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                 odocstringstream ods;
1336                 // we have to provide all the optional arguments here, even though
1337                 // the last one is the only one we care about.
1338                 owner_->latex(bp, f, ods, tr, features.runparams(), 0, -1, true);
1339                 docstring const d = ods.str();
1340                 if (!d.empty()) {
1341                         // this will have "{" at the beginning, but not at the end
1342                         string const content = to_utf8(d);
1343                         string const cmd = layout_->latexname();
1344                         features.addPreambleSnippet("\\" + cmd + content + "}");
1345                 }
1346         }
1347
1348         if (features.runparams().flavor == OutputParams::HTML
1349             && layout_->htmltitle()) {
1350                 features.setHTMLTitle(owner_->asString(AS_STR_INSETS));
1351         }
1352
1353         // check the params.
1354         if (!params_.spacing().isDefault())
1355                 features.require("setspace");
1356
1357         // then the layouts
1358         features.useLayout(layout_->name());
1359
1360         // then the fonts
1361         fontlist_.validate(features);
1362
1363         // then the indentation
1364         if (!params_.leftIndent().zero())
1365                 features.require("ParagraphLeftIndent");
1366
1367         // then the insets
1368         InsetList::const_iterator icit = insetlist_.begin();
1369         InsetList::const_iterator iend = insetlist_.end();
1370         for (; icit != iend; ++icit) {
1371                 if (icit->inset) {
1372                         icit->inset->validate(features);
1373                         if (layout_->needprotect &&
1374                             icit->inset->lyxCode() == FOOT_CODE)
1375                                 features.require("NeedLyXFootnoteCode");
1376                 }
1377         }
1378
1379         // then the contents
1380         for (pos_type i = 0; i < int(text_.size()) ; ++i) {
1381                 for (size_t pnr = 0; pnr < phrases_nr; ++pnr) {
1382                         if (!special_phrases[pnr].builtin
1383                             && isTextAt(special_phrases[pnr].phrase, i)) {
1384                                 features.require(special_phrases[pnr].phrase);
1385                                 break;
1386                         }
1387                 }
1388                 Encodings::validate(text_[i], features);
1389         }
1390 }
1391
1392 /////////////////////////////////////////////////////////////////////
1393 //
1394 // Paragraph
1395 //
1396 /////////////////////////////////////////////////////////////////////
1397
1398 namespace {
1399         Layout const emptyParagraphLayout;
1400 }
1401
1402 Paragraph::Paragraph()
1403         : d(new Paragraph::Private(this, emptyParagraphLayout))
1404 {
1405         itemdepth = 0;
1406         d->params_.clear();
1407 }
1408
1409
1410 Paragraph::Paragraph(Paragraph const & par)
1411         : itemdepth(par.itemdepth),
1412         d(new Paragraph::Private(*par.d, this))
1413 {
1414         registerWords();
1415 }
1416
1417
1418 Paragraph::Paragraph(Paragraph const & par, pos_type beg, pos_type end)
1419         : itemdepth(par.itemdepth),
1420         d(new Paragraph::Private(*par.d, this, beg, end))
1421 {
1422         registerWords();
1423 }
1424
1425
1426 Paragraph & Paragraph::operator=(Paragraph const & par)
1427 {
1428         // needed as we will destroy the private part before copying it
1429         if (&par != this) {
1430                 itemdepth = par.itemdepth;
1431
1432                 deregisterWords();
1433                 delete d;
1434                 d = new Private(*par.d, this);
1435                 registerWords();
1436         }
1437         return *this;
1438 }
1439
1440
1441 Paragraph::~Paragraph()
1442 {
1443         deregisterWords();
1444         delete d;
1445 }
1446
1447
1448 namespace {
1449
1450 // this shall be called just before every "os << ..." action.
1451 void flushString(ostream & os, docstring & s)
1452 {
1453         os << to_utf8(s);
1454         s.erase();
1455 }
1456
1457 }
1458
1459
1460 void Paragraph::write(ostream & os, BufferParams const & bparams,
1461         depth_type & dth) const
1462 {
1463         // The beginning or end of a deeper (i.e. nested) area?
1464         if (dth != d->params_.depth()) {
1465                 if (d->params_.depth() > dth) {
1466                         while (d->params_.depth() > dth) {
1467                                 os << "\n\\begin_deeper";
1468                                 ++dth;
1469                         }
1470                 } else {
1471                         while (d->params_.depth() < dth) {
1472                                 os << "\n\\end_deeper";
1473                                 --dth;
1474                         }
1475                 }
1476         }
1477
1478         // First write the layout
1479         os << "\n\\begin_layout " << to_utf8(d->layout_->name()) << '\n';
1480
1481         d->params_.write(os);
1482
1483         Font font1(inherit_font, bparams.language);
1484
1485         Change running_change = Change(Change::UNCHANGED);
1486
1487         // this string is used as a buffer to avoid repetitive calls
1488         // to to_utf8(), which turn out to be expensive (JMarc)
1489         docstring write_buffer;
1490
1491         int column = 0;
1492         for (pos_type i = 0; i <= size(); ++i) {
1493
1494                 Change const change = lookupChange(i);
1495                 if (change != running_change)
1496                         flushString(os, write_buffer);
1497                 Changes::lyxMarkChange(os, bparams, column, running_change, change);
1498                 running_change = change;
1499
1500                 if (i == size())
1501                         break;
1502
1503                 // Write font changes
1504                 Font font2 = getFontSettings(bparams, i);
1505                 if (font2 != font1) {
1506                         flushString(os, write_buffer);
1507                         font2.lyxWriteChanges(font1, os);
1508                         column = 0;
1509                         font1 = font2;
1510                 }
1511
1512                 char_type const c = d->text_[i];
1513                 switch (c) {
1514                 case META_INSET:
1515                         if (Inset const * inset = getInset(i)) {
1516                                 flushString(os, write_buffer);
1517                                 if (inset->directWrite()) {
1518                                         // international char, let it write
1519                                         // code directly so it's shorter in
1520                                         // the file
1521                                         inset->write(os);
1522                                 } else {
1523                                         if (i)
1524                                                 os << '\n';
1525                                         os << "\\begin_inset ";
1526                                         inset->write(os);
1527                                         os << "\n\\end_inset\n\n";
1528                                         column = 0;
1529                                 }
1530                         }
1531                         break;
1532                 case '\\':
1533                         flushString(os, write_buffer);
1534                         os << "\n\\backslash\n";
1535                         column = 0;
1536                         break;
1537                 case '.':
1538                         flushString(os, write_buffer);
1539                         if (i + 1 < size() && d->text_[i + 1] == ' ') {
1540                                 os << ".\n";
1541                                 column = 0;
1542                         } else
1543                                 os << '.';
1544                         break;
1545                 default:
1546                         if ((column > 70 && c == ' ')
1547                             || column > 79) {
1548                                 flushString(os, write_buffer);
1549                                 os << '\n';
1550                                 column = 0;
1551                         }
1552                         // this check is to amend a bug. LyX sometimes
1553                         // inserts '\0' this could cause problems.
1554                         if (c != '\0')
1555                                 write_buffer.push_back(c);
1556                         else
1557                                 LYXERR0("NUL char in structure.");
1558                         ++column;
1559                         break;
1560                 }
1561         }
1562
1563         flushString(os, write_buffer);
1564         os << "\n\\end_layout\n";
1565 }
1566
1567
1568 void Paragraph::validate(LaTeXFeatures & features) const
1569 {
1570         d->validate(features);
1571 }
1572
1573
1574 void Paragraph::insert(pos_type start, docstring const & str,
1575                        Font const & font, Change const & change)
1576 {
1577         for (size_t i = 0, n = str.size(); i != n ; ++i)
1578                 insertChar(start + i, str[i], font, change);
1579 }
1580
1581
1582 void Paragraph::appendChar(char_type c, Font const & font,
1583                 Change const & change)
1584 {
1585         // track change
1586         d->changes_.insert(change, d->text_.size());
1587         // when appending characters, no need to update tables
1588         d->text_.push_back(c);
1589         setFont(d->text_.size() - 1, font);
1590 }
1591
1592
1593 void Paragraph::appendString(docstring const & s, Font const & font,
1594                 Change const & change)
1595 {
1596         pos_type end = s.size();
1597         size_t oldsize = d->text_.size();
1598         size_t newsize = oldsize + end;
1599         size_t capacity = d->text_.capacity();
1600         if (newsize >= capacity)
1601                 d->text_.reserve(max(capacity + 100, newsize));
1602
1603         // when appending characters, no need to update tables
1604         d->text_.append(s);
1605
1606         // FIXME: Optimize this!
1607         for (size_t i = oldsize; i != newsize; ++i) {
1608                 // track change
1609                 d->changes_.insert(change, i);
1610         }
1611         d->fontlist_.set(oldsize, font);
1612         d->fontlist_.set(newsize - 1, font);
1613 }
1614
1615
1616 void Paragraph::insertChar(pos_type pos, char_type c,
1617                            bool trackChanges)
1618 {
1619         d->insertChar(pos, c, Change(trackChanges ?
1620                            Change::INSERTED : Change::UNCHANGED));
1621 }
1622
1623
1624 void Paragraph::insertChar(pos_type pos, char_type c,
1625                            Font const & font, bool trackChanges)
1626 {
1627         d->insertChar(pos, c, Change(trackChanges ?
1628                            Change::INSERTED : Change::UNCHANGED));
1629         setFont(pos, font);
1630 }
1631
1632
1633 void Paragraph::insertChar(pos_type pos, char_type c,
1634                            Font const & font, Change const & change)
1635 {
1636         d->insertChar(pos, c, change);
1637         setFont(pos, font);
1638 }
1639
1640
1641 bool Paragraph::insertInset(pos_type pos, Inset * inset,
1642                             Font const & font, Change const & change)
1643 {
1644         bool const success = insertInset(pos, inset, change);
1645         // Set the font/language of the inset...
