src/lyxfind.cpp

   1 /**
   2  * \file lyxfind.cpp
   3  * This file is part of LyX, the document processor.
   4  * License details can be found in the file COPYING.
   5  *
   6  * \author Lars Gullik Bjønnes
   7  * \author John Levon
   8  * \author Jürgen Vigna
   9  * \author Alfredo Braunstein
  10  * \author Tommaso Cucinotta
  11  * \author Kornel Benko
  12  *
  13  * Full author contact details are available in file CREDITS.
  14  */
  15
  16 #include <config.h>
  17
  18 #include "lyxfind.h"
  19
  20 #include "Buffer.h"
  21 #include "BufferList.h"
  22 #include "BufferParams.h"
  23 #include "BufferView.h"
  24 #include "Changes.h"
  25 #include "Cursor.h"
  26 #include "CutAndPaste.h"
  27 #include "FuncRequest.h"
  28 #include "LyX.h"
  29 #include "output_latex.h"
  30 #include "OutputParams.h"
  31 #include "Paragraph.h"
  32 #include "Text.h"
  33 #include "Encoding.h"
  34 #include "Language.h"
  35
  36 #include "frontends/Application.h"
  37 #include "frontends/alert.h"
  38
  39 #include "mathed/InsetMath.h"
  40 #include "mathed/InsetMathHull.h"
  41 #include "mathed/MathData.h"
  42 #include "mathed/MathStream.h"
  43 #include "mathed/MathSupport.h"
  44
  45 #include "support/debug.h"
  46 #include "support/docstream.h"
  47 #include "support/FileName.h"
  48 #include "support/gettext.h"
  49 #include "support/lassert.h"
  50 #include "support/lstrings.h"
  51 #include "support/textutils.h"
  52
  53 #include <unordered_map>
  54 #include <regex>
  55
  56 //#define ResultsDebug
  57 #define USE_QT_FOR_SEARCH
  58 #if defined(USE_QT_FOR_SEARCH)
  59         #include <QtCore>       // sets QT_VERSION
  60         #if (QT_VERSION >= 0x050000)
  61                 #include <QRegularExpression>
  62                 #define QTSEARCH 1
  63         #else
  64                 #define QTSEARCH 0
  65         #endif
  66 #else
  67         #define QTSEARCH 0
  68 #endif
  69
  70 using namespace std;
  71 using namespace lyx::support;
  72
  73 namespace lyx {
  74
  75 typedef unordered_map<string, string> AccentsMap;
  76 typedef unordered_map<string,string>::const_iterator AccentsIterator;
  77 static AccentsMap accents = unordered_map<string, string>();
  78
  79 // Helper class for deciding what should be ignored
  80 class IgnoreFormats {
  81  public:
  82         ///
  83         IgnoreFormats() = default;
  84         ///
  85         bool getFamily() const { return ignoreFamily_; }
  86         ///
  87         bool getSeries() const { return ignoreSeries_; }
  88         ///
  89         bool getShape() const { return ignoreShape_; }
  90         ///
  91         bool getSize() const { return ignoreSize_; }
  92         ///
  93         bool getUnderline() const { return ignoreUnderline_; }
  94         ///
  95         bool getMarkUp() const { return ignoreMarkUp_; }
  96         ///
  97         bool getStrikeOut() const { return ignoreStrikeOut_; }
  98         ///
  99         bool getSectioning() const { return ignoreSectioning_; }
 100         ///
 101         bool getFrontMatter() const { return ignoreFrontMatter_; }
 102         ///
 103         bool getColor() const { return ignoreColor_; }
 104         ///
 105         bool getLanguage() const { return ignoreLanguage_; }
 106         ///
 107         bool getDeleted() const { return ignoreDeleted_; }
 108         ///
 109         void setIgnoreDeleted(bool value);
 110         ///
 111         bool getNonContent() const { return searchNonContent_; }
 112         ///
 113         void setIgnoreFormat(string const & type, bool value, bool fromUser = true);
 114
 115 private:
 116         ///
 117         bool ignoreFamily_ = false;
 118         ///
 119         bool ignoreSeries_ = false;
 120         ///
 121         bool ignoreShape_ = false;
 122         ///
 123         bool ignoreSize_ = true;
 124         ///
 125         bool ignoreUnderline_ = false;
 126         ///
 127         bool ignoreMarkUp_ = false;
 128         ///
 129         bool ignoreStrikeOut_ = false;
 130         ///
 131         bool ignoreSectioning_ = false;
 132         ///
 133         bool ignoreFrontMatter_ = false;
 134         ///
 135         bool ignoreColor_ = false;
 136         ///
 137         bool ignoreLanguage_ = false;
 138         bool userSelectedIgnoreLanguage_ = false;
 139         ///
 140         bool ignoreDeleted_ = true;
 141         ///
 142         bool searchNonContent_ = true;
 143 };
 144
 145 void IgnoreFormats::setIgnoreFormat(string const & type, bool value, bool fromUser)
 146 {
 147         if (type == "color") {
 148                 ignoreColor_ = value;
 149         }
 150         else if (type == "language") {
 151                 if (fromUser) {
 152                         userSelectedIgnoreLanguage_ = value;
 153                         ignoreLanguage_ = value;
 154                 }
 155                 else
 156                         ignoreLanguage_ = (value || userSelectedIgnoreLanguage_);
 157         }
 158         else if (type == "sectioning") {
 159                 ignoreSectioning_ = value;
 160                 ignoreFrontMatter_ = value;
 161         }
 162         else if (type == "font") {
 163                 ignoreSeries_ = value;
 164                 ignoreShape_ = value;
 165                 ignoreFamily_ = value;
 166         }
 167         else if (type == "series") {
 168                 ignoreSeries_ = value;
 169         }
 170         else if (type == "shape") {
 171                 ignoreShape_ = value;
 172         }
 173         else if (type == "size") {
 174                 ignoreSize_ = value;
 175         }
 176         else if (type == "family") {
 177                 ignoreFamily_ = value;
 178         }
 179         else if (type == "markup") {
 180                 ignoreMarkUp_ = value;
 181         }
 182         else if (type == "underline") {
 183                 ignoreUnderline_ = value;
 184         }
 185         else if (type == "strike") {
 186                 ignoreStrikeOut_ = value;
 187         }
 188         else if (type == "deleted") {
 189                 ignoreDeleted_ = value;
 190         }
 191         else if (type == "non-output-content") {
 192                 searchNonContent_ = !value;
 193         }
 194 }
 195
 196 // The global variable that can be changed from outside
 197 IgnoreFormats ignoreFormats;
 198
 199
 200 void setIgnoreFormat(string const & type, bool value, bool fromUser)
 201 {
 202         ignoreFormats.setIgnoreFormat(type, value, fromUser);
 203 }
 204
 205
 206 namespace {
 207
 208 bool parse_bool(docstring & howto, bool const defvalue = false)
 209 {
 210         if (howto.empty())
 211                 return defvalue;
 212         docstring var;
 213         howto = split(howto, var, ' ');
 214         return var == "1";
 215 }
 216
 217
 218 class MatchString
 219 {
 220 public:
 221         MatchString(docstring const & s, bool cs, bool mw)
 222                 : str(s), case_sens(cs), whole_words(mw)
 223         {}
 224
 225         // returns true if the specified string is at the specified position
 226         // del specifies whether deleted strings in ct mode will be considered
 227         int operator()(Paragraph const & par, pos_type pos, bool del = true) const
 228         {
 229                 return par.find(str, case_sens, whole_words, pos, del);
 230         }
 231
 232 private:
 233         // search string
 234         docstring str;
 235         // case sensitive
 236         bool case_sens;
 237         // match whole words only
 238         bool whole_words;
 239 };
 240
 241
 242 int findForward(DocIterator & cur, DocIterator const endcur,
 243                 MatchString const & match,
 244                 bool find_del = true, bool onlysel = false)
 245 {
 246         for (; cur; cur.forwardChar()) {
 247                 if (onlysel && endcur.pit() == cur.pit()
 248                     && endcur.idx() == cur.idx() && endcur.pos() < cur.pos())
 249                         break;
 250                 if (cur.inTexted()) {
 251                         int len = match(cur.paragraph(), cur.pos(), find_del);
 252                         if (len > 0)
 253                                 return len;
 254                 }
 255         }
 256         return 0;
 257 }
 258
 259
 260 int findBackwards(DocIterator & cur, DocIterator const endcur,
 261                   MatchString const & match,
 262                   bool find_del = true, bool onlysel = false)
 263 {
 264         while (cur) {
 265                 cur.backwardChar();
 266                 if (onlysel && endcur.pit() == cur.pit()
 267                     && endcur.idx() == cur.idx() && endcur.pos() > cur.pos())
 268                         break;
 269                 if (cur.inTexted()) {
 270                         int len = match(cur.paragraph(), cur.pos(), find_del);
 271                         if (len > 0)
 272                                 return len;
 273                 }
 274         }
 275         return 0;
 276 }
 277
 278
 279 bool searchAllowed(docstring const & str)
 280 {
 281         if (str.empty()) {
 282                 frontend::Alert::error(_("Search error"), _("Search string is empty"));
 283                 return false;
 284         }
 285         return true;
 286 }
 287
 288 } // namespace
 289
 290
 291 bool findOne(BufferView * bv, docstring const & searchstr,
 292              bool case_sens, bool whole, bool forward,
 293              bool find_del, bool check_wrap, bool const auto_wrap,
 294              bool instant, bool onlysel)
 295 {
