                         os.put(c);
2483                                         os << "'>";
2484                                 } else {
2485                                         // add location information and throw again.
2486                                         e.par_id = id();
2487                                         e.pos = i;
2488                                         throw(e);
2489                                 }
2490                         }
2491                 }
2492                 }
2493
2494                 // Set the encoding to that returned from latexSpecialChar (see
2495                 // comment for encoding member in OutputParams.h)
2496                 runparams.encoding = rp.encoding;
2497         }
2498
2499         // If we have an open font definition, we have to close it
2500         if (open_font) {
2501 #ifdef FIXED_LANGUAGE_END_DETECTION
2502                 if (next_) {
2503                         running_font.latexWriteEndChanges(os, bparams,
2504                                         runparams, basefont,
2505                                         next_->getFont(bparams, 0, outerfont));
2506                 } else {
2507                         running_font.latexWriteEndChanges(os, bparams,
2508                                         runparams, basefont, basefont);
2509                 }
2510 #else
2511 //FIXME: For now we ALWAYS have to close the foreign font settings if they are
2512 //FIXME: there as we start another \selectlanguage with the next paragraph if
2513 //FIXME: we are in need of this. This should be fixed sometime (Jug)
2514                 running_font.latexWriteEndChanges(os, bparams, runparams,
2515                                 basefont, basefont);
2516 #endif
2517         }
2518
2519         column += Changes::latexMarkChange(os, bparams, runningChange,
2520                                            Change(Change::UNCHANGED), runparams);
2521
2522         // Needed if there is an optional argument but no contents.
2523         if (body_pos > 0 && body_pos == size()) {
2524                 os << "}]~";
2525         }
2526
2527         if (allowcust && d->endTeXParParams(bparams, os, texrow, runparams)
2528             && runparams.encoding != prev_encoding) {
2529                 runparams.encoding = prev_encoding;
2530                 if (!runparams.isFullUnicode())
2531                         os << setEncoding(prev_encoding->iconvName());
2532         }
2533
2534         LYXERR(Debug::LATEX, "Paragraph::latex... done " << this);
2535 }
2536
2537
2538 bool Paragraph::emptyTag() const
2539 {
2540         for (pos_type i = 0; i < size(); ++i) {
2541                 if (Inset const * inset = getInset(i)) {
2542                         InsetCode lyx_code = inset->lyxCode();
2543                         // FIXME testing like that is wrong. What is
2544                         // the intent?
2545                         if (lyx_code != TOC_CODE &&
2546                             lyx_code != INCLUDE_CODE &&
2547                             lyx_code != GRAPHICS_CODE &&
2548                             lyx_code != ERT_CODE &&
2549                             lyx_code != LISTINGS_CODE &&
2550                             lyx_code != FLOAT_CODE &&
2551                             lyx_code != TABULAR_CODE) {
2552                                 return false;
2553                         }
2554                 } else {
2555                         char_type c = d->text_[i];
2556                         if (c != ' ' && c != '\t')
2557                                 return false;
2558                 }
2559         }
2560         return true;
2561 }
2562
2563
2564 string Paragraph::getID(Buffer const & buf, OutputParams const & runparams)
2565         const
2566 {
2567         for (pos_type i = 0; i < size(); ++i) {
2568                 if (Inset const * inset = getInset(i)) {
2569                         InsetCode lyx_code = inset->lyxCode();
2570                         if (lyx_code == LABEL_CODE) {
2571                                 InsetLabel const * const il = static_cast<InsetLabel const *>(inset);
2572                                 docstring const & id = il->getParam("name");
2573                                 return "id='" + to_utf8(sgml::cleanID(buf, runparams, id)) + "'";
2574                         }
2575                 }
2576         }
2577         return string();
2578 }
2579
2580
2581 pos_type Paragraph::firstWordDocBook(odocstream & os, OutputParams const & runparams)
2582         const
2583 {
2584         pos_type i;
2585         for (i = 0; i < size(); ++i) {
2586                 if (Inset const * inset = getInset(i)) {
2587                         inset->docbook(os, runparams);
2588                 } else {
2589                         char_type c = d->text_[i];
2590                         if (c == ' ')
2591                                 break;
2592                         os << sgml::escapeChar(c);
2593                 }
2594         }
2595         return i;
2596 }
2597
2598
2599 pos_type Paragraph::firstWordLyXHTML(XHTMLStream & xs, OutputParams const & runparams)
2600         const
2601 {
2602         pos_type i;
2603         for (i = 0; i < size(); ++i) {
2604                 if (Inset const * inset = getInset(i)) {
2605                         inset->xhtml(xs, runparams);
2606                 } else {
2607                         char_type c = d->text_[i];
2608                         if (c == ' ')
2609                                 break;
2610                         xs << c;
2611                 }
2612         }
2613         return i;
2614 }
2615
2616
2617 bool Paragraph::Private::onlyText(Buffer const & buf, Font const & outerfont, pos_type initial) const
2618 {
2619         Font font_old;
2620         pos_type size = text_.size();
2621         for (pos_type i = initial; i < size; ++i) {
2622                 Font font = owner_->getFont(buf.params(), i, outerfont);
2623                 if (text_[i] == META_INSET)
2624                         return false;
2625                 if (i != initial && font != font_old)
2626                         return false;
2627                 font_old = font;
2628         }
2629
2630         return true;
2631 }
2632
2633
2634 void Paragraph::simpleDocBookOnePar(Buffer const & buf,
