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 "WordLangTuple.h"
  47 #include "WordList.h"
  48
  49 #include "frontends/alert.h"
  50
  51 #include "insets/InsetBibitem.h"
  52 #include "insets/InsetLabel.h"
  53 #include "insets/InsetSpecialChar.h"
  54
  55 #include "support/debug.h"
  56 #include "support/docstring_list.h"
  57 #include "support/ExceptionMessage.h"
  58 #include "support/gettext.h"
  59 #include "support/lassert.h"
  60 #include "support/lstrings.h"
  61 #include "support/textutils.h"
  62
  63 #include <sstream>
  64 #include <vector>
  65
  66 using namespace std;
  67 using namespace lyx::support;
  68
  69 namespace lyx {
  70
  71 namespace {
  72
  73 /// Inset identifier (above 0x10ffff, for ucs-4)
  74 char_type const META_INSET = 0x200001;
  75
  76 }
  77
  78
  79 /////////////////////////////////////////////////////////////////////
  80 //
  81 // SpellResultRange
  82 //
  83 /////////////////////////////////////////////////////////////////////
  84
  85 class SpellResultRange {
  86 public:
  87         SpellResultRange(FontSpan range, SpellChecker::Result result)
  88         : range_(range), result_(result)
  89         {}
  90         ///
  91         FontSpan const & range() const { return range_; }
  92         ///
  93         void range(FontSpan const & r) { range_ = r; }
  94         ///
  95         SpellChecker::Result result() const { return result_; }
  96         ///
  97         void result(SpellChecker::Result r) { result_ = r; }
  98         ///
  99         bool inside(pos_type pos) const { return range_.inside(pos); }
 100         ///
 101         bool covered(FontSpan const & r) const
 102         {
 103                 // 1. first of new range inside current range or
 104                 // 2. last of new range inside current range or
 105                 // 3. first of current range inside new range or
 106                 // 4. last of current range inside new range
 107                 return range_.inside(r.first) || range_.inside(r.last) ||
 108                         r.inside(range_.first) || r.inside(range_.last);
 109         }
 110         ///
 111         void shift(pos_type pos, int offset)
 112         {
 113                 if (range_.first > pos) {
 114                         range_.first += offset;
 115                         range_.last += offset;
 116                 } else if (range_.last >= pos) {
 117                         range_.last += offset;
 118                 }
 119         }
 120 private:
 121         FontSpan range_ ;
 122         SpellChecker::Result result_ ;
 123 };
 124
 125
 126 /////////////////////////////////////////////////////////////////////
 127 //
 128 // SpellCheckerState
 129 //
 130 /////////////////////////////////////////////////////////////////////
 131
 132 class SpellCheckerState {
 133 public:
 134         SpellCheckerState()
 135         {
 136                 needs_refresh_ = true;
 137                 current_change_number_ = 0;
 138         }
 139
 140         void setRange(FontSpan const & fp, SpellChecker::Result state)
 141         {
 142                 Ranges result;
 143                 RangesIterator et = ranges_.end();
 144                 RangesIterator it = ranges_.begin();
 145                 for (; it != et; ++it) {
 146                         if (!it->covered(fp))
 147                                 result.push_back(SpellResultRange(it->range(), it->result()));
 148                         else if (state == SpellChecker::WORD_OK) {
 149                                 // trim or split the current misspelled range
 150                                 // store misspelled ranges only
 151                                 FontSpan range = it->range();
 152                                 if (fp.first > range.first) {
 153                                         // misspelled area in front of WORD_OK
 154                                         range.last = fp.first - 1;
 155                                         result.push_back(SpellResultRange(range, it->result()));
 156                                         range = it->range();
 157                                 }
 158                                 if (fp.last < range.last) {
 159                                         // misspelled area after WORD_OK range
 160                                         range.first = fp.last + 1;
 161                                         result.push_back(SpellResultRange(range, it->result()));
 162                                 }
 163                         }
 164                 }
 165                 ranges_ = result;
 166                 if (state != SpellChecker::WORD_OK)
 167                         ranges_.push_back(SpellResultRange(fp, state));
 168         }
 169
 170         void increasePosAfterPos(pos_type pos)
 171         {
 172                 correctRangesAfterPos(pos, 1);
 173                 needsRefresh(pos);
 174         }
 175
 176         void decreasePosAfterPos(pos_type pos)
 177         {
 178                 correctRangesAfterPos(pos, -1);
 179                 needsRefresh(pos);
 180         }
 181
 182         void refreshLast(pos_type pos)
 183         {
 184                 if (pos < refresh_.last)
 185                         refresh_.last = pos;
 186         }
 187
 188         SpellChecker::Result getState(pos_type pos) const
 189         {
 190                 SpellChecker::Result result = SpellChecker::WORD_OK;
 191                 RangesIterator et = ranges_.end();
 192                 RangesIterator it = ranges_.begin();
 193                 for (; it != et; ++it) {
 194                         if(it->inside(pos)) {
 195                                 return it->result();
 196                         }
 197                 }
 198                 return result;
 199         }
 200
 201         FontSpan const & getRange(pos_type pos) const
 202         {
 203                 /// empty span to indicate mismatch
 204                 static FontSpan empty_;
 205                 RangesIterator et = ranges_.end();
 206                 RangesIterator it = ranges_.begin();
 207                 for (; it != et; ++it) {
 208                         if(it->inside(pos)) {
 209                                 return it->range();
 210                         }
 211                 }
 212                 return empty_;
 213         }
 214
 215         bool needsRefresh() const
 216         {
 217                 return needs_refresh_;
 218         }
 219
 220         SpellChecker::ChangeNumber currentChangeNumber() const
 221         {
 222                 return current_change_number_;
 223         }
 224
 225         void refreshRange(pos_type & first, pos_type & last) const
 226         {
 227                 first = refresh_.first;
 228                 last = refresh_.last;
 229         }
 230
 231         void needsRefresh(pos_type pos)
 232         {
 233                 if (needs_refresh_ && pos != -1) {
 234                         if (pos < refresh_.first)
 235                                 refresh_.first = pos;
 236                         if (pos > refresh_.last)
 237                                 refresh_.last = pos;
 238                 } else if (pos != -1) {
 239                         // init request check for neighbour positions too
 240                         refresh_.first = pos > 0 ? pos - 1 : 0;
 241                         // no need for special end of paragraph check
 242                         refresh_.last = pos + 1;
 243                 }
 244                 needs_refresh_ = pos != -1;
 245         }
 246
 247         void needsCompleteRefresh(SpellChecker::ChangeNumber change_number)
 248         {
 249                 needs_refresh_ = true;
 250                 refresh_.first = 0;
 251                 refresh_.last = -1;
 252                 current_change_number_ = change_number;
 253         }
 254 private:
 255         typedef vector<SpellResultRange> Ranges;
 256         typedef Ranges::const_iterator RangesIterator;
 257         Ranges ranges_;
 258         /// the area of the paragraph with pending spell check
 259         FontSpan refresh_;
 260         bool needs_refresh_;
 261         /// spell state cache version number
 262         SpellChecker::ChangeNumber current_change_number_;
 263
 264
 265         void correctRangesAfterPos(pos_type pos, int offset)
 266         {
 267                 RangesIterator et = ranges_.end();
 268                 Ranges::iterator it = ranges_.begin();
 269                 for (; it != et; ++it) {
 270                         it->shift(pos, offset);
 271                 }
 272         }
 273
 274 };
 275
 276 /////////////////////////////////////////////////////////////////////
 277 //
 278 // Paragraph::Private
 279 //
 280 /////////////////////////////////////////////////////////////////////
 281
 282 class Paragraph::Private
 283 {
 284 public:
 285         ///
 286         Private(Paragraph * owner, Layout const & layout);
 287         /// "Copy constructor"
 288         Private(Private const &, Paragraph * owner);
 289         /// Copy constructor from \p beg  to \p end
 290         Private(Private const &, Paragraph * owner, pos_type beg, pos_type end);
 291
 292         ///
 293         void insertChar(pos_type pos, char_type c, Change const & change);
 294
 295         /// Output the surrogate pair formed by \p c and \p next to \p os.
 296         /// \return the number of characters written.
 297         int latexSurrogatePair(otexstream & os, char_type c, char_type next,
 298                                OutputParams const &);
 299
 300         /// Output a space in appropriate formatting (or a surrogate pair
 301         /// if the next character is a combining character).
 302         /// \return whether a surrogate pair was output.
 303         bool simpleTeXBlanks(OutputParams const &,
 304                              otexstream &,
 305                              pos_type i,
 306                              unsigned int & column,
 307                              Font const & font,
 308                              Layout const & style);
 309
 310         /// Output consecutive unicode chars, belonging to the same script as
 311         /// specified by the latex macro \p ltx, to \p os starting from \p i.
 312         /// \return the number of characters written.
 313         int writeScriptChars(otexstream & os, docstring const & ltx,
 314                            Change const &, Encoding const &, pos_type & i);
 315
 316         /// This could go to ParagraphParameters if we want to.
 317         int startTeXParParams(BufferParams const &, otexstream &,
 318                               OutputParams const &) const;
 319
 320         /// This could go to ParagraphParameters if we want to.
 321         bool endTeXParParams(BufferParams const &, otexstream &,
 322                              OutputParams const &) const;
 323
 324         ///
 325         void latexInset(BufferParams const &,
 326                                    otexstream &,
 327                                    OutputParams &,
 328                                    Font & running_font,
 329                                    Font & basefont,
 330                                    Font const & outerfont,
 331                                    bool & open_font,
 332                                    Change & running_change,
 333                                    Layout const & style,
 334                                    pos_type & i,
 335                                    unsigned int & column);
 336
 337         ///
 338         void latexSpecialChar(
 339                                    otexstream & os,
 340                                    BufferParams const & bparams,
 341                                    OutputParams const & runparams,
 342                                    Font const & running_font,
 343                                    Change const & running_change,
 344                                    Layout const & style,
 345                                    pos_type & i,
 346                                    pos_type end_pos,
 347                                    unsigned int & column);
 348
 349         ///
 350         bool latexSpecialT1(
 351                 char_type const c,
 352                 otexstream & os,
 353                 pos_type i,
 354                 unsigned int & column);
 355         ///
 356         bool latexSpecialT3(
 357                 char_type const c,
 358                 otexstream & os,
 359                 pos_type i,
 360                 unsigned int & column);
 361         ///
 362         bool latexSpecialTypewriter(
 363                 char_type const c,
 364                 otexstream & os,
 365                 pos_type i,
 366                 unsigned int & column);
 367         ///
 368         bool latexSpecialPhrase(
 369                 otexstream & os,
 370                 pos_type & i,
 371                 pos_type end_pos,
 372                 unsigned int & column,
 373                 OutputParams const & runparams);
 374
 375         ///
 376         void validate(LaTeXFeatures & features) const;
 377
 378         /// Checks if the paragraph contains only text and no inset or font change.
 379         bool onlyText(Buffer const & buf, Font const & outerfont,
 380                       pos_type initial) const;
 381
 382         /// match a string against a particular point in the paragraph
 383         bool isTextAt(string const & str, pos_type pos) const;
 384
 385         /// a vector of speller skip positions
 386         typedef vector<FontSpan> SkipPositions;
 387         typedef SkipPositions::const_iterator SkipPositionsIterator;
 388
 389         void appendSkipPosition(SkipPositions & skips, pos_type const pos) const;
 390
 391         Language * getSpellLanguage(pos_type const from) const;
 392
 393         Language * locateSpellRange(pos_type & from, pos_type & to,
 394                                     SkipPositions & skips) const;
 395
 396         bool hasSpellerChange() const
 397         {
 398                 SpellChecker::ChangeNumber speller_change_number = 0;
 399                 if (theSpellChecker())
 400                         speller_change_number = theSpellChecker()->changeNumber();
 401                 return speller_change_number > speller_state_.currentChangeNumber();
 402         }
 403
 404         bool ignoreWord(docstring const & word) const ;
 405
 406         void setMisspelled(pos_type from, pos_type to, SpellChecker::Result state)
 407         {
 408                 pos_type textsize = owner_->size();
 409                 // check for sane arguments
 410                 if (to <= from || from >= textsize)
 411                         return;
 412                 FontSpan fp = FontSpan(from, to - 1);
 413                 speller_state_.setRange(fp, state);
 414         }
 415
 416         void requestSpellCheck(pos_type pos)
 417         {
 418                 if (pos == -1)
 419                         speller_state_.needsCompleteRefresh(speller_state_.currentChangeNumber());
 420                 else
 421                         speller_state_.needsRefresh(pos);
 422         }
 423
 424         void readySpellCheck()
 425         {
 426                 speller_state_.needsRefresh(-1);
 427         }
 428
 429         bool needsSpellCheck() const
 430         {
 431                 return speller_state_.needsRefresh();
 432         }
 433
 434         void rangeOfSpellCheck(pos_type & first, pos_type & last) const
 435         {
 436                 speller_state_.refreshRange(first, last);
 437                 if (last == -1) {
 438                         last = owner_->size();
 439                         return;
 440                 }
 441                 pos_type endpos = last;
 442                 owner_->locateWord(first, endpos, WHOLE_WORD);
 443                 if (endpos < last) {
 444                         endpos = last;
 445                         owner_->locateWord(last, endpos, WHOLE_WORD);
 446                 }
 447                 last = endpos;
 448         }
 449
 450         int countSkips(SkipPositionsIterator & it, SkipPositionsIterator const et,
 451                             int & start) const
 452         {
 453                 int numskips = 0;
 454                 while (it != et && it->first < start) {
 455                         int skip = it->last - it->first + 1;
 456                         start += skip;
 457                         numskips += skip;
 458                         ++it;
 459                 }
 460                 return numskips;
 461         }
 462
 463         void markMisspelledWords(pos_type const & first, pos_type const & last,
 464                                                          SpellChecker::Result result,
 465                                                          docstring const & word,
 466                                                          SkipPositions const & skips);
 467
 468         InsetCode ownerCode() const
 469         {
 470                 return inset_owner_ ? inset_owner_->lyxCode() : NO_CODE;
 471         }
 472
 473         /// Which Paragraph owns us?
 474         Paragraph * owner_;
 475
 476         /// In which Inset?
 477         Inset const * inset_owner_;
 478
 479         ///
 480         FontList fontlist_;
 481
 482         ///
 483         int id_;
 484
 485         ///
 486         ParagraphParameters params_;
 487
 488         /// for recording and looking up changes
 489         Changes changes_;
 490
 491         ///
 492         InsetList insetlist_;
 493
 494         /// end of label
 495         pos_type begin_of_body_;
 496
 497         typedef docstring TextContainer;
 498         ///
 499         TextContainer text_;
 500
 501         typedef set<docstring> Words;
 502         typedef map<Language, Words> LangWordsMap;
 503         ///
 504         LangWordsMap words_;
 505         ///
 506         Layout const * layout_;
 507         ///
 508         SpellCheckerState speller_state_;
 509 };
 510
 511
 512 namespace {
 513
 514 struct special_phrase {
 515         string phrase;
 516         docstring macro;
 517         bool builtin;
 518 };
 519
 520 special_phrase const special_phrases[] = {
 521         { "LyX", from_ascii("\\LyX{}"), false },
 522         { "TeX", from_ascii("\\TeX{}"), true },
 523         { "LaTeX2e", from_ascii("\\LaTeXe{}"), true },
 524         { "LaTeX", from_ascii("\\LaTeX{}"), true },
 525 };
 526
 527 size_t const phrases_nr = sizeof(special_phrases)/sizeof(special_phrase);
 528
 529 } // namespace anon
 530
 531
 532 Paragraph::Private::Private(Paragraph * owner, Layout const & layout)
 533         : owner_(owner), inset_owner_(0), id_(-1), begin_of_body_(0), layout_(&layout)
 534 {
 535         text_.reserve(100);
 536 }
 537
 538
 539 // Initialization of the counter for the paragraph id's,
 540 //
 541 // FIXME: There should be a more intelligent way to generate and use the
 542 // paragraph ids per buffer instead a global static counter for all InsetText
 543 // in the running program.