 296         // Clean up previous selections with empty searchstr on instant
 297         if (searchstr.empty() && instant) {
 298                 if (bv->cursor().selection()) {
 299                         bv->setCursor(bv->cursor().selectionBegin());
 300                         bv->clearSelection();
 301                 }
 302                 return true;
 303         }
 304
 305         if (!searchAllowed(searchstr))
 306                 return false;
 307
 308         DocIterator const endcur = forward ? bv->cursor().selectionEnd() : bv->cursor().selectionBegin();
 309
 310         if (onlysel && bv->cursor().selection()) {
 311                 docstring const matchstring = bv->cursor().selectionAsString(false);
 312                 docstring const lcmatchsting = support::lowercase(matchstring);
 313                 if (matchstring == searchstr || (!case_sens && lcmatchsting == lowercase(searchstr))) {
 314                         docstring q = _("The search string matches the selection, and search is limited to selection.\n"
 315                                         "Continue search outside?");
 316                         int search_answer = frontend::Alert::prompt(_("Search outside selection?"),
 317                                 q, 0, 1, _("&Yes"), _("&No"));
 318                         if (search_answer == 0) {
 319                                 bv->clearSelection();
 320                                 if (findOne(bv, searchstr, case_sens, whole, forward,
 321                                             find_del, check_wrap, auto_wrap, false, false))
 322                                         return true;
 323                         }
 324                         return false;
 325                 }
 326         }
 327
 328         DocIterator cur = forward
 329                 ? ((instant || onlysel) ? bv->cursor().selectionBegin() : bv->cursor().selectionEnd())
 330                 : ((instant || onlysel) ? bv->cursor().selectionEnd() : bv->cursor().selectionBegin());
 331
 332         MatchString const match(searchstr, case_sens, whole);
 333
 334         int match_len = forward
 335                 ? findForward(cur, endcur, match, find_del, onlysel)
 336                 : findBackwards(cur, endcur, match, find_del, onlysel);
 337
 338         if (match_len > 0)
 339                 bv->putSelectionAt(cur, match_len, !forward);
 340         else if (onlysel) {
 341                 docstring q = _("The search string was not found within the selection.\n"
 342                                 "Continue search outside?");
 343                 int search_answer = frontend::Alert::prompt(_("Search outside selection?"),
 344                         q, 0, 1, _("&Yes"), _("&No"));
 345                 if (search_answer == 0) {
 346                         bv->clearSelection();
 347                         if (findOne(bv, searchstr, case_sens, whole, forward,
 348                                     find_del, check_wrap, auto_wrap, false, false))
 349                                 return true;
 350                 }
 351                 return false;
 352         }
 353         else if (check_wrap) {
 354                 DocIterator cur_orig(bv->cursor());
 355                 bool wrap = auto_wrap;
 356                 if (!auto_wrap) {
 357                         docstring q;
 358                         if (forward)
 359                                 q = _("End of file reached while searching forward.\n"
 360                                   "Continue searching from the beginning?");
 361                         else
 362                                 q = _("Beginning of file reached while searching backward.\n"
 363                                   "Continue searching from the end?");
 364                         int wrap_answer = frontend::Alert::prompt(_("Wrap search?"),
 365                                 q, 0, 1, _("&Yes"), _("&No"));
 366                         wrap = wrap_answer == 0;
 367                 }
 368                 if (wrap) {
 369                         if (forward) {
 370                                 bv->cursor().clear();
 371                                 bv->cursor().push_back(CursorSlice(bv->buffer().inset()));
 372                         } else {
 373                                 bv->cursor().setCursor(doc_iterator_end(&bv->buffer()));
 374                                 bv->cursor().backwardPos();
 375                         }
 376                         if (auto_wrap) {
 377                                 docstring const msg = forward
 378                                   ? _("Search reached end of document, continuing from beginning.")
 379                                   : _("Search reached beginning of document, continuing from end.");
 380                                 bv->message(msg);
 381                         }
 382                         bv->clearSelection();
 383                         if (findOne(bv, searchstr, case_sens, whole, forward,
 384                                     find_del, false, false, false, false))
 385                                 return true;
 386                 }
 387                 bv->cursor().setCursor(cur_orig);
 388                 return false;
 389         }
 390
 391         return match_len > 0;
 392 }
 393
 394
 395 namespace {
 396
 397 int replaceAll(BufferView * bv,
 398                docstring const & searchstr, docstring const & replacestr,
 399                bool case_sens, bool whole, bool onlysel)
 400 {
 401         Buffer & buf = bv->buffer();
 402
 403         if (!searchAllowed(searchstr) || buf.isReadonly())
 404                 return 0;
 405
 406         DocIterator startcur = bv->cursor().selectionBegin();
 407         DocIterator endcur = bv->cursor().selectionEnd();
 408         bool const had_selection = bv->cursor().selection();
 409
 410         MatchString const match(searchstr, case_sens, whole);
 411         int num = 0;
 412
 413         int const rsize = replacestr.size();
 414         int const ssize = searchstr.size();
 415
 416         Cursor cur(*bv);
 417         cur.setCursor(doc_iterator_begin(&buf));
 418         int match_len = findForward(cur, endcur, match, false, onlysel);
 419         while (match_len > 0) {
 420                 // Backup current cursor position and font.
 421                 pos_type const pos = cur.pos();
 422                 Font const font = cur.paragraph().getFontSettings(buf.params(), pos);
 423                 cur.recordUndo();
 424                 int ct_deleted_text = ssize -
 425                         cur.paragraph().eraseChars(pos, pos + match_len,
 426                                                    buf.params().track_changes);
 427                 cur.paragraph().insert(pos, replacestr, font,
 428                                        Change(buf.params().track_changes
 429                                               ? Change::INSERTED
 430                                               : Change::UNCHANGED));
 431                 for (int i = 0; i < rsize + ct_deleted_text
 432                      && cur.pos() < cur.lastpos(); ++i)
 433                         cur.forwardPos();
 434                 if (onlysel && cur.pit() == endcur.pit() && cur.idx() == endcur.idx()) {
 435                         // Adjust end of selection for replace-all in selection
 436                         if (rsize > ssize) {
 437                                 int const offset = rsize - ssize;
 438                                 for (int i = 0; i < offset + ct_deleted_text
 439                                      && endcur.pos() < endcur.lastpos(); ++i)
 440                                         endcur.forwardPos();
 441                         } else {
 442                                 int const offset = ssize - rsize;
 443                                 for (int i = 0; i < offset && endcur.pos() > 0; ++i)
 444                                         endcur.backwardPos();
 445                                 for (int i = 0; i < ct_deleted_text
 446                                      && endcur.pos() < endcur.lastpos(); ++i)
 447                                         endcur.forwardPos();
 448                         }
 449                 }
 450                 ++num;
 451                 match_len = findForward(cur, endcur, match, false, onlysel);
 452         }
 453
 454         bv->putSelectionAt(doc_iterator_begin(&buf), 0, false);
 455
 456         startcur.fixIfBroken();
 457         bv->setCursor(startcur);
 458
 459         // Reset selection, accounting for changes in selection
 460         if (had_selection) {
 461                 endcur.fixIfBroken();
 462                 bv->cursor().resetAnchor();
 463                 bv->setCursorSelectionTo(endcur);
 464         }
 465
 466         return num;
 467 }
 468
 469
 470 // the idea here is that we are going to replace the string that
 471 // is selected IF it is the search string.
 472 // if there is a selection, but it is not the search string, then
 473 // we basically ignore it. (FIXME We ought to replace only within
 474 // the selection.)
 475 // if there is no selection, then:
 476 //  (i) if some search string has been provided, then we find it.
 477 //      (think of how the dialog works when you hit "replace" the
 478 //      first time.)
 479 // (ii) if no search string has been provided, then we treat the
 480 //      word the cursor is in as the search string. (why? i have no
 481 //      idea.) but this only works in text?
 482 //
 483 // returns the number of replacements made (one, if any) and
 484 // whether anything at all was done.