1646         setFont(pos, font);
1647         return success;
1648 }
1649
1650
1651 void Paragraph::resetFonts(Font const & font)
1652 {
1653         d->fontlist_.clear();
1654         d->fontlist_.set(0, font);
1655         d->fontlist_.set(d->text_.size() - 1, font);
1656 }
1657
1658 // Gets uninstantiated font setting at position.
1659 Font const & Paragraph::getFontSettings(BufferParams const & bparams,
1660                                          pos_type pos) const
1661 {
1662         if (pos > size()) {
1663                 LYXERR0("pos: " << pos << " size: " << size());
1664                 LASSERT(pos <= size(), /**/);
1665         }
1666
1667         FontList::const_iterator cit = d->fontlist_.fontIterator(pos);
1668         if (cit != d->fontlist_.end())
1669                 return cit->font();
1670
1671         if (pos == size() && !empty())
1672                 return getFontSettings(bparams, pos - 1);
1673
1674         // Optimisation: avoid a full font instantiation if there is no
1675         // language change from previous call.
1676         static Font previous_font;
1677         static Language const * previous_lang = 0;
1678         Language const * lang = getParLanguage(bparams);
1679         if (lang != previous_lang) {
1680                 previous_lang = lang;
1681                 previous_font = Font(inherit_font, lang);
1682         }
1683         return previous_font;
1684 }
1685
1686
1687 FontSpan Paragraph::fontSpan(pos_type pos) const
1688 {
1689         LASSERT(pos <= size(), /**/);
1690         pos_type start = 0;
1691
1692         FontList::const_iterator cit = d->fontlist_.begin();
1693         FontList::const_iterator end = d->fontlist_.end();
1694         for (; cit != end; ++cit) {
1695                 if (cit->pos() >= pos) {
1696                         if (pos >= beginOfBody())
1697                                 return FontSpan(max(start, beginOfBody()),
1698                                                 cit->pos());
1699                         else
1700                                 return FontSpan(start,
1701                                                 min(beginOfBody() - 1,
1702                                                          cit->pos()));
1703                 }
1704                 start = cit->pos() + 1;
1705         }
1706
1707         // This should not happen, but if so, we take no chances.
1708         // LYXERR0("Paragraph::getEndPosOfFontSpan: This should not happen!");
1709         return FontSpan(pos, pos);
1710 }
1711
1712
1713 // Gets uninstantiated font setting at position 0
1714 Font const & Paragraph::getFirstFontSettings(BufferParams const & bparams) const
1715 {
1716         if (!empty() && !d->fontlist_.empty())
1717                 return d->fontlist_.begin()->font();
1718
1719         // Optimisation: avoid a full font instantiation if there is no
1720         // language change from previous call.
1721         static Font previous_font;
1722         static Language const * previous_lang = 0;
1723         if (bparams.language != previous_lang) {
1724                 previous_lang = bparams.language;
1725                 previous_font = Font(inherit_font, bparams.language);
1726         }
1727
1728         return previous_font;
1729 }
1730
1731
1732 // Gets the fully instantiated font at a given position in a paragraph
1733 // This is basically the same function as Text::GetFont() in text2.cpp.
1734 // The difference is that this one is used for generating the LaTeX file,
1735 // and thus cosmetic "improvements" are disallowed: This has to deliver
1736 // the true picture of the buffer. (Asger)
1737 Font const Paragraph::getFont(BufferParams const & bparams, pos_type pos,
1738                                  Font const & outerfont) const
1739 {
1740         LASSERT(pos >= 0, /**/);
1741
1742         Font font = getFontSettings(bparams, pos);
1743
1744         pos_type const body_pos = beginOfBody();
1745         FontInfo & fi = font.fontInfo();
1746         if (pos < body_pos)
1747                 fi.realize(d->layout_->labelfont);
1748         else
1749                 fi.realize(d->layout_->font);
1750
1751         fi.realize(outerfont.fontInfo());
1752         fi.realize(bparams.getFont().fontInfo());
1753
1754         return font;
1755 }
1756
1757
1758 Font const Paragraph::getLabelFont
1759         (BufferParams const & bparams, Font const & outerfont) const
1760 {
1761         FontInfo tmpfont = d->layout_->labelfont;
1762         tmpfont.realize(outerfont.fontInfo());
1763         tmpfont.realize(bparams.getFont().fontInfo());
1764         return Font(tmpfont, getParLanguage(bparams));
1765 }
1766
1767
1768 Font const Paragraph::getLayoutFont
1769         (BufferParams const & bparams, Font const & outerfont) const
1770 {
1771         FontInfo tmpfont = d->layout_->font;
1772         tmpfont.realize(outerfont.fontInfo());
1773         tmpfont.realize(bparams.getFont().fontInfo());
1774         return Font(tmpfont, getParLanguage(bparams));
1775 }
1776
1777
1778 /// Returns the height of the highest font in range
1779 FontSize Paragraph::highestFontInRange
1780         (pos_type startpos, pos_type endpos, FontSize def_size) const
1781 {
1782         return d->fontlist_.highestInRange(startpos, endpos, def_size);
1783 }
1784
1785
1786 char_type Paragraph::getUChar(BufferParams const & bparams, pos_type pos) const
1787 {
1788         char_type c = d->text_[pos];
1789         if (!lyxrc.rtl_support)
1790                 return c;
1791
1792         char_type uc = c;
1793         switch (c) {
1794         case '(':
1795                 uc = ')';
1796                 break;
1797         case ')':
1798                 uc = '(';
1799                 break;
1800         case '[':
1801                 uc = ']';
1802                 break;
1803         case ']':
1804                 uc = '[';
1805                 break;
1806         case '{':
1807                 uc = '}';
1808                 break;
1809         case '}':
1810                 uc = '{';
1811                 break;
1812         case '<':
1813                 uc = '>';
1814                 break;
1815         case '>':
1816                 uc = '<';
1817                 break;
1818         }
1819         if (uc != c && getFontSettings(bparams, pos).isRightToLeft())
1820                 return uc;
1821         return c;
1822 }
1823
1824
1825 void Paragraph::setFont(pos_type pos, Font const & font)
1826 {
1827         LASSERT(pos <= size(), /**/);
1828
1829         // First, reduce font against layout/label font
1830         // Update: The setCharFont() routine in text2.cpp already
1831         // reduces font, so we don't need to do that here. (Asger)
1832
1833         d->fontlist_.set(pos, font);
1834 }
1835
1836
1837 void Paragraph::makeSameLayout(Paragraph const & par)
1838 {
1839         d->layout_ = par.d->layout_;
1840         d->params_ = par.d->params_;
1841 }
1842
1843
1844 bool Paragraph::stripLeadingSpaces(bool trackChanges)
1845 {
1846         if (isFreeSpacing())
1847                 return false;
1848
1849         int pos = 0;
1850         int count = 0;
1851
1852         while (pos < size() && (isNewline(pos) || isLineSeparator(pos))) {
1853                 if (eraseChar(pos, trackChanges))
1854                         ++count;
1855                 else
1856                         ++pos;
1857         }
1858
1859         return count > 0 || pos > 0;
1860 }
1861
1862
1863 bool Paragraph::hasSameLayout(Paragraph const & par) const
1864 {
1865         return par.d->layout_ == d->layout_
1866                 && d->params_.sameLayout(par.d->params_);
1867 }
1868
1869
1870 depth_type Paragraph::getDepth() const
1871 {
1872         return d->params_.depth();
1873 }
1874
1875
1876 depth_type Paragraph::getMaxDepthAfter() const
1877 {
1878         if (d->layout_->isEnvironment())
1879                 return d->params_.depth() + 1;
1880         else
1881                 return d->params_.depth();
1882 }
1883
1884
1885 char Paragraph::getAlign() const
1886 {
1887         if (d->params_.align() == LYX_ALIGN_LAYOUT)
1888                 return d->layout_->align;
1889         else
1890                 return d->params_.align();
1891 }
1892
1893
1894 docstring const & Paragraph::labelString() const
1895 {
1896         return d->params_.labelString();
1897 }
1898
1899
1900 // the next two functions are for the manual labels
1901 docstring const Paragraph::getLabelWidthString() const
1902 {
1903         if (d->layout_->margintype == MARGIN_MANUAL
1904             || d->layout_->latextype == LATEX_BIB_ENVIRONMENT)
1905                 return d->params_.labelWidthString();
1906         else
1907                 return _("Senseless with this layout!");
1908 }
1909
1910
1911 void Paragraph::setLabelWidthString(docstring const & s)
1912 {
1913         d->params_.labelWidthString(s);
1914 }
1915
1916
1917 docstring Paragraph::expandLabel(Layout const & layout,
1918                 BufferParams const & bparams) const
1919 {
1920         return expandParagraphLabel(layout, bparams, true);
1921 }
1922
1923
1924 docstring Paragraph::expandDocBookLabel(Layout const & layout,
1925                 BufferParams const & bparams) const
1926 {
1927         return expandParagraphLabel(layout, bparams, false);
1928 }
1929
1930
1931 docstring Paragraph::expandParagraphLabel(Layout const & layout,
1932                 BufferParams const & bparams, bool process_appendix) const
1933 {
1934         DocumentClass const & tclass = bparams.documentClass();
1935         string const & lang = getParLanguage(bparams)->code();
1936         bool const in_appendix = process_appendix && d->params_.appendix();
1937         docstring fmt = translateIfPossible(layout.labelstring(in_appendix), lang);
1938
1939         if (fmt.empty() && layout.labeltype == LABEL_COUNTER
1940             && !layout.counter.empty())
1941                 return tclass.counters().theCounter(layout.counter, lang);
1942
1943         // handle 'inherited level parts' in 'fmt',
1944         // i.e. the stuff between '@' in   '@Section@.\arabic{subsection}'
1945         size_t const i = fmt.find('@', 0);
1946         if (i != docstring::npos) {
1947                 size_t const j = fmt.find('@', i + 1);
1948                 if (j != docstring::npos) {
1949                         docstring parent(fmt, i + 1, j - i - 1);
1950                         docstring label = from_ascii("??");
1951                         if (tclass.hasLayout(parent))
1952                                 docstring label = expandParagraphLabel(tclass[parent], bparams,