 544 static int paragraph_id = -1;
 545
 546 Paragraph::Private::Private(Private const & p, Paragraph * owner)
 547         : owner_(owner), inset_owner_(p.inset_owner_), fontlist_(p.fontlist_),
 548           params_(p.params_), changes_(p.changes_), insetlist_(p.insetlist_),
 549           begin_of_body_(p.begin_of_body_), text_(p.text_), words_(p.words_),
 550           layout_(p.layout_)
 551 {
 552         id_ = ++paragraph_id;
 553         requestSpellCheck(p.text_.size());
 554 }
 555
 556
 557 Paragraph::Private::Private(Private const & p, Paragraph * owner,
 558         pos_type beg, pos_type end)
 559         : owner_(owner), inset_owner_(p.inset_owner_),
 560           params_(p.params_), changes_(p.changes_),
 561           insetlist_(p.insetlist_, beg, end),
 562           begin_of_body_(p.begin_of_body_), words_(p.words_),
 563           layout_(p.layout_)
 564 {
 565         id_ = ++paragraph_id;
 566         if (beg >= pos_type(p.text_.size()))
 567                 return;
 568         text_ = p.text_.substr(beg, end - beg);
 569
 570         FontList::const_iterator fcit = fontlist_.begin();
 571         FontList::const_iterator fend = fontlist_.end();
 572         for (; fcit != fend; ++fcit) {
 573                 if (fcit->pos() < beg)
 574                         continue;
 575                 if (fcit->pos() >= end) {
 576                         // Add last entry in the fontlist_.
 577                         fontlist_.set(text_.size() - 1, fcit->font());
 578                         break;
 579                 }
 580                 // Add a new entry in the fontlist_.
 581                 fontlist_.set(fcit->pos() - beg, fcit->font());
 582         }
 583         requestSpellCheck(p.text_.size());
 584 }
 585
 586
 587 void Paragraph::addChangesToToc(DocIterator const & cdit,
 588         Buffer const & buf) const
 589 {
 590         d->changes_.addToToc(cdit, buf);
 591 }
 592
 593
 594 bool Paragraph::isDeleted(pos_type start, pos_type end) const
 595 {
 596         LASSERT(start >= 0 && start <= size(), /**/);
 597         LASSERT(end > start && end <= size() + 1, /**/);
 598
 599         return d->changes_.isDeleted(start, end);
 600 }
 601
 602
 603 bool Paragraph::isChanged(pos_type start, pos_type end) const
 604 {
 605         LASSERT(start >= 0 && start <= size(), /**/);
 606         LASSERT(end > start && end <= size() + 1, /**/);
 607
 608         return d->changes_.isChanged(start, end);
 609 }
 610
 611
 612 bool Paragraph::isMergedOnEndOfParDeletion(bool trackChanges) const
 613 {
 614         // keep the logic here in sync with the logic of eraseChars()
 615         if (!trackChanges)
 616                 return true;
 617
 618         Change const change = d->changes_.lookup(size());
 619         return change.inserted() && change.currentAuthor();
 620 }
 621
 622
 623 void Paragraph::setChange(Change const & change)
 624 {
 625         // beware of the imaginary end-of-par character!
 626         d->changes_.set(change, 0, size() + 1);
 627
 628         /*
 629          * Propagate the change recursively - but not in case of DELETED!
 630          *
 631          * Imagine that your co-author makes changes in an existing inset. He
 632          * sends your document to you and you come to the conclusion that the
 633          * inset should go completely. If you erase it, LyX must not delete all
 634          * text within the inset. Otherwise, the change tracked insertions of
 635          * your co-author get lost and there is no way to restore them later.
 636          *
 637          * Conclusion: An inset's content should remain untouched if you delete it
 638          */
 639
 640         if (!change.deleted()) {
 641                 for (pos_type pos = 0; pos < size(); ++pos) {
 642                         if (Inset * inset = getInset(pos))
 643                                 inset->setChange(change);
 644                 }
 645         }
 646 }
 647
 648
 649 void Paragraph::setChange(pos_type pos, Change const & change)
 650 {
 651         LASSERT(pos >= 0 && pos <= size(), /**/);
 652         d->changes_.set(change, pos);
 653
 654         // see comment in setChange(Change const &) above
 655         if (!change.deleted() && pos < size())
 656                         if (Inset * inset = getInset(pos))
 657                                 inset->setChange(change);
 658 }
 659
 660
 661 Change const & Paragraph::lookupChange(pos_type pos) const
 662 {
 663         LASSERT(pos >= 0 && pos <= size(), /**/);
 664         return d->changes_.lookup(pos);
 665 }
 666
 667
 668 void Paragraph::acceptChanges(pos_type start, pos_type end)
 669 {
 670         LASSERT(start >= 0 && start <= size(), /**/);
 671         LASSERT(end > start && end <= size() + 1, /**/);
 672
 673         for (pos_type pos = start; pos < end; ++pos) {
 674                 switch (lookupChange(pos).type) {
 675                         case Change::UNCHANGED:
 676                                 // accept changes in nested inset
 677                                 if (Inset * inset = getInset(pos))
 678                                         inset->acceptChanges();
 679                                 break;
 680
 681                         case Change::INSERTED:
 682                                 d->changes_.set(Change(Change::UNCHANGED), pos);
 683                                 // also accept changes in nested inset
 684                                 if (Inset * inset = getInset(pos))
 685                                         inset->acceptChanges();
 686                                 break;
 687
 688                         case Change::DELETED:
 689                                 // Suppress access to non-existent
 690                                 // "end-of-paragraph char"
 691                                 if (pos < size()) {
 692                                         eraseChar(pos, false);
 693                                         --end;
 694                                         --pos;
 695                                 }
 696                                 break;
 697                 }
 698
 699         }
 700 }
 701
 702
 703 void Paragraph::rejectChanges(pos_type start, pos_type end)
 704 {
 705         LASSERT(start >= 0 && start <= size(), /**/);
 706         LASSERT(end > start && end <= size() + 1, /**/);
 707
 708         for (pos_type pos = start; pos < end; ++pos) {
 709                 switch (lookupChange(pos).type) {
 710                         case Change::UNCHANGED:
 711                                 // reject changes in nested inset
 712                                 if (Inset * inset = getInset(pos))
 713                                                 inset->rejectChanges();
 714                                 break;
 715
 716                         case Change::INSERTED:
 717                                 // Suppress access to non-existent
 718                                 // "end-of-paragraph char"
 719                                 if (pos < size()) {
 720                                         eraseChar(pos, false);
 721                                         --end;
 722                                         --pos;
 723                                 }
 724                                 break;
 725
 726                         case Change::DELETED:
 727                                 d->changes_.set(Change(Change::UNCHANGED), pos);
 728
 729                                 // Do NOT reject changes within a deleted inset!
 730                                 // There may be insertions of a co-author inside of it!
 731
 732                                 break;
 733                 }
 734         }
 735 }
 736
 737
 738 void Paragraph::Private::insertChar(pos_type pos, char_type c,
 739                 Change const & change)
 740 {
 741         LASSERT(pos >= 0 && pos <= int(text_.size()), /**/);
 742
 743         // track change
 744         changes_.insert(change, pos);
 745
 746         // This is actually very common when parsing buffers (and
 747         // maybe inserting ascii text)
 748         if (pos == pos_type(text_.size())) {
 749                 // when appending characters, no need to update tables
 750                 text_.push_back(c);
 751                 // but we want spell checking
 752                 requestSpellCheck(pos);
 753                 return;
 754         }
 755
 756         text_.insert(text_.begin() + pos, c);
 757
 758         // Update the font table.
 759         fontlist_.increasePosAfterPos(pos);
 760
 761         // Update the insets
 762         insetlist_.increasePosAfterPos(pos);
 763
 764         // Update list of misspelled positions
 765         speller_state_.increasePosAfterPos(pos);
 766 }
 767
 768
 769 bool Paragraph::insertInset(pos_type pos, Inset * inset,
 770                                    Change const & change)
 771 {
 772         LASSERT(inset, /**/);
 773         LASSERT(pos >= 0 && pos <= size(), /**/);
 774
 775         // Paragraph::insertInset() can be used in cut/copy/paste operation where
 776         // d->inset_owner_ is not set yet.
 777         if (d->inset_owner_ && !d->inset_owner_->insetAllowed(inset->lyxCode()))
 778                 return false;
 779
 780         d->insertChar(pos, META_INSET, change);
 781         LASSERT(d->text_[pos] == META_INSET, /**/);
 782
 783         // Add a new entry in the insetlist_.
 784         d->insetlist_.insert(inset, pos);
 785
 786         // Some insets require run of spell checker
 787         requestSpellCheck(pos);
 788         return true;
 789 }
 790
 791
 792 bool Paragraph::eraseChar(pos_type pos, bool trackChanges)
 793 {
 794         LASSERT(pos >= 0 && pos <= size(), return false);
 795
 796         // keep the logic here in sync with the logic of isMergedOnEndOfParDeletion()
 797
 798         if (trackChanges) {
 799                 Change change = d->changes_.lookup(pos);
 800
 801                 // set the character to DELETED if
 802                 //  a) it was previously unchanged or
 803                 //  b) it was inserted by a co-author
 804
 805                 if (!change.changed() ||
 806                       (change.inserted() && !change.currentAuthor())) {
 807                         setChange(pos, Change(Change::DELETED));
 808                         // request run of spell checker
 809                         requestSpellCheck(pos);
 810                         return false;
 811                 }
 812
 813                 if (change.deleted())
 814                         return false;
 815         }
 816
 817         // Don't physically access the imaginary end-of-paragraph character.
 818         // eraseChar() can only mark it as DELETED. A physical deletion of
 819         // end-of-par must be handled externally.
 820         if (pos == size()) {
 821                 return false;
 822         }
 823
 824         // track change
 825         d->changes_.erase(pos);
 826
 827         // if it is an inset, delete the inset entry
 828         if (d->text_[pos] == META_INSET)
 829                 d->insetlist_.erase(pos);
 830
 831         d->text_.erase(d->text_.begin() + pos);
 832
 833         // Update the fontlist_
 834         d->fontlist_.erase(pos);
 835
 836         // Update the insetlist_
 837         d->insetlist_.decreasePosAfterPos(pos);
 838
 839         // Update list of misspelled positions
 840         d->speller_state_.decreasePosAfterPos(pos);
 841         d->speller_state_.refreshLast(size());
 842
 843         return true;
 844 }
 845
 846
 847 int Paragraph::eraseChars(pos_type start, pos_type end, bool trackChanges)
 848 {
 849         LASSERT(start >= 0 && start <= size(), /**/);
 850         LASSERT(end >= start && end <= size() + 1, /**/);
 851
 852         pos_type i = start;
 853         for (pos_type count = end - start; count; --count) {
 854                 if (!eraseChar(i, trackChanges))
 855                         ++i;
 856         }
 857         return end - i;
 858 }
 859
 860
 861 int Paragraph::Private::latexSurrogatePair(otexstream & os, char_type c,
 862                 char_type next, OutputParams const & runparams)
 863 {
 864         // Writing next here may circumvent a possible font change between
 865         // c and next. Since next is only output if it forms a surrogate pair
 866         // with c we can ignore this:
 867         // A font change inside a surrogate pair does not make sense and is
 868         // hopefully impossible to input.
 869         // FIXME: change tracking
 870         // Is this correct WRT change tracking?
 871         Encoding const & encoding = *(runparams.encoding);
 872         docstring latex1 = encoding.latexChar(next).first;
 873         if (runparams.inIPA) {
 874                 string const tipashortcut = Encodings::TIPAShortcut(next);
 875                 if (!tipashortcut.empty()) {
 876                         latex1 = from_ascii(tipashortcut);
 877                 }
 878         }
 879         docstring const latex2 = encoding.latexChar(c).first;
 880         if (docstring(1, next) == latex1) {
 881                 // the encoding supports the combination
 882                 os << latex2 << latex1;
 883                 return latex1.length() + latex2.length();
 884         } else if (runparams.local_font &&
 885                    runparams.local_font->language()->lang() == "polutonikogreek") {
 886                 // polutonikogreek only works without the brackets
 887                 os << latex1 << latex2;
 888                 return latex1.length() + latex2.length();
 889         } else
 890                 os << latex1 << '{' << latex2 << '}';
 891         return latex1.length() + latex2.length() + 2;
 892 }
 893
 894
 895 bool Paragraph::Private::simpleTeXBlanks(OutputParams const & runparams,
 896                                        otexstream & os,
 897                                        pos_type i,
 898                                        unsigned int & column,
 899                                        Font const & font,
 900                                        Layout const & style)
 901 {
 902         if (style.pass_thru || runparams.pass_thru)
 903                 return false;
 904
 905         if (i + 1 < int(text_.size())) {
 906                 char_type next = text_[i + 1];
 907                 if (Encodings::isCombiningChar(next)) {
 908                         // This space has an accent, so we must always output it.
 909                         column += latexSurrogatePair(os, ' ', next, runparams) - 1;
 910                         return true;
 911                 }
 912         }
 913
 914         if (runparams.linelen > 0
 915             && column > runparams.linelen
 916             && i
 917             && text_[i - 1] != ' '
 918             && (i + 1 < int(text_.size()))
 919             // same in FreeSpacing mode
 920             && !owner_->isFreeSpacing()
 921             // In typewriter mode, we want to avoid
 922             // ! . ? : at the end of a line
 923             && !(font.fontInfo().family() == TYPEWRITER_FAMILY
 924                  && (text_[i - 1] == '.'
 925                      || text_[i - 1] == '?'
 926                      || text_[i - 1] == ':'
 927                      || text_[i - 1] == '!'))) {
 928                 os << '\n';
 929                 os.texrow().start(owner_->id(), i + 1);
 930                 column = 0;
 931         } else if (style.free_spacing) {
 932                 os << '~';
 933         } else {
 934                 os << ' ';
 935         }
 936         return false;
 937 }
 938
 939
 940 int Paragraph::Private::writeScriptChars(otexstream & os,
 941                                          docstring const & ltx,
 942                                          Change const & runningChange,
 943                                          Encoding const & encoding,
 944                                          pos_type & i)
 945 {
 946         // FIXME: modifying i here is not very nice...
 947
 948         // We only arrive here when a proper language for character text_[i] has
 949         // not been specified (i.e., it could not be translated in the current
 950         // latex encoding) or its latex translation has been forced, and it
 951         // belongs to a known script.
 952         // Parameter ltx contains the latex translation of text_[i] as specified
 953         // in the unicodesymbols file and is something like "\textXXX{<spec>}".
 954         // The latex macro name "textXXX" specifies the script to which text_[i]
 955         // belongs and we use it in order to check whether characters from the
 956         // same script immediately follow, such that we can collect them in a
 957         // single "\textXXX" macro. So, we have to retain "\textXXX{<spec>"
 958         // for the first char but only "<spec>" for all subsequent chars.
 959         docstring::size_type const brace1 = ltx.find_first_of(from_ascii("{"));
 960         docstring::size_type const brace2 = ltx.find_last_of(from_ascii("}"));
 961         string script = to_ascii(ltx.substr(1, brace1 - 1));
 962         int pos = 0;
 963         int length = brace2;
 964         bool closing_brace = true;
 965         if (script == "textgreek" && encoding.latexName() == "iso-8859-7") {
 966                 // Correct encoding is being used, so we can avoid \textgreek.
 967                 pos = brace1 + 1;
 968                 length -= pos;
 969                 closing_brace = false;
 970         }
 971         os << ltx.substr(pos, length);
 972         int size = text_.size();
 973         while (i + 1 < size) {
 974                 char_type const next = text_[i + 1];
 975                 // Stop here if next character belongs to another script
 976                 // or there is a change in change tracking status.
 977                 if (!Encodings::isKnownScriptChar(next, script) ||
 978                     runningChange != owner_->lookupChange(i + 1))
 979                         break;
 980                 Font prev_font;
 981                 bool found = false;
 982                 FontList::const_iterator cit = fontlist_.begin();
 983                 FontList::const_iterator end = fontlist_.end();
 984                 for (; cit != end; ++cit) {
 985                         if (cit->pos() >= i && !found) {
 986                                 prev_font = cit->font();
 987                                 found = true;
 988                         }
 989                         if (cit->pos() >= i + 1)
 990                                 break;
 991                 }
 992                 // Stop here if there is a font attribute or encoding change.
 993                 if (found && cit != end && prev_font != cit->font())
 994                         break;
 995                 docstring const latex = encoding.latexChar(next).first;
 996                 docstring::size_type const b1 =
 997                                         latex.find_first_of(from_ascii("{"));
 998                 docstring::size_type const b2 =
 999                                         latex.find_last_of(from_ascii("}"));
1000                 int const len = b2 - b1 - 1;
1001                 os << latex.substr(b1 + 1, len);
1002                 length += len;
1003                 ++i;
1004         }
1005         if (closing_brace) {
1006                 os << '}';
1007                 ++length;
1008         }
1009         return length;
1010 }
1011
1012
1013 bool Paragraph::Private::isTextAt(string const & str, pos_type pos) const
1014 {
1015         pos_type const len = str.length();
1016
1017         // is the paragraph large enough?