 485 pair<bool, int> replaceOne(BufferView * bv, docstring searchstr,
 486                            docstring const & replacestr, bool case_sens,
 487                            bool whole, bool forward, bool findnext, bool wrap,
 488                            bool onlysel)
 489 {
 490         Cursor & cur = bv->cursor();
 491         if (!cur.selection() || onlysel) {
 492                 // no selection, non-empty search string: find it
 493                 if (!searchstr.empty()) {
 494                         bool const found = findOne(bv, searchstr, case_sens, whole,
 495                                                    forward, true, findnext, wrap, false, onlysel);
 496                         return make_pair(found, 0);
 497                 }
 498                 // empty search string
 499                 if (!cur.inTexted())
 500                         // bail in math
 501                         return make_pair(false, 0);
 502                 // select current word and treat it as the search string.
 503                 // This causes a minor bug as undo will restore this selection,
 504                 // which the user did not create (#8986).
 505                 cur.innerText()->selectWord(cur, WHOLE_WORD);
 506                 searchstr = cur.selectionAsString(false, true);
 507         }
 508
 509         // if we still don't have a search string, report the error
 510         // and abort.
 511         if (!searchAllowed(searchstr))
 512                 return make_pair(false, 0);
 513
 514         bool have_selection = cur.selection();
 515         docstring const selected = cur.selectionAsString(false, true);
 516         bool match =
 517                 case_sens
 518                 ? searchstr == selected
 519                 : compare_no_case(searchstr, selected) == 0;
 520
 521         // no selection or current selection is not search word:
 522         // just find the search word
 523         if (!have_selection || !match) {
 524                 bool const found = findOne(bv, searchstr, case_sens, whole, forward,
 525                                            true, findnext, wrap, false, onlysel);
 526                 return make_pair(found, 0);
 527         }
 528
 529         // we're now actually ready to replace. if the buffer is
 530         // read-only, we can't, though.
 531         if (bv->buffer().isReadonly())
 532                 return make_pair(false, 0);
 533
 534         cap::replaceSelectionWithString(cur, replacestr);
 535         if (forward) {
 536                 cur.pos() += replacestr.length();
 537                 LASSERT(cur.pos() <= cur.lastpos(),
 538                         cur.pos() = cur.lastpos());
 539         }
 540         if (findnext)
 541                 findOne(bv, searchstr, case_sens, whole,
 542                         forward, false, findnext, wrap, false, onlysel);
 543
 544         return make_pair(true, 1);
 545 }
 546
 547 } // namespace
 548
 549
 550 docstring const find2string(docstring const & search,
 551                             bool casesensitive, bool matchword,
 552                             bool forward, bool wrap, bool instant,
 553                             bool onlysel)
 554 {
 555         odocstringstream ss;
 556         ss << search << '\n'
 557            << int(casesensitive) << ' '
 558            << int(matchword) << ' '
 559            << int(forward) << ' '
 560            << int(wrap) << ' '
 561            << int(instant) << ' '
 562            << int(onlysel);
 563         return ss.str();
 564 }
 565
 566
 567 docstring const replace2string(docstring const & replace,
 568                                docstring const & search,
 569                                bool casesensitive, bool matchword,
 570                                bool all, bool forward, bool findnext,
 571                                bool wrap, bool onlysel)
 572 {
 573         odocstringstream ss;
 574         ss << replace << '\n'
 575            << search << '\n'
 576            << int(casesensitive) << ' '
 577            << int(matchword) << ' '
 578            << int(all) << ' '
 579            << int(forward) << ' '
 580            << int(findnext) << ' '
 581            << int(wrap) << ' '
 582            << int(onlysel);
 583         return ss.str();
 584 }
 585
 586
 587 docstring const string2find(docstring const & argument,
 588                               bool &casesensitive,
 589                               bool &matchword,
 590                               bool &forward,
 591                               bool &wrap,
 592                               bool &instant,
 593                               bool &onlysel)
 594 {
 595         // data is of the form
 596         // "<search>
 597         //  <casesensitive> <matchword> <forward> <wrap> <onlysel>"
 598         docstring search;
 599         docstring howto = split(argument, search, '\n');
 600
 601         casesensitive = parse_bool(howto);
 602         matchword     = parse_bool(howto);
 603         forward       = parse_bool(howto, true);
 604         wrap          = parse_bool(howto);
 605         instant       = parse_bool(howto);
 606         onlysel       = parse_bool(howto);
 607
 608         return search;
 609 }
 610
 611
 612 bool lyxfind(BufferView * bv, FuncRequest const & ev)
 613 {
 614         if (!bv || ev.action() != LFUN_WORD_FIND)
 615                 return false;
 616
 617         //lyxerr << "find called, cmd: " << ev << endl;
 618         bool casesensitive;
 619         bool matchword;
 620         bool forward;
 621         bool wrap;
 622         bool instant;
 623         bool onlysel;
 624
 625         docstring search = string2find(ev.argument(), casesensitive,
 626                                        matchword, forward, wrap, instant, onlysel);
 627
 628         return findOne(bv, search, casesensitive, matchword, forward,
 629                        false, true, wrap, instant, onlysel);
 630 }
 631
 632
 633 bool lyxreplace(BufferView * bv, FuncRequest const & ev)
 634 {
 635         if (!bv || ev.action() != LFUN_WORD_REPLACE)
 636                 return false;
 637
 638         // data is of the form
 639         // "<search>
 640         //  <replace>
 641         //  <casesensitive> <matchword> <all> <forward> <findnext> <wrap> <onlysel>"
 642         docstring search;
 643         docstring rplc;
 644         docstring howto = split(ev.argument(), rplc, '\n');
 645         howto = split(howto, search, '\n');
 646
 647         bool casesensitive = parse_bool(howto);
 648         bool matchword     = parse_bool(howto);
 649         bool all           = parse_bool(howto);
 650         bool forward       = parse_bool(howto, true);
 651         bool findnext      = parse_bool(howto, true);
 652         bool wrap          = parse_bool(howto);
 653         bool onlysel       = parse_bool(howto);
 654
 655         if (!bv->cursor().selection())
 656                 // only selection only makes sense with selection
 657                 onlysel = false;
 658
 659         bool update = false;
 660
 661         int replace_count = 0;
 662         if (all) {
 663                 replace_count = replaceAll(bv, search, rplc, casesensitive,
 664                                            matchword, onlysel);
 665                 update = replace_count > 0;
 666         } else {
 667                 pair<bool, int> rv =
 668                         replaceOne(bv, search, rplc, casesensitive, matchword,
 669                                    forward, findnext, wrap, onlysel);
 670                 update = rv.first;
 671                 replace_count = rv.second;
 672         }
 673
 674         Buffer const & buf = bv->buffer();
 675         if (!update) {
 676                 // emit message signal.