2635                                     odocstream & os,
2636                                     OutputParams const & runparams,
2637                                     Font const & outerfont,
2638                                     pos_type initial) const
2639 {
2640         bool emph_flag = false;
2641
2642         Layout const & style = *d->layout_;
2643         FontInfo font_old =
2644                 style.labeltype == LABEL_MANUAL ? style.labelfont : style.font;
2645
2646         if (style.pass_thru && !d->onlyText(buf, outerfont, initial))
2647                 os << "]]>";
2648
2649         // parsing main loop
2650         for (pos_type i = initial; i < size(); ++i) {
2651                 Font font = getFont(buf.params(), i, outerfont);
2652
2653                 // handle <emphasis> tag
2654                 if (font_old.emph() != font.fontInfo().emph()) {
2655                         if (font.fontInfo().emph() == FONT_ON) {
2656                                 os << "<emphasis>";
2657                                 emph_flag = true;
2658                         } else if (i != initial) {
2659                                 os << "</emphasis>";
2660                                 emph_flag = false;
2661                         }
2662                 }
2663
2664                 if (Inset const * inset = getInset(i)) {
2665                         inset->docbook(os, runparams);
2666                 } else {
2667                         char_type c = d->text_[i];
2668
2669                         if (style.pass_thru)
2670                                 os.put(c);
2671                         else
2672                                 os << sgml::escapeChar(c);
2673                 }
2674                 font_old = font.fontInfo();
2675         }
2676
2677         if (emph_flag) {
2678                 os << "</emphasis>";
2679         }
2680
2681         if (style.free_spacing)
2682                 os << '\n';
2683         if (style.pass_thru && !d->onlyText(buf, outerfont, initial))
2684                 os << "<![CDATA[";
2685 }
2686
2687
2688 docstring Paragraph::simpleLyXHTMLOnePar(Buffer const & buf,
2689                                     XHTMLStream & xs,
2690                                     OutputParams const & runparams,
2691                                     Font const & outerfont,
2692                                     pos_type initial) const
2693 {
2694         docstring retval;
2695
2696         bool emph_flag = false;
2697         bool bold_flag = false;
2698         string closing_tag;
2699
2700         Layout const & style = *d->layout_;
2701
2702         if (!runparams.for_toc && runparams.html_make_pars) {
2703                 // generate a magic label for this paragraph
2704                 string const attr = "id='" + magicLabel() + "'";
2705                 xs << html::CompTag("a", attr);
2706         }
2707
2708         FontInfo font_old =
2709                 style.labeltype == LABEL_MANUAL ? style.labelfont : style.font;
2710
2711         // parsing main loop
2712         for (pos_type i = initial; i < size(); ++i) {
2713                 // let's not show deleted material in the output
2714                 if (isDeleted(i))
2715                         continue;
2716
2717                 Font font = getFont(buf.params(), i, outerfont);
2718
2719                 // emphasis
2720                 if (font_old.emph() != font.fontInfo().emph()) {
2721                         if (font.fontInfo().emph() == FONT_ON) {
2722                                 xs << html::StartTag("em");
2723                                 emph_flag = true;
2724                         } else if (emph_flag && i != initial) {
2725                                 xs << html::EndTag("em");
2726                                 emph_flag = false;
2727                         }
2728                 }
2729                 // bold
2730                 if (font_old.series() != font.fontInfo().series()) {
2731                         if (font.fontInfo().series() == BOLD_SERIES) {
2732                                 xs << html::StartTag("strong");
2733                                 bold_flag = true;
2734                         } else if (bold_flag && i != initial) {
2735                                 xs << html::EndTag("strong");
2736                                 bold_flag = false;
2737                         }
2738                 }
2739                 // FIXME XHTML
2740                 // Other such tags? What about the other text ranges?
2741
2742                 Inset const * inset = getInset(i);
2743                 if (inset) {
2744                         if (!runparams.for_toc || inset->isInToc()) {
2745                                 OutputParams np = runparams;
2746                                 if (!inset->getLayout().htmlisblock())
2747                                         np.html_in_par = true;
2748                                 retval += inset->xhtml(xs, np);
2749                         }
2750                 } else {
2751                         char_type c = d->text_[i];
2752
2753                         if (style.pass_thru)
2754                                 xs << c;
2755                         else if (c == '-') {
2756                                 docstring str;
2757                                 int j = i + 1;
2758                                 if (j < size() && d->text_[j] == '-') {
2759                                         j += 1;
2760                                         if (j < size() && d->text_[j] == '-') {
2761                                                 str += from_ascii("&mdash;");
2762                                                 i += 2;
2763                                         } else {
2764                                                 str += from_ascii("&ndash;");
2765                                                 i += 1;
2766                                         }
2767                                 }
2768                                 else
2769                                         str += c;
2770                                 // We don't want to escape the entities. Note that
2771                                 // it is safe to do this, since str can otherwise
2772                                 // only be "-". E.g., it can't be "<".