1018         if (pos + len > int(text_.size()))
1019                 return false;
1020
1021         // does the wanted text start at point?
1022         for (string::size_type i = 0; i < str.length(); ++i) {
1023                 // Caution: direct comparison of characters works only
1024                 // because str is pure ASCII.
1025                 if (str[i] != text_[pos + i])
1026                         return false;
1027         }
1028
1029         return fontlist_.hasChangeInRange(pos, len);
1030 }
1031
1032
1033 void Paragraph::Private::latexInset(BufferParams const & bparams,
1034                                     otexstream & os,
1035                                     OutputParams & runparams,
1036                                     Font & running_font,
1037                                     Font & basefont,
1038                                     Font const & outerfont,
1039                                     bool & open_font,
1040                                     Change & running_change,
1041                                     Layout const & style,
1042                                     pos_type & i,
1043                                     unsigned int & column)
1044 {
1045         Inset * inset = owner_->getInset(i);
1046         LASSERT(inset, /**/);
1047
1048         if (style.pass_thru) {
1049                 inset->plaintext(os.os(), runparams);
1050                 return;
1051         }
1052
1053         // FIXME: move this to InsetNewline::latex
1054         if (inset->lyxCode() == NEWLINE_CODE) {
1055                 // newlines are handled differently here than
1056                 // the default in simpleTeXSpecialChars().
1057                 if (!style.newline_allowed) {
1058                         os << '\n';
1059                 } else {
1060                         if (open_font) {
1061                                 column += running_font.latexWriteEndChanges(
1062                                         os, bparams, runparams,
1063                                         basefont, basefont);
1064                                 open_font = false;
1065                         }
1066
1067                         if (running_font.fontInfo().family() == TYPEWRITER_FAMILY)
1068                                 os << '~';
1069
1070                         basefont = owner_->getLayoutFont(bparams, outerfont);
1071                         running_font = basefont;
1072
1073                         if (runparams.moving_arg)
1074                                 os << "\\protect ";
1075
1076                 }
1077                 os.texrow().start(owner_->id(), i + 1);
1078                 column = 0;
1079         }
1080
1081         if (owner_->isDeleted(i)) {
1082                 if( ++runparams.inDeletedInset == 1)
1083                         runparams.changeOfDeletedInset = owner_->lookupChange(i);
1084         }
1085
1086         if (inset->canTrackChanges()) {
1087                 column += Changes::latexMarkChange(os, bparams, running_change,
1088                         Change(Change::UNCHANGED), runparams);
1089                 running_change = Change(Change::UNCHANGED);
1090         }
1091
1092         bool close = false;
1093         odocstream::pos_type const len = os.os().tellp();
1094
1095         if (inset->forceLTR()
1096             && !runparams.use_polyglossia
1097             && running_font.isRightToLeft()
1098             // ERT is an exception, it should be output with no
1099             // decorations at all
1100             && inset->lyxCode() != ERT_CODE) {
1101                 if (running_font.language()->lang() == "farsi")
1102                         os << "\\beginL{}";
1103                 else
1104                         os << "\\L{";
1105                 close = true;
1106         }
1107
1108         // FIXME: Bug: we can have an empty font change here!
1109         // if there has just been a font change, we are going to close it
1110         // right now, which means stupid latex code like \textsf{}. AFAIK,
1111         // this does not harm dvi output. A minor bug, thus (JMarc)
1112
1113         // Some insets cannot be inside a font change command.
1114         // However, even such insets *can* be placed in \L or \R
1115         // or their equivalents (for RTL language switches), so we don't
1116         // close the language in those cases.
1117         // ArabTeX, though, cannot handle this special behavior, it seems.
1118         bool arabtex = basefont.language()->lang() == "arabic_arabtex"
1119                 || running_font.language()->lang() == "arabic_arabtex";
1120         if (open_font && !inset->inheritFont()) {
1121                 bool closeLanguage = arabtex
1122                         || basefont.isRightToLeft() == running_font.isRightToLeft();
1123                 unsigned int count = running_font.latexWriteEndChanges(os,
1124                         bparams, runparams, basefont, basefont, closeLanguage);
1125                 column += count;
1126                 // if any font properties were closed, update the running_font,
1127                 // making sure, however, to leave the language as it was
1128                 if (count > 0) {
1129                         // FIXME: probably a better way to keep track of the old
1130                         // language, than copying the entire font?
1131                         Font const copy_font(running_font);
1132                         basefont = owner_->getLayoutFont(bparams, outerfont);
1133                         running_font = basefont;
1134                         if (!closeLanguage)
1135                                 running_font.setLanguage(copy_font.language());
1136                         // leave font open if language is still open
1137                         open_font = (running_font.language() == basefont.language());
1138                         if (closeLanguage)
1139                                 runparams.local_font = &basefont;
1140                 }
1141         }
1142
1143         int prev_rows = os.texrow().rows();
1144
1145         try {
1146                 runparams.lastid = id_;
1147                 runparams.lastpos = i;
1148                 inset->latex(os, runparams);
1149         } catch (EncodingException & e) {
1150                 // add location information and throw again.
1151                 e.par_id = id_;
1152                 e.pos = i;
1153                 throw(e);
1154         }
1155
1156         if (close) {
1157                 if (running_font.language()->lang() == "farsi")
1158                                 os << "\\endL{}";
1159                         else
1160                                 os << '}';
1161         }
1162
1163         if (os.texrow().rows() > prev_rows) {
1164                 os.texrow().start(owner_->id(), i + 1);
1165                 column = 0;
1166         } else {
1167                 column += (unsigned int)(os.os().tellp() - len);
1168         }
1169
1170         if (owner_->isDeleted(i))
1171                 --runparams.inDeletedInset;
1172 }
1173
1174
1175 void Paragraph::Private::latexSpecialChar(otexstream & os,
1176                                           BufferParams const & bparams,
1177                                           OutputParams const & runparams,
1178                                           Font const & running_font,
1179                                           Change const & running_change,
1180                                           Layout const & style,
1181                                           pos_type & i,
1182                                           pos_type end_pos,
1183                                           unsigned int & column)
1184 {
1185         // With polyglossia, brackets and stuff need not be reversed
1186         // in RTL scripts (see bug #8251)
1187         char_type const c = (runparams.use_polyglossia) ?
1188                 owner_->getUChar(bparams, i) : text_[i];
1189
1190         if (style.pass_thru || runparams.pass_thru) {
1191                 if (c != '\0') {
1192                         Encoding const * const enc = runparams.encoding;
1193                         if (enc && !enc->encodable(c))
1194                                 throw EncodingException(c);
1195                         os.put(c);
1196                 }
1197                 return;
1198         }
1199
1200         // TIPA uses its own T3 encoding
1201         if (runparams.inIPA && latexSpecialT3(c, os, i, column))
1202                 return;
1203         // If T1 font encoding is used, use the special
1204         // characters it provides.
1205         // NOTE: some languages reset the font encoding
1206         // internally
1207         if (!runparams.inIPA && !running_font.language()->internalFontEncoding()
1208             && lyxrc.fontenc == "T1" && latexSpecialT1(c, os, i, column))
1209                 return;
1210
1211         // \tt font needs special treatment
1212         if (!runparams.inIPA
1213              && running_font.fontInfo().family() == TYPEWRITER_FAMILY
1214              && latexSpecialTypewriter(c, os, i, column))
1215                 return;
1216
1217         // Otherwise, we use what LaTeX provides us.
1218         switch (c) {
1219         case '\\':
1220                 os << "\\textbackslash{}";
1221                 column += 15;
1222                 break;
1223         case '<':
1224                 os << "\\textless{}";
1225                 column += 10;
1226                 break;
1227         case '>':
1228                 os << "\\textgreater{}";
1229                 column += 13;
1230                 break;
1231         case '|':
1232                 os << "\\textbar{}";
1233                 column += 9;
1234                 break;
1235         case '-':
1236                 os << '-';
1237                 break;
1238         case '\"':
1239                 os << "\\char`\\\"{}";
1240                 column += 9;
1241                 break;
1242
1243         case '$': case '&':
1244         case '%': case '#': case '{':
1245         case '}': case '_':
1246                 os << '\\';
1247                 os.put(c);
1248                 column += 1;
1249                 break;
1250
1251         case '~':
1252                 os << "\\textasciitilde{}";
1253                 column += 16;
1254                 break;
1255
1256         case '^':
1257                 os << "\\textasciicircum{}";
1258                 column += 17;
1259                 break;
1260
1261         case '*':
1262         case '[':
1263         case ']':
1264                 // avoid being mistaken for optional arguments
1265                 os << '{';
1266                 os.put(c);
1267                 os << '}';
1268                 column += 2;
1269                 break;
1270
1271         case ' ':
1272                 // Blanks are printed before font switching.
1273                 // Sure? I am not! (try nice-latex)
1274                 // I am sure it's correct. LyX might be smarter
1275                 // in the future, but for now, nothing wrong is
1276                 // written. (Asger)
1277                 break;
1278
1279         default:
1280                 // LyX, LaTeX etc.
1281                 if (latexSpecialPhrase(os, i, end_pos, column, runparams))
1282                         return;
1283
1284                 if (c == '\0')
1285                         return;
1286
1287                 Encoding const & encoding = *(runparams.encoding);
1288                 char_type next = '\0';
1289                 if (i + 1 < int(text_.size())) {
1290                         next = text_[i + 1];
1291                         if (Encodings::isCombiningChar(next)) {
1292                                 column += latexSurrogatePair(os, c, next, runparams) - 1;
1293                                 ++i;
1294                                 break;
1295                         }
1296                 }
1297                 string script;
1298                 pair<docstring, bool> latex = encoding.latexChar(c);
1299                 docstring nextlatex;
1300                 bool nexttipas = false;
1301                 string nexttipashortcut;
1302                 if (next != '\0' && next != META_INSET) {
1303                         nextlatex = encoding.latexChar(next).first;
1304                         if (runparams.inIPA) {
1305                                 nexttipashortcut = Encodings::TIPAShortcut(next);
1306                                 nexttipas = !nexttipashortcut.empty();
1307                         }
1308                 }
1309                 bool tipas = false;
1310                 if (runparams.inIPA) {
1311                         string const tipashortcut = Encodings::TIPAShortcut(c);
1312                         if (!tipashortcut.empty()) {
1313                                 latex.first = from_ascii(tipashortcut);
1314                                 latex.second = false;
1315                                 tipas = true;
1316                         }
1317                 }
1318                 if (Encodings::isKnownScriptChar(c, script)
1319                     && prefixIs(latex.first, from_ascii("\\" + script)))
1320                         column += writeScriptChars(os, latex.first,
1321                                         running_change, encoding, i) - 1;
1322                 else if (latex.second
1323                          && ((!prefixIs(nextlatex, '\\')
1324                                && !prefixIs(nextlatex, '{')
1325                                && !prefixIs(nextlatex, '}'))
1326                              || (nexttipas
1327                                  && !prefixIs(from_ascii(nexttipashortcut), '\\')))
1328                          && !tipas) {
1329                         // Prevent eating of a following
1330                         // space or command corruption by
1331                         // following characters
1332                         if (next == ' ' || next == '\0') {
1333                                 column += latex.first.length() + 1;
1334                                 os << latex.first << "{}";
1335                         } else {
1336                                 column += latex.first.length();
1337                                 os << latex.first << " ";
1338                         }
1339                 } else {
1340                         column += latex.first.length() - 1;
1341                         os << latex.first;
1342                 }
1343                 break;
1344         }
1345 }
1346
1347
1348 bool Paragraph::Private::latexSpecialT1(char_type const c, otexstream & os,
1349         pos_type i, unsigned int & column)
1350 {
1351         switch (c) {
1352         case '>':
1353         case '<':
1354                 os.put(c);
1355                 // In T1 encoding, these characters exist
1356                 // but we should avoid ligatures
1357                 if (i + 1 >= int(text_.size()) || text_[i + 1] != c)
1358                         return true;
1359                 os << "\\textcompwordmark{}";
1360                 column += 19;
1361                 return true;
1362         case '|':
1363                 os.put(c);
1364                 return true;
1365         case '\"':
1366                 // soul.sty breaks with \char`\"
1367                 os << "\\textquotedbl{}";
1368                 column += 14;
1369                 return true;
1370         default:
1371                 return false;
1372         }
1373 }
1374
1375
1376 bool Paragraph::Private::latexSpecialT3(char_type const c, otexstream & os,
1377         pos_type /*i*/, unsigned int & column)
1378 {
1379         switch (c) {
1380         case '*':
1381         case '[':
1382         case ']':
1383         case '\"':
1384                 os.put(c);
1385                 return true;
1386         case '|':
1387                 os << "\\textvertline{}";
1388                 column += 14;
1389                 return true;
1390         default:
1391                 return false;
1392         }
1393 }
1394
1395
1396 bool Paragraph::Private::latexSpecialTypewriter(char_type const c, otexstream & os,
1397         pos_type i, unsigned int & column)
1398 {
1399         switch (c) {
1400         case '-':
1401                 // within \ttfamily, "--" is merged to "-" (no endash)
1402                 // so we avoid this rather irritating ligature
1403                 if (i + 1 < int(text_.size()) && text_[i + 1] == '-') {
1404                         os << "-{}";
1405                         column += 2;
1406                 } else
1407                         os << '-';
1408                 return true;
1409
1410         // everything else has to be checked separately
1411         // (depending on the encoding)
1412         default:
1413                 return false;
1414         }
1415 }
1416
1417
1418 /// \param end_pos
1419 ///   If [start_pos, end_pos) does not include entirely the special phrase, then
1420 ///   do not apply the macro transformation.
1421 bool Paragraph::Private::latexSpecialPhrase(otexstream & os, pos_type & i, pos_type end_pos,
1422         unsigned int & column, OutputParams const & runparams)
1423 {
1424         // FIXME: if we have "LaTeX" with a font
1425         // change in the middle (before the 'T', then
1426         // the "TeX" part is still special cased.
1427         // Really we should only operate this on
1428         // "words" for some definition of word
1429
1430         for (size_t pnr = 0; pnr < phrases_nr; ++pnr) {
1431                 if (!isTextAt(special_phrases[pnr].phrase, i)
1432                     || (end_pos != -1 && i + int(special_phrases[pnr].phrase.size()) > end_pos))
1433                         continue;
1434                 if (runparams.moving_arg)
1435                         os << "\\protect";
1436                 os << special_phrases[pnr].macro;
1437                 i += special_phrases[pnr].phrase.length() - 1;
1438                 column += special_phrases[pnr].macro.length() - 1;
1439                 return true;
1440         }
1441         return false;
1442 }
1443
1444
1445 void Paragraph::Private::validate(LaTeXFeatures & features) const
1446 {
1447         if (layout_->inpreamble && inset_owner_) {
1448                 bool const is_command = layout_->latextype == LATEX_COMMAND;
1449                 Buffer const & buf = inset_owner_->buffer();
1450                 BufferParams const & bp = buf.params();
1451                 Font f;
1452                 TexRow texrow;
1453                 // Using a string stream here circumvents the encoding
1454                 // switching machinery of odocstream. Therefore the
1455                 // output is wrong if this paragraph contains content
1456                 // that needs to switch encoding.
1457                 odocstringstream ods;
1458                 otexstream os(ods, texrow);
1459                 if (is_command) {
1460                         os << '\\' << from_ascii(layout_->latexname());
1461                         // we have to provide all the optional arguments here, even though
1462                         // the last one is the only one we care about.
1463                         // Separate handling of optional argument inset.
1464                         if (layout_->optargs != 0 || layout_->reqargs != 0)
1465                                 latexArgInsets(*owner_, os, features.runparams(),
1466                                         layout_->reqargs, layout_->optargs);
1467                         else
1468                                 os << from_ascii(layout_->latexparam());
1469                 }
1470                 docstring::size_type const length = ods.str().length();
1471                 // this will output "{" at the beginning, but not at the end
1472                 owner_->latex(bp, f, os, features.runparams(), 0, -1, true);
1473                 if (ods.str().length() > length) {
1474                         if (is_command)
1475                                 ods << '}';
1476                         string const snippet = to_utf8(ods.str());
1477                         features.addPreambleSnippet(snippet);
1478                 }
1479         }
1480
1481         if (features.runparams().flavor == OutputParams::HTML
1482             && layout_->htmltitle()) {
1483                 features.setHTMLTitle(owner_->asString(AS_STR_INSETS | AS_STR_SKIPDELETE));
1484         }
1485
1486         // check the params.