 677                 if (onlysel)
 678                         buf.message(_("String not found in selection."));
 679                 else
 680                         buf.message(_("String not found."));
 681         } else {
 682                 if (replace_count == 0) {
 683                         buf.message(_("String found."));
 684                 } else if (replace_count == 1) {
 685                         buf.message(_("String has been replaced."));
 686                 } else {
 687                         docstring const str = onlysel
 688                                         ? bformat(_("%1$d strings have been replaced in the selection."), replace_count)
 689                                         : bformat(_("%1$d strings have been replaced."), replace_count);
 690                         buf.message(str);
 691                 }
 692         }
 693         return update;
 694 }
 695
 696
 697 bool findNextChange(BufferView * bv, Cursor & cur, bool const check_wrap)
 698 {
 699         for (; cur; cur.forwardPos())
 700                 if (cur.inTexted() && cur.paragraph().isChanged(cur.pos()))
 701                         return true;
 702
 703         if (check_wrap) {
 704                 DocIterator cur_orig(bv->cursor());
 705                 docstring q = _("End of file reached while searching forward.\n"
 706                           "Continue searching from the beginning?");
 707                 int wrap_answer = frontend::Alert::prompt(_("Wrap search?"),
 708                         q, 0, 1, _("&Yes"), _("&No"));
 709                 if (wrap_answer == 0) {
 710                         bv->cursor().clear();
 711                         bv->cursor().push_back(CursorSlice(bv->buffer().inset()));
 712                         bv->clearSelection();
 713                         cur.setCursor(bv->cursor().selectionBegin());
 714                         if (findNextChange(bv, cur, false))
 715                                 return true;
 716                 }
 717                 bv->cursor().setCursor(cur_orig);
 718         }
 719
 720         return false;
 721 }
 722
 723
 724 bool findPreviousChange(BufferView * bv, Cursor & cur, bool const check_wrap)
 725 {
 726         for (cur.backwardPos(); cur; cur.backwardPos()) {
 727                 if (cur.inTexted() && cur.paragraph().isChanged(cur.pos()))
 728                         return true;
 729         }
 730
 731         if (check_wrap) {
 732                 DocIterator cur_orig(bv->cursor());
 733                 docstring q = _("Beginning of file reached while searching backward.\n"
 734                           "Continue searching from the end?");
 735                 int wrap_answer = frontend::Alert::prompt(_("Wrap search?"),
 736                         q, 0, 1, _("&Yes"), _("&No"));
 737                 if (wrap_answer == 0) {
 738                         bv->cursor().setCursor(doc_iterator_end(&bv->buffer()));
 739                         bv->cursor().backwardPos();
 740                         bv->clearSelection();
 741                         cur.setCursor(bv->cursor().selectionBegin());
 742                         if (findPreviousChange(bv, cur, false))
 743                                 return true;
 744                 }
 745                 bv->cursor().setCursor(cur_orig);
 746         }
 747
 748         return false;
 749 }
 750
 751
 752 bool selectChange(Cursor & cur, bool forward)
 753 {
 754         if (!cur.inTexted() || !cur.paragraph().isChanged(cur.pos()))
 755                 return false;
 756         Change ch = cur.paragraph().lookupChange(cur.pos());
 757
 758         CursorSlice tip1 = cur.top();
 759         for (; tip1.pit() < tip1.lastpit() || tip1.pos() < tip1.lastpos(); tip1.forwardPos()) {
 760                 Change ch2 = tip1.paragraph().lookupChange(tip1.pos());
 761                 if (!ch2.isSimilarTo(ch))
 762                         break;
 763         }
 764         CursorSlice tip2 = cur.top();
 765         for (; tip2.pit() > 0 || tip2.pos() > 0;) {
 766                 tip2.backwardPos();
 767                 Change ch2 = tip2.paragraph().lookupChange(tip2.pos());
 768                 if (!ch2.isSimilarTo(ch)) {
 769                         // take a step forward to correctly set the selection
 770                         tip2.forwardPos();
 771                         break;
 772                 }
 773         }
 774         if (forward)
 775                 swap(tip1, tip2);
 776         cur.top() = tip1;
 777         cur.bv().mouseSetCursor(cur, false);
 778         cur.top() = tip2;
 779         cur.bv().mouseSetCursor(cur, true);
 780         return true;
 781 }
 782
 783
 784 namespace {
 785
 786
 787 bool findChange(BufferView * bv, bool forward)
 788 {
 789         Cursor cur(*bv);
 790         cur.setCursor(forward ? bv->cursor().selectionEnd()
 791                       : bv->cursor().selectionBegin());
 792         forward ? findNextChange(bv, cur, true) : findPreviousChange(bv, cur, true);
 793         return selectChange(cur, forward);
 794 }
 795
 796 } // namespace
 797
 798 bool findNextChange(BufferView * bv)
 799 {
 800         return findChange(bv, true);
 801 }
 802
 803
 804 bool findPreviousChange(BufferView * bv)
 805 {
 806         return findChange(bv, false);
 807 }
 808
 809
 810
 811 namespace {
 812
 813 typedef vector<pair<string, string> > Escapes;
 814
 815 static string getRegexSpaceCount(int count)
 816 {
 817         if (count > 0) {
 818                 if (count > 1)
 819                         return "\\s{" + std::to_string(count) + "}";
 820                 else
 821                         return "\\s";
 822         }
 823         return "";
 824 }
 825
 826 string string2regex(string in)
 827 {
 828         static std::regex specialChars { R"([-[\]{}()*+?.,\^$|#\$\\])" };
 829         string tempx = std::regex_replace(in, specialChars,  R"(\$&)" );
 830         // Special handling for ' '
 831         string temp("");
 832         int blanks = 0;
 833         for (unsigned i = 0; i < tempx.size(); i++) {
 834                 if (tempx[i] == ' ' || tempx[i] == '~' ) {
 835                         // normal blanks
 836                         blanks++;
 837                 }
 838                 else if ((tempx[i] == '\302' && tempx[i+1] == '\240')
 839                         || (tempx[i] == '\342' && tempx[i+1] == '\200')) {
 840                         // protected space
 841                         // thin space
 842                         blanks++;
 843                         i++;
 844                 }
 845                 else {
 846                         if (blanks > 0) {
 847                                 temp += getRegexSpaceCount(blanks);
 848                         }
 849                         temp += tempx[i];
 850                         blanks = 0;
 851                 }
 852         }
 853         if (blanks > 0) {
 854                 temp += getRegexSpaceCount(blanks);
 855         }
 856
 857         string temp2("");
 858         size_t lastpos = 0;
 859         size_t fl_pos = 0;
 860         int offset = 1;
 861         while (fl_pos < temp.size()) {
 862                 fl_pos = temp.find("\\\\foreignlanguage", lastpos + offset);
 863                 if (fl_pos == string::npos)
 864                         break;
 865                 offset = 16;
 866                 temp2 += temp.substr(lastpos, fl_pos - lastpos);
 867                 temp2 += "\\n";
 868                 lastpos = fl_pos;
 869         }
 870         if (lastpos == 0)
 871                 return(temp);
 872         if (lastpos < temp.size()) {
 873                 temp2 += temp.substr(lastpos, temp.size() - lastpos);
 874         }
 875         return temp2;
 876 }
 877
 878 static void buildAccentsMap();
 879
 880 string correctRegex(string t, bool withformat)
 881 {
 882         /* Convert \backslash => \
 883          * and \{, \}, \[, \] => {, }, [, ]
 884          */
 885         string s("");
 886         static std::regex wordre("(\\\\)*(\\\\(( |[A-Za-z]+|[\\{\\}%])( |\\{\\})?|[\\[\\]\\{\\}]))");
 887         static std::regex protectedSpace { R"(~)" };
 888         size_t lastpos = 0;
 889         smatch sub;
 890         bool backslashed = false;
 891         if (accents.empty())
 892                 buildAccentsMap();
 893
 894         //LYXERR0("correctRegex input '" << t << "'");
 895         int skip = 0;
 896         for (sregex_iterator it(t.begin(), t.end(), wordre), end; it != end; ++it) {
 897                 sub = *it;
 898                 string replace;
 899                 if ((sub.position(2) - sub.position(0)) % 2 == 1) {
 900                         continue;
 901                 }
 902                 else {
 903                         if (sub.str(4) == "backslash") {
 904                                 replace = "\\";
 905                                 {
 906                                         // transforms '\backslash \{' into '\{'
 907                                         string next = t.substr(sub.position(2) + sub.str(2).length(), 2);
 908                                         if ((next == "\\{") || (next == "\\}") || (next == "\\ ")) {
 909                                                 replace = "";
 910                                                 backslashed = true;
 911                                         }
 912                                         else if (withformat && next[0] == '$') {
 913                                                 replace = accents["lyxdollar"];
 914                                                 skip = 1;       // Skip following '$'
 915                                         }
 916                                 }
 917                         }
 918                         else if (sub.str(4) == "mathcircumflex")
 919                                 replace = "^";
 920                         else if (backslashed) {
 921                                 backslashed = false;
 922                                 if (withformat) {
 923                                         if (sub.str(3) == "{")
 924                                                 replace = accents["braceleft"];
 925                                         else if (sub.str(3) == "}")
 926                                                 replace = accents["braceright"];
 927                                         else if (sub.str(3) == " ")
 928                                                 replace = "\\ ";
 929                                         else {
 930                                                 // else part should not exist
 931                                                 LASSERT(0, /**/);
 932                                         }
 933                                 }
 934                                 else {
 935                                         if (sub.str(3) == "{")
 936                                                 replace = "\\{";
 937                                         else if (sub.str(3) == "}")
 938                                                 replace = "\\}";
 939                                         else if (sub.str(3) == " ")
 940                                                 replace = "\\ ";
 941                                         else {
 942                                                 // else part should not exist
 943                                                 LASSERT(0, /**/);
 944                                         }
 945                                 }
 946                         }
 947                         else if (sub.str(4) == "{") // transforms '\{' into '{'
 948                                 replace = "{";
 949                         else if (sub.str(4) == "}")
 950                                 replace = "}";
 951                         else if (sub.str(4) == "%")
 952                                 replace = "%";
 953                         else if (sub.str(4) == " ")
 954                                 replace = " ";
 955                         else {
 956                                 AccentsIterator it_ac = accents.find(sub.str(4));
 957                                 if (it_ac == accents.end()) {
 958                                         replace = sub.str(2);
 959                                 }
 960                                 else {
 961                                         replace = it_ac->second;
 962                                 }
 963                         }
 964                 }
 965                 if (lastpos < (size_t) sub.position(2))
 966                         s += std::regex_replace(t.substr(lastpos, sub.position(2) - lastpos), protectedSpace, R"( )");
 967                 s += replace;
 968                 lastpos = sub.position(2) + sub.length(2) + skip;
 969                 skip = 0;
 970         }
 971         if (lastpos == 0)
 972                 s = std::regex_replace(t, protectedSpace, R"( )");
 973         else if (lastpos < t.length())
 974                 s += std::regex_replace(t.substr(lastpos, t.length() - lastpos), protectedSpace, R"( )");
 975         // Handle quotes in regex
 976         // substitute all '„', '“', '»', '«' with '"'
 977         // and all '‚', '‘', '›', '‹' with "\'"
 978         static std::regex plainquotes { R"(„|“|»|«)" };
 979         static std::regex innerquotes { R"(‚|‘|›|‹)" };
 980         t = std::regex_replace(s, plainquotes, R"(")");
 981         s = std::regex_replace(t, innerquotes, R"(')");
 982         //LYXERR0("correctRegex output '" << s << "'");
 983         return s;
 984 }
 985
 986 /// Within \regexp{} apply get_lyx_unescapes() only (i.e., preserve regexp semantics of the string),
 987 /// while outside apply get_lyx_unescapes()+get_regexp_escapes().