2773                                 xs << XHTMLStream::ESCAPE_NONE << str;
2774                         } else
2775                                 xs << c;
2776                 }
2777                 font_old = font.fontInfo();
2778         }
2779
2780         xs.closeFontTags();
2781         return retval;
2782 }
2783
2784
2785 bool Paragraph::isHfill(pos_type pos) const
2786 {
2787         Inset const * inset = getInset(pos);
2788         return inset && (inset->lyxCode() == SPACE_CODE &&
2789                          inset->isStretchableSpace());
2790 }
2791
2792
2793 bool Paragraph::isNewline(pos_type pos) const
2794 {
2795         Inset const * inset = getInset(pos);
2796         return inset && inset->lyxCode() == NEWLINE_CODE;
2797 }
2798
2799
2800 bool Paragraph::isLineSeparator(pos_type pos) const
2801 {
2802         char_type const c = d->text_[pos];
2803         if (isLineSeparatorChar(c))
2804                 return true;
2805         Inset const * inset = getInset(pos);
2806         return inset && inset->isLineSeparator();
2807 }
2808
2809
2810 bool Paragraph::isWordSeparator(pos_type pos) const
2811 {
2812         if (Inset const * inset = getInset(pos))
2813                 return !inset->isLetter();
2814         char_type const c = d->text_[pos];
2815         // We want to pass the ' and escape chars to the spellchecker
2816         static docstring const quote = from_utf8(lyxrc.spellchecker_esc_chars + '\'');
2817         return (!isLetterChar(c) && !isDigit(c) && !contains(quote, c))
2818                 || pos == size();
2819 }
2820
2821
2822 bool Paragraph::isChar(pos_type pos) const
2823 {
2824         if (Inset const * inset = getInset(pos))
2825                 return inset->isChar();
2826         char_type const c = d->text_[pos];
2827         return !isLetterChar(c) && !isDigit(c) && !lyx::isSpace(c);
2828 }
2829
2830
2831 bool Paragraph::isSpace(pos_type pos) const
2832 {
2833         if (Inset const * inset = getInset(pos))
2834                 return inset->isSpace();
2835         char_type const c = d->text_[pos];
2836         return lyx::isSpace(c);
2837 }
2838
2839
2840 Language const *
2841 Paragraph::getParLanguage(BufferParams const & bparams) const
2842 {
2843         if (!empty())
2844                 return getFirstFontSettings(bparams).language();
2845         // FIXME: we should check the prev par as well (Lgb)
2846         return bparams.language;
2847 }
2848
2849
2850 bool Paragraph::isRTL(BufferParams const & bparams) const
2851 {
2852         return lyxrc.rtl_support
2853                 && getParLanguage(bparams)->rightToLeft()
2854                 && !inInset().getLayout().forceLTR();
2855 }
2856
2857
2858 void Paragraph::changeLanguage(BufferParams const & bparams,
2859                                Language const * from, Language const * to)
2860 {
2861         // change language including dummy font change at the end
2862         for (pos_type i = 0; i <= size(); ++i) {
2863                 Font font = getFontSettings(bparams, i);
2864                 if (font.language() == from) {
2865                         font.setLanguage(to);
2866                         setFont(i, font);
2867                 }
2868         }
2869         d->requestSpellCheck(size());
2870 }
2871
2872
2873 bool Paragraph::isMultiLingual(BufferParams const & bparams) const
2874 {
2875         Language const * doc_language = bparams.language;
2876         FontList::const_iterator cit = d->fontlist_.begin();
2877         FontList::const_iterator end = d->fontlist_.end();
2878
2879         for (; cit != end; ++cit)
2880                 if (cit->font().language() != ignore_language &&
2881                     cit->font().language() != latex_language &&
2882                     cit->font().language() != doc_language)
2883                         return true;
2884         return false;
2885 }
2886
2887
2888 void Paragraph::getLanguages(std::set<Language const *> & languages) const
2889 {
2890         FontList::const_iterator cit = d->fontlist_.begin();
2891         FontList::const_iterator end = d->fontlist_.end();
2892
2893         for (; cit != end; ++cit) {
2894                 Language const * lang = cit->font().language();
2895                 if (lang != ignore_language &&
2896                     lang != latex_language)
2897                         languages.insert(lang);
2898         }
2899 }
2900
2901
2902 docstring Paragraph::asString(int options) const
2903 {
2904         return asString(0, size(), options);
2905 }
2906
2907
2908 docstring Paragraph::asString(pos_type beg, pos_type end, int options) const
2909 {
2910         odocstringstream os;
2911
2912         if (beg == 0
2913             && options & AS_STR_LABEL
2914             && !d->params_.labelString().empty())
2915                 os << d->params_.labelString() << ' ';
2916
2917         for (pos_type i = beg; i < end; ++i) {
2918                 if ((options & AS_STR_SKIPDELETE) && isDeleted(i))
2919                         continue;
2920                 char_type const c = d->text_[i];
2921                 if (isPrintable(c) || c == '\t'
2922                     || (c == '\n' && (options & AS_STR_NEWLINES)))
2923                         os.put(c);
2924                 else if (c == META_INSET && (options & AS_STR_INSETS)) {
2925                         getInset(i)->tocString(os);
2926                         if (getInset(i)->asInsetMath())
2927                                 os << " ";
2928                 }
2929         }
2930