1487         if (!params_.spacing().isDefault())
1488                 features.require("setspace");
1489
1490         // then the layouts
1491         features.useLayout(layout_->name());
1492
1493         // then the fonts
1494         fontlist_.validate(features);
1495
1496         // then the indentation
1497         if (!params_.leftIndent().zero())
1498                 features.require("ParagraphLeftIndent");
1499
1500         // then the insets
1501         InsetList::const_iterator icit = insetlist_.begin();
1502         InsetList::const_iterator iend = insetlist_.end();
1503         for (; icit != iend; ++icit) {
1504                 if (icit->inset) {
1505                         icit->inset->validate(features);
1506                         if (layout_->needprotect &&
1507                             icit->inset->lyxCode() == FOOT_CODE)
1508                                 features.require("NeedLyXFootnoteCode");
1509                 }
1510         }
1511
1512         // then the contents
1513         for (pos_type i = 0; i < int(text_.size()) ; ++i) {
1514                 for (size_t pnr = 0; pnr < phrases_nr; ++pnr) {
1515                         if (!special_phrases[pnr].builtin
1516                             && isTextAt(special_phrases[pnr].phrase, i)) {
1517                                 features.require(special_phrases[pnr].phrase);
1518                                 break;
1519                         }
1520                 }
1521                 Encodings::validate(text_[i], features);
1522         }
1523 }
1524
1525 /////////////////////////////////////////////////////////////////////
1526 //
1527 // Paragraph
1528 //
1529 /////////////////////////////////////////////////////////////////////
1530
1531 namespace {
1532         Layout const emptyParagraphLayout;
1533 }
1534
1535 Paragraph::Paragraph()
1536         : d(new Paragraph::Private(this, emptyParagraphLayout))
1537 {
1538         itemdepth = 0;
1539         d->params_.clear();
1540 }
1541
1542
1543 Paragraph::Paragraph(Paragraph const & par)
1544         : itemdepth(par.itemdepth),
1545         d(new Paragraph::Private(*par.d, this))
1546 {
1547         registerWords();
1548 }
1549
1550
1551 Paragraph::Paragraph(Paragraph const & par, pos_type beg, pos_type end)
1552         : itemdepth(par.itemdepth),
1553         d(new Paragraph::Private(*par.d, this, beg, end))
1554 {
1555         registerWords();
1556 }
1557
1558
1559 Paragraph & Paragraph::operator=(Paragraph const & par)
1560 {
1561         // needed as we will destroy the private part before copying it
1562         if (&par != this) {
1563                 itemdepth = par.itemdepth;
1564
1565                 deregisterWords();
1566                 delete d;
1567                 d = new Private(*par.d, this);
1568                 registerWords();
1569         }
1570         return *this;
1571 }
1572
1573
1574 Paragraph::~Paragraph()
1575 {
1576         deregisterWords();
1577         delete d;
1578 }
1579
1580
1581 namespace {
1582
1583 // this shall be called just before every "os << ..." action.
1584 void flushString(ostream & os, docstring & s)
1585 {
1586         os << to_utf8(s);
1587         s.erase();
1588 }
1589
1590 }
1591
1592
1593 void Paragraph::write(ostream & os, BufferParams const & bparams,
1594         depth_type & dth) const
1595 {
1596         // The beginning or end of a deeper (i.e. nested) area?
1597         if (dth != d->params_.depth()) {
1598                 if (d->params_.depth() > dth) {
1599                         while (d->params_.depth() > dth) {
1600                                 os << "\n\\begin_deeper";
1601                                 ++dth;
1602                         }
1603                 } else {
1604                         while (d->params_.depth() < dth) {
1605                                 os << "\n\\end_deeper";
1606                                 --dth;
1607                         }
1608                 }
1609         }
1610
1611         // First write the layout
1612         os << "\n\\begin_layout " << to_utf8(d->layout_->name()) << '\n';
1613
1614         d->params_.write(os);
1615
1616         Font font1(inherit_font, bparams.language);
1617
1618         Change running_change = Change(Change::UNCHANGED);
1619
1620         // this string is used as a buffer to avoid repetitive calls
1621         // to to_utf8(), which turn out to be expensive (JMarc)
1622         docstring write_buffer;
1623
1624         int column = 0;
1625         for (pos_type i = 0; i <= size(); ++i) {
1626
1627                 Change const change = lookupChange(i);
1628                 if (change != running_change)
1629                         flushString(os, write_buffer);
1630                 Changes::lyxMarkChange(os, bparams, column, running_change, change);
1631                 running_change = change;
1632
1633                 if (i == size())
1634                         break;
1635
1636                 // Write font changes
1637                 Font font2 = getFontSettings(bparams, i);
1638                 if (font2 != font1) {
1639                         flushString(os, write_buffer);
1640                         font2.lyxWriteChanges(font1, os);
1641                         column = 0;
1642                         font1 = font2;
1643                 }
1644
1645                 char_type const c = d->text_[i];
1646                 switch (c) {
1647                 case META_INSET:
1648                         if (Inset const * inset = getInset(i)) {
1649                                 flushString(os, write_buffer);
1650                                 if (inset->directWrite()) {
1651                                         // international char, let it write
1652                                         // code directly so it's shorter in
1653                                         // the file
1654                                         inset->write(os);
1655                                 } else {
1656                                         if (i)
1657                                                 os << '\n';
1658                                         os << "\\begin_inset ";
1659                                         inset->write(os);
1660                                         os << "\n\\end_inset\n\n";
1661                                         column = 0;
1662                                 }
1663                         }
1664                         break;
1665                 case '\\':
1666                         flushString(os, write_buffer);
1667                         os << "\n\\backslash\n";
1668                         column = 0;
1669                         break;
1670                 case '.':
1671                         flushString(os, write_buffer);
1672                         if (i + 1 < size() && d->text_[i + 1] == ' ') {
1673                                 os << ".\n";
1674                                 column = 0;
1675                         } else
1676                                 os << '.';
1677                         break;
1678                 default:
1679                         if ((column > 70 && c == ' ')
1680                             || column > 79) {
1681                                 flushString(os, write_buffer);
1682                                 os << '\n';
1683                                 column = 0;
1684                         }
1685                         // this check is to amend a bug. LyX sometimes
1686                         // inserts '\0' this could cause problems.
1687                         if (c != '\0')
1688                                 write_buffer.push_back(c);
1689                         else
1690                                 LYXERR0("NUL char in structure.");
1691                         ++column;
1692                         break;
1693                 }
1694         }
1695
1696         flushString(os, write_buffer);
1697         os << "\n\\end_layout\n";
1698 }
1699
1700
1701 void Paragraph::validate(LaTeXFeatures & features) const
1702 {
1703         d->validate(features);
1704 }
1705
1706
1707 void Paragraph::insert(pos_type start, docstring const & str,
1708                        Font const & font, Change const & change)
1709 {
1710         for (size_t i = 0, n = str.size(); i != n ; ++i)
1711                 insertChar(start + i, str[i], font, change);
1712 }
1713
1714
1715 void Paragraph::appendChar(char_type c, Font const & font,
1716                 Change const & change)
1717 {
1718         // track change
1719         d->changes_.insert(change, d->text_.size());
1720         // when appending characters, no need to update tables
1721         d->text_.push_back(c);
1722         setFont(d->text_.size() - 1, font);
1723         d->requestSpellCheck(d->text_.size() - 1);
1724 }
1725
1726
1727 void Paragraph::appendString(docstring const & s, Font const & font,
1728                 Change const & change)
1729 {
1730         pos_type end = s.size();
1731         size_t oldsize = d->text_.size();
1732         size_t newsize = oldsize + end;
1733         size_t capacity = d->text_.capacity();
1734         if (newsize >= capacity)
1735                 d->text_.reserve(max(capacity + 100, newsize));
1736
1737         // when appending characters, no need to update tables
1738         d->text_.append(s);
1739
1740         // FIXME: Optimize this!
1741         for (size_t i = oldsize; i != newsize; ++i) {
1742                 // track change
1743                 d->changes_.insert(change, i);
1744                 d->requestSpellCheck(i);
1745         }
1746         d->fontlist_.set(oldsize, font);
1747         d->fontlist_.set(newsize - 1, font);
1748 }
1749
1750
1751 void Paragraph::insertChar(pos_type pos, char_type c,
1752                            bool trackChanges)
1753 {
1754         d->insertChar(pos, c, Change(trackChanges ?
1755                            Change::INSERTED : Change::UNCHANGED));
1756 }
1757
1758
1759 void Paragraph::insertChar(pos_type pos, char_type c,
1760                            Font const & font, bool trackChanges)
1761 {
1762         d->insertChar(pos, c, Change(trackChanges ?
1763                            Change::INSERTED : Change::UNCHANGED));
1764         setFont(pos, font);
1765 }
1766
1767
1768 void Paragraph::insertChar(pos_type pos, char_type c,
1769                            Font const & font, Change const & change)
1770 {
1771         d->insertChar(pos, c, change);
1772         setFont(pos, font);
1773 }
1774
1775
1776 bool Paragraph::insertInset(pos_type pos, Inset * inset,
1777                             Font const & font, Change const & change)
1778 {
1779         bool const success = insertInset(pos, inset, change);
1780         // Set the font/language of the inset...
1781         setFont(pos, font);
1782         return success;
1783 }
1784
1785
1786 void Paragraph::resetFonts(Font const & font)
1787 {
1788         d->fontlist_.clear();
1789         d->fontlist_.set(0, font);
1790         d->fontlist_.set(d->text_.size() - 1, font);
1791 }
1792
1793 // Gets uninstantiated font setting at position.
1794 Font const & Paragraph::getFontSettings(BufferParams const & bparams,
1795                                          pos_type pos) const
1796 {
1797         if (pos > size()) {
1798                 LYXERR0("pos: " << pos << " size: " << size());
1799                 LASSERT(pos <= size(), /**/);
1800         }
1801
1802         FontList::const_iterator cit = d->fontlist_.fontIterator(pos);
1803         if (cit != d->fontlist_.end())
1804                 return cit->font();
1805
1806         if (pos == size() && !empty())
1807                 return getFontSettings(bparams, pos - 1);
1808
1809         // Optimisation: avoid a full font instantiation if there is no
1810         // language change from previous call.
1811         static Font previous_font;
1812         static Language const * previous_lang = 0;
1813         Language const * lang = getParLanguage(bparams);
1814         if (lang != previous_lang) {
1815                 previous_lang = lang;
1816                 previous_font = Font(inherit_font, lang);
1817         }
1818         return previous_font;
1819 }
1820
1821
1822 FontSpan Paragraph::fontSpan(pos_type pos) const
1823 {
1824         LASSERT(pos <= size(), /**/);
1825         pos_type start = 0;
1826
1827         FontList::const_iterator cit = d->fontlist_.begin();
1828         FontList::const_iterator end = d->fontlist_.end();
1829         for (; cit != end; ++cit) {
1830                 if (cit->pos() >= pos) {
1831                         if (pos >= beginOfBody())
1832                                 return FontSpan(max(start, beginOfBody()),
1833                                                 cit->pos());
1834                         else
1835                                 return FontSpan(start,
1836                                                 min(beginOfBody() - 1,
1837                                                          cit->pos()));
1838                 }
1839                 start = cit->pos() + 1;
1840         }
1841
1842         // This should not happen, but if so, we take no chances.
1843         // LYXERR0("Paragraph::getEndPosOfFontSpan: This should not happen!");
1844         return FontSpan(pos, pos);
1845 }
1846
1847
1848 // Gets uninstantiated font setting at position 0
1849 Font const & Paragraph::getFirstFontSettings(BufferParams const & bparams) const
1850 {
1851         if (!empty() && !d->fontlist_.empty())
1852                 return d->fontlist_.begin()->font();
1853
1854         // Optimisation: avoid a full font instantiation if there is no
1855         // language change from previous call.
1856         static Font previous_font;
1857         static Language const * previous_lang = 0;
1858         if (bparams.language != previous_lang) {
1859                 previous_lang = bparams.language;
1860                 previous_font = Font(inherit_font, bparams.language);
1861         }
1862
1863         return previous_font;
1864 }
1865
1866
1867 // Gets the fully instantiated font at a given position in a paragraph
1868 // This is basically the same function as Text::GetFont() in text2.cpp.
1869 // The difference is that this one is used for generating the LaTeX file,
1870 // and thus cosmetic "improvements" are disallowed: This has to deliver
1871 // the true picture of the buffer. (Asger)
1872 Font const Paragraph::getFont(BufferParams const & bparams, pos_type pos,
1873                                  Font const & outerfont) const
1874 {
1875         LASSERT(pos >= 0, /**/);
1876
1877         Font font = getFontSettings(bparams, pos);
1878
1879         pos_type const body_pos = beginOfBody();
1880         FontInfo & fi = font.fontInfo();
1881         if (pos < body_pos)
1882                 fi.realize(d->layout_->labelfont);
1883         else
1884                 fi.realize(d->layout_->font);
1885
1886         fi.realize(outerfont.fontInfo());
1887         fi.realize(bparams.getFont().fontInfo());
1888
1889         return font;
1890 }
1891
1892
1893 Font const Paragraph::getLabelFont
1894         (BufferParams const & bparams, Font const & outerfont) const
1895 {
1896         FontInfo tmpfont = d->layout_->labelfont;
1897         tmpfont.realize(outerfont.fontInfo());
1898         tmpfont.realize(bparams.getFont().fontInfo());
1899         return Font(tmpfont, getParLanguage(bparams));
1900 }
1901
1902
1903 Font const Paragraph::getLayoutFont
1904         (BufferParams const & bparams, Font const & outerfont) const
1905 {
1906         FontInfo tmpfont = d->layout_->font;
1907         tmpfont.realize(outerfont.fontInfo());
1908         tmpfont.realize(bparams.getFont().fontInfo());
1909         return Font(tmpfont, getParLanguage(bparams));
1910 }
1911
1912
1913 /// Returns the height of the highest font in range
1914 FontSize Paragraph::highestFontInRange
1915         (pos_type startpos, pos_type endpos, FontSize def_size) const
1916 {
1917         return d->fontlist_.highestInRange(startpos, endpos, def_size);
1918 }
1919
1920
1921 char_type Paragraph::getUChar(BufferParams const & bparams, pos_type pos) const
1922 {
1923         char_type c = d->text_[pos];
1924         if (!lyxrc.rtl_support || !getFontSettings(bparams, pos).isRightToLeft())
1925                 return c;
1926
1927         // FIXME: The arabic special casing is due to the difference of arabic
1928         // round brackets input introduced in r18599. Check if this should be
1929         // unified with Hebrew or at least if all bracket types should be
1930         // handled the same (file format change in either case).