 988 /// If match_latex is true, then apply regexp_latex_escapes() to \regexp{} contents as well.
 989 string escape_for_regex(string s, bool withformat)
 990 {
 991         size_t lastpos = 0;
 992         string result = "";
 993         while (lastpos < s.size()) {
 994                 size_t regex_pos = s.find("\\regexp{", lastpos);
 995                 if (regex_pos == string::npos) {
 996                         regex_pos = s.size();
 997                 }
 998                 if (regex_pos > lastpos) {
 999                         result += string2regex(s.substr(lastpos, regex_pos-lastpos));
1000                         lastpos = regex_pos;
1001                         if (lastpos == s.size())
1002                                 break;
1003                 }
1004                 size_t end_pos = s.find("\\endregexp", regex_pos + 8);
1005                 result += correctRegex(s.substr(regex_pos + 8, end_pos -(regex_pos + 8)), withformat);
1006                 lastpos = end_pos + 13;
1007         }
1008         return result;
1009 }
1010
1011
1012 /// Wrapper for lyx::regex_replace with simpler interface
1013 bool regex_replace(string const & s, string & t, string const & searchstr,
1014                    string const & replacestr)
1015 {
1016         regex e(searchstr, regex_constants::ECMAScript);
1017         ostringstream oss;
1018         ostream_iterator<char, char> it(oss);
1019         regex_replace(it, s.begin(), s.end(), e, replacestr);
1020         // tolerate t and s be references to the same variable
1021         bool rv = (s != oss.str());
1022         t = oss.str();
1023         return rv;
1024 }
1025
1026 class MatchResult {
1027 public:
1028         enum range {
1029                 newIsTooFar,
1030                 newIsBetter,
1031                 newIsInvalid
1032         };
1033         int match_len;
1034         int match_prefix;
1035         int match2end;
1036         int pos;
1037         int leadsize;
1038         int pos_len;
1039         int searched_size;
1040         vector <string> result = vector <string>();
1041         MatchResult(int len = 0): match_len(len),match_prefix(0),match2end(0), pos(0),leadsize(0),pos_len(-1),searched_size(0) {}
1042 };
1043
1044 static MatchResult::range interpretMatch(MatchResult &oldres, MatchResult &newres)
1045 {
1046         if (newres.match2end < oldres.match2end)
1047                 return MatchResult::newIsTooFar;
1048         if (newres.match_len < oldres.match_len)
1049                 return MatchResult::newIsTooFar;
1050
1051         if (newres.match_len == oldres.match_len) {
1052                 if (newres.match2end == oldres.match2end)
1053                         return MatchResult::newIsBetter;
1054         }
1055         return MatchResult::newIsInvalid;
1056 }
1057
1058 /** The class performing a match between a position in the document and the FindAdvOptions.
1059  **/
1060
1061 class MatchStringAdv {
1062 public:
1063         MatchStringAdv(lyx::Buffer & buf, FindAndReplaceOptions & opt);
1064
1065         /** Tests if text starting at the supplied position matches with the one provided to the MatchStringAdv
1066          ** constructor as opt.search, under the opt.* options settings.
1067          **
1068          ** @param at_begin
1069          **     If set to MatchStringAdv::MatchFromStart,
1070          **       then match is searched only against beginning of text starting at cur.
1071          **     Otherwise the match is searched anywhere in text starting at cur.
1072          **
1073          ** @return
1074          ** The length of the matching text, or zero if no match was found.
1075          **/
1076         enum matchType {
1077                 MatchAnyPlace,
1078                 MatchFromStart
1079         };
1080         string matchTypeAsString(matchType const x) const { return (x == MatchFromStart ? "MatchFromStart" : "MatchAnyPlace"); }
1081         MatchResult operator()(DocIterator const & cur, int len, matchType at_begin) const;
1082 #if QTSEARCH
1083         bool regexIsValid;
1084         string regexError;
1085 #endif
1086
1087 public:
1088         /// buffer
1089         lyx::Buffer * p_buf;
1090         /// first buffer on which search was started
1091         lyx::Buffer * const p_first_buf;
1092         /// options
1093         FindAndReplaceOptions const & opt;
1094
1095 private:
1096         /// Auxiliary find method (does not account for opt.matchword)
1097         MatchResult findAux(DocIterator const & cur, int len, matchType at_begin) const;
1098         void CreateRegexp(FindAndReplaceOptions const & opt, string regexp_str, string regexp2_str, string par_as_string = "");
1099
1100         /** Normalize a stringified or latexified LyX paragraph.
1101          **
1102          ** Normalize means:
1103          ** <ul>
1104          **   <li>if search is not casesensitive, then lowercase the string;
1105          **   <li>remove any newline at begin or end of the string;
1106          **   <li>replace any newline in the middle of the string with a simple space;
1107          **   <li>remove stale empty styles and environments, like \emph{} and \textbf{}.
1108          ** </ul>
1109          **
1110          ** @todo Normalization should also expand macros, if the corresponding
1111          ** search option was checked.
1112          **/
1113         string normalize(docstring const & s, bool ignore_fomat) const;
1114         // normalized string to search
1115         string par_as_string;
1116         // regular expression to use for searching
1117         // regexp2 is same as regexp, but prefixed with a ".*?"
1118 #if QTSEARCH
1119         QRegularExpression regexp;
1120         QRegularExpression regexp2;
1121 #else
1122         regex regexp;
1123         regex regexp2;
1124 #endif
1125         // leading format material as string
1126         string lead_as_string;
1127         // par_as_string after removal of lead_as_string
1128         string par_as_string_nolead;
1129         // unmatched open braces in the search string/regexp
1130         int open_braces;
1131         // number of (.*?) subexpressions added at end of search regexp for closing
1132         // environments, math mode, styles, etc...
1133         int close_wildcards;
1134 public:
1135         // Are we searching with regular expressions ?