2931         return os.str();
2932 }
2933
2934
2935 docstring Paragraph::stringify(pos_type beg, pos_type end, int options, OutputParams & runparams) const
2936 {
2937         odocstringstream os;
2938
2939         if (beg == 0
2940                 && options & AS_STR_LABEL
2941                 && !d->params_.labelString().empty())
2942                 os << d->params_.labelString() << ' ';
2943
2944         for (pos_type i = beg; i < end; ++i) {
2945                 char_type const c = d->text_[i];
2946                 if (isPrintable(c) || c == '\t'
2947                     || (c == '\n' && (options & AS_STR_NEWLINES)))
2948                         os.put(c);
2949                 else if (c == META_INSET && (options & AS_STR_INSETS)) {
2950                         getInset(i)->plaintext(os, runparams);
2951                 }
2952         }
2953
2954         return os.str();
2955 }
2956
2957
2958 void Paragraph::setInsetOwner(Inset const * inset)
2959 {
2960         d->inset_owner_ = inset;
2961 }
2962
2963
2964 int Paragraph::id() const
2965 {
2966         return d->id_;
2967 }
2968
2969
2970 void Paragraph::setId(int id)
2971 {
2972         d->id_ = id;
2973 }
2974
2975
2976 Layout const & Paragraph::layout() const
2977 {
2978         return *d->layout_;
2979 }
2980
2981
2982 void Paragraph::setLayout(Layout const & layout)
2983 {
2984         d->layout_ = &layout;
2985 }
2986
2987
2988 void Paragraph::setDefaultLayout(DocumentClass const & tc)
2989 {
2990         setLayout(tc.defaultLayout());
2991 }
2992
2993
2994 void Paragraph::setPlainLayout(DocumentClass const & tc)
2995 {
2996         setLayout(tc.plainLayout());
2997 }
2998
2999
3000 void Paragraph::setPlainOrDefaultLayout(DocumentClass const & tclass)
3001 {
3002         if (usePlainLayout())
3003                 setPlainLayout(tclass);
3004         else
3005                 setDefaultLayout(tclass);
3006 }
3007
3008
3009 Inset const & Paragraph::inInset() const
3010 {
3011         LASSERT(d->inset_owner_, throw ExceptionMessage(BufferException,
3012                 _("Memory problem"), _("Paragraph not properly initialized")));
3013         return *d->inset_owner_;
3014 }
3015
3016
3017 ParagraphParameters & Paragraph::params()
3018 {
3019         return d->params_;
3020 }
3021
3022
3023 ParagraphParameters const & Paragraph::params() const
3024 {
3025         return d->params_;
3026 }
3027
3028
3029 bool Paragraph::isFreeSpacing() const
3030 {
3031         if (d->layout_->free_spacing)
3032                 return true;
3033         return d->inset_owner_ && d->inset_owner_->isFreeSpacing();
3034 }
3035
3036
3037 bool Paragraph::allowEmpty() const
3038 {
3039         if (d->layout_->keepempty)
3040                 return true;
3041         return d->inset_owner_ && d->inset_owner_->allowEmpty();
3042 }
3043
3044
3045 char_type Paragraph::transformChar(char_type c, pos_type pos) const
3046 {
3047         if (!Encodings::isArabicChar(c))
3048                 return c;
3049
3050         char_type prev_char = ' ';
3051         char_type next_char = ' ';
3052
3053         for (pos_type i = pos - 1; i >= 0; --i) {
3054                 char_type const par_char = d->text_[i];
3055                 if (!Encodings::isArabicComposeChar(par_char)) {
3056                         prev_char = par_char;
3057                         break;
3058                 }
3059         }
3060
3061         for (pos_type i = pos + 1, end = size(); i < end; ++i) {
3062                 char_type const par_char = d->text_[i];
3063                 if (!Encodings::isArabicComposeChar(par_char)) {
3064                         next_char = par_char;
3065                         break;
3066                 }
3067         }
3068
3069         if (Encodings::isArabicChar(next_char)) {
3070                 if (Encodings::isArabicChar(prev_char) &&
3071                         !Encodings::isArabicSpecialChar(prev_char))
3072                         return Encodings::transformChar(c, Encodings::FORM_MEDIAL);
3073                 else
3074                         return Encodings::transformChar(c, Encodings::FORM_INITIAL);
3075         } else {
3076                 if (Encodings::isArabicChar(prev_char) &&
3077                         !Encodings::isArabicSpecialChar(prev_char))
3078                         return Encodings::transformChar(c, Encodings::FORM_FINAL);
3079                 else
3080                         return Encodings::transformChar(c, Encodings::FORM_ISOLATED);
3081         }
3082 }
3083
3084
3085 int Paragraph::checkBiblio(Buffer const & buffer)
3086 {
3087         // FIXME From JS:
3088         // This is getting more and more a mess. ...We really should clean
3089         // up this bibitem issue for 1.6.
3090
3091         // Add bibitem insets if necessary
3092         if (d->layout_->labeltype != LABEL_BIBLIO)
3093                 return 0;
3094
3095         bool hasbibitem = !d->insetlist_.empty()
3096                 // Insist on it being in pos 0
3097                 && d->text_[0] == META_INSET
3098                 && d->insetlist_.begin()->inset->lyxCode() == BIBITEM_CODE;
3099
3100         bool track_changes = buffer.params().trackChanges;
3101
3102         docstring oldkey;
3103         docstring oldlabel;
3104
3105         // remove a bibitem in pos != 0
3106         // restore it later in pos 0 if necessary
3107         // (e.g. if a user inserts contents _before_ the item)
3108         // we're assuming there's only one of these, which there
3109         // should be.