1931         string const & lang = getFontSettings(bparams, pos).language()->lang();
1932         bool const arabic = lang == "arabic_arabtex" || lang == "arabic_arabi"
1933                 || lang == "farsi";
1934         char_type uc = c;
1935         switch (c) {
1936         case '(':
1937                 uc = arabic ? c : ')';
1938                 break;
1939         case ')':
1940                 uc = arabic ? c : '(';
1941                 break;
1942         case '[':
1943                 uc = ']';
1944                 break;
1945         case ']':
1946                 uc = '[';
1947                 break;
1948         case '{':
1949                 uc = '}';
1950                 break;
1951         case '}':
1952                 uc = '{';
1953                 break;
1954         case '<':
1955                 uc = '>';
1956                 break;
1957         case '>':
1958                 uc = '<';
1959                 break;
1960         }
1961
1962         return uc;
1963 }
1964
1965
1966 void Paragraph::setFont(pos_type pos, Font const & font)
1967 {
1968         LASSERT(pos <= size(), /**/);
1969
1970         // First, reduce font against layout/label font
1971         // Update: The setCharFont() routine in text2.cpp already
1972         // reduces font, so we don't need to do that here. (Asger)
1973
1974         d->fontlist_.set(pos, font);
1975 }
1976
1977
1978 void Paragraph::makeSameLayout(Paragraph const & par)
1979 {
1980         d->layout_ = par.d->layout_;
1981         d->params_ = par.d->params_;
1982 }
1983
1984
1985 bool Paragraph::stripLeadingSpaces(bool trackChanges)
1986 {
1987         if (isFreeSpacing())
1988                 return false;
1989
1990         int pos = 0;
1991         int count = 0;
1992
1993         while (pos < size() && (isNewline(pos) || isLineSeparator(pos))) {
1994                 if (eraseChar(pos, trackChanges))
1995                         ++count;
1996                 else
1997                         ++pos;
1998         }
1999
2000         return count > 0 || pos > 0;
2001 }
2002
2003
2004 bool Paragraph::hasSameLayout(Paragraph const & par) const
2005 {
2006         return par.d->layout_ == d->layout_
2007                 && d->params_.sameLayout(par.d->params_);
2008 }
2009
2010
2011 depth_type Paragraph::getDepth() const
2012 {
2013         return d->params_.depth();
2014 }
2015
2016
2017 depth_type Paragraph::getMaxDepthAfter() const
2018 {
2019         if (d->layout_->isEnvironment())
2020                 return d->params_.depth() + 1;
2021         else
2022                 return d->params_.depth();
2023 }
2024
2025
2026 char Paragraph::getAlign() const
2027 {
2028         if (d->params_.align() == LYX_ALIGN_LAYOUT)
2029                 return d->layout_->align;
2030         else
2031                 return d->params_.align();
2032 }
2033
2034
2035 docstring const & Paragraph::labelString() const
2036 {
2037         return d->params_.labelString();
2038 }
2039
2040
2041 // the next two functions are for the manual labels
2042 docstring const Paragraph::getLabelWidthString() const
2043 {
2044         if (d->layout_->margintype == MARGIN_MANUAL
2045             || d->layout_->latextype == LATEX_BIB_ENVIRONMENT)
2046                 return d->params_.labelWidthString();
2047         else
2048                 return _("Senseless with this layout!");
2049 }
2050
2051
2052 void Paragraph::setLabelWidthString(docstring const & s)
2053 {
2054         d->params_.labelWidthString(s);
2055 }
2056
2057
2058 docstring Paragraph::expandLabel(Layout const & layout,
2059                 BufferParams const & bparams) const
2060 {
2061         return expandParagraphLabel(layout, bparams, true);
2062 }
2063
2064
2065 docstring Paragraph::expandDocBookLabel(Layout const & layout,
2066                 BufferParams const & bparams) const
2067 {
2068         return expandParagraphLabel(layout, bparams, false);
2069 }
2070
2071
2072 docstring Paragraph::expandParagraphLabel(Layout const & layout,
2073                 BufferParams const & bparams, bool process_appendix) const
2074 {
2075         DocumentClass const & tclass = bparams.documentClass();
2076         string const & lang = getParLanguage(bparams)->code();
2077         bool const in_appendix = process_appendix && d->params_.appendix();
2078         docstring fmt = translateIfPossible(layout.labelstring(in_appendix), lang);
2079
2080         if (fmt.empty() && layout.labeltype == LABEL_COUNTER
2081             && !layout.counter.empty())
2082                 return tclass.counters().theCounter(layout.counter, lang);
2083
2084         // handle 'inherited level parts' in 'fmt',
2085         // i.e. the stuff between '@' in   '@Section@.\arabic{subsection}'
2086         size_t const i = fmt.find('@', 0);
2087         if (i != docstring::npos) {
2088                 size_t const j = fmt.find('@', i + 1);
2089                 if (j != docstring::npos) {
2090                         docstring parent(fmt, i + 1, j - i - 1);
2091                         docstring label = from_ascii("??");
2092                         if (tclass.hasLayout(parent))
2093                                 docstring label = expandParagraphLabel(tclass[parent], bparams,
2094                                                       process_appendix);
2095                         fmt = docstring(fmt, 0, i) + label
2096                                 + docstring(fmt, j + 1, docstring::npos);
2097                 }
2098         }
2099
2100         return tclass.counters().counterLabel(fmt, lang);
2101 }
2102
2103
2104 void Paragraph::applyLayout(Layout const & new_layout)
2105 {
2106         d->layout_ = &new_layout;
2107         LyXAlignment const oldAlign = d->params_.align();
2108
2109         if (!(oldAlign & d->layout_->alignpossible)) {
2110                 frontend::Alert::warning(_("Alignment not permitted"),
2111                         _("The new layout does not permit the alignment previously used.\nSetting to default."));
2112                 d->params_.align(LYX_ALIGN_LAYOUT);
2113         }
2114 }
2115
2116
2117 pos_type Paragraph::beginOfBody() const
2118 {
2119         return d->begin_of_body_;
2120 }
2121
2122
2123 void Paragraph::setBeginOfBody()
2124 {
2125         if (d->layout_->labeltype != LABEL_MANUAL) {
2126                 d->begin_of_body_ = 0;
2127                 return;
2128         }
2129
2130         // Unroll the first two cycles of the loop
2131         // and remember the previous character to
2132         // remove unnecessary getChar() calls
2133         pos_type i = 0;
2134         pos_type end = size();
2135         if (i < end && !isNewline(i)) {
2136                 ++i;
2137                 char_type previous_char = 0;
2138                 char_type temp = 0;
2139                 if (i < end) {
2140                         previous_char = d->text_[i];
2141                         if (!isNewline(i)) {
2142                                 ++i;
2143                                 while (i < end && previous_char != ' ') {
2144                                         temp = d->text_[i];
2145                                         if (isNewline(i))
2146                                                 break;
2147                                         ++i;
2148                                         previous_char = temp;
2149                                 }
2150                         }
2151                 }
2152         }
2153
2154         d->begin_of_body_ = i;
2155 }
2156
2157
2158 bool Paragraph::allowParagraphCustomization() const
2159 {
2160         return inInset().allowParagraphCustomization();
2161 }
2162
2163
2164 bool Paragraph::usePlainLayout() const
2165 {
2166         return inInset().usePlainLayout();
2167 }
2168
2169
2170 bool Paragraph::isPassThru() const
2171 {
2172         return inInset().getLayout().isPassThru() || d->layout_->pass_thru;
2173 }
2174
2175 namespace {
2176
2177 // paragraphs inside floats need different alignment tags to avoid
2178 // unwanted space
2179
2180 bool noTrivlistCentering(InsetCode code)
2181 {
2182         return code == FLOAT_CODE
2183                || code == WRAP_CODE
2184                || code == CELL_CODE;
2185 }
2186
2187
2188 string correction(string const & orig)
2189 {
2190         if (orig == "flushleft")
2191                 return "raggedright";
2192         if (orig == "flushright")
2193                 return "raggedleft";
2194         if (orig == "center")
2195                 return "centering";
2196         return orig;
2197 }
2198
2199
2200 string const corrected_env(string const & suffix, string const & env,
2201         InsetCode code, bool const lastpar)
2202 {
2203         string output = suffix + "{";
2204         if (noTrivlistCentering(code)) {
2205                 if (lastpar) {
2206                         // the last paragraph in non-trivlist-aligned
2207                         // context is special (to avoid unwanted whitespace)
2208                         if (suffix == "\\begin")
2209                                 return "\\" + correction(env) + "{}";
2210                         return string();
2211                 }
2212                 output += correction(env);
2213         } else
2214                 output += env;
2215         output += "}";
2216         if (suffix == "\\begin")
2217                 output += "\n";
2218         return output;
2219 }
2220
2221
2222 void adjust_column(string const & str, int & column)
2223 {
2224         if (!contains(str, "\n"))
2225                 column += str.size();
2226         else {
2227                 string tmp;
2228                 column = rsplit(str, tmp, '\n').size();
2229         }
2230 }
2231
2232 } // namespace anon
2233
2234
2235 int Paragraph::Private::startTeXParParams(BufferParams const & bparams,
2236                         otexstream & os, OutputParams const & runparams) const
2237 {
2238         int column = 0;
2239
2240         if (params_.noindent() && !layout_->pass_thru) {
2241                 os << "\\noindent ";
2242                 column += 10;
2243         }
2244
2245         LyXAlignment const curAlign = params_.align();
2246
2247         if (curAlign == layout_->align)
2248                 return column;
2249
2250         switch (curAlign) {
2251         case LYX_ALIGN_NONE:
2252         case LYX_ALIGN_BLOCK:
2253         case LYX_ALIGN_LAYOUT:
2254         case LYX_ALIGN_SPECIAL:
2255         case LYX_ALIGN_DECIMAL:
2256                 break;
2257         case LYX_ALIGN_LEFT:
2258         case LYX_ALIGN_RIGHT:
2259         case LYX_ALIGN_CENTER:
2260                 if (runparams.moving_arg) {
2261                         os << "\\protect";
2262                         column += 8;
2263                 }
2264                 break;
2265         }
2266
2267         string const begin_tag = "\\begin";
2268         InsetCode code = ownerCode();
2269         bool const lastpar = runparams.isLastPar;
2270
2271         switch (curAlign) {
2272         case LYX_ALIGN_NONE:
2273         case LYX_ALIGN_BLOCK:
2274         case LYX_ALIGN_LAYOUT:
2275         case LYX_ALIGN_SPECIAL:
2276         case LYX_ALIGN_DECIMAL:
2277                 break;
2278         case LYX_ALIGN_LEFT: {
2279                 string output;
2280                 if (owner_->getParLanguage(bparams)->babel() != "hebrew")
2281                         output = corrected_env(begin_tag, "flushleft", code, lastpar);
2282                 else
2283                         output = corrected_env(begin_tag, "flushright", code, lastpar);
2284                 os << from_ascii(output);
2285                 adjust_column(output, column);
2286                 break;
2287         } case LYX_ALIGN_RIGHT: {
2288                 string output;
2289                 if (owner_->getParLanguage(bparams)->babel() != "hebrew")
2290                         output = corrected_env(begin_tag, "flushright", code, lastpar);
2291                 else
2292                         output = corrected_env(begin_tag, "flushleft", code, lastpar);
2293                 os << from_ascii(output);
2294                 adjust_column(output, column);
2295                 break;
2296         } case LYX_ALIGN_CENTER: {
2297                 string output;
2298                 output = corrected_env(begin_tag, "center", code, lastpar);
2299                 os << from_ascii(output);
2300                 adjust_column(output, column);
2301                 break;
2302         }
2303         }
2304
2305         return column;
2306 }
2307
2308
2309 bool Paragraph::Private::endTeXParParams(BufferParams const & bparams,
2310                         otexstream & os, OutputParams const & runparams) const
2311 {
2312         LyXAlignment const curAlign = params_.align();
2313
2314         if (curAlign == layout_->align)
2315                 return false;
2316
2317         switch (curAlign) {
2318         case LYX_ALIGN_NONE:
2319         case LYX_ALIGN_BLOCK:
2320         case LYX_ALIGN_LAYOUT:
2321         case LYX_ALIGN_SPECIAL:
2322         case LYX_ALIGN_DECIMAL:
2323                 break;
2324         case LYX_ALIGN_LEFT:
2325         case LYX_ALIGN_RIGHT:
2326         case LYX_ALIGN_CENTER:
2327                 if (runparams.moving_arg)
2328                         os << "\\protect";
2329                 break;
2330         }
2331
2332         string output;
2333         string const end_tag = "\n\\par\\end";
2334         InsetCode code = ownerCode();
2335         bool const lastpar = runparams.isLastPar;
2336
2337         switch (curAlign) {
2338         case LYX_ALIGN_NONE:
2339         case LYX_ALIGN_BLOCK:
2340         case LYX_ALIGN_LAYOUT:
2341         case LYX_ALIGN_SPECIAL:
2342         case LYX_ALIGN_DECIMAL:
2343                 break;
2344         case LYX_ALIGN_LEFT: {
2345                 if (owner_->getParLanguage(bparams)->babel() != "hebrew")
2346                         output = corrected_env(end_tag, "flushleft", code, lastpar);
2347                 else
2348                         output = corrected_env(end_tag, "flushright", code, lastpar);
2349                 os << from_ascii(output);
2350                 break;
2351         } case LYX_ALIGN_RIGHT: {
2352                 if (owner_->getParLanguage(bparams)->babel() != "hebrew")
2353                         output = corrected_env(end_tag, "flushright", code, lastpar);
2354                 else
2355                         output = corrected_env(end_tag, "flushleft", code, lastpar);
2356                 os << from_ascii(output);
2357                 break;
2358         } case LYX_ALIGN_CENTER: {
2359                 output = corrected_env(end_tag, "center", code, lastpar);
2360                 os << from_ascii(output);
2361                 break;
2362         }
2363         }
2364
2365         return !output.empty() || lastpar;
2366 }
2367
2368
2369 // This one spits out the text of the paragraph
2370 void Paragraph::latex(BufferParams const & bparams,
2371         Font const & outerfont,
2372         otexstream & os,
2373         OutputParams const & runparams,
2374         int start_pos, int end_pos, bool force) const
2375 {
2376         LYXERR(Debug::LATEX, "Paragraph::latex...     " << this);
2377
2378         // FIXME This check should not be needed. Perhaps issue an
2379         // error if it triggers.
2380         Layout const & style = inInset().forcePlainLayout() ?
2381                 bparams.documentClass().plainLayout() : *d->layout_;
2382
2383         if (!force && style.inpreamble)
2384                 return;
2385
2386         bool const allowcust = allowParagraphCustomization();
2387
2388         // Current base font for all inherited font changes, without any
2389         // change caused by an individual character, except for the language:
2390         // It is set to the language of the first character.
2391         // As long as we are in the label, this font is the base font of the
2392         // label. Before the first body character it is set to the base font
2393         // of the body.
2394         Font basefont;
2395
2396         // Maybe we have to create a optional argument.
2397         pos_type body_pos = beginOfBody();
2398         unsigned int column = 0;
2399
2400         if (body_pos > 0) {
2401                 // the optional argument is kept in curly brackets in
2402                 // case it contains a ']'
2403                 // This is not strictly needed, but if this is changed it
2404                 // would be a file format change, and tex2lyx would need
2405                 // to be adjusted, since it unconditionally removes the
2406                 // braces when it parses \item.
2407                 os << "[{";
2408                 column += 2;
2409                 basefont = getLabelFont(bparams, outerfont);
2410         } else {
2411                 basefont = getLayoutFont(bparams, outerfont);
2412         }
2413
2414         // Which font is currently active?
2415         Font running_font(basefont);
2416         // Do we have an open font change?
2417         bool open_font = false;
2418
2419         Change runningChange = Change(Change::UNCHANGED);
2420
2421         Encoding const * const prev_encoding = runparams.encoding;
2422
2423         os.texrow().start(id(), 0);
2424
2425         // if the paragraph is empty, the loop will not be entered at all
2426         if (empty()) {
2427                 if (style.isCommand()) {
2428                         os << '{';
2429                         ++column;
2430                 }
2431                 if (allowcust)
2432                         column += d->startTeXParParams(bparams, os, runparams);
2433         }
2434
2435         for (pos_type i = 0; i < size(); ++i) {
2436                 // First char in paragraph or after label?
2437                 if (i == body_pos) {
2438                         if (body_pos > 0) {
2439                                 if (open_font) {
2440                                         column += running_font.latexWriteEndChanges(
2441                                                 os, bparams, runparams,
2442                                                 basefont, basefont);
2443                                         open_font = false;
2444                                 }
2445                                 basefont = getLayoutFont(bparams, outerfont);
2446                                 running_font = basefont;
2447
2448                                 column += Changes::latexMarkChange(os, bparams,
2449                                                 runningChange, Change(Change::UNCHANGED),
2450                                                 runparams);
2451                                 runningChange = Change(Change::UNCHANGED);
2452
2453                                 os << "}] ";
2454                                 column +=3;
2455                         }
2456                         if (style.isCommand()) {
2457                                 os << '{';
2458                                 ++column;
2459                         }
2460
2461                         if (allowcust)
2462                                 column += d->startTeXParParams(bparams, os,
2463                                                             runparams);
2464                 }
2465
2466                 Change const & change = runparams.inDeletedInset
2467                         ? runparams.changeOfDeletedInset : lookupChange(i);
2468
2469                 if (bparams.outputChanges && runningChange != change) {
2470                         if (open_font) {
2471                                 column += running_font.latexWriteEndChanges(
2472                                                 os, bparams, runparams, basefont, basefont);
2473                                 open_font = false;
2474                         }
2475                         basefont = getLayoutFont(bparams, outerfont);
2476                         running_font = basefont;
2477
2478                         column += Changes::latexMarkChange(os, bparams, runningChange,
2479                                                            change, runparams);
2480                         runningChange = change;
2481                 }
2482
2483                 // do not output text which is marked deleted
2484                 // if change tracking output is disabled
2485                 if (!bparams.outputChanges && change.deleted()) {
2486                         continue;
2487                 }
2488
2489                 ++column;
2490
2491                 // Fully instantiated font
2492                 Font const font = getFont(bparams, i, outerfont);
2493
2494                 Font const last_font = running_font;
2495
2496                 // Do we need to close the previous font?