1136         bool use_regexp = false;
1137         static int valid_matches;
1138         static vector <string> matches;
1139         void FillResults(MatchResult &found_mr);
1140 };
1141
1142 int MatchStringAdv::valid_matches = 0;
1143 vector <string> MatchStringAdv::matches = vector <string>(10);
1144
1145 void MatchStringAdv::FillResults(MatchResult &found_mr)
1146 {
1147         if (found_mr.match_len > 0) {
1148                 valid_matches = found_mr.result.size();
1149                 for (size_t i = 0; i < found_mr.result.size(); i++)
1150                         matches[i] = found_mr.result[i];
1151         } else
1152                 valid_matches = 0;
1153 }
1154
1155 static void setFindParams(OutputParams &runparams)
1156 {
1157         runparams.flavor = Flavor::XeTeX;
1158         //runparams.use_polyglossia = true;
1159         runparams.linelen = 10000; //lyxrc.plaintext_linelen;
1160         // No side effect of file copying and image conversion
1161         runparams.dryrun = true;
1162 }
1163
1164 static docstring buffer_to_latex(Buffer & buffer)
1165 {
1166         //OutputParams runparams(&buffer.params().encoding());
1167         OutputParams runparams(encodings.fromLyXName("utf8"));
1168         odocstringstream ods;
1169         otexstream os(ods);
1170         runparams.nice = true;
1171         setFindParams(runparams);
1172         if (ignoreFormats.getDeleted())
1173                 runparams.find_set_feature(OutputParams::SearchWithoutDeleted);
1174         else
1175                 runparams.find_set_feature(OutputParams::SearchWithDeleted);
1176         if (ignoreFormats.getNonContent()) {
1177                 runparams.find_add_feature(OutputParams::SearchNonOutput);
1178         }
1179         pit_type const endpit = buffer.paragraphs().size();
1180         for (pit_type pit = 0; pit != endpit; ++pit) {
1181                 TeXOnePar(buffer, buffer.text(), pit, os, runparams, string(), -1, -1, true);
1182                 LYXERR(Debug::FINDVERBOSE, "searchString up to here: " << ods.str());
1183         }
1184         return ods.str();
1185 }
1186
1187 static string latexNamesToUtf8(docstring strIn, bool withformat)
1188 {
1189         string addtmp = to_utf8(strIn);
1190         static regex const rmAcc("(\\\\)*("
1191                                          "\\\\([A-Za-z]+\\{.\\})"       // e.g. "ddot{A}" == sub.str(3)
1192                                         "|\\\\([A-Za-z]+)( |\\{\\})?"   // e.g. "LyX", "LyX{}", "LyX " == sub.str(4)
1193                                         ")"
1194                                 );
1195         size_t lastpos = 0;
1196         smatch sub;
1197         string replace;
1198         string add("");
1199         if (accents.empty())
1200                 buildAccentsMap();
1201         for (sregex_iterator it_add(addtmp.begin(), addtmp.end(), rmAcc), end; it_add != end; ++it_add) {
1202                 sub = *it_add;
1203                 if ((sub.position(2) - sub.position(0)) % 2 == 1) {
1204                         continue;
1205                 }
1206                 else {
1207                         string key;
1208                         if (sub.length(3) > 0)
1209                                 key = sub.str(3);
1210                         else
1211                                 key = sub.str(4);
1212                         AccentsIterator it_ac = accents.find(key);
1213                         if (it_ac == accents.end()) {
1214                                 replace = sub.str(2);
1215                         }
1216                         else {
1217                                 replace = it_ac->second;
1218                         }
1219                 }
1220                 if (lastpos < (size_t) sub.position(2))
1221                         add += addtmp.substr(lastpos, sub.position(2) - lastpos);
1222                 add += replace;
1223                 lastpos = sub.position(2) + sub.length(2);
1224         }
1225         if (lastpos == 0)
1226                 add = addtmp;
1227         else if (addtmp.length() > lastpos)
1228                 add += addtmp.substr(lastpos, addtmp.length() - lastpos);
1229         if (!withformat) {
1230                 static std::regex repltilde { R"(~)" };
1231                 add = std::regex_replace(add, repltilde, accents["lyxtilde"]);
1232         }
1233         LYXERR(Debug::FINDVERBOSE, "Adding to search string: '"
1234                         << add << "'");
1235         return add;
1236 }
1237
1238 static docstring stringifySearchBuffer(Buffer & buffer, FindAndReplaceOptions const & opt)
1239 {
1240         docstring str;
1241         if (!opt.ignoreformat) {
1242                 str = buffer_to_latex(buffer);
1243         } else {
1244                 // OutputParams runparams(&buffer.params().encoding());
1245                 OutputParams runparams(encodings.fromLyXName("utf8"));
1246                 runparams.nice = true;
1247                 setFindParams(runparams);
1248                 int option = AS_STR_INSETS |AS_STR_PLAINTEXT;
1249                 if (ignoreFormats.getDeleted()) {
1250                         option |= AS_STR_SKIPDELETE;
1251                         runparams.find_set_feature(OutputParams::SearchWithoutDeleted);
1252                 }
1253                 else {
1254                         runparams.find_set_feature(OutputParams::SearchWithDeleted);
1255                 }
1256                 if (ignoreFormats.getNonContent()) {
1257                         runparams.find_add_feature(OutputParams::SearchNonOutput);
1258                 }
1259                 string t("");
1260                 for (pos_type pit = pos_type(0); pit < (pos_type)buffer.paragraphs().size(); ++pit) {
1261                         Paragraph const & par = buffer.paragraphs().at(pit);
1262                         string add = latexNamesToUtf8(par.asString(pos_type(0), par.size(),
1263                                                                 option,
1264                                                                 &runparams), !opt.ignoreformat);
1265                         LYXERR(Debug::FINDVERBOSE, "Adding to search string: '"
1266                                 << add << "'");
1267                         t += add;
1268                 }
1269                 // Even in ignore-format we have to remove "\text{}, \lyxmathsym{}" parts
1270                 while (regex_replace(t, t, "\\\\(text|lyxmathsym|ensuremath)\\{([^\\}]*)\\}", "$2"));
1271                 str = from_utf8(t);
1272         }
1273         return str;
1274 }
1275
1276
1277 /// Return separation pos between the leading material and the rest
1278 static size_t identifyLeading(string const & s)
1279 {
1280         string t = s;
1281         // @TODO Support \item[text]
1282         // Kornel: Added textsl, textsf, textit, texttt and noun
1283         // + allow to search for colored text too
1284         while (regex_replace(t, t, "^\\\\(("
1285                              "(author|title|subtitle|subject|publishers|dedication|uppertitleback|lowertitleback|extratitle|"
1286                              "lyxaddress|lyxrightaddress|"
1287                              "footnotesize|tiny|scriptsize|small|large|Large|LARGE|huge|Huge|"
1288                              "emph|noun|minisec|text(bf|md|sl|sf|it|tt))|"
1289                              "((textcolor|foreignlanguage|latexenvironment)\\{[a-z]+\\*?\\})|"
1290                              "(u|uu)line|(s|x)out|uwave)|((sub)?(((sub)?section)|paragraph)|part|chapter)\\*?)\\{", "")
1291                || regex_replace(t, t, "^\\$", "")
1292                || regex_replace(t, t, "^\\\\\\[", "")
1293                || regex_replace(t, t, "^ ?\\\\item\\{[a-z]+\\}", "")
1294                || regex_replace(t, t, "^\\\\begin\\{[a-zA-Z_]*\\*?\\}", ""))
1295                ;
1296         LYXERR(Debug::FINDVERBOSE, "  after removing leading $, \\[ , \\emph{, \\textbf{, etc.: '" << t << "'");
1297         return s.size() - t.size();
1298 }
1299
1300 /*
1301  * Given a latexified string, retrieve some handled features
1302  * The features of the regex will later be compared with the features
1303  * of the searched text. If the regex features are not a
1304  * subset of the analized, then, in not format ignoring search
1305  * we can early stop the search in the relevant inset.
1306  */
1307 typedef map<string, bool> Features;
1308
1309 static Features identifyFeatures(string const & s)
1310 {
1311         static regex const feature("\\\\(([a-zA-Z]+(\\{([a-z]+\\*?)\\}|\\*)?))\\{");
1312         static regex const valid("^("
1313                 "("
1314                         "(footnotesize|tiny|scriptsize|small|large|Large|LARGE|huge|Huge|"
1315                                 "emph|noun|text(bf|md|sl|sf|it|tt)|"
1316                                 "(textcolor|foreignlanguage|item|listitem|latexenvironment)\\{[a-z]+\\*?\\})|"
1317                         "(u|uu)line|(s|x)out|uwave|"
1318                         "(sub|extra)?title|author|subject|publishers|dedication|(upper|lower)titleback|lyx(right)?address)|"
1319                 "((sub)?(((sub)?section)|paragraph)|part|chapter|lyxslide)\\*?)$");
1320         smatch sub;
1321         bool displ = true;
1322         Features info;
1323
1324         for (sregex_iterator it(s.begin(), s.end(), feature), end; it != end; ++it) {
1325                 sub = *it;
1326                 if (displ) {
1327                         if (sub.str(1).compare("regexp") == 0) {
1328                                 displ = false;
1329                                 continue;
1330                         }
1331                         string token = sub.str(1);
1332                         smatch sub2;
1333                         if (regex_match(token, sub2, valid)) {
1334                                 info[token] = true;
1335                         }
1336                         else {
1337                                 // ignore
1338                         }
1339                 }
1340                 else {
1341                         if (sub.str(1).compare("endregexp") == 0) {
1342                                 displ = true;
1343                                 continue;
1344                         }
1345                 }
1346         }
1347         return info;
1348 }
1349
1350 /*
1351  * defines values features of a key "\\[a-z]+{"
1352  */
1353 class KeyInfo {
1354 public:
1355         enum KeyType {
1356                 /* Char type with content discarded
1357                  * like \hspace{1cm} */
1358                 noContent,
1359                 /* Char, like \backslash */
1360                 isChar,
1361                 /* replace starting backslash with '#' */
1362                 isText,
1363                 /* \part, \section*, ... */
1364                 isSectioning,
1365                 /* title, author etc */
1366                 isTitle,
1367                 /* \foreignlanguage{ngerman}, ... */
1368                 isMain,
1369                 /* inside \code{}
1370                  * to discard language in content */
1371                 noMain,
1372                 isRegex,
1373                 /* \begin{eqnarray}...\end{eqnarray}, ... $...$ */
1374                 isMath,
1375                 /* fonts, colors, markups, ... */
1376                 isStandard,
1377                 /* footnotesize, ... large, ...