1953                                                       process_appendix);
1954                         fmt = docstring(fmt, 0, i) + label
1955                                 + docstring(fmt, j + 1, docstring::npos);
1956                 }
1957         }
1958
1959         return tclass.counters().counterLabel(fmt, lang);
1960 }
1961
1962
1963 void Paragraph::applyLayout(Layout const & new_layout)
1964 {
1965         d->layout_ = &new_layout;
1966         LyXAlignment const oldAlign = d->params_.align();
1967
1968         if (!(oldAlign & d->layout_->alignpossible)) {
1969                 frontend::Alert::warning(_("Alignment not permitted"),
1970                         _("The new layout does not permit the alignment previously used.\nSetting to default."));
1971                 d->params_.align(LYX_ALIGN_LAYOUT);
1972         }
1973 }
1974
1975
1976 pos_type Paragraph::beginOfBody() const
1977 {
1978         return d->begin_of_body_;
1979 }
1980
1981
1982 void Paragraph::setBeginOfBody()
1983 {
1984         if (d->layout_->labeltype != LABEL_MANUAL) {
1985                 d->begin_of_body_ = 0;
1986                 return;
1987         }
1988
1989         // Unroll the first two cycles of the loop
1990         // and remember the previous character to
1991         // remove unnecessary getChar() calls
1992         pos_type i = 0;
1993         pos_type end = size();
1994         if (i < end && !isNewline(i)) {
1995                 ++i;
1996                 char_type previous_char = 0;
1997                 char_type temp = 0;
1998                 if (i < end) {
1999                         previous_char = d->text_[i];
2000                         if (!isNewline(i)) {
2001                                 ++i;
2002                                 while (i < end && previous_char != ' ') {
2003                                         temp = d->text_[i];
2004                                         if (isNewline(i))
2005                                                 break;
2006                                         ++i;
2007                                         previous_char = temp;
2008                                 }
2009                         }
2010                 }
2011         }
2012
2013         d->begin_of_body_ = i;
2014 }
2015
2016
2017 bool Paragraph::allowParagraphCustomization() const
2018 {
2019         return inInset().allowParagraphCustomization();
2020 }
2021
2022
2023 bool Paragraph::usePlainLayout() const
2024 {
2025         return inInset().usePlainLayout();
2026 }
2027
2028
2029 namespace {
2030
2031 // paragraphs inside floats need different alignment tags to avoid
2032 // unwanted space
2033
2034 bool noTrivlistCentering(InsetCode code)
2035 {
2036         return code == FLOAT_CODE
2037                || code == WRAP_CODE
2038                || code == CELL_CODE;
2039 }
2040
2041
2042 string correction(string const & orig)
2043 {
2044         if (orig == "flushleft")
2045                 return "raggedright";
2046         if (orig == "flushright")
2047                 return "raggedleft";
2048         if (orig == "center")
2049                 return "centering";
2050         return orig;
2051 }
2052
2053
2054 string const corrected_env(string const & suffix, string const & env,
2055         InsetCode code, bool const lastpar)
2056 {
2057         string output = suffix + "{";
2058         if (noTrivlistCentering(code)) {
2059                 if (lastpar) {
2060                         // the last paragraph in non-trivlist-aligned
2061                         // context is special (to avoid unwanted whitespace)
2062                         if (suffix == "\\begin")
2063                                 return "\\" + correction(env) + "{}";
2064                         return string();
2065                 }
2066                 output += correction(env);
2067         } else
2068                 output += env;
2069         output += "}";
2070         if (suffix == "\\begin")
2071                 output += "\n";
2072         return output;
2073 }
2074
2075
2076 void adjust_row_column(string const & str, TexRow & texrow, int & column)
2077 {
2078         if (!contains(str, "\n"))
2079                 column += str.size();
2080         else {
2081                 string tmp;
2082                 texrow.newline();
2083                 column = rsplit(str, tmp, '\n').size();
2084         }
2085 }
2086
2087 } // namespace anon
2088
2089
2090 int Paragraph::Private::startTeXParParams(BufferParams const & bparams,
2091                                  odocstream & os, TexRow & texrow,
2092                                  OutputParams const & runparams) const
2093 {
2094         int column = 0;
2095
2096         if (params_.noindent()) {
2097                 os << "\\noindent ";
2098                 column += 10;
2099         }
2100
2101         LyXAlignment const curAlign = params_.align();
2102
2103         if (curAlign == layout_->align)
2104                 return column;
2105
2106         switch (curAlign) {
2107         case LYX_ALIGN_NONE:
2108         case LYX_ALIGN_BLOCK:
2109         case LYX_ALIGN_LAYOUT:
2110         case LYX_ALIGN_SPECIAL:
2111         case LYX_ALIGN_DECIMAL:
2112                 break;
2113         case LYX_ALIGN_LEFT:
2114         case LYX_ALIGN_RIGHT:
2115         case LYX_ALIGN_CENTER:
2116                 if (runparams.moving_arg) {
2117                         os << "\\protect";
2118                         column += 8;
2119                 }
2120                 break;
2121         }
2122
2123         string const begin_tag = "\\begin";
2124         InsetCode code = ownerCode();
2125         bool const lastpar = runparams.isLastPar;
2126
2127         switch (curAlign) {
2128         case LYX_ALIGN_NONE:
2129         case LYX_ALIGN_BLOCK:
2130         case LYX_ALIGN_LAYOUT:
2131         case LYX_ALIGN_SPECIAL:
2132         case LYX_ALIGN_DECIMAL:
2133                 break;
2134         case LYX_ALIGN_LEFT: {
2135                 string output;
2136                 if (owner_->getParLanguage(bparams)->babel() != "hebrew")
2137                         output = corrected_env(begin_tag, "flushleft", code, lastpar);
2138                 else
2139                         output = corrected_env(begin_tag, "flushright", code, lastpar);
2140                 os << from_ascii(output);
2141                 adjust_row_column(output, texrow, column);
2142                 break;
2143         } case LYX_ALIGN_RIGHT: {
2144                 string output;
2145                 if (owner_->getParLanguage(bparams)->babel() != "hebrew")
2146                         output = corrected_env(begin_tag, "flushright", code, lastpar);
2147                 else
2148                         output = corrected_env(begin_tag, "flushleft", code, lastpar);
2149                 os << from_ascii(output);
2150                 adjust_row_column(output, texrow, column);
2151                 break;
2152         } case LYX_ALIGN_CENTER: {
2153                 string output;
2154                 output = corrected_env(begin_tag, "center", code, lastpar);
2155                 os << from_ascii(output);
2156                 adjust_row_column(output, texrow, column);
2157                 break;
2158         }
2159         }
2160
2161         return column;
2162 }
2163
2164
2165 int Paragraph::Private::endTeXParParams(BufferParams const & bparams,
2166                                odocstream & os, TexRow & texrow,
2167                                OutputParams const & runparams) const
2168 {
2169         int column = 0;
2170
2171         LyXAlignment const curAlign = params_.align();
2172
2173         if (curAlign == layout_->align)
2174                 return column;
2175
2176         switch (curAlign) {
2177         case LYX_ALIGN_NONE:
2178         case LYX_ALIGN_BLOCK:
2179         case LYX_ALIGN_LAYOUT:
2180         case LYX_ALIGN_SPECIAL:
2181         case LYX_ALIGN_DECIMAL:
2182                 break;
2183         case LYX_ALIGN_LEFT:
2184         case LYX_ALIGN_RIGHT:
2185         case LYX_ALIGN_CENTER:
2186                 if (runparams.moving_arg) {
2187                         os << "\\protect";
2188                         column = 8;
2189                 }
2190                 break;
2191         }
2192
2193         string const end_tag = "\n\\par\\end";
2194         InsetCode code = ownerCode();
2195         bool const lastpar = runparams.isLastPar;
2196
2197         switch (curAlign) {
2198         case LYX_ALIGN_NONE:
2199         case LYX_ALIGN_BLOCK:
2200         case LYX_ALIGN_LAYOUT:
2201         case LYX_ALIGN_SPECIAL:
2202         case LYX_ALIGN_DECIMAL:
2203                 break;
2204         case LYX_ALIGN_LEFT: {
2205                 string output;
2206                 if (owner_->getParLanguage(bparams)->babel() != "hebrew")
2207                         output = corrected_env(end_tag, "flushleft", code, lastpar);
2208                 else
2209                         output = corrected_env(end_tag, "flushright", code, lastpar);
2210                 os << from_ascii(output);
2211                 adjust_row_column(output, texrow, column);
2212                 break;
2213         } case LYX_ALIGN_RIGHT: {
2214                 string output;
2215                 if (owner_->getParLanguage(bparams)->babel() != "hebrew")
2216                         output = corrected_env(end_tag, "flushright", code, lastpar);
2217                 else
2218                         output = corrected_env(end_tag, "flushleft", code, lastpar);
2219                 os << from_ascii(output);
2220                 adjust_row_column(output, texrow, column);
2221                 break;
2222         } case LYX_ALIGN_CENTER: {
2223                 string output;
2224                 output = corrected_env(end_tag, "center", code, lastpar);
2225                 os << from_ascii(output);
2226                 adjust_row_column(output, texrow, column);
2227                 break;
2228         }
2229         }
2230
2231         return column;
2232 }
2233
2234
2235 // This one spits out the text of the paragraph
2236 void Paragraph::latex(BufferParams const & bparams,
2237         Font const & outerfont,
2238         odocstream & os, TexRow & texrow,
2239         OutputParams const & runparams,
2240         int start_pos, int end_pos, bool force) const
2241 {
2242         LYXERR(Debug::LATEX, "Paragraph::latex...     " << this);
2243
2244         // FIXME This check should not be needed. Perhaps issue an
2245         // error if it triggers.