2497                 if (open_font &&
2498                     (font != running_font ||
2499                      font.language() != running_font.language()))
2500                 {
2501                         column += running_font.latexWriteEndChanges(
2502                                         os, bparams, runparams, basefont,
2503                                         (i == body_pos-1) ? basefont : font);
2504                         running_font = basefont;
2505                         open_font = false;
2506                 }
2507
2508                 string const running_lang = runparams.use_polyglossia ?
2509                         running_font.language()->polyglossia() : running_font.language()->babel();
2510                 // close babel's font environment before opening CJK.
2511                 string const lang_end_command = runparams.use_polyglossia ?
2512                         "\\end{$$lang}" : lyxrc.language_command_end;
2513                 if (!running_lang.empty() &&
2514                     font.language()->encoding()->package() == Encoding::CJK) {
2515                                 string end_tag = subst(lang_end_command,
2516                                                         "$$lang",
2517                                                         running_lang);
2518                                 os << from_ascii(end_tag);
2519                                 column += end_tag.length();
2520                 }
2521
2522                 // Switch file encoding if necessary (and allowed)
2523                 if (!runparams.pass_thru && !style.pass_thru &&
2524                     runparams.encoding->package() != Encoding::none &&
2525                     font.language()->encoding()->package() != Encoding::none) {
2526                         pair<bool, int> const enc_switch =
2527                                 switchEncoding(os.os(), bparams, runparams,
2528                                         *(font.language()->encoding()));
2529                         if (enc_switch.first) {
2530                                 column += enc_switch.second;
2531                                 runparams.encoding = font.language()->encoding();
2532                         }
2533                 }
2534
2535                 char_type const c = d->text_[i];
2536
2537                 // Do we need to change font?
2538                 if ((font != running_font ||
2539                      font.language() != running_font.language()) &&
2540                         i != body_pos - 1)
2541                 {
2542                         odocstringstream ods;
2543                         column += font.latexWriteStartChanges(ods, bparams,
2544                                                               runparams, basefont,
2545                                                               last_font);
2546                         running_font = font;
2547                         open_font = true;
2548                         docstring fontchange = ods.str();
2549                         // check whether the fontchange ends with a \\textcolor
2550                         // modifier and the text starts with a space (bug 4473)
2551                         docstring const last_modifier = rsplit(fontchange, '\\');
2552                         if (prefixIs(last_modifier, from_ascii("textcolor")) && c == ' ')
2553                                 os << fontchange << from_ascii("{}");
2554                         // check if the fontchange ends with a trailing blank
2555                         // (like "\small " (see bug 3382)
2556                         else if (suffixIs(fontchange, ' ') && c == ' ')
2557                                 os << fontchange.substr(0, fontchange.size() - 1)
2558                                    << from_ascii("{}");
2559                         else
2560                                 os << fontchange;
2561                 }
2562
2563                 // FIXME: think about end_pos implementation...
2564                 if (c == ' ' && i >= start_pos && (end_pos == -1 || i < end_pos)) {
2565                         // FIXME: integrate this case in latexSpecialChar
2566                         // Do not print the separation of the optional argument
2567                         // if style.pass_thru is false. This works because
2568                         // latexSpecialChar ignores spaces if
2569                         // style.pass_thru is false.
2570                         if (i != body_pos - 1) {
2571                                 if (d->simpleTeXBlanks(runparams, os,
2572                                                 i, column, font, style)) {
2573                                         // A surrogate pair was output. We
2574                                         // must not call latexSpecialChar
2575                                         // in this iteration, since it would output
2576                                         // the combining character again.
2577                                         ++i;
2578                                         continue;
2579                                 }
2580                         }
2581                 }
2582
2583                 OutputParams rp = runparams;
2584                 rp.free_spacing = style.free_spacing;
2585                 rp.local_font = &font;
2586                 rp.intitle = style.intitle;
2587
2588                 // Two major modes:  LaTeX or plain
2589                 // Handle here those cases common to both modes
2590                 // and then split to handle the two modes separately.
2591                 if (c == META_INSET) {
2592                         if (i >= start_pos && (end_pos == -1 || i < end_pos)) {
2593                                 d->latexInset(bparams, os, rp, running_font,
2594                                                 basefont, outerfont, open_font,
2595                                                 runningChange, style, i, column);
2596                         }
2597                 } else {
2598                         if (i >= start_pos && (end_pos == -1 || i < end_pos)) {
2599                                 try {
2600                                         d->latexSpecialChar(os, bparams, rp, running_font, runningChange,
2601                                                             style, i, end_pos, column);
2602                                 } catch (EncodingException & e) {
2603                                 if (runparams.dryrun) {
2604                                         os << "<" << _("LyX Warning: ")
2605                                            << _("uncodable character") << " '";
2606                                         os.put(c);
2607                                         os << "'>";
2608                                 } else {
2609                                         // add location information and throw again.
2610                                         e.par_id = id();
2611                                         e.pos = i;
2612                                         throw(e);
2613                                 }
2614                         }
2615                 }
2616                 }
2617
2618                 // Set the encoding to that returned from latexSpecialChar (see
2619                 // comment for encoding member in OutputParams.h)
2620                 runparams.encoding = rp.encoding;
2621         }
2622
2623         // If we have an open font definition, we have to close it
2624         if (open_font) {
2625 #ifdef FIXED_LANGUAGE_END_DETECTION
2626                 if (next_) {
2627                         running_font.latexWriteEndChanges(os, bparams,
2628                                         runparams, basefont,
2629                                         next_->getFont(bparams, 0, outerfont));
2630                 } else {
2631                         running_font.latexWriteEndChanges(os, bparams,
2632                                         runparams, basefont, basefont);
2633                 }
2634 #else
2635 //FIXME: For now we ALWAYS have to close the foreign font settings if they are
2636 //FIXME: there as we start another \selectlanguage with the next paragraph if
2637 //FIXME: we are in need of this. This should be fixed sometime (Jug)
2638                 running_font.latexWriteEndChanges(os, bparams, runparams,
2639                                 basefont, basefont);
2640 #endif
2641         }
2642
2643         column += Changes::latexMarkChange(os, bparams, runningChange,
2644                                            Change(Change::UNCHANGED), runparams);
2645
2646         // Needed if there is an optional argument but no contents.
2647         if (body_pos > 0 && body_pos == size()) {
2648                 os << "}]~";
2649         }
2650
2651         if (allowcust && d->endTeXParParams(bparams, os, runparams)
2652             && runparams.encoding != prev_encoding) {
2653                 runparams.encoding = prev_encoding;
2654                 if (!runparams.isFullUnicode())
2655                         os << setEncoding(prev_encoding->iconvName());
2656         }
2657
2658         LYXERR(Debug::LATEX, "Paragraph::latex... done " << this);
2659 }
2660
2661
2662 bool Paragraph::emptyTag() const
2663 {
2664         for (pos_type i = 0; i < size(); ++i) {
2665                 if (Inset const * inset = getInset(i)) {
2666                         InsetCode lyx_code = inset->lyxCode();
2667                         // FIXME testing like that is wrong. What is
2668                         // the intent?
2669                         if (lyx_code != TOC_CODE &&
2670                             lyx_code != INCLUDE_CODE &&
2671                             lyx_code != GRAPHICS_CODE &&
2672                             lyx_code != ERT_CODE &&
2673                             lyx_code != LISTINGS_CODE &&
2674                             lyx_code != FLOAT_CODE &&
2675                             lyx_code != TABULAR_CODE) {
2676                                 return false;
2677                         }
2678                 } else {
2679                         char_type c = d->text_[i];
2680                         if (c != ' ' && c != '\t')
2681                                 return false;
2682                 }
2683         }
2684         return true;
2685 }
2686
2687
2688 string Paragraph::getID(Buffer const & buf, OutputParams const & runparams)
2689         const
2690 {
2691         for (pos_type i = 0; i < size(); ++i) {
2692                 if (Inset const * inset = getInset(i)) {
2693                         InsetCode lyx_code = inset->lyxCode();
2694                         if (lyx_code == LABEL_CODE) {
2695                                 InsetLabel const * const il = static_cast<InsetLabel const *>(inset);
2696                                 docstring const & id = il->getParam("name");
2697                                 return "id='" + to_utf8(sgml::cleanID(buf, runparams, id)) + "'";
2698                         }
2699                 }
2700         }
2701         return string();
2702 }
2703
2704
2705 pos_type Paragraph::firstWordDocBook(odocstream & os, OutputParams const & runparams)
2706         const
2707 {
2708         pos_type i;
2709         for (i = 0; i < size(); ++i) {
2710                 if (Inset const * inset = getInset(i)) {
2711                         inset->docbook(os, runparams);
2712                 } else {
2713                         char_type c = d->text_[i];
2714                         if (c == ' ')
2715                                 break;
2716                         os << sgml::escapeChar(c);
2717                 }
2718         }
2719         return i;
2720 }
2721
2722
2723 pos_type Paragraph::firstWordLyXHTML(XHTMLStream & xs, OutputParams const & runparams)
2724         const
2725 {
2726         pos_type i;
2727         for (i = 0; i < size(); ++i) {
2728                 if (Inset const * inset = getInset(i)) {
2729                         inset->xhtml(xs, runparams);
2730                 } else {
2731                         char_type c = d->text_[i];
2732                         if (c == ' ')
2733                                 break;
2734                         xs << c;
2735                 }
2736         }
2737         return i;
2738 }
2739
2740
2741 bool Paragraph::Private::onlyText(Buffer const & buf, Font const & outerfont, pos_type initial) const
2742 {
2743         Font font_old;
2744         pos_type size = text_.size();
2745         for (pos_type i = initial; i < size; ++i) {
2746                 Font font = owner_->getFont(buf.params(), i, outerfont);
2747                 if (text_[i] == META_INSET)
2748                         return false;
2749                 if (i != initial && font != font_old)
2750                         return false;
2751                 font_old = font;
2752         }
2753
2754         return true;
2755 }
2756
2757
2758 void Paragraph::simpleDocBookOnePar(Buffer const & buf,
2759                                     odocstream & os,
2760                                     OutputParams const & runparams,
2761                                     Font const & outerfont,
2762                                     pos_type initial) const
2763 {
2764         bool emph_flag = false;
2765
2766         Layout const & style = *d->layout_;
2767         FontInfo font_old =
2768                 style.labeltype == LABEL_MANUAL ? style.labelfont : style.font;
2769
2770         if (style.pass_thru && !d->onlyText(buf, outerfont, initial))
2771                 os << "]]>";
2772
2773         // parsing main loop
2774         for (pos_type i = initial; i < size(); ++i) {
2775                 Font font = getFont(buf.params(), i, outerfont);
2776
2777                 // handle <emphasis> tag
2778                 if (font_old.emph() != font.fontInfo().emph()) {
2779                         if (font.fontInfo().emph() == FONT_ON) {
2780                                 os << "<emphasis>";
2781                                 emph_flag = true;
2782                         } else if (i != initial) {
2783                                 os << "</emphasis>";
2784                                 emph_flag = false;
2785                         }
2786                 }
2787
2788                 if (Inset const * inset = getInset(i)) {
2789                         inset->docbook(os, runparams);
2790                 } else {
2791                         char_type c = d->text_[i];
2792
2793                         if (style.pass_thru)
2794                                 os.put(c);
2795                         else
2796                                 os << sgml::escapeChar(c);
2797                 }
2798                 font_old = font.fontInfo();
2799         }
2800
2801         if (emph_flag) {
2802                 os << "</emphasis>";
2803         }
2804
2805         if (style.free_spacing)
2806                 os << '\n';
2807         if (style.pass_thru && !d->onlyText(buf, outerfont, initial))
2808                 os << "<![CDATA[";
2809 }
2810
2811
2812 docstring Paragraph::simpleLyXHTMLOnePar(Buffer const & buf,
2813                                     XHTMLStream & xs,
2814                                     OutputParams const & runparams,
2815                                     Font const & outerfont,
2816                                     pos_type initial) const
2817 {
2818         docstring retval;
2819
2820         bool emph_flag = false;
2821         bool bold_flag = false;
2822
2823         Layout const & style = *d->layout_;
2824
2825         xs.startParagraph(allowEmpty());
2826
2827         if (!runparams.for_toc && runparams.html_make_pars) {
2828                 // generate a magic label for this paragraph
2829                 string const attr = "id='" + magicLabel() + "'";
2830                 xs << html::CompTag("a", attr);
2831         }
2832
2833         FontInfo font_old =
2834                 style.labeltype == LABEL_MANUAL ? style.labelfont : style.font;
2835
2836         // parsing main loop
2837         for (pos_type i = initial; i < size(); ++i) {
2838                 // let's not show deleted material in the output
2839                 if (isDeleted(i))
2840                         continue;
2841
2842                 Font font = getFont(buf.params(), i, outerfont);
2843
2844                 // emphasis
2845                 if (font_old.emph() != font.fontInfo().emph()) {
2846                         if (font.fontInfo().emph() == FONT_ON) {
2847                                 xs << html::StartTag("em");
2848                                 emph_flag = true;
2849                         } else if (emph_flag && i != initial) {
2850                                 xs << html::EndTag("em");
2851                                 emph_flag = false;
2852                         }
2853                 }
2854                 // bold
2855                 if (font_old.series() != font.fontInfo().series()) {
2856                         if (font.fontInfo().series() == BOLD_SERIES) {
2857                                 xs << html::StartTag("strong");
2858                                 bold_flag = true;
2859                         } else if (bold_flag && i != initial) {
2860                                 xs << html::EndTag("strong");
2861                                 bold_flag = false;
2862                         }
2863                 }
2864                 // FIXME XHTML
2865                 // Other such tags? What about the other text ranges?
2866
2867                 Inset const * inset = getInset(i);
2868                 if (inset) {
2869                         if (!runparams.for_toc || inset->isInToc()) {
2870                                 OutputParams np = runparams;
2871                                 if (!inset->getLayout().htmlisblock())
2872                                         np.html_in_par = true;
2873                                 retval += inset->xhtml(xs, np);
2874                         }
2875                 } else {
2876                         char_type c = getUChar(buf.params(), i);
2877
2878                         if (style.pass_thru)
2879                                 xs << c;
2880                         else if (c == '-') {
2881                                 docstring str;
2882                                 int j = i + 1;
2883                                 if (j < size() && d->text_[j] == '-') {
2884                                         j += 1;
2885                                         if (j < size() && d->text_[j] == '-') {
2886                                                 str += from_ascii("&mdash;");
2887                                                 i += 2;
2888                                         } else {
2889                                                 str += from_ascii("&ndash;");
2890                                                 i += 1;
2891                                         }
2892                                 }
2893                                 else
2894                                         str += c;
2895                                 // We don't want to escape the entities. Note that
2896                                 // it is safe to do this, since str can otherwise
2897                                 // only be "-". E.g., it can't be "<".
2898                                 xs << XHTMLStream::ESCAPE_NONE << str;
2899                         } else
2900                                 xs << c;
2901                 }
2902                 font_old = font.fontInfo();
2903         }
2904
2905         xs.closeFontTags();
2906         xs.endParagraph();
2907         return retval;
2908 }
2909
2910
2911 bool Paragraph::isHfill(pos_type pos) const
2912 {
2913         Inset const * inset = getInset(pos);
2914         return inset && (inset->lyxCode() == SPACE_CODE &&
2915                          inset->isStretchableSpace());
2916 }
2917
2918
2919 bool Paragraph::isNewline(pos_type pos) const
2920 {
2921         Inset const * inset = getInset(pos);
2922         return inset && inset->lyxCode() == NEWLINE_CODE;
2923 }
2924
2925
2926 bool Paragraph::isLineSeparator(pos_type pos) const
2927 {
2928         char_type const c = d->text_[pos];
2929         if (isLineSeparatorChar(c))
2930                 return true;
2931         Inset const * inset = getInset(pos);
2932         return inset && inset->isLineSeparator();
2933 }
2934
2935
2936 bool Paragraph::isWordSeparator(pos_type pos) const
2937 {
2938         if (pos == size())
2939                 return true;
2940         if (Inset const * inset = getInset(pos))
2941                 return !inset->isLetter();
2942         // if we have a hard hyphen (no en- or emdash) or apostrophe
2943         // we pass this to the spell checker
2944         // FIXME: this method is subject to change, visit
2945         // https://bugzilla.mozilla.org/show_bug.cgi?id=355178
2946         // to get an impression how complex this is.