3110         int erasedInsetPosition = -1;
3111         InsetList::iterator it = d->insetlist_.begin();
3112         InsetList::iterator end = d->insetlist_.end();
3113         for (; it != end; ++it)
3114                 if (it->inset->lyxCode() == BIBITEM_CODE
3115                       && it->pos > 0) {
3116                         InsetCommand * olditem = it->inset->asInsetCommand();
3117                         oldkey = olditem->getParam("key");
3118                         oldlabel = olditem->getParam("label");
3119                         erasedInsetPosition = it->pos;
3120                         eraseChar(erasedInsetPosition, track_changes);
3121                         break;
3122         }
3123
3124         // There was an InsetBibitem at the beginning, and we didn't
3125         // have to erase one.
3126         if (hasbibitem && erasedInsetPosition < 0)
3127                         return 0;
3128
3129         // There was an InsetBibitem at the beginning and we did have to
3130         // erase one. So we give its properties to the beginning inset.
3131         if (hasbibitem) {
3132                 InsetCommand * inset = d->insetlist_.begin()->inset->asInsetCommand();
3133                 if (!oldkey.empty())
3134                         inset->setParam("key", oldkey);
3135                 inset->setParam("label", oldlabel);
3136                 return -erasedInsetPosition;
3137         }
3138
3139         // There was no inset at the beginning, so we need to create one with
3140         // the key and label of the one we erased.
3141         InsetBibitem * inset =
3142                 new InsetBibitem(const_cast<Buffer *>(&buffer), InsetCommandParams(BIBITEM_CODE));
3143         // restore values of previously deleted item in this par.
3144         if (!oldkey.empty())
3145                 inset->setParam("key", oldkey);
3146         inset->setParam("label", oldlabel);
3147         insertInset(0, inset,
3148                     Change(track_changes ? Change::INSERTED : Change::UNCHANGED));
3149
3150         return 1;
3151 }
3152
3153
3154 void Paragraph::checkAuthors(AuthorList const & authorList)
3155 {
3156         d->changes_.checkAuthors(authorList);
3157 }
3158
3159
3160 bool Paragraph::isChanged(pos_type pos) const
3161 {
3162         return lookupChange(pos).changed();
3163 }
3164
3165
3166 bool Paragraph::isInserted(pos_type pos) const
3167 {
3168         return lookupChange(pos).inserted();
3169 }
3170
3171
3172 bool Paragraph::isDeleted(pos_type pos) const
3173 {
3174         return lookupChange(pos).deleted();
3175 }
3176
3177
3178 InsetList const & Paragraph::insetList() const
3179 {
3180         return d->insetlist_;
3181 }
3182
3183
3184 void Paragraph::setBuffer(Buffer & b)
3185 {
3186         d->insetlist_.setBuffer(b);
3187 }
3188
3189
3190 Inset * Paragraph::releaseInset(pos_type pos)
3191 {
3192         Inset * inset = d->insetlist_.release(pos);
3193         /// does not honour change tracking!
3194         eraseChar(pos, false);
3195         return inset;
3196 }
3197
3198
3199 Inset * Paragraph::getInset(pos_type pos)
3200 {
3201         return (pos < pos_type(d->text_.size()) && d->text_[pos] == META_INSET)
3202                  ? d->insetlist_.get(pos) : 0;
3203 }
3204
3205
3206 Inset const * Paragraph::getInset(pos_type pos) const
3207 {
3208         return (pos < pos_type(d->text_.size()) && d->text_[pos] == META_INSET)
3209                  ? d->insetlist_.get(pos) : 0;
3210 }
3211
3212
3213 void Paragraph::changeCase(BufferParams const & bparams, pos_type pos,
3214                 pos_type & right, TextCase action)
3215 {
3216         // process sequences of modified characters; in change
3217         // tracking mode, this approach results in much better
3218         // usability than changing case on a char-by-char basis
3219         docstring changes;
3220
3221         bool const trackChanges = bparams.trackChanges;
3222
3223         bool capitalize = true;
3224
3225         for (; pos < right; ++pos) {
3226                 char_type oldChar = d->text_[pos];
3227                 char_type newChar = oldChar;
3228
3229                 // ignore insets and don't play with deleted text!