2246         Layout const & style = inInset().forcePlainLayout() ?
2247                 bparams.documentClass().plainLayout() : *d->layout_;
2248
2249         if (!force && style.inpreamble)
2250                 return;
2251
2252         bool const allowcust = allowParagraphCustomization();
2253
2254         // Current base font for all inherited font changes, without any
2255         // change caused by an individual character, except for the language:
2256         // It is set to the language of the first character.
2257         // As long as we are in the label, this font is the base font of the
2258         // label. Before the first body character it is set to the base font
2259         // of the body.
2260         Font basefont;
2261
2262         // Maybe we have to create a optional argument.
2263         pos_type body_pos = beginOfBody();
2264         unsigned int column = 0;
2265
2266         if (body_pos > 0) {
2267                 // the optional argument is kept in curly brackets in
2268                 // case it contains a ']'
2269                 os << "[{";
2270                 column += 2;
2271                 basefont = getLabelFont(bparams, outerfont);
2272         } else {
2273                 basefont = getLayoutFont(bparams, outerfont);
2274         }
2275
2276         // Which font is currently active?
2277         Font running_font(basefont);
2278         // Do we have an open font change?
2279         bool open_font = false;
2280
2281         Change runningChange = Change(Change::UNCHANGED);
2282
2283         Encoding const * const prev_encoding = runparams.encoding;
2284
2285         texrow.start(id(), 0);
2286
2287         // if the paragraph is empty, the loop will not be entered at all
2288         if (empty()) {
2289                 if (style.isCommand()) {
2290                         os << '{';
2291                         ++column;
2292                 }
2293                 if (allowcust)
2294                         column += d->startTeXParParams(bparams, os, texrow,
2295                                                     runparams);
2296         }
2297
2298         for (pos_type i = 0; i < size(); ++i) {
2299                 // First char in paragraph or after label?
2300                 if (i == body_pos) {
2301                         if (body_pos > 0) {
2302                                 if (open_font) {
2303                                         column += running_font.latexWriteEndChanges(
2304                                                 os, bparams, runparams,
2305                                                 basefont, basefont);
2306                                         open_font = false;
2307                                 }
2308                                 basefont = getLayoutFont(bparams, outerfont);
2309                                 running_font = basefont;
2310
2311                                 column += Changes::latexMarkChange(os, bparams,
2312                                                 runningChange, Change(Change::UNCHANGED),
2313                                                 runparams);
2314                                 runningChange = Change(Change::UNCHANGED);
2315
2316                                 os << "}] ";
2317                                 column +=3;
2318                         }
2319                         if (style.isCommand()) {
2320                                 os << '{';
2321                                 ++column;
2322                         }
2323
2324                         if (allowcust)
2325                                 column += d->startTeXParParams(bparams, os,
2326                                                             texrow,
2327                                                             runparams);
2328                 }
2329
2330                 Change const & change = runparams.inDeletedInset ? runparams.changeOfDeletedInset
2331                                                                  : lookupChange(i);
2332
2333                 if (bparams.outputChanges && runningChange != change) {
2334                         if (open_font) {
2335                                 column += running_font.latexWriteEndChanges(
2336                                                 os, bparams, runparams, basefont, basefont);
2337                                 open_font = false;
2338                         }
2339                         basefont = getLayoutFont(bparams, outerfont);
2340                         running_font = basefont;
2341
2342                         column += Changes::latexMarkChange(os, bparams, runningChange,
2343                                                            change, runparams);
2344                         runningChange = change;
2345                 }
2346
2347                 // do not output text which is marked deleted
2348                 // if change tracking output is disabled
2349                 if (!bparams.outputChanges && change.deleted()) {
2350                         continue;
2351                 }
2352
2353                 ++column;
2354
2355                 // Fully instantiated font
2356                 Font const font = getFont(bparams, i, outerfont);
2357
2358                 Font const last_font = running_font;
2359
2360                 // Do we need to close the previous font?
2361                 if (open_font &&
2362                     (font != running_font ||
2363                      font.language() != running_font.language()))
2364                 {
2365                         column += running_font.latexWriteEndChanges(
2366                                         os, bparams, runparams, basefont,
2367                                         (i == body_pos-1) ? basefont : font);
2368                         running_font = basefont;
2369                         open_font = false;
2370                 }
2371
2372                 // close babel's font environment before opening CJK.
2373                 if (!running_font.language()->babel().empty() &&
2374                     font.language()->encoding()->package() == Encoding::CJK) {
2375                                 string end_tag = subst(lyxrc.language_command_end,
2376                                                         "$$lang",
2377                                                         running_font.language()->babel());
2378                                 os << from_ascii(end_tag);
2379                                 column += end_tag.length();
2380                 }
2381
2382                 // Switch file encoding if necessary (and allowed)
2383                 if (!runparams.pass_thru && !style.pass_thru &&
2384                     runparams.encoding->package() != Encoding::none &&
2385                     font.language()->encoding()->package() != Encoding::none) {
2386                         pair<bool, int> const enc_switch = switchEncoding(os, bparams,
2387                                         runparams, *(font.language()->encoding()));
2388                         if (enc_switch.first) {
2389                                 column += enc_switch.second;
2390                                 runparams.encoding = font.language()->encoding();
2391                         }
2392                 }
2393
2394                 char_type const c = d->text_[i];
2395
2396                 // Do we need to change font?
2397                 if ((font != running_font ||
2398                      font.language() != running_font.language()) &&
2399                         i != body_pos - 1)
2400                 {
2401                         odocstringstream ods;
2402                         column += font.latexWriteStartChanges(ods, bparams,
2403                                                               runparams, basefont,
2404                                                               last_font);
2405                         running_font = font;
2406                         open_font = true;
2407                         docstring fontchange = ods.str();
2408                         // check whether the fontchange ends with a \\textcolor
2409                         // modifier and the text starts with a space (bug 4473)
2410                         docstring const last_modifier = rsplit(fontchange, '\\');
2411                         if (prefixIs(last_modifier, from_ascii("textcolor")) && c == ' ')
2412                                 os << fontchange << from_ascii("{}");
2413                         // check if the fontchange ends with a trailing blank
2414                         // (like "\small " (see bug 3382)
2415                         else if (suffixIs(fontchange, ' ') && c == ' ')
2416                                 os << fontchange.substr(0, fontchange.size() - 1)
2417                                    << from_ascii("{}");
2418                         else
2419                                 os << fontchange;
2420                 }
2421
2422                 // FIXME: think about end_pos implementation...
2423                 if (c == ' ' && i >= start_pos && (end_pos == -1 || i < end_pos)) {
2424                         // FIXME: integrate this case in latexSpecialChar
2425                         // Do not print the separation of the optional argument
2426                         // if style.pass_thru is false. This works because
2427                         // latexSpecialChar ignores spaces if
2428                         // style.pass_thru is false.
2429                         if (i != body_pos - 1) {
2430                                 if (d->simpleTeXBlanks(
2431                                                 runparams, os, texrow,
2432                                                 i, column, font, style)) {
2433                                         // A surrogate pair was output. We
2434                                         // must not call latexSpecialChar
2435                                         // in this iteration, since it would output
2436                                         // the combining character again.
2437                                         ++i;
2438                                         continue;
2439                                 }
2440                         }
2441                 }
2442
2443                 OutputParams rp = runparams;
2444                 rp.free_spacing = style.free_spacing;
2445                 rp.local_font = &font;
2446                 rp.intitle = style.intitle;
2447
2448                 // Two major modes:  LaTeX or plain
2449                 // Handle here those cases common to both modes
2450                 // and then split to handle the two modes separately.
2451                 if (c == META_INSET) {
2452                         if (i >= start_pos && (end_pos == -1 || i < end_pos)) {
2453                                 d->latexInset(bparams, os,
2454                                                 texrow, rp, running_font,
2455                                                 basefont, outerfont, open_font,
2456                                                 runningChange, style, i, column);
2457                         }
2458                 } else {
2459                         if (i >= start_pos && (end_pos == -1 || i < end_pos)) {
2460                                 try {
2461                                         d->latexSpecialChar(os, rp, running_font, runningChange,
2462                                                 style, i, column);
2463                                 } catch (EncodingException & e) {
2464                                 if (runparams.dryrun) {
2465                                         os << "<" << _("LyX Warning: ")
2466                                            << _("uncodable character") << " '";
2467                                         os.put(c);
2468                                         os << "'>";
2469                                 } else {
2470                                         // add location information and throw again.
2471                                         e.par_id = id();
2472                                         e.pos = i;
2473                                         throw(e);
2474                                 }
2475                         }
2476                 }
2477                 }
2478
2479                 // Set the encoding to that returned from latexSpecialChar (see
2480                 // comment for encoding member in OutputParams.h)
2481                 runparams.encoding = rp.encoding;
2482         }
2483
2484         // If we have an open font definition, we have to close it
2485         if (open_font) {
2486 #ifdef FIXED_LANGUAGE_END_DETECTION
2487                 if (next_) {
2488                         running_font.latexWriteEndChanges(os, bparams,
2489                                         runparams, basefont,
2490                                         next_->getFont(bparams, 0, outerfont));
2491                 } else {
2492                         running_font.latexWriteEndChanges(os, bparams,
2493                                         runparams, basefont, basefont);
2494                 }
2495 #else
2496 //FIXME: For now we ALWAYS have to close the foreign font settings if they are
2497 //FIXME: there as we start another \selectlanguage with the next paragraph if
2498 //FIXME: we are in need of this. This should be fixed sometime (Jug)
2499                 running_font.latexWriteEndChanges(os, bparams, runparams,
2500                                 basefont, basefont);
2501 #endif
2502         }
2503
2504         column += Changes::latexMarkChange(os, bparams, runningChange,
2505                                            Change(Change::UNCHANGED), runparams);
2506
2507         // Needed if there is an optional argument but no contents.