1378                  * Ignore all of them */
1379                 isSize,
1380                 invalid,
1381                 /* inputencoding, ...
1382                  * Discard also content, because they do not help in search */
1383                 doRemove,
1384                 /* twocolumns, ...
1385                  * like remove, but also all arguments */
1386                 removeWithArg,
1387                 /* item, listitem */
1388                 isList,
1389                 /* tex, latex, ... like isChar */
1390                 isIgnored,
1391                 /* like \lettrine[lines=5]{}{} */
1392                 cleanToStart,
1393                 // like isStandard, but always remove head
1394                 headRemove,
1395                 /* End of arguments marker for lettrine,
1396                  * so that they can be ignored */
1397                 endArguments
1398         };
1399         KeyInfo() = default;
1400         KeyInfo(KeyType type, int parcount, bool disable)
1401                 : keytype(type),
1402                   parenthesiscount(parcount),
1403                   disabled(disable) {}
1404         KeyType keytype = invalid;
1405         string head;
1406         int _tokensize = -1;
1407         int _tokenstart = -1;
1408         int _dataStart = -1;
1409         int _dataEnd = -1;
1410         int parenthesiscount = 1;
1411         bool disabled = false;
1412         bool used = false; /* by pattern */
1413 };
1414
1415 class Border {
1416 public:
1417         Border(int l=0, int u=0) : low(l), upper(u) {}
1418         int low;
1419         int upper;
1420 };
1421
1422 #define MAXOPENED 30
1423 class Intervall {
1424         bool isPatternString_;
1425 public:
1426         explicit Intervall(bool isPattern, string const & p)
1427                 : isPatternString_(isPattern), par(p), ignoreidx(-1),
1428                   actualdeptindex(0), hasTitle(false), langcount(0)
1429         {
1430                 depts[0] = 0;
1431                 closes[0] = 0;
1432         }
1433
1434         string par;
1435         int ignoreidx;
1436         static vector<Border> borders;
1437         static vector<int> depts;
1438         static vector<int> closes;
1439         int actualdeptindex;
1440         int previousNotIgnored(int) const;
1441         int nextNotIgnored(int) const;
1442         void handleOpenP(int i);
1443         void handleCloseP(int i, bool closingAllowed);
1444         void resetOpenedP(int openPos);
1445         void addIntervall(int upper);
1446         void addIntervall(int low, int upper); /* if explicit */
1447         void removeAccents();
1448         void setForDefaultLang(KeyInfo const & defLang) const;
1449         int findclosing(int start, int end, char up, char down, int repeat);
1450         void removeInvalidClosings(void);
1451         void handleParentheses(int lastpos, bool closingAllowed);
1452         bool hasTitle;
1453         // Number of disabled language specs up
1454         // to current position in actual interval
1455         int langcount;
1456         int isOpeningPar(int pos) const;
1457         string titleValue;
1458         void output(ostringstream &os, int lastpos);
1459         // string show(int lastpos);
1460 };
1461
1462 vector<Border> Intervall::borders = vector<Border>(30);
1463 vector<int> Intervall::depts = vector<int>(30);
1464 vector<int> Intervall::closes = vector<int>(30);
1465
1466 int Intervall::isOpeningPar(int pos) const
1467 {
1468         if ((pos < 0) || (size_t(pos) >= par.size()))
1469                 return 0;
1470         if (par[pos] != '{')
1471                 return 0;
1472         if (size_t(pos) + 2 >= par.size())
1473                 return 1;
1474         if (par[pos+2] != '}')
1475                 return 1;
1476         if (par[pos+1] == '[' || par[pos+1] == ']')
1477                 return 3;
1478         return 1;
1479 }
1480
1481 void Intervall::setForDefaultLang(KeyInfo const & defLang) const
1482 {
1483         // Enable the use of first token again
1484         if (ignoreidx >= 0) {
1485                 int value = defLang._tokenstart + defLang._tokensize;
1486                 int borderidx = 0;
1487                 if (hasTitle)
1488                         borderidx = 1;
1489                 if (value > 0) {
1490                         if (borders[borderidx].low < value)
1491                                 borders[borderidx].low = value;
1492                         if (borders[borderidx].upper < value)
1493                                 borders[borderidx].upper = value;
1494                 }
1495         }
1496 }
1497
1498 #if 0
1499 // Not needed, because dpts and closes are now dynamically expanded
1500 static void checkDepthIndex(int val)
1501 {
1502         static int maxdepthidx = MAXOPENED-2;
1503         static int lastmaxdepth = 0;
1504         if (val > lastmaxdepth) {
1505                 LYXERR(Debug::INFO, "Depth reached " << val);
1506                 lastmaxdepth = val;
1507         }
1508         if (val > maxdepthidx) {
1509                 maxdepthidx = val;
1510                 LYXERR(Debug::INFO, "maxdepthidx now " << val);
1511         }
1512 }
1513 #endif
1514
1515 #if 0
1516 // Not needed, because borders are now dynamically expanded
1517 static void checkIgnoreIdx(int val)
1518 {
1519         static int lastmaxignore = -1;
1520         if ((lastmaxignore < val) && (size_t(val+1) >= borders.size())) {
1521                 LYXERR(Debug::INFO, "IgnoreIdx reached " << val);
1522                 lastmaxignore = val;
1523         }
1524 }
1525 #endif
1526
1527 /*
1528  * Expand the region of ignored parts of the input latex string
1529  * The region is only relevant in output()
1530  */
1531 void Intervall::addIntervall(int low, int upper)
1532 {
1533         int idx;
1534         if (low == upper) return;
1535         for (idx = ignoreidx+1; idx > 0; --idx) {
1536                 if (low > borders[idx-1].upper) {
1537                         break;
1538                 }
1539         }
1540         Border br(low, upper);
1541         if (idx > ignoreidx) {
1542                 if (borders.size() <= size_t(idx)) {
1543                         borders.push_back(br);
1544                 }
1545                 else {
1546                         borders[idx] = br;
1547                 }
1548                 ignoreidx = idx;
1549                 // checkIgnoreIdx(ignoreidx);
1550                 return;
1551         }
1552         else {
1553                 // Expand only if one of the new bound is inside the interwall
1554                 // We know here that br.low > borders[idx-1].upper
1555                 if (br.upper < borders[idx].low) {
1556                         // We have to insert at this pos
1557                         if (size_t(ignoreidx+1) >= borders.size()) {
1558                                 borders.push_back(borders[ignoreidx]);
1559                         }
1560                         else {
1561                                 borders[ignoreidx+1] = borders[ignoreidx];
1562                         }
1563                         for (int i = ignoreidx; i > idx; --i) {
1564                                 borders[i] = borders[i-1];
1565                         }
1566                         borders[idx] = br;
1567                         ignoreidx += 1;
1568                         // checkIgnoreIdx(ignoreidx);
1569                         return;
1570                 }
1571                 // Here we know, that we are overlapping
1572                 if (br.low > borders[idx].low)
1573                         br.low = borders[idx].low;
1574                 // check what has to be concatenated
1575                 int count = 0;
1576                 for (int i = idx; i <= ignoreidx; i++) {
1577                         if (br.upper >= borders[i].low) {
1578                                 count++;
1579                                 if (br.upper < borders[i].upper)
1580                                         br.upper = borders[i].upper;
1581                         }
1582                         else {
1583                                 break;
1584                         }
1585                 }
1586                 // count should be >= 1 here
1587                 borders[idx] = br;
1588                 if (count > 1) {
1589                         for (int i = idx + count; i <= ignoreidx; i++) {
1590                                 borders[i-count+1] = borders[i];
1591                         }
1592                         ignoreidx -= count - 1;
1593                         return;
1594                 }
1595         }
1596 }
1597
1598 static void buildaccent(string n, string param, string values)
1599 {
1600         stringstream s(n);
1601         string name;
1602         const char delim = '|';
1603         while (getline(s, name, delim)) {
1604                 size_t start = 0;