2947         if (isHardHyphenOrApostrophe(pos))
2948                 return false;
2949         char_type const c = d->text_[pos];
2950         // We want to pass the escape chars to the spellchecker
2951         docstring const escape_chars = from_utf8(lyxrc.spellchecker_esc_chars);
2952         return !isLetterChar(c) && !isDigitASCII(c) && !contains(escape_chars, c);
2953 }
2954
2955
2956 bool Paragraph::isHardHyphenOrApostrophe(pos_type pos) const
2957 {
2958         pos_type const psize = size();
2959         if (pos >= psize)
2960                 return false;
2961         char_type const c = d->text_[pos];
2962         if (c != '-' && c != '\'')
2963                 return false;
2964         int nextpos = pos + 1;
2965         int prevpos = pos > 0 ? pos - 1 : 0;
2966         if ((nextpos == psize || isSpace(nextpos))
2967                 && (pos == 0 || isSpace(prevpos)))
2968                 return false;
2969         return c == '\''
2970                 || ((nextpos == psize || d->text_[nextpos] != '-')
2971                 && (pos == 0 || d->text_[prevpos] != '-'));
2972 }
2973
2974
2975 bool Paragraph::isSameSpellRange(pos_type pos1, pos_type pos2) const
2976 {
2977         return pos1 == pos2
2978                 || d->speller_state_.getRange(pos1) == d->speller_state_.getRange(pos2);
2979 }
2980
2981
2982 bool Paragraph::isChar(pos_type pos) const
2983 {
2984         if (Inset const * inset = getInset(pos))
2985                 return inset->isChar();
2986         char_type const c = d->text_[pos];
2987         return !isLetterChar(c) && !isDigitASCII(c) && !lyx::isSpace(c);
2988 }
2989
2990
2991 bool Paragraph::isSpace(pos_type pos) const
2992 {
2993         if (Inset const * inset = getInset(pos))
2994                 return inset->isSpace();
2995         char_type const c = d->text_[pos];
2996         return lyx::isSpace(c);
2997 }
2998
2999
3000 Language const *
3001 Paragraph::getParLanguage(BufferParams const & bparams) const
3002 {
3003         if (!empty())
3004                 return getFirstFontSettings(bparams).language();
3005         // FIXME: we should check the prev par as well (Lgb)
3006         return bparams.language;
3007 }
3008
3009
3010 bool Paragraph::isRTL(BufferParams const & bparams) const
3011 {
3012         return lyxrc.rtl_support
3013                 && getParLanguage(bparams)->rightToLeft()
3014                 && !inInset().getLayout().forceLTR();
3015 }
3016
3017
3018 void Paragraph::changeLanguage(BufferParams const & bparams,
3019                                Language const * from, Language const * to)
3020 {
3021         // change language including dummy font change at the end
3022         for (pos_type i = 0; i <= size(); ++i) {
3023                 Font font = getFontSettings(bparams, i);
3024                 if (font.language() == from) {
3025                         font.setLanguage(to);
3026                         setFont(i, font);
3027                         d->requestSpellCheck(i);
3028                 }
3029         }
3030 }
3031
3032
3033 bool Paragraph::isMultiLingual(BufferParams const & bparams) const
3034 {
3035         Language const * doc_language = bparams.language;
3036         FontList::const_iterator cit = d->fontlist_.begin();
3037         FontList::const_iterator end = d->fontlist_.end();
3038
3039         for (; cit != end; ++cit)
3040                 if (cit->font().language() != ignore_language &&
3041                     cit->font().language() != latex_language &&
3042                     cit->font().language() != doc_language)
3043                         return true;
3044         return false;
3045 }
3046
3047
3048 void Paragraph::getLanguages(std::set<Language const *> & languages) const
3049 {
3050         FontList::const_iterator cit = d->fontlist_.begin();
3051         FontList::const_iterator end = d->fontlist_.end();
3052
3053         for (; cit != end; ++cit) {
3054                 Language const * lang = cit->font().language();
3055                 if (lang != ignore_language &&
3056                     lang != latex_language)
3057                         languages.insert(lang);
3058         }
3059 }
3060
3061
3062 docstring Paragraph::asString(int options) const
3063 {
3064         return asString(0, size(), options);
3065 }
3066
3067
3068 docstring Paragraph::asString(pos_type beg, pos_type end, int options) const
3069 {
3070         odocstringstream os;
3071
3072         if (beg == 0
3073             && options & AS_STR_LABEL
3074             && !d->params_.labelString().empty())
3075                 os << d->params_.labelString() << ' ';
3076
3077         for (pos_type i = beg; i < end; ++i) {
3078                 if ((options & AS_STR_SKIPDELETE) && isDeleted(i))
3079                         continue;
3080                 char_type const c = d->text_[i];
3081                 if (isPrintable(c) || c == '\t'
3082                     || (c == '\n' && (options & AS_STR_NEWLINES)))
3083                         os.put(c);
3084                 else if (c == META_INSET && (options & AS_STR_INSETS)) {
3085                         getInset(i)->toString(os);
3086                         if (getInset(i)->asInsetMath())
3087                                 os << " ";
3088                 }
3089         }
3090
3091         return os.str();
3092 }
3093
3094
3095 void Paragraph::forToc(docstring & os, size_t maxlen) const
3096 {
3097         if (!d->params_.labelString().empty())
3098                 os += d->params_.labelString() + ' ';
3099         for (pos_type i = 0; i < size() && os.length() < maxlen; ++i) {
3100                 if (isDeleted(i))
3101                         continue;
3102                 char_type const c = d->text_[i];
3103                 if (isPrintable(c))
3104                         os += c;
3105                 else if (c == '\t' || c == '\n')
3106                         os += ' ';
3107                 else if (c == META_INSET)
3108                         getInset(i)->forToc(os, maxlen);
3109         }
3110 }
3111
3112
3113 docstring Paragraph::stringify(pos_type beg, pos_type end, int options, OutputParams & runparams) const
3114 {
3115         odocstringstream os;
3116
3117         if (beg == 0
3118                 && options & AS_STR_LABEL
3119                 && !d->params_.labelString().empty())
3120                 os << d->params_.labelString() << ' ';
3121
3122         for (pos_type i = beg; i < end; ++i) {
3123                 char_type const c = d->text_[i];
3124                 if (isPrintable(c) || c == '\t'
3125                     || (c == '\n' && (options & AS_STR_NEWLINES)))
3126                         os.put(c);
3127                 else if (c == META_INSET && (options & AS_STR_INSETS)) {
3128                         getInset(i)->plaintext(os, runparams);
3129                 }
3130         }
3131
3132         return os.str();
3133 }
3134
3135
3136 void Paragraph::setInsetOwner(Inset const * inset)
3137 {
3138         d->inset_owner_ = inset;
3139 }
3140
3141
3142 int Paragraph::id() const
3143 {
3144         return d->id_;
3145 }
3146
3147
3148 void Paragraph::setId(int id)
3149 {
3150         d->id_ = id;
3151 }
3152
3153
3154 Layout const & Paragraph::layout() const
3155 {
3156         return *d->layout_;
3157 }
3158
3159
3160 void Paragraph::setLayout(Layout const & layout)
3161 {
3162         d->layout_ = &layout;
3163 }
3164
3165
3166 void Paragraph::setDefaultLayout(DocumentClass const & tc)
3167 {
3168         setLayout(tc.defaultLayout());
3169 }
3170
3171
3172 void Paragraph::setPlainLayout(DocumentClass const & tc)
3173 {
3174         setLayout(tc.plainLayout());
3175 }
3176
3177
3178 void Paragraph::setPlainOrDefaultLayout(DocumentClass const & tclass)
3179 {
3180         if (usePlainLayout())
3181                 setPlainLayout(tclass);
3182         else
3183                 setDefaultLayout(tclass);
3184 }
3185
3186
3187 Inset const & Paragraph::inInset() const
3188 {
3189         LASSERT(d->inset_owner_, throw ExceptionMessage(BufferException,
3190                 _("Memory problem"), _("Paragraph not properly initialized")));
3191         return *d->inset_owner_;
3192 }
3193
3194
3195 ParagraphParameters & Paragraph::params()
3196 {
3197         return d->params_;
3198 }
3199
3200
3201 ParagraphParameters const & Paragraph::params() const
3202 {
3203         return d->params_;
3204 }
3205
3206
3207 bool Paragraph::isFreeSpacing() const
3208 {
3209         if (d->layout_->free_spacing)
3210                 return true;
3211         return d->inset_owner_ && d->inset_owner_->isFreeSpacing();
3212 }
3213
3214
3215 bool Paragraph::allowEmpty() const
3216 {
3217         if (d->layout_->keepempty)
3218                 return true;
3219         return d->inset_owner_ && d->inset_owner_->allowEmpty();
3220 }
3221
3222
3223 char_type Paragraph::transformChar(char_type c, pos_type pos) const
3224 {
3225         if (!Encodings::isArabicChar(c))
3226                 return c;
3227
3228         char_type prev_char = ' ';
3229         char_type next_char = ' ';
3230
3231         for (pos_type i = pos - 1; i >= 0; --i) {
3232                 char_type const par_char = d->text_[i];
3233                 if (!Encodings::isArabicComposeChar(par_char)) {
3234                         prev_char = par_char;
3235                         break;
3236                 }
3237         }
3238
3239         for (pos_type i = pos + 1, end = size(); i < end; ++i) {
3240                 char_type const par_char = d->text_[i];
3241                 if (!Encodings::isArabicComposeChar(par_char)) {
3242                         next_char = par_char;
3243                         break;
3244                 }
3245         }
3246
3247         if (Encodings::isArabicChar(next_char)) {
3248                 if (Encodings::isArabicChar(prev_char) &&
3249                         !Encodings::isArabicSpecialChar(prev_char))
3250                         return Encodings::transformChar(c, Encodings::FORM_MEDIAL);
3251                 else
3252                         return Encodings::transformChar(c, Encodings::FORM_INITIAL);
3253         } else {
3254                 if (Encodings::isArabicChar(prev_char) &&
3255                         !Encodings::isArabicSpecialChar(prev_char))
3256                         return Encodings::transformChar(c, Encodings::FORM_FINAL);
3257                 else
3258                         return Encodings::transformChar(c, Encodings::FORM_ISOLATED);
3259         }
3260 }
3261
3262
3263 bool Paragraph::brokenBiblio() const
3264 {
3265         // there is a problem if there is no bibitem at position 0 or
3266         // if there is another bibitem in the paragraph.
3267         return d->layout_->labeltype == LABEL_BIBLIO
3268                 && (d->insetlist_.find(BIBITEM_CODE) != 0
3269                     || d->insetlist_.find(BIBITEM_CODE, 1) > 0);
3270 }
3271
3272
3273 int Paragraph::fixBiblio(Buffer const & buffer)
3274 {
3275         // FIXME: What about the case where paragraph is not BIBLIO
3276         // but there is an InsetBibitem?
3277         // FIXME: when there was already an inset at 0, the return value is 1,
3278         // which does not tell whether another inset has been remove; the
3279         // cursor cannot be correctly updated.
3280
3281         if (d->layout_->labeltype != LABEL_BIBLIO)
3282                 return 0;
3283
3284         bool const track_changes = buffer.params().trackChanges;
3285         int bibitem_pos = d->insetlist_.find(BIBITEM_CODE);
3286         bool const hasbibitem0 = bibitem_pos == 0;
3287
3288         if (hasbibitem0) {
3289                 bibitem_pos = d->insetlist_.find(BIBITEM_CODE, 1);
3290                 // There was an InsetBibitem at pos 0, and no other one => OK
3291                 if (bibitem_pos == -1)
3292                         return 0;
3293                 // there is a bibitem at the 0 position, but since
3294                 // there is a second one, we copy the second on the
3295                 // first. We're assuming there are at most two of
3296                 // these, which there should be.
3297                 // FIXME: why does it make sense to do that rather
3298                 // than keep the first? (JMarc)
3299                 Inset * inset = d->insetlist_.release(bibitem_pos);
3300                 eraseChar(bibitem_pos, track_changes);
3301                 d->insetlist_.begin()->inset = inset;
3302                 return -bibitem_pos;
3303         }
3304
3305         // We need to create an inset at the beginning
3306         Inset * inset = 0;
3307         if (bibitem_pos > 0) {
3308                 // there was one somewhere in the paragraph, let's move it
3309                 inset = d->insetlist_.release(bibitem_pos);
3310                 eraseChar(bibitem_pos, track_changes);
3311         } else
3312                 // make a fresh one
3313                 inset = new InsetBibitem(const_cast<Buffer *>(&buffer),
3314                                          InsetCommandParams(BIBITEM_CODE));
3315
3316         insertInset(0, inset, Change(track_changes ? Change::INSERTED
3317                                                    : Change::UNCHANGED));
3318
3319         return 1;
3320 }
3321
3322
3323 void Paragraph::checkAuthors(AuthorList const & authorList)
3324 {
3325         d->changes_.checkAuthors(authorList);
3326 }
3327
3328
3329 bool Paragraph::isChanged(pos_type pos) const
3330 {
3331         return lookupChange(pos).changed();
3332 }
3333
3334
3335 bool Paragraph::isInserted(pos_type pos) const
3336 {
3337         return lookupChange(pos).inserted();
3338 }
3339
3340
3341 bool Paragraph::isDeleted(pos_type pos) const
3342 {
3343         return lookupChange(pos).deleted();
3344 }
3345
3346
3347 InsetList const & Paragraph::insetList() const
3348 {
3349         return d->insetlist_;
3350 }
3351
3352
3353 void Paragraph::setBuffer(Buffer & b)
3354 {
3355         d->insetlist_.setBuffer(b);
3356 }
3357
3358
3359 Inset * Paragraph::releaseInset(pos_type pos)
3360 {
3361         Inset * inset = d->insetlist_.release(pos);
3362         /// does not honour change tracking!
3363         eraseChar(pos, false);
3364         return inset;
3365 }
3366
3367
3368 Inset * Paragraph::getInset(pos_type pos)
3369 {
3370         return (pos < pos_type(d->text_.size()) && d->text_[pos] == META_INSET)
3371                  ? d->insetlist_.get(pos) : 0;
3372 }
3373
3374
3375 Inset const * Paragraph::getInset(pos_type pos) const
3376 {
3377         return (pos < pos_type(d->text_.size()) && d->text_[pos] == META_INSET)
3378                  ? d->insetlist_.get(pos) : 0;
3379 }
3380
3381
3382 void Paragraph::changeCase(BufferParams const & bparams, pos_type pos,
3383                 pos_type & right, TextCase action)
3384 {
3385         // process sequences of modified characters; in change
3386         // tracking mode, this approach results in much better
3387         // usability than changing case on a char-by-char basis
3388         docstring changes;
3389
3390         bool const trackChanges = bparams.trackChanges;
3391
3392         bool capitalize = true;
3393
3394         for (; pos < right; ++pos) {
3395                 char_type oldChar = d->text_[pos];
3396                 char_type newChar = oldChar;
3397
3398                 // ignore insets and don't play with deleted text!