2508         if (body_pos > 0 && body_pos == size()) {
2509                 os << "}]~";
2510         }
2511
2512         if (allowcust && d->endTeXParParams(bparams, os, texrow, runparams)
2513             && runparams.encoding != prev_encoding) {
2514                 runparams.encoding = prev_encoding;
2515                 if (!bparams.useXetex)
2516                         os << setEncoding(prev_encoding->iconvName());
2517         }
2518
2519         LYXERR(Debug::LATEX, "Paragraph::latex... done " << this);
2520 }
2521
2522
2523 bool Paragraph::emptyTag() const
2524 {
2525         for (pos_type i = 0; i < size(); ++i) {
2526                 if (Inset const * inset = getInset(i)) {
2527                         InsetCode lyx_code = inset->lyxCode();
2528                         // FIXME testing like that is wrong. What is
2529                         // the intent?
2530                         if (lyx_code != TOC_CODE &&
2531                             lyx_code != INCLUDE_CODE &&
2532                             lyx_code != GRAPHICS_CODE &&
2533                             lyx_code != ERT_CODE &&
2534                             lyx_code != LISTINGS_CODE &&
2535                             lyx_code != FLOAT_CODE &&
2536                             lyx_code != TABULAR_CODE) {
2537                                 return false;
2538                         }
2539                 } else {
2540                         char_type c = d->text_[i];
2541                         if (c != ' ' && c != '\t')
2542                                 return false;
2543                 }
2544         }
2545         return true;
2546 }
2547
2548
2549 string Paragraph::getID(Buffer const & buf, OutputParams const & runparams)
2550         const
2551 {
2552         for (pos_type i = 0; i < size(); ++i) {
2553                 if (Inset const * inset = getInset(i)) {
2554                         InsetCode lyx_code = inset->lyxCode();
2555                         if (lyx_code == LABEL_CODE) {
2556                                 InsetLabel const * const il = static_cast<InsetLabel const *>(inset);
2557                                 docstring const & id = il->getParam("name");
2558                                 return "id='" + to_utf8(sgml::cleanID(buf, runparams, id)) + "'";
2559                         }
2560                 }
2561         }
2562         return string();
2563 }
2564
2565
2566 pos_type Paragraph::firstWordDocBook(odocstream & os, OutputParams const & runparams)
2567         const
2568 {
2569         pos_type i;
2570         for (i = 0; i < size(); ++i) {
2571                 if (Inset const * inset = getInset(i)) {
2572                         inset->docbook(os, runparams);
2573                 } else {
2574                         char_type c = d->text_[i];
2575                         if (c == ' ')
2576                                 break;
2577                         os << sgml::escapeChar(c);
2578                 }
2579         }
2580         return i;
2581 }
2582
2583
2584 pos_type Paragraph::firstWordLyXHTML(XHTMLStream & xs, OutputParams const & runparams)
2585         const
2586 {
2587         pos_type i;
2588         for (i = 0; i < size(); ++i) {
2589                 if (Inset const * inset = getInset(i)) {
2590                         inset->xhtml(xs, runparams);
2591                 } else {
2592                         char_type c = d->text_[i];
2593                         if (c == ' ')
2594                                 break;
2595                         xs << c;
2596                 }
2597         }
2598         return i;
2599 }
2600
2601
2602 bool Paragraph::Private::onlyText(Buffer const & buf, Font const & outerfont, pos_type initial) const
2603 {
2604         Font font_old;
2605         pos_type size = text_.size();
2606         for (pos_type i = initial; i < size; ++i) {
2607                 Font font = owner_->getFont(buf.params(), i, outerfont);
2608                 if (text_[i] == META_INSET)
2609                         return false;
2610                 if (i != initial && font != font_old)
2611                         return false;
2612                 font_old = font;
2613         }
2614
2615         return true;
2616 }
2617
2618
2619 void Paragraph::simpleDocBookOnePar(Buffer const & buf,
2620                                     odocstream & os,
2621                                     OutputParams const & runparams,
2622                                     Font const & outerfont,
2623                                     pos_type initial) const
2624 {
2625         bool emph_flag = false;
2626
2627         Layout const & style = *d->layout_;
2628         FontInfo font_old =
2629                 style.labeltype == LABEL_MANUAL ? style.labelfont : style.font;
2630
2631         if (style.pass_thru && !d->onlyText(buf, outerfont, initial))
2632                 os << "]]>";
2633
2634         // parsing main loop
2635         for (pos_type i = initial; i < size(); ++i) {
2636                 Font font = getFont(buf.params(), i, outerfont);
2637
2638                 // handle <emphasis> tag
2639                 if (font_old.emph() != font.fontInfo().emph()) {
2640                         if (font.fontInfo().emph() == FONT_ON) {
2641                                 os << "<emphasis>";
2642                                 emph_flag = true;
2643                         } else if (i != initial) {
2644                                 os << "</emphasis>";
2645                                 emph_flag = false;
2646                         }
2647                 }
2648
2649                 if (Inset const * inset = getInset(i)) {
2650                         inset->docbook(os, runparams);
2651                 } else {
2652                         char_type c = d->text_[i];
2653
2654                         if (style.pass_thru)
2655                                 os.put(c);
2656                         else
2657                                 os << sgml::escapeChar(c);
2658                 }
2659                 font_old = font.fontInfo();
2660         }
2661
2662         if (emph_flag) {
2663                 os << "</emphasis>";
2664         }
2665
2666         if (style.free_spacing)
2667                 os << '\n';
2668         if (style.pass_thru && !d->onlyText(buf, outerfont, initial))
2669                 os << "<![CDATA[";
2670 }
2671
2672
2673 docstring Paragraph::simpleLyXHTMLOnePar(Buffer const & buf,
2674                                     XHTMLStream & xs,
2675                                     OutputParams const & runparams,
2676                                     Font const & outerfont,
2677                                     pos_type initial) const
2678 {
2679         docstring retval;
2680
2681         bool emph_flag = false;
2682         bool bold_flag = false;
2683         string closing_tag;
2684
2685         Layout const & style = *d->layout_;
2686
2687         if (!runparams.for_toc && runparams.html_make_pars) {
2688                 // generate a magic label for this paragraph
2689                 string const attr = "id='" + magicLabel() + "'";
2690                 xs << html::CompTag("a", attr);
2691         }
2692
2693         FontInfo font_old =
2694                 style.labeltype == LABEL_MANUAL ? style.labelfont : style.font;
2695
2696         // parsing main loop
2697         for (pos_type i = initial; i < size(); ++i) {
2698                 // let's not show deleted material in the output
2699                 if (isDeleted(i))
2700                         continue;
2701
2702                 Font font = getFont(buf.params(), i, outerfont);
2703
2704                 // emphasis
2705                 if (font_old.emph() != font.fontInfo().emph()) {
2706                         if (font.fontInfo().emph() == FONT_ON) {
2707                                 xs << html::StartTag("em");
2708                                 emph_flag = true;
2709                         } else if (emph_flag && i != initial) {
2710                                 xs << html::EndTag("em");
2711                                 emph_flag = false;
2712                         }
2713                 }
2714                 // bold
2715                 if (font_old.series() != font.fontInfo().series()) {
2716                         if (font.fontInfo().series() == BOLD_SERIES) {
2717                                 xs << html::StartTag("strong");
2718                                 bold_flag = true;
2719                         } else if (bold_flag && i != initial) {
2720                                 xs << html::EndTag("strong");
2721                                 bold_flag = false;
2722                         }
2723                 }
2724                 // FIXME XHTML
2725                 // Other such tags? What about the other text ranges?
2726
2727                 Inset const * inset = getInset(i);
2728                 if (inset) {
2729                         if (!runparams.for_toc || inset->isInToc()) {
2730                                 OutputParams np = runparams;
2731                                 if (!inset->getLayout().htmlisblock())
2732                                         np.html_in_par = true;
2733                                 retval += inset->xhtml(xs, np);
2734                         }
2735                 } else {
2736                         char_type c = d->text_[i];
2737
2738                         if (style.pass_thru)
2739                                 xs << c;
2740                         else if (c == '-') {
2741                                 docstring str;
2742                                 int j = i + 1;
2743                                 if (j < size() && d->text_[j] == '-') {
2744                                         j += 1;
2745                                         if (j < size() && d->text_[j] == '-') {
2746                                                 str += from_ascii("&mdash;");
2747                                                 i += 2;
2748                                         } else {
2749                                                 str += from_ascii("&ndash;");
2750                                                 i += 1;
2751                                         }
2752                                 }
2753                                 else
2754                                         str += c;
2755                                 // We don't want to escape the entities. Note that
2756                                 // it is safe to do this, since str can otherwise
2757                                 // only be "-". E.g., it can't be "<".