3230                 if (oldChar != META_INSET && !isDeleted(pos)) {
3231                         switch (action) {
3232                                 case text_lowercase:
3233                                         newChar = lowercase(oldChar);
3234                                         break;
3235                                 case text_capitalization:
3236                                         if (capitalize) {
3237                                                 newChar = uppercase(oldChar);
3238                                                 capitalize = false;
3239                                         }
3240                                         break;
3241                                 case text_uppercase:
3242                                         newChar = uppercase(oldChar);
3243                                         break;
3244                         }
3245                 }
3246
3247                 if (isWordSeparator(pos) || isDeleted(pos)) {
3248                         // permit capitalization again
3249                         capitalize = true;
3250                 }
3251
3252                 if (oldChar != newChar) {
3253                         changes += newChar;
3254                         if (pos != right - 1)
3255                                 continue;
3256                         // step behind the changing area
3257                         pos++;
3258                 }
3259
3260                 int erasePos = pos - changes.size();
3261                 for (size_t i = 0; i < changes.size(); i++) {
3262                         insertChar(pos, changes[i],
3263                                    getFontSettings(bparams,
3264                                                    erasePos),
3265                                    trackChanges);
3266                         if (!eraseChar(erasePos, trackChanges)) {
3267                                 ++erasePos;
3268                                 ++pos; // advance
3269                                 ++right; // expand selection
3270                         }
3271                 }
3272                 changes.clear();
3273         }
3274 }
3275
3276
3277 int Paragraph::find(docstring const & str, bool cs, bool mw,
3278                 pos_type start_pos, bool del) const
3279 {
3280         pos_type pos = start_pos;
3281         int const strsize = str.length();
3282         int i = 0;
3283         pos_type const parsize = d->text_.size();
3284         for (i = 0; i < strsize && pos < parsize; ++i, ++pos) {
3285                 // Ignore ligature break and hyphenation chars while searching
3286                 while (pos < parsize - 1 && isInset(pos)) {
3287                         const InsetSpecialChar *isc = dynamic_cast<const InsetSpecialChar*>(getInset(pos));
3288                         if (isc == 0
3289                             || (isc->kind() != InsetSpecialChar::HYPHENATION
3290                                 && isc->kind() != InsetSpecialChar::LIGATURE_BREAK))
3291                                 break;
3292                         pos++;
3293                 }
3294                 if (cs && str[i] != d->text_[pos])
3295                         break;
3296                 if (!cs && uppercase(str[i]) != uppercase(d->text_[pos]))
3297                         break;
3298                 if (!del && isDeleted(pos))
3299                         break;
3300         }
3301
3302         if (i != strsize)
3303                 return 0;
3304
3305         // if necessary, check whether string matches word
3306         if (mw) {
3307                 if (start_pos > 0 && !isWordSeparator(start_pos - 1))
3308                         return 0;
3309                 if (pos < parsize
3310                         && !isWordSeparator(pos))
3311                         return 0;
3312         }
3313
3314         return pos - start_pos;
3315 }
3316
3317
3318 char_type Paragraph::getChar(pos_type pos) const
3319 {
3320         return d->text_[pos];
3321 }
3322
3323
3324 pos_type Paragraph::size() const
3325 {
3326         return d->text_.size();
3327 }
3328
3329
3330 bool Paragraph::empty() const
3331 {
3332         return d->text_.empty();
3333 }
3334
3335
3336 bool Paragraph::isInset(pos_type pos) const
3337 {
3338         return d->text_[pos] == META_INSET;
3339 }
3340
3341
3342 bool Paragraph::isSeparator(pos_type pos) const
3343 {
3344         //FIXME: Are we sure this can be the only separator?
3345         return d->text_[pos] == ' ';
3346 }
3347
3348
3349 void Paragraph::deregisterWords()
3350 {
3351         Private::LangWordsMap::const_iterator itl = d->words_.begin();
3352         Private::LangWordsMap::const_iterator ite = d->words_.end();
3353         for (; itl != ite; ++itl) {
3354                 WordList * wl = theWordList(itl->first);
3355                 Private::Words::const_iterator it = (itl->second).begin();
3356                 Private::Words::const_iterator et = (itl->second).end();
3357                 for (; it != et; ++it)
3358                         wl->remove(*it);
3359         }
3360         d->words_.clear();
3361 }
3362
3363
3364 void Paragraph::locateWord(pos_type & from, pos_type & to,
3365         word_location const loc) const
3366 {
3367         switch (loc) {
3368         case WHOLE_WORD_STRICT:
3369                 if (from == 0 || from == size()
3370                     || isWordSeparator(from)
3371                     || isWordSeparator(from - 1)) {
3372                         to = from;
3373                         return;
3374                 }
3375                 // no break here, we go to the next
3376
3377         case WHOLE_WORD:
3378                 // If we are already at the beginning of a word, do nothing
3379                 if (!from || isWordSeparator(from - 1))
3380                         break;
3381                 // no break here, we go to the next
3382
3383         case PREVIOUS_WORD:
3384                 // always move the cursor to the beginning of previous word
3385                 while (from && !isWordSeparator(from - 1))
3386                         --from;
3387                 break;
3388         case NEXT_WORD:
3389                 LYXERR0("Paragraph::locateWord: NEXT_WORD not implemented yet");
3390                 break;
3391         case PARTIAL_WORD:
3392                 // no need to move the 'from' cursor
3393                 break;
3394         }
3395         to = from;
3396         while (to < size() && !isWordSeparator(to))
3397                 ++to;
3398 }
3399
3400
3401 void Paragraph::collectWords()
3402 {
3403         // This is the value that needs to be exposed in the preferences
3404         // to resolve bug #6760.