3399                 if (oldChar != META_INSET && !isDeleted(pos)) {
3400                         switch (action) {
3401                                 case text_lowercase:
3402                                         newChar = lowercase(oldChar);
3403                                         break;
3404                                 case text_capitalization:
3405                                         if (capitalize) {
3406                                                 newChar = uppercase(oldChar);
3407                                                 capitalize = false;
3408                                         }
3409                                         break;
3410                                 case text_uppercase:
3411                                         newChar = uppercase(oldChar);
3412                                         break;
3413                         }
3414                 }
3415
3416                 if (isWordSeparator(pos) || isDeleted(pos)) {
3417                         // permit capitalization again
3418                         capitalize = true;
3419                 }
3420
3421                 if (oldChar != newChar) {
3422                         changes += newChar;
3423                         if (pos != right - 1)
3424                                 continue;
3425                         // step behind the changing area
3426                         pos++;
3427                 }
3428
3429                 int erasePos = pos - changes.size();
3430                 for (size_t i = 0; i < changes.size(); i++) {
3431                         insertChar(pos, changes[i],
3432                                    getFontSettings(bparams,
3433                                                    erasePos),
3434                                    trackChanges);
3435                         if (!eraseChar(erasePos, trackChanges)) {
3436                                 ++erasePos;
3437                                 ++pos; // advance
3438                                 ++right; // expand selection
3439                         }
3440                 }
3441                 changes.clear();
3442         }
3443 }
3444
3445
3446 int Paragraph::find(docstring const & str, bool cs, bool mw,
3447                 pos_type start_pos, bool del) const
3448 {
3449         pos_type pos = start_pos;
3450         int const strsize = str.length();
3451         int i = 0;
3452         pos_type const parsize = d->text_.size();
3453         for (i = 0; i < strsize && pos < parsize; ++i, ++pos) {
3454                 // Ignore "invisible" letters such as ligature breaks
3455                 // and hyphenation chars while searching
3456                 while (pos < parsize - 1 && isInset(pos)) {
3457                         odocstringstream os;
3458                         getInset(pos)->toString(os);
3459                         if (!getInset(pos)->isLetter() || !os.str().empty())
3460                                 break;
3461                         pos++;
3462                 }
3463                 if (cs && str[i] != d->text_[pos])
3464                         break;
3465                 if (!cs && uppercase(str[i]) != uppercase(d->text_[pos]))
3466                         break;
3467                 if (!del && isDeleted(pos))
3468                         break;
3469         }
3470
3471         if (i != strsize)
3472                 return 0;
3473
3474         // if necessary, check whether string matches word
3475         if (mw) {
3476                 if (start_pos > 0 && !isWordSeparator(start_pos - 1))
3477                         return 0;
3478                 if (pos < parsize
3479                         && !isWordSeparator(pos))
3480                         return 0;
3481         }
3482
3483         return pos - start_pos;
3484 }
3485
3486
3487 char_type Paragraph::getChar(pos_type pos) const
3488 {
3489         return d->text_[pos];
3490 }
3491
3492
3493 pos_type Paragraph::size() const
3494 {
3495         return d->text_.size();
3496 }
3497
3498
3499 bool Paragraph::empty() const
3500 {
3501         return d->text_.empty();
3502 }
3503
3504
3505 bool Paragraph::isInset(pos_type pos) const
3506 {
3507         return d->text_[pos] == META_INSET;
3508 }
3509
3510
3511 bool Paragraph::isSeparator(pos_type pos) const
3512 {
3513         //FIXME: Are we sure this can be the only separator?
3514         return d->text_[pos] == ' ';
3515 }
3516
3517
3518 void Paragraph::deregisterWords()
3519 {
3520         Private::LangWordsMap::const_iterator itl = d->words_.begin();
3521         Private::LangWordsMap::const_iterator ite = d->words_.end();
3522         for (; itl != ite; ++itl) {
3523                 WordList * wl = theWordList(itl->first);
3524                 Private::Words::const_iterator it = (itl->second).begin();
3525                 Private::Words::const_iterator et = (itl->second).end();
3526                 for (; it != et; ++it)
3527                         wl->remove(*it);
3528         }
3529         d->words_.clear();
3530 }
3531
3532
3533 void Paragraph::locateWord(pos_type & from, pos_type & to,
3534         word_location const loc) const
3535 {
3536         switch (loc) {
3537         case WHOLE_WORD_STRICT:
3538                 if (from == 0 || from == size()
3539                     || isWordSeparator(from)
3540                     || isWordSeparator(from - 1)) {
3541                         to = from;
3542                         return;
3543                 }
3544                 // no break here, we go to the next
3545
3546         case WHOLE_WORD:
3547                 // If we are already at the beginning of a word, do nothing
3548                 if (!from || isWordSeparator(from - 1))
3549                         break;
3550                 // no break here, we go to the next
3551
3552         case PREVIOUS_WORD:
3553                 // always move the cursor to the beginning of previous word
3554                 while (from && !isWordSeparator(from - 1))
3555                         --from;
3556                 break;
3557         case NEXT_WORD:
3558                 LYXERR0("Paragraph::locateWord: NEXT_WORD not implemented yet");
3559                 break;
3560         case PARTIAL_WORD:
3561                 // no need to move the 'from' cursor
3562                 break;
3563         }
3564         to = from;
3565         while (to < size() && !isWordSeparator(to))
3566                 ++to;
3567 }
3568
3569
3570 void Paragraph::collectWords()
3571 {
3572         pos_type n = size();
3573         for (pos_type pos = 0; pos < n; ++pos) {
3574                 if (isWordSeparator(pos))
3575                         continue;
3576                 pos_type from = pos;
3577                 locateWord(from, pos, WHOLE_WORD);
3578                 if ((pos - from) >= (int)lyxrc.completion_minlength) {
3579                         docstring word = asString(from, pos, AS_STR_NONE);
3580                         FontList::const_iterator cit = d->fontlist_.fontIterator(pos);
3581                         if (cit == d->fontlist_.end())
3582                                 return;
3583                         Language const * lang = cit->font().language();
3584                         d->words_[*lang].insert(word);
3585                 }
3586         }
3587 }
3588
3589
3590 void Paragraph::registerWords()
3591 {
3592         Private::LangWordsMap::const_iterator itl = d->words_.begin();
3593         Private::LangWordsMap::const_iterator ite = d->words_.end();
3594         for (; itl != ite; ++itl) {
3595                 WordList * wl = theWordList(itl->first);
3596                 Private::Words::const_iterator it = (itl->second).begin();
3597                 Private::Words::const_iterator et = (itl->second).end();
3598                 for (; it != et; ++it)
3599                         wl->insert(*it);
3600         }
3601 }
3602
3603
3604 void Paragraph::updateWords()
3605 {
3606         deregisterWords();
3607         collectWords();
3608         registerWords();
3609 }
3610
3611
3612 void Paragraph::Private::appendSkipPosition(SkipPositions & skips, pos_type const pos) const
3613 {
3614         SkipPositionsIterator begin = skips.begin();
3615         SkipPositions::iterator end = skips.end();
3616         if (pos > 0 && begin < end) {
3617                 --end;
3618                 if (end->last == pos - 1) {
3619                         end->last = pos;
3620                         return;
3621                 }
3622         }
3623         skips.insert(end, FontSpan(pos, pos));
3624 }
3625
3626
3627 Language * Paragraph::Private::locateSpellRange(
3628         pos_type & from, pos_type & to,
3629         SkipPositions & skips) const
3630 {
3631         // skip leading white space
3632         while (from < to && owner_->isWordSeparator(from))
3633                 ++from;
3634         // don't check empty range
3635         if (from >= to)
3636                 return 0;
3637         // get current language
3638         Language * lang = getSpellLanguage(from);
3639         pos_type last = from;
3640         bool samelang = true;
3641         bool sameinset = true;
3642         while (last < to && samelang && sameinset) {
3643                 // hop to end of word
3644                 while (last < to && !owner_->isWordSeparator(last)) {
3645                         if (owner_->getInset(last)) {
3646                                 appendSkipPosition(skips, last);
3647                         } else if (owner_->isDeleted(last)) {
3648                                 appendSkipPosition(skips, last);
3649                         }
3650                         ++last;
3651                 }
3652                 // hop to next word while checking for insets
3653                 while (sameinset && last < to && owner_->isWordSeparator(last)) {
3654                         if (Inset const * inset = owner_->getInset(last))
3655                                 sameinset = inset->isChar() && inset->isLetter();
3656                         if (sameinset && owner_->isDeleted(last)) {
3657                                 appendSkipPosition(skips, last);
3658                         }
3659                         if (sameinset)
3660                                 last++;
3661                 }
3662                 if (sameinset && last < to) {
3663                         // now check for language change
3664                         samelang = lang == getSpellLanguage(last);
3665                 }
3666         }
3667         // if language change detected backstep is needed
3668         if (!samelang)
3669                 --last;
3670         to = last;
3671         return lang;
3672 }
3673
3674
3675 Language * Paragraph::Private::getSpellLanguage(pos_type const from) const
3676 {
3677         Language * lang =
3678                 const_cast<Language *>(owner_->getFontSettings(
3679                         inset_owner_->buffer().params(), from).language());
3680         if (lang == inset_owner_->buffer().params().language
3681                 && !lyxrc.spellchecker_alt_lang.empty()) {
3682                 string lang_code;
3683                 string const lang_variety =
3684                         split(lyxrc.spellchecker_alt_lang, lang_code, '-');
3685                 lang->setCode(lang_code);
3686                 lang->setVariety(lang_variety);
3687         }
3688         return lang;
3689 }
3690
3691
3692 void Paragraph::requestSpellCheck(pos_type pos)
3693 {
3694         d->requestSpellCheck(pos);
3695 }
3696
3697
3698 bool Paragraph::needsSpellCheck() const
3699 {
3700         SpellChecker::ChangeNumber speller_change_number = 0;
3701         if (theSpellChecker())
3702                 speller_change_number = theSpellChecker()->changeNumber();
3703         if (speller_change_number > d->speller_state_.currentChangeNumber()) {
3704                 d->speller_state_.needsCompleteRefresh(speller_change_number);
3705         }
3706         return d->needsSpellCheck();
3707 }
3708
3709
3710 bool Paragraph::Private::ignoreWord(docstring const & word) const
3711 {
3712         // Ignore words with digits
3713         // FIXME: make this customizable
3714         // (note that some checkers ignore words with digits by default)
3715         docstring::const_iterator cit = word.begin();
3716         docstring::const_iterator const end = word.end();
3717         for (; cit != end; ++cit) {
3718                 if (isNumber((*cit)))
3719                         return true;
3720         }
3721         return false;
3722 }
3723
3724
3725 SpellChecker::Result Paragraph::spellCheck(pos_type & from, pos_type & to,
3726         WordLangTuple & wl, docstring_list & suggestions,
3727         bool do_suggestion, bool check_learned) const
3728 {
3729         SpellChecker::Result result = SpellChecker::WORD_OK;
3730         SpellChecker * speller = theSpellChecker();
3731         if (!speller)
3732                 return result;
3733
3734         if (!d->layout_->spellcheck || !inInset().allowSpellCheck())
3735                 return result;
3736
3737         locateWord(from, to, WHOLE_WORD);
3738         if (from == to || from >= size())
3739                 return result;
3740
3741         docstring word = asString(from, to, AS_STR_INSETS | AS_STR_SKIPDELETE);
3742         Language * lang = d->getSpellLanguage(from);
3743
3744         wl = WordLangTuple(word, lang);
3745
3746         if (word.empty())
3747                 return result;
3748
3749         if (needsSpellCheck() || check_learned) {
3750                 pos_type end = to;
3751                 if (!d->ignoreWord(word)) {
3752                         bool const trailing_dot = to < size() && d->text_[to] == '.';
3753                         result = speller->check(wl);
3754                         if (SpellChecker::misspelled(result) && trailing_dot) {
3755                                 wl = WordLangTuple(word.append(from_ascii(".")), lang);
3756                                 result = speller->check(wl);
3757                                 if (!SpellChecker::misspelled(result)) {
3758                                         LYXERR(Debug::GUI, "misspelled word is correct with dot: \"" <<
3759                                            word << "\" [" <<
3760                                            from << ".." << to << "]");
3761                                 } else {
3762                                         // spell check with dot appended failed too
3763                                         // restore original word/lang value
3764                                         word = asString(from, to, AS_STR_INSETS | AS_STR_SKIPDELETE);
3765                                         wl = WordLangTuple(word, lang);
3766                                 }
3767                         }
3768                 }
3769                 if (!SpellChecker::misspelled(result)) {
3770                         // area up to the begin of the next word is not misspelled
3771                         while (end < size() && isWordSeparator(end))
3772                                 ++end;
3773                 }
3774                 d->setMisspelled(from, end, result);
3775         } else {
3776                 result = d->speller_state_.getState(from);
3777         }
3778
3779         if (do_suggestion)
3780                 suggestions.clear();
3781
3782         if (SpellChecker::misspelled(result)) {
3783                 LYXERR(Debug::GUI, "misspelled word: \"" <<
3784                            word << "\" [" <<
3785                            from << ".." << to << "]");
3786                 if (do_suggestion)
3787                         speller->suggest(wl, suggestions);
3788         }
3789         return result;
3790 }
3791
3792
3793 void Paragraph::Private::markMisspelledWords(
3794         pos_type const & first, pos_type const & last,
3795         SpellChecker::Result result,
3796         docstring const & word,
3797         SkipPositions const & skips)
3798 {
3799         if (!SpellChecker::misspelled(result)) {
3800                 setMisspelled(first, last, SpellChecker::WORD_OK);
3801                 return;
3802         }
3803         int snext = first;
3804         SpellChecker * speller = theSpellChecker();
3805         // locate and enumerate the error positions
3806         int nerrors = speller->numMisspelledWords();
3807         int numskipped = 0;
3808         SkipPositionsIterator it = skips.begin();
3809         SkipPositionsIterator et = skips.end();
3810         for (int index = 0; index < nerrors; ++index) {
3811                 int wstart;
3812                 int wlen = 0;
3813                 speller->misspelledWord(index, wstart, wlen);
3814                 /// should not happen if speller supports range checks
3815                 if (!wlen) continue;
3816                 docstring const misspelled = word.substr(wstart, wlen);
3817                 wstart += first + numskipped;
3818                 if (snext < wstart) {
3819                         /// mark the range of correct spelling
3820                         numskipped += countSkips(it, et, wstart);
3821                         setMisspelled(snext,
3822                                 wstart - 1, SpellChecker::WORD_OK);
3823                 }
3824                 snext = wstart + wlen;
3825                 numskipped += countSkips(it, et, snext);
3826                 /// mark the range of misspelling
3827                 setMisspelled(wstart, snext, result);
3828                 LYXERR(Debug::GUI, "misspelled word: \"" <<
3829                            misspelled << "\" [" <<
3830                            wstart << ".." << (snext-1) << "]");
3831                 ++snext;
3832         }
3833         if (snext <= last) {
3834                 /// mark the range of correct spelling at end
3835                 setMisspelled(snext, last, SpellChecker::WORD_OK);
3836         }
3837 }
3838
3839
3840 void Paragraph::spellCheck() const
3841 {
3842         SpellChecker * speller = theSpellChecker();
3843         if (!speller || empty() ||!needsSpellCheck())
3844                 return;
3845         pos_type start;
3846         pos_type endpos;
3847         d->rangeOfSpellCheck(start, endpos);
3848         if (speller->canCheckParagraph()) {
3849                 // loop until we leave the range
3850                 for (pos_type first = start; first < endpos; ) {
3851                         pos_type last = endpos;
3852                         Private::SkipPositions skips;
3853                         Language * lang = d->locateSpellRange(first, last, skips);
3854                         if (first >= endpos)
3855                                 break;
3856                         // start the spell checker on the unit of meaning
3857                         docstring word = asString(first, last, AS_STR_INSETS + AS_STR_SKIPDELETE);
3858                         WordLangTuple wl = WordLangTuple(word, lang);
3859                         SpellChecker::Result result = word.size() ?
3860                                 speller->check(wl) : SpellChecker::WORD_OK;
3861                         d->markMisspelledWords(first, last, result, word, skips);
3862                         first = ++last;
3863                 }
3864         } else {
3865                 static docstring_list suggestions;
3866                 pos_type to = endpos;
3867                 while (start < endpos) {
3868                         WordLangTuple wl;
3869                         spellCheck(start, to, wl, suggestions, false);
3870                         start = to + 1;
3871                 }
3872         }
3873         d->readySpellCheck();
3874 }
3875
3876
3877 bool Paragraph::isMisspelled(pos_type pos, bool check_boundary) const
3878 {
3879         bool result = SpellChecker::misspelled(d->speller_state_.getState(pos));
3880         if (result || pos <= 0 || pos > size())
3881                 return result;
3882         if (check_boundary && (pos == size() || isWordSeparator(pos)))
3883                 result = SpellChecker::misspelled(d->speller_state_.getState(pos - 1));
3884         return result;
3885 }
3886
3887
3888 string Paragraph::magicLabel() const
3889 {
3890         stringstream ss;
3891         ss << "magicparlabel-" << id();
3892         return ss.str();
3893 }
3894
3895
3896 } // namespace lyx