2758                                 xs << XHTMLStream::NextRaw() << str;
2759                         } else
2760                                 xs << c;
2761                 }
2762                 font_old = font.fontInfo();
2763         }
2764
2765         xs.closeFontTags();
2766         return retval;
2767 }
2768
2769
2770 bool Paragraph::isHfill(pos_type pos) const
2771 {
2772         Inset const * inset = getInset(pos);
2773         return inset && (inset->lyxCode() == SPACE_CODE &&
2774                          inset->isStretchableSpace());
2775 }
2776
2777
2778 bool Paragraph::isNewline(pos_type pos) const
2779 {
2780         Inset const * inset = getInset(pos);
2781         return inset && inset->lyxCode() == NEWLINE_CODE;
2782 }
2783
2784
2785 bool Paragraph::isLineSeparator(pos_type pos) const
2786 {
2787         char_type const c = d->text_[pos];
2788         if (isLineSeparatorChar(c))
2789                 return true;
2790         Inset const * inset = getInset(pos);
2791         return inset && inset->isLineSeparator();
2792 }
2793
2794
2795 bool Paragraph::isWordSeparator(pos_type pos) const
2796 {
2797         if (Inset const * inset = getInset(pos))
2798                 return !inset->isLetter();
2799         char_type const c = d->text_[pos];
2800         // We want to pass the ' and escape chars to the spellchecker
2801         static docstring const quote = from_utf8(lyxrc.spellchecker_esc_chars + '\'');
2802         return (!isLetterChar(c) && !isDigit(c) && !contains(quote, c))
2803                 || pos == size();
2804 }
2805
2806
2807 bool Paragraph::isChar(pos_type pos) const
2808 {
2809         if (Inset const * inset = getInset(pos))
2810                 return inset->isChar();
2811         char_type const c = d->text_[pos];
2812         return !isLetterChar(c) && !isDigit(c) && !lyx::isSpace(c);
2813 }
2814
2815
2816 bool Paragraph::isSpace(pos_type pos) const
2817 {
2818         if (Inset const * inset = getInset(pos))
2819                 return inset->isSpace();
2820         char_type const c = d->text_[pos];
2821         return lyx::isSpace(c);
2822 }
2823
2824
2825 Language const *
2826 Paragraph::getParLanguage(BufferParams const & bparams) const
2827 {
2828         if (!empty())
2829                 return getFirstFontSettings(bparams).language();
2830         // FIXME: we should check the prev par as well (Lgb)
2831         return bparams.language;
2832 }
2833
2834
2835 bool Paragraph::isRTL(BufferParams const & bparams) const
2836 {
2837         return lyxrc.rtl_support
2838                 && getParLanguage(bparams)->rightToLeft()
2839                 && !inInset().getLayout().forceLTR();
2840 }
2841
2842
2843 void Paragraph::changeLanguage(BufferParams const & bparams,
2844                                Language const * from, Language const * to)
2845 {
2846         // change language including dummy font change at the end
2847         for (pos_type i = 0; i <= size(); ++i) {
2848                 Font font = getFontSettings(bparams, i);
2849                 if (font.language() == from) {
2850                         font.setLanguage(to);
2851                         setFont(i, font);
2852                 }
2853         }
2854         d->requestSpellCheck(size());
2855 }
2856
2857
2858 bool Paragraph::isMultiLingual(BufferParams const & bparams) const
2859 {
2860         Language const * doc_language = bparams.language;
2861         FontList::const_iterator cit = d->fontlist_.begin();
2862         FontList::const_iterator end = d->fontlist_.end();
2863
2864         for (; cit != end; ++cit)
2865                 if (cit->font().language() != ignore_language &&
2866                     cit->font().language() != latex_language &&
2867                     cit->font().language() != doc_language)
2868                         return true;
2869         return false;
2870 }
2871
2872
2873 void Paragraph::getLanguages(std::set<Language const *> & languages) const
2874 {
2875         FontList::const_iterator cit = d->fontlist_.begin();
2876         FontList::const_iterator end = d->fontlist_.end();
2877
2878         for (; cit != end; ++cit) {
2879                 Language const * lang = cit->font().language();
2880                 if (lang != ignore_language &&
2881                     lang != latex_language)
2882                         languages.insert(lang);
2883         }
2884 }
2885
2886
2887 docstring Paragraph::asString(int options) const
2888 {
2889         return asString(0, size(), options);
2890 }
2891
2892
2893 docstring Paragraph::asString(pos_type beg, pos_type end, int options) const
2894 {
2895         odocstringstream os;
2896
2897         if (beg == 0
2898             && options & AS_STR_LABEL
2899             && !d->params_.labelString().empty())
2900                 os << d->params_.labelString() << ' ';
2901
2902         for (pos_type i = beg; i < end; ++i) {
2903                 if ((options & AS_STR_SKIPDELETE) && isDeleted(i))
2904                         continue;
2905                 char_type const c = d->text_[i];
2906                 if (isPrintable(c) || c == '\t'
2907                     || (c == '\n' && (options & AS_STR_NEWLINES)))
2908                         os.put(c);
2909                 else if (c == META_INSET && (options & AS_STR_INSETS)) {
2910                         getInset(i)->tocString(os);
2911                         if (getInset(i)->asInsetMath())
2912                                 os << " ";
2913                 }
2914         }
2915
2916         return os.str();
2917 }
2918
2919
2920 docstring Paragraph::stringify(pos_type beg, pos_type end, int options, OutputParams & runparams) const
2921 {
2922         odocstringstream os;
2923
2924         if (beg == 0
2925                 && options & AS_STR_LABEL
2926                 && !d->params_.labelString().empty())
2927                 os << d->params_.labelString() << ' ';
2928
2929         for (pos_type i = beg; i < end; ++i) {
2930                 char_type const c = d->text_[i];
2931                 if (isPrintable(c) || c == '\t'
2932                     || (c == '\n' && (options & AS_STR_NEWLINES)))
2933                         os.put(c);
2934                 else if (c == META_INSET && (options & AS_STR_INSETS)) {
2935                         getInset(i)->plaintext(os, runparams);
2936                 }
2937         }
2938
2939         return os.str();
2940 }
2941
2942
2943 void Paragraph::setInsetOwner(Inset const * inset)
2944 {
2945         d->inset_owner_ = inset;
2946 }
2947
2948
2949 int Paragraph::id() const
2950 {
2951         return d->id_;
2952 }
2953
2954
2955 void Paragraph::setId(int id)
2956 {
2957         d->id_ = id;
2958 }
2959
2960
2961 Layout const & Paragraph::layout() const
2962 {
2963         return *d->layout_;
2964 }
2965
2966
2967 void Paragraph::setLayout(Layout const & layout)
2968 {
2969         d->layout_ = &layout;
2970 }
2971
2972
2973 void Paragraph::setDefaultLayout(DocumentClass const & tc)
2974 {
2975         setLayout(tc.defaultLayout());
2976 }
2977
2978
2979 void Paragraph::setPlainLayout(DocumentClass const & tc)
2980 {
2981         setLayout(tc.plainLayout());
2982 }
2983
2984
2985 void Paragraph::setPlainOrDefaultLayout(DocumentClass const & tclass)
2986 {
2987         if (usePlainLayout())
2988                 setPlainLayout(tclass);
2989         else
2990                 setDefaultLayout(tclass);
2991 }
2992
2993
2994 Inset const & Paragraph::inInset() const
2995 {
2996         LASSERT(d->inset_owner_, throw ExceptionMessage(BufferException,
2997                 _("Memory problem"), _("Paragraph not properly initialized")));
2998         return *d->inset_owner_;
2999 }
3000
3001
3002 ParagraphParameters & Paragraph::params()
3003 {
3004         return d->params_;
3005 }
3006
3007
3008 ParagraphParameters const & Paragraph::params() const
3009 {
3010         return d->params_;
3011 }
3012
3013
3014 bool Paragraph::isFreeSpacing() const
3015 {
3016         if (d->layout_->free_spacing)
3017                 return true;
3018         return d->inset_owner_ && d->inset_owner_->isFreeSpacing();
3019 }
3020
3021
3022 bool Paragraph::allowEmpty() const
3023 {
3024         if (d->layout_->keepempty)
3025                 return true;
3026         return d->inset_owner_ && d->inset_owner_->allowEmpty();
3027 }
3028
3029
3030 char_type Paragraph::transformChar(char_type c, pos_type pos) const
3031 {
3032         if (!Encodings::isArabicChar(c))
3033                 return c;
3034
3035         char_type prev_char = ' ';
3036         char_type next_char = ' ';
3037
3038         for (pos_type i = pos - 1; i >= 0; --i) {
3039                 char_type const par_char = d->text_[i];
3040                 if (!Encodings::isArabicComposeChar(par_char)) {
3041                         prev_char = par_char;
3042                         break;
3043                 }
3044         }
3045
3046         for (pos_type i = pos + 1, end = size(); i < end; ++i) {
3047                 char_type const par_char = d->text_[i];
3048                 if (!Encodings::isArabicComposeChar(par_char)) {
3049                         next_char = par_char;
3050                         break;
3051                 }
3052         }
3053
3054         if (Encodings::isArabicChar(next_char)) {
3055                 if (Encodings::isArabicChar(prev_char) &&
3056                         !Encodings::isArabicSpecialChar(prev_char))
3057                         return Encodings::transformChar(c, Encodings::FORM_MEDIAL);
3058                 else
3059                         return Encodings::transformChar(c, Encodings::FORM_INITIAL);
3060         } else {
3061                 if (Encodings::isArabicChar(prev_char) &&
3062                         !Encodings::isArabicSpecialChar(prev_char))
3063                         return Encodings::transformChar(c, Encodings::FORM_FINAL);
3064                 else
3065                         return Encodings::transformChar(c, Encodings::FORM_ISOLATED);
3066         }
3067 }
3068
3069
3070 int Paragraph::checkBiblio(Buffer const & buffer)
3071 {
3072         // FIXME From JS:
3073         // This is getting more and more a mess. ...We really should clean
3074         // up this bibitem issue for 1.6. See also bug 2743.
3075
3076         // Add bibitem insets if necessary
3077         if (d->layout_->labeltype != LABEL_BIBLIO)
3078                 return 0;
3079
3080         bool hasbibitem = !d->insetlist_.empty()
3081                 // Insist on it being in pos 0
3082                 && d->text_[0] == META_INSET
3083                 && d->insetlist_.begin()->inset->lyxCode() == BIBITEM_CODE;
3084
3085         bool track_changes = buffer.params().trackChanges;
3086
3087         docstring oldkey;
3088         docstring oldlabel;
3089
3090         // remove a bibitem in pos != 0
3091         // restore it later in pos 0 if necessary
3092         // (e.g. if a user inserts contents _before_ the item)
3093         // we're assuming there's only one of these, which there
3094         // should be.
3095         int erasedInsetPosition = -1;
3096         InsetList::iterator it = d->insetlist_.begin();
3097         InsetList::iterator end = d->insetlist_.end();
3098         for (; it != end; ++it)
3099                 if (it->inset->lyxCode() == BIBITEM_CODE
3100                       && it->pos > 0) {
3101                         InsetCommand * olditem = it->inset->asInsetCommand();
3102                         oldkey = olditem->getParam("key");
3103                         oldlabel = olditem->getParam("label");
3104                         erasedInsetPosition = it->pos;
3105                         eraseChar(erasedInsetPosition, track_changes);
3106                         break;
3107         }
3108
3109         // There was an InsetBibitem at the beginning, and we didn't
3110         // have to erase one.