3405         static int minlength = 6;
3406         pos_type n = size();
3407         for (pos_type pos = 0; pos < n; ++pos) {
3408                 if (isWordSeparator(pos))
3409                         continue;
3410                 pos_type from = pos;
3411                 locateWord(from, pos, WHOLE_WORD);
3412                 if (pos - from >= minlength) {
3413                         docstring word = asString(from, pos, AS_STR_NONE);
3414                         FontList::const_iterator cit = d->fontlist_.fontIterator(pos);
3415                         if (cit == d->fontlist_.end())
3416                                 return;
3417                         Language const * lang = cit->font().language();
3418                         d->words_[*lang].insert(word);
3419                 }
3420         }
3421 }
3422
3423
3424 void Paragraph::registerWords()
3425 {
3426         Private::LangWordsMap::const_iterator itl = d->words_.begin();
3427         Private::LangWordsMap::const_iterator ite = d->words_.end();
3428         for (; itl != ite; ++itl) {
3429                 WordList * wl = theWordList(itl->first);
3430                 Private::Words::const_iterator it = (itl->second).begin();
3431                 Private::Words::const_iterator et = (itl->second).end();
3432                 for (; it != et; ++it)
3433                         wl->insert(*it);
3434         }
3435 }
3436
3437
3438 void Paragraph::updateWords()
3439 {
3440         deregisterWords();
3441         collectWords();
3442         registerWords();
3443 }
3444
3445
3446 void Paragraph::Private::appendSkipPosition(SkipPositions & skips, pos_type const pos) const
3447 {
3448         SkipPositionsIterator begin = skips.begin();
3449         SkipPositions::iterator end = skips.end();
3450         if (pos > 0 && begin < end) {
3451                 --end;
3452                 if (end->last == pos - 1) {
3453                         end->last = pos;
3454                         return;
3455                 }
3456         }
3457         skips.insert(end, FontSpan(pos, pos));
3458 }
3459
3460
3461 Language * Paragraph::Private::locateSpellRange(
3462         pos_type & from, pos_type & to,
3463         SkipPositions & skips) const
3464 {
3465         // skip leading white space
3466         while (from < to && owner_->isWordSeparator(from))
3467                 ++from;
3468         // don't check empty range
3469         if (from >= to)
3470                 return 0;
3471         // get current language
3472         Language * lang = getSpellLanguage(from);
3473         pos_type last = from;
3474         bool samelang = true;
3475         bool sameinset = true;
3476         while (last < to && samelang && sameinset) {
3477                 // hop to end of word
3478                 while (last < to && !owner_->isWordSeparator(last)) {
3479                         if (owner_->getInset(last)) {
3480                                 appendSkipPosition(skips, last);
3481                         } else if (owner_->isDeleted(last)) {
3482                                 appendSkipPosition(skips, last);
3483                         }
3484                         ++last;
3485                 }
3486                 // hop to next word while checking for insets
3487                 while (sameinset && last < to && owner_->isWordSeparator(last)) {
3488                         if (Inset const * inset = owner_->getInset(last))
3489                                 sameinset = inset->isChar() && inset->isLetter();
3490                         if (sameinset && owner_->isDeleted(last)) {
3491                                 appendSkipPosition(skips, last);
3492                         }
3493                         if (sameinset)
3494                                 last++;
3495                 }
3496                 if (sameinset && last < to) {
3497                         // now check for language change
3498                         samelang = lang == getSpellLanguage(last);
3499                 }
3500         }
3501         // if language change detected backstep is needed
3502         if (!samelang)
3503                 --last;
3504         to = last;
3505         return lang;
3506 }
3507
3508
3509 Language * Paragraph::Private::getSpellLanguage(pos_type const from) const
3510 {
3511         Language * lang =
3512                 const_cast<Language *>(owner_->getFontSettings(
3513                         inset_owner_->buffer().params(), from).language());
3514         if (lang == inset_owner_->buffer().params().language
3515                 && !lyxrc.spellchecker_alt_lang.empty()) {
3516                 string lang_code;
3517                 string const lang_variety =
3518                         split(lyxrc.spellchecker_alt_lang, lang_code, '-');
3519                 lang->setCode(lang_code);
3520                 lang->setVariety(lang_variety);
3521         }
3522         return lang;
3523 }
3524
3525
3526 void Paragraph::requestSpellCheck(pos_type pos)
3527 {
3528         d->requestSpellCheck(pos == -1 ? size() : pos);
3529 }
3530
3531
3532 bool Paragraph::needsSpellCheck() const
3533 {
3534         SpellChecker::ChangeNumber speller_change_number = 0;
3535         if (theSpellChecker())
3536                 speller_change_number = theSpellChecker()->changeNumber();
3537         if (speller_change_number > d->speller_state_.currentChangeNumber()) {
3538                 d->speller_state_.needsCompleteRefresh(speller_change_number);
3539         }
3540         return d->needsSpellCheck();
3541 }
3542
3543
3544 SpellChecker::Result Paragraph::spellCheck(pos_type & from, pos_type & to,
3545         WordLangTuple & wl, docstring_list & suggestions,
3546         bool do_suggestion, bool check_learned) const
3547 {
3548         SpellChecker::Result result = SpellChecker::WORD_OK;
3549         SpellChecker * speller = theSpellChecker();
3550         if (!speller)
3551                 return result;
3552
3553         if (!d->layout_->spellcheck || !inInset().allowSpellCheck())
3554                 return result;
3555
3556         locateWord(from, to, WHOLE_WORD);
3557         if (from == to || from >= size())
3558                 return result;
3559
3560         docstring word = asString(from, to, AS_STR_INSETS + AS_STR_SKIPDELETE);