1605                 for (char c : param) {
1606                         string key = name + "{" + c + "}";
1607                         // get the corresponding utf8-value
1608                         if ((values[start] & 0xc0) != 0xc0) {
1609                                 // should not happen, utf8 encoding starts at least with 11xxxxxx
1610                                 // but value for '\dot{i}' is 'i', which is ascii
1611                                 if ((values[start] & 0x80) == 0) {
1612                                         // is ascii
1613                                         accents[key] = values.substr(start, 1);
1614                                         // LYXERR(Debug::INFO, "" << key << "=" << accents[key]);
1615                                 }
1616                                 start++;
1617                                 continue;
1618                         }
1619                         for (int j = 1; ;j++) {
1620                                 if (start + j >= values.size()) {
1621                                         accents[key] = values.substr(start, j);
1622                                         start = values.size() - 1;
1623                                         break;
1624                                 }
1625                                 else if ((values[start+j] & 0xc0) != 0x80) {
1626                                         // This is the first byte of following utf8 char
1627                                         accents[key] = values.substr(start, j);
1628                                         start += j;
1629                                         // LYXERR(Debug::INFO, "" << key << "=" << accents[key]);
1630                                         break;
1631                                 }
1632                         }
1633                 }
1634         }
1635 }
1636
1637 // Helper function
1638 static string getutf8(unsigned uchar)
1639 {
1640 #define maxc 5
1641         string ret = string();
1642         char c[maxc] = {0};
1643         if (uchar <= 0x7f) {
1644                 c[maxc-1] = uchar & 0x7f;
1645         }
1646         else {
1647                 unsigned char rest = 0x40;
1648                 unsigned char first = 0x80;
1649                 int start = maxc-1;
1650                 for (int i = start; i >=0; --i) {
1651                         if (uchar < rest) {
1652                                 c[i] = first + uchar;
1653                                 break;
1654                         }
1655                         c[i] = 0x80 | (uchar &  0x3f);
1656                         uchar >>= 6;
1657                         rest >>= 1;
1658                         first >>= 1;
1659                         first |= 0x80;
1660                 }
1661         }
1662         for (int i = 0; i < maxc; i++) {
1663                 if (c[i] == 0) continue;
1664                 ret += c[i];
1665         }
1666         return(ret);
1667 }
1668
1669 static void addAccents(string latex_in, string unicode_out)
1670 {
1671         latex_in = latex_in.substr(1);
1672         AccentsIterator it_ac = accents.find(latex_in);
1673         if (it_ac == accents.end()) {
1674                 accents[latex_in] = unicode_out;
1675         }
1676         else {
1677                 LYXERR0("Key " << latex_in  << " already set");
1678         }
1679 }
1680
1681 void static fillMissingUnicodesymbols()
1682 {
1683         addAccents("\\pounds", getutf8(0x00a3));
1684         addAccents("\\textsterling", getutf8(0x00a3));
1685         addAccents("\\textyen", getutf8(0x00a5));
1686         addAccents("\\yen", getutf8(0x00a5));
1687         addAccents("\\textsection", getutf8(0x00a7));
1688         addAccents("\\mathsection", getutf8(0x00a7));
1689         addAccents("\\textcopyright", getutf8(0x00a9));
1690         addAccents("\\copyright", getutf8(0x00a9));
1691         addAccents("\\textlnot", getutf8(0x00ac));
1692         addAccents("\\neg", getutf8(0x00ac));
1693         addAccents("\\textregistered", getutf8(0x00ae));
1694         addAccents("\\circledR", getutf8(0x00ae));
1695         addAccents("\\textpm", getutf8(0x00b1));
1696         addAccents("\\pm", getutf8(0x00b1));
1697         addAccents("\\textparagraph", getutf8(0x00b6));
1698         addAccents("\\mathparagraph", getutf8(0x00b6));
1699         addAccents("\\textperiodcentered", getutf8(0x00b7));
1700         addAccents("\\texttimes", getutf8(0x00d7));
1701         addAccents("\\times", getutf8(0x00d7));
1702         addAccents("\\O", getutf8(0x00d8));
1703         addAccents("\\dh", getutf8(0x00f0));
1704         addAccents("\\eth", getutf8(0x00f0));
1705         addAccents("\\textdiv", getutf8(0x00f7));
1706         addAccents("\\div", getutf8(0x00f7));
1707         addAccents("\\o", getutf8(0x00f8));
1708         addAccents("\\textcrlambda", getutf8(0x019b));
1709         addAccents("\\j", getutf8(0x0237));
1710         addAccents("\\textrevepsilon", getutf8(0x025c));
1711         addAccents("\\textbaru", getutf8(0x0289));
1712         addAccents("\\textquoteleft", getutf8(0x02bb));
1713         addAccents("\\textGamma", getutf8(0x0393));
1714         addAccents("\\Gamma", getutf8(0x0393));
1715         addAccents("\\textDelta", getutf8(0x0394));
1716         addAccents("\\Delta", getutf8(0x0394));
1717         addAccents("\\textTheta", getutf8(0x0398));
1718         addAccents("\\Theta", getutf8(0x0398));
1719         addAccents("\\textLambda", getutf8(0x039b));
1720         addAccents("\\Lambda", getutf8(0x039b));
1721         addAccents("\\textXi", getutf8(0x039e));
1722         addAccents("\\Xi", getutf8(0x039e));
1723         addAccents("\\textPi", getutf8(0x03a0));
1724         addAccents("\\Pi", getutf8(0x03a0));
1725         addAccents("\\textSigma", getutf8(0x03a3));
1726         addAccents("\\Sigma", getutf8(0x03a3));
1727         addAccents("\\textUpsilon", getutf8(0x03a5));
1728         addAccents("\\Upsilon", getutf8(0x03a5));
1729         addAccents("\\textPhi", getutf8(0x03a6));
1730         addAccents("\\Phi", getutf8(0x03a6));
1731         addAccents("\\textPsi", getutf8(0x03a8));
1732         addAccents("\\Psi", getutf8(0x03a8));
1733         addAccents("\\textOmega", getutf8(0x03a9));
1734         addAccents("\\Omega", getutf8(0x03a9));
1735         addAccents("\\textalpha", getutf8(0x03b1));
1736         addAccents("\\alpha", getutf8(0x03b1));
1737         addAccents("\\textbeta", getutf8(0x03b2));
1738         addAccents("\\beta", getutf8(0x03b2));
1739         addAccents("\\textgamma", getutf8(0x03b3));
1740         addAccents("\\gamma", getutf8(0x03b3));
1741         addAccents("\\textdelta", getutf8(0x03b4));
1742         addAccents("\\delta", getutf8(0x03b4));
1743         addAccents("\\textepsilon", getutf8(0x03b5));
1744         addAccents("\\varepsilon", getutf8(0x03b5));
1745         addAccents("\\textzeta", getutf8(0x03b6));
1746         addAccents("\\zeta", getutf8(0x03b6));
1747         addAccents("\\texteta", getutf8(0x03b7));
1748         addAccents("\\eta", getutf8(0x03b7));
1749         addAccents("\\texttheta", getutf8(0x03b8));
1750         addAccents("\\theta", getutf8(0x03b8));
1751         addAccents("\\textiota", getutf8(0x03b9));
1752         addAccents("\\iota", getutf8(0x03b9));
1753         addAccents("\\textkappa", getutf8(0x03ba));
1754         addAccents("\\kappa", getutf8(0x03ba));
1755         addAccents("\\textlambda", getutf8(0x03bb));
1756         addAccents("\\lambda", getutf8(0x03bb));
1757         addAccents("\\textmu", getutf8(0x03bc));
1758         addAccents("\\mu", getutf8(0x03bc));
1759         addAccents("\\textnu", getutf8(0x03bd));
1760         addAccents("\\nu", getutf8(0x03bd));
1761         addAccents("\\textxi", getutf8(0x03be));
1762         addAccents("\\xi", getutf8(0x03be));
1763         addAccents("\\textpi", getutf8(0x03c0));
1764         addAccents("\\pi", getutf8(0x03c0));
1765         addAccents("\\textrho", getutf8(0x03c1));
1766         addAccents("\\rho", getutf8(0x03c1));
1767         addAccents("\\textfinalsigma", getutf8(0x03c2));
1768         addAccents("\\varsigma", getutf8(0x03c2));
1769         addAccents("\\textsigma", getutf8(0x03c3));
1770         addAccents("\\sigma", getutf8(0x03c3));
1771         addAccents("\\texttau", getutf8(0x03c4));
1772         addAccents("\\tau", getutf8(0x03c4));
1773         addAccents("\\textupsilon", getutf8(0x03c5));
1774         addAccents("\\upsilon", getutf8(0x03c5));
1775         addAccents("\\textphi", getutf8(0x03c6));
1776         addAccents("\\varphi", getutf8(0x03c6));
1777         addAccents("\\textchi", getutf8(0x03c7));
1778         addAccents("\\chi", getutf8(0x03c7));
1779         addAccents("\\textpsi", getutf8(0x03c8));
1780         addAccents("\\psi", getutf8(0x03c8));
1781         addAccents("\\textomega", getutf8(0x03c9));
1782         addAccents("\\omega", getutf8(0x03c9));
1783         addAccents("\\textdigamma", getutf8