3111         if (hasbibitem && erasedInsetPosition < 0)
3112                         return 0;
3113
3114         // There was an InsetBibitem at the beginning and we did have to
3115         // erase one. So we give its properties to the beginning inset.
3116         if (hasbibitem) {
3117                 InsetCommand * inset = d->insetlist_.begin()->inset->asInsetCommand();
3118                 if (!oldkey.empty())
3119                         inset->setParam("key", oldkey);
3120                 inset->setParam("label", oldlabel);
3121                 return -erasedInsetPosition;
3122         }
3123
3124         // There was no inset at the beginning, so we need to create one with
3125         // the key and label of the one we erased.
3126         InsetBibitem * inset =
3127                 new InsetBibitem(const_cast<Buffer *>(&buffer), InsetCommandParams(BIBITEM_CODE));
3128         // restore values of previously deleted item in this par.
3129         if (!oldkey.empty())
3130                 inset->setParam("key", oldkey);
3131         inset->setParam("label", oldlabel);
3132         insertInset(0, inset,
3133                     Change(track_changes ? Change::INSERTED : Change::UNCHANGED));
3134
3135         return 1;
3136 }
3137
3138
3139 void Paragraph::checkAuthors(AuthorList const & authorList)
3140 {
3141         d->changes_.checkAuthors(authorList);
3142 }
3143
3144
3145 bool Paragraph::isChanged(pos_type pos) const
3146 {
3147         return lookupChange(pos).changed();
3148 }
3149
3150
3151 bool Paragraph::isInserted(pos_type pos) const
3152 {
3153         return lookupChange(pos).inserted();
3154 }
3155
3156
3157 bool Paragraph::isDeleted(pos_type pos) const
3158 {
3159         return lookupChange(pos).deleted();
3160 }
3161
3162
3163 InsetList const & Paragraph::insetList() const
3164 {
3165         return d->insetlist_;
3166 }
3167
3168
3169 void Paragraph::setBuffer(Buffer & b)
3170 {
3171         d->insetlist_.setBuffer(b);
3172 }
3173
3174
3175 Inset * Paragraph::releaseInset(pos_type pos)
3176 {
3177         Inset * inset = d->insetlist_.release(pos);
3178         /// does not honour change tracking!
3179         eraseChar(pos, false);
3180         return inset;
3181 }
3182
3183
3184 Inset * Paragraph::getInset(pos_type pos)
3185 {
3186         return (pos < pos_type(d->text_.size()) && d->text_[pos] == META_INSET)
3187                  ? d->insetlist_.get(pos) : 0;
3188 }
3189
3190
3191 Inset const * Paragraph::getInset(pos_type pos) const
3192 {
3193         return (pos < pos_type(d->text_.size()) && d->text_[pos] == META_INSET)
3194                  ? d->insetlist_.get(pos) : 0;
3195 }
3196
3197
3198 void Paragraph::changeCase(BufferParams const & bparams, pos_type pos,
3199                 pos_type & right, TextCase action)
3200 {
3201         // process sequences of modified characters; in change
3202         // tracking mode, this approach results in much better
3203         // usability than changing case on a char-by-char basis
3204         docstring changes;
3205
3206         bool const trackChanges = bparams.trackChanges;
3207
3208         bool capitalize = true;
3209
3210         for (; pos < right; ++pos) {
3211                 char_type oldChar = d->text_[pos];
3212                 char_type newChar = oldChar;
3213
3214                 // ignore insets and don't play with deleted text!
3215                 if (oldChar != META_INSET && !isDeleted(pos)) {
3216                         switch (action) {
3217                                 case text_lowercase:
3218                                         newChar = lowercase(oldChar);
3219                                         break;
3220                                 case text_capitalization:
3221                                         if (capitalize) {
3222                                                 newChar = uppercase(oldChar);
3223                                                 capitalize = false;
3224                                         }
3225                                         break;
3226                                 case text_uppercase:
3227                                         newChar = uppercase(oldChar);
3228                                         break;
3229                         }
3230                 }
3231
3232                 if (isWordSeparator(pos) || isDeleted(pos)) {
3233                         // permit capitalization again
3234                         capitalize = true;
3235                 }
3236
3237                 if (oldChar != newChar) {
3238                         changes += newChar;
3239                         if (pos != right - 1)
3240                                 continue;
3241                         // step behind the changing area
3242                         pos++;
3243                 }
3244
3245                 int erasePos = pos - changes.size();
3246                 for (size_t i = 0; i < changes.size(); i++) {
3247                         insertChar(pos, changes[i],
3248                                    getFontSettings(bparams,
3249                                                    erasePos),
3250                                    trackChanges);
3251                         if (!eraseChar(erasePos, trackChanges)) {
3252                                 ++erasePos;
3253                                 ++pos; // advance
3254                                 ++right; // expand selection
3255                         }
3256                 }
3257                 changes.clear();
3258         }
3259 }
3260
3261
3262 int Paragraph::find(docstring const & str, bool cs, bool mw,
3263                 pos_type start_pos, bool del) const
3264 {
3265         pos_type pos = start_pos;
3266         int const strsize = str.length();
3267         int i = 0;
3268         pos_type const parsize = d->text_.size();
3269         for (i = 0; i < strsize && pos < parsize; ++i, ++pos) {
3270                 // Ignore ligature break and hyphenation chars while searching
3271                 while (pos < parsize - 1 && isInset(pos)) {
3272                         const InsetSpecialChar *isc = dynamic_cast<const InsetSpecialChar*>(getInset(pos));
3273                         if (isc == 0
3274                             || (isc->kind() != InsetSpecialChar::HYPHENATION
3275                                 && isc->kind() != InsetSpecialChar::LIGATURE_BREAK))
3276                                 break;
3277                         pos++;
3278                 }
3279                 if (cs && str[i] != d->text_[pos])
3280                         break;
3281                 if (!cs && uppercase(str[i]) != uppercase(d->text_[pos]))
3282                         break;
3283                 if (!del && isDeleted(pos))
3284                         break;
3285         }
3286
3287         if (i != strsize)
3288                 return 0;
3289
3290         // if necessary, check whether string matches word
3291         if (mw) {
3292                 if (start_pos > 0 && !isWordSeparator(start_pos - 1))
3293                         return 0;
3294                 if (pos < parsize
3295                         && !isWordSeparator(pos))
3296                         return 0;
3297         }
3298
3299         return pos - start_pos;
3300 }
3301
3302
3303 char_type Paragraph::getChar(pos_type pos) const
3304 {
3305         return d->text_[pos];
3306 }
3307
3308
3309 pos_type Paragraph::size() const
3310 {
3311         return d->text_.size();
3312 }
3313
3314
3315 bool Paragraph::empty() const
3316 {
3317         return d->text_.empty();
3318 }
3319
3320
3321 bool Paragraph::isInset(pos_type pos) const
3322 {
3323         return d->text_[pos] == META_INSET;
3324 }
3325
3326
3327 bool Paragraph::isSeparator(pos_type pos) const
3328 {
3329         //FIXME: Are we sure this can be the only separator?
3330         return d->text_[pos] == ' ';
3331 }
3332
3333
3334 void Paragraph::deregisterWords()
3335 {
3336         Private::LangWordsMap::const_iterator itl = d->words_.begin();
3337         Private::LangWordsMap::const_iterator ite = d->words_.end();
3338         for (; itl != ite; ++itl) {
3339                 WordList * wl = theWordList(itl->first);
3340                 Private::Words::const_iterator it = (itl->second).begin();
3341                 Private::Words::const_iterator et = (itl->second).end();
3342                 for (; it != et; ++it)
3343                         wl->remove(*it);
3344         }
3345         d->words_.clear();
3346 }
3347
3348
3349 void Paragraph::locateWord(pos_type & from, pos_type & to,
3350         word_location const loc) const
3351 {
3352         switch (loc) {
3353         case WHOLE_WORD_STRICT:
3354                 if (from == 0 || from == size()
3355                     || isWordSeparator(from)
3356                     || isWordSeparator(from - 1)) {
3357                         to = from;
3358                         return;
3359                 }
3360                 // no break here, we go to the next
3361
3362         case WHOLE_WORD:
3363                 // If we are already at the beginning of a word, do nothing
3364                 if (!from || isWordSeparator(from - 1))
3365                         break;
3366                 // no break here, we go to the next
3367
3368         case PREVIOUS_WORD:
3369                 // always move the cursor to the beginning of previous word
3370                 while (from && !isWordSeparator(from - 1))
3371                         --from;
3372                 break;
3373         case NEXT_WORD:
3374                 LYXERR0("Paragraph::locateWord: NEXT_WORD not implemented yet");
3375                 break;
3376         case PARTIAL_WORD:
3377                 // no need to move the 'from' cursor
3378                 break;
3379         }
3380         to = from;
3381         while (to < size() && !isWordSeparator(to))
3382                 ++to;
3383 }
3384
3385
3386 void Paragraph::collectWords()
3387 {
3388         // This is the value that needs to be exposed in the preferences
3389         // to resolve bug #6760.