3561         Language * lang = d->getSpellLanguage(from);
3562
3563         wl = WordLangTuple(word, lang);
3564
3565         if (!word.size())
3566                 return result;
3567
3568         if (needsSpellCheck() || check_learned) {
3569                 // Ignore words with digits
3570                 // FIXME: make this customizable
3571                 // (note that some checkers ignore words with digits by default)
3572                 if (!hasDigit(word)) {
3573                         bool const trailing_dot = to < size() && d->text_[to] == '.';
3574                         result = speller->check(wl);
3575                         if (SpellChecker::misspelled(result) && trailing_dot) {
3576                                 wl = WordLangTuple(word.append(from_ascii(".")), lang);
3577                                 result = speller->check(wl);
3578                                 if (!SpellChecker::misspelled(result)) {
3579                                         LYXERR(Debug::GUI, "misspelled word is correct with dot: \"" <<
3580                                            word << "\" [" <<
3581                                            from << ".." << to << "]");
3582                                 }
3583                         }
3584                 }
3585                 d->setMisspelled(from, to, result);
3586         } else {
3587                 result = d->speller_state_.getState(from);
3588         }
3589
3590         bool const misspelled_ = SpellChecker::misspelled(result) ;
3591         if (misspelled_ && do_suggestion)
3592                 speller->suggest(wl, suggestions);
3593         else if (misspelled_)
3594                 LYXERR(Debug::GUI, "misspelled word: \"" <<
3595                            word << "\" [" <<
3596                            from << ".." << to << "]");
3597         else
3598                 suggestions.clear();
3599
3600         return result;
3601 }
3602
3603
3604 void Paragraph::Private::markMisspelledWords(
3605         pos_type const & first, pos_type const & last,
3606         SpellChecker::Result result,
3607         docstring const & word,
3608         SkipPositions const & skips)
3609 {
3610         if (!SpellChecker::misspelled(result)) {
3611                 setMisspelled(first, last, SpellChecker::WORD_OK);
3612                 return;
3613         }
3614         int snext = first;
3615         SpellChecker * speller = theSpellChecker();
3616         // locate and enumerate the error positions
3617         int nerrors = speller->numMisspelledWords();
3618         int numskipped = 0;
3619         SkipPositionsIterator it = skips.begin();
3620         SkipPositionsIterator et = skips.end();
3621         for (int index = 0; index < nerrors; ++index) {
3622                 int wstart;
3623                 int wlen = 0;
3624                 speller->misspelledWord(index, wstart, wlen);
3625                 /// should not happen if speller supports range checks
3626                 if (!wlen) continue;
3627                 docstring const misspelled = word.substr(wstart, wlen);
3628                 wstart += first + numskipped;
3629                 if (snext < wstart) {
3630                         /// mark the range of correct spelling
3631                         numskipped += countSkips(it, et, wstart);
3632                         setMisspelled(snext,
3633                                 wstart - 1, SpellChecker::WORD_OK);
3634                 }
3635                 snext = wstart + wlen;
3636                 numskipped += countSkips(it, et, snext);
3637                 /// mark the range of misspelling
3638                 setMisspelled(wstart, snext, result);
3639                 LYXERR(Debug::GUI, "misspelled word: \"" <<
3640                            misspelled << "\" [" <<
3641                            wstart << ".." << (snext-1) << "]");
3642                 ++snext;
3643         }
3644         if (snext <= last) {
3645                 /// mark the range of correct spelling at end
3646                 setMisspelled(snext, last, SpellChecker::WORD_OK);
3647         }
3648 }
3649
3650
3651 void Paragraph::spellCheck() const
3652 {
3653         SpellChecker * speller = theSpellChecker();
3654         if (!speller || !size() ||!needsSpellCheck())
3655                 return;
3656         pos_type start;
3657         pos_type endpos;
3658         d->rangeOfSpellCheck(start, endpos);
3659         if (speller->canCheckParagraph()) {
3660                 // loop until we leave the range
3661                 for (pos_type first = start; first < endpos; ) {
3662                         pos_type last = endpos;
3663                         Private::SkipPositions skips;
3664                         Language * lang = d->locateSpellRange(first, last, skips);
3665                         if (first >= endpos)
3666                                 break;
3667                         // start the spell checker on the unit of meaning
3668                         docstring word = asString(first, last, AS_STR_INSETS + AS_STR_SKIPDELETE);
3669                         WordLangTuple wl = WordLangTuple(word, lang);
3670                         SpellChecker::Result result = word.size() ?
3671                                 speller->check(wl) : SpellChecker::WORD_OK;
3672                         d->markMisspelledWords(first, last, result, word, skips);
3673                         first = ++last;
3674                 }
3675         } else {
3676                 static docstring_list suggestions;
3677                 pos_type to = endpos;
3678                 while (start < endpos) {
3679                         WordLangTuple wl;
3680                         spellCheck(start, to, wl, suggestions, false);
3681                         start = to + 1;
3682                 }
3683         }
3684         d->readySpellCheck();
3685 }
3686
3687
3688 bool Paragraph::isMisspelled(pos_type pos) const
3689 {
3690         return SpellChecker::misspelled(d->speller_state_.getState(pos));
3691 }
3692
3693
3694 string Paragraph::magicLabel() const
3695 {
3696         stringstream ss;
3697         ss << "magicparlabel-" << id();
3698         return ss.str();
3699 }
3700
3701
3702 } // namespace lyx