3390         static int minlength = 6;
3391         pos_type n = size();
3392         for (pos_type pos = 0; pos < n; ++pos) {
3393                 if (isWordSeparator(pos))
3394                         continue;
3395                 pos_type from = pos;
3396                 locateWord(from, pos, WHOLE_WORD);
3397                 if (pos - from >= minlength) {
3398                         docstring word = asString(from, pos, AS_STR_NONE);
3399                         FontList::const_iterator cit = d->fontlist_.fontIterator(pos);
3400                         if (cit == d->fontlist_.end())
3401                                 return;
3402                         Language const * lang = cit->font().language();
3403                         d->words_[*lang].insert(word);
3404                 }
3405         }
3406 }
3407
3408
3409 void Paragraph::registerWords()
3410 {
3411         Private::LangWordsMap::const_iterator itl = d->words_.begin();
3412         Private::LangWordsMap::const_iterator ite = d->words_.end();
3413         for (; itl != ite; ++itl) {
3414                 WordList * wl = theWordList(itl->first);
3415                 Private::Words::const_iterator it = (itl->second).begin();
3416                 Private::Words::const_iterator et = (itl->second).end();
3417                 for (; it != et; ++it)
3418                         wl->insert(*it);
3419         }
3420 }
3421
3422
3423 void Paragraph::updateWords()
3424 {
3425         deregisterWords();
3426         collectWords();
3427         registerWords();
3428 }
3429
3430
3431 void Paragraph::Private::appendSkipPosition(SkipPositions & skips, pos_type const pos) const
3432 {
3433         SkipPositionsIterator begin = skips.begin();
3434         SkipPositions::iterator end = skips.end();
3435         if (pos > 0 && begin < end) {
3436                 --end;
3437                 if (end->last == pos - 1) {
3438                         end->last = pos;
3439                         return;
3440                 }
3441         }
3442         skips.insert(end, FontSpan(pos, pos));
3443 }
3444
3445
3446 Language * Paragraph::Private::locateSpellRange(
3447         pos_type & from, pos_type & to,
3448         SkipPositions & skips) const
3449 {
3450         // skip leading white space
3451         while (from < to && owner_->isWordSeparator(from))
3452                 ++from;
3453         // don't check empty range
3454         if (from >= to)
3455                 return 0;
3456         // get current language
3457         Language * lang = getSpellLanguage(from);
3458         pos_type last = from;
3459         bool samelang = true;
3460         bool sameinset = true;
3461         while (last < to && samelang && sameinset) {
3462                 // hop to end of word
3463                 while (last < to && !owner_->isWordSeparator(last)) {
3464                         if (owner_->getInset(last)) {
3465                                 appendSkipPosition(skips, last);
3466                         } else if (owner_->isDeleted(last)) {
3467                                 appendSkipPosition(skips, last);
3468                         }
3469                         ++last;
3470                 }
3471                 // hop to next word while checking for insets
3472                 while (sameinset && last < to && owner_->isWordSeparator(last)) {
3473                         if (Inset const * inset = owner_->getInset(last))
3474                                 sameinset = inset->isChar() && inset->isLetter();
3475                         if (sameinset && owner_->isDeleted(last)) {
3476                                 appendSkipPosition(skips, last);
3477                         }
3478                         if (sameinset)
3479                                 last++;
3480                 }
3481                 if (sameinset && last < to) {
3482                         // now check for language change
3483                         samelang = lang == getSpellLanguage(last);
3484                 }
3485         }
3486         // if language change detected backstep is needed
3487         if (!samelang)
3488                 --last;
3489         to = last;
3490         return lang;
3491 }
3492
3493
3494 Language * Paragraph::Private::getSpellLanguage(pos_type const from) const
3495 {
3496         Language * lang =
3497                 const_cast<Language *>(owner_->getFontSettings(
3498                         inset_owner_->buffer().params(), from).language());
3499         if (lang == inset_owner_->buffer().params().language
3500                 && !lyxrc.spellchecker_alt_lang.empty()) {
3501                 string lang_code;
3502                 string const lang_variety =
3503                         split(lyxrc.spellchecker_alt_lang, lang_code, '-');
3504                 lang->setCode(lang_code);
3505                 lang->setVariety(lang_variety);
3506         }
3507         return lang;
3508 }
3509
3510
3511 void Paragraph::requestSpellCheck(pos_type pos)
3512 {
3513         d->requestSpellCheck(pos == -1 ? size() : pos);
3514 }
3515
3516
3517 bool Paragraph::needsSpellCheck() const
3518 {
3519         SpellChecker::ChangeNumber speller_change_number = 0;
3520         if (theSpellChecker())
3521                 speller_change_number = theSpellChecker()->changeNumber();
3522         if (speller_change_number > d->speller_state_.currentChangeNumber()) {
3523                 d->speller_state_.needsCompleteRefresh(speller_change_number);
3524         }
3525         return d->needsSpellCheck();
3526 }
3527
3528
3529 SpellChecker::Result Paragraph::spellCheck(pos_type & from, pos_type & to,
3530         WordLangTuple & wl, docstring_list & suggestions,
3531         bool do_suggestion, bool check_learned) const
3532 {
3533         SpellChecker::Result result = SpellChecker::WORD_OK;
3534         SpellChecker * speller = theSpellChecker();
3535         if (!speller)
3536                 return result;
3537
3538         if (!d->layout_->spellcheck || !inInset().allowSpellCheck())
3539                 return result;
3540
3541         locateWord(from, to, WHOLE_WORD);
3542         if (from == to || from >= size())
3543                 return result;
3544
3545         docstring word = asString(from, to, AS_STR_INSETS + AS_STR_SKIPDELETE);
3546         Language * lang = d->getSpellLanguage(from);
3547
3548         wl = WordLangTuple(word, lang);
3549
3550         if (!word.size())
3551                 return result;
3552
3553         if (needsSpellCheck() || check_learned) {
3554                 // Ignore words with digits
3555                 // FIXME: make this customizable
3556                 // (note that some checkers ignore words with digits by default)
3557                 if (!hasDigit(word)) {
3558                         bool const trailing_dot = to < size() && d->text_[to] == '.';
3559                         result = speller->check(wl);
3560                         if (SpellChecker::misspelled(result) && trailing_dot) {
3561                                 wl = WordLangTuple(word.append(from_ascii(".")), lang);
3562                                 result = speller->check(wl);
3563                                 if (!SpellChecker::misspelled(result)) {
3564                                         LYXERR(Debug::GUI, "misspelled word is correct with dot: \"" <<
3565                                            word << "\" [" <<
3566                                            from << ".." << to << "]");
3567                                 }
3568                         }
3569                 }
3570                 d->setMisspelled(from, to, result);
3571         } else {
3572                 result = d->speller_state_.getState(from);
3573         }
3574
3575         bool const misspelled_ = SpellChecker::misspelled(result) ;
3576         if (misspelled_ && do_suggestion)
3577                 speller->suggest(wl, suggestions);
3578         else if (misspelled_)
3579                 LYXERR(Debug::GUI, "misspelled word: \"" <<
3580                            word << "\" [" <<
3581                            from << ".." << to << "]");
3582         else
3583                 suggestions.clear();
3584
3585         return result;
3586 }
3587
3588
3589 void Paragraph::Private::markMisspelledWords(
3590         pos_type const & first, pos_type const & last,
3591         SpellChecker::Result result,
3592         docstring const & word,
3593         SkipPositions const & skips)
3594 {
3595         if (!SpellChecker::misspelled(result)) {
3596                 setMisspelled(first, last, SpellChecker::WORD_OK);
3597                 return;
3598         }
3599         int snext = first;
3600         SpellChecker * speller = theSpellChecker();
3601         // locate and enumerate the error positions
3602         int nerrors = speller->numMisspelledWords();
3603         int numskipped = 0;
3604         SkipPositionsIterator it = skips.begin();
3605         SkipPositionsIterator et = skips.end();
3606         for (int index = 0; index < nerrors; ++index) {
3607                 int wstart;
3608                 int wlen = 0;
3609                 speller->misspelledWord(index, wstart, wlen);
3610                 /// should not happen if speller supports range checks
3611                 if (!wlen) continue;
3612                 docstring const misspelled = word.substr(wstart, wlen);
3613                 wstart += first + numskipped;
3614                 if (snext < wstart) {
3615                         /// mark the range of correct spelling
3616                         numskipped += countSkips(it, et, wstart);
3617                         setMisspelled(snext,
3618                                 wstart - 1, SpellChecker::WORD_OK);
3619                 }
3620                 snext = wstart + wlen;
3621                 numskipped += countSkips(it, et, snext);
3622                 /// mark the range of misspelling
3623                 setMisspelled(wstart, snext, result);
3624                 LYXERR(Debug::GUI, "misspelled word: \"" <<
3625                            misspelled << "\" [" <<
3626                            wstart << ".." << (snext-1) << "]");
3627                 ++snext;
3628         }
3629         if (snext <= last) {
3630                 /// mark the range of correct spelling at end
3631                 setMisspelled(snext, last, SpellChecker::WORD_OK);
3632         }
3633 }
3634
3635
3636 void Paragraph::spellCheck() const
3637 {
3638         SpellChecker * speller = theSpellChecker();
3639         if (!speller || !size() ||!needsSpellCheck())
3640                 return;
3641         pos_type start;
3642         pos_type endpos;
3643         d->rangeOfSpellCheck(start, endpos);
3644         if (speller->canCheckParagraph()) {
3645                 // loop until we leave the range
3646                 for (pos_type first = start; first < endpos; ) {
3647                         pos_type last = endpos;
3648                         Private::SkipPositions skips;
3649                         Language * lang = d->locateSpellRange(first, last, skips);
3650                         if (first >= endpos)
3651                                 break;
3652                         // start the spell checker on the unit of meaning
3653                         docstring word = asString(first, last, AS_STR_INSETS + AS_STR_SKIPDELETE);
3654                         WordLangTuple wl = WordLangTuple(word, lang);
3655                         SpellChecker::Result result = word.size() ?
3656                                 speller->check(wl) : SpellChecker::WORD_OK;
3657                         d->markMisspelledWords(first, last, result, word, skips);
3658                         first = ++last;
3659                 }
3660         } else {
3661                 static docstring_list suggestions;
3662                 pos_type to = endpos;
3663                 while (start < endpos) {
3664                         WordLangTuple wl;
3665                         spellCheck(start, to, wl, suggestions, false);
3666                         start = to + 1;
3667                 }
3668         }
3669         d->readySpellCheck();
3670 }
3671
3672
3673 bool Paragraph::isMisspelled(pos_type pos) const
3674 {
3675         return SpellChecker::misspelled(d->speller_state_.getState(pos));
3676 }
3677
3678
3679 string Paragraph::magicLabel() const
3680 {
3681         stringstream ss;
3682         ss << "magicparlabel-" << id();
3683         return ss.str();
3684 }
3685
3686
3687 } // namespace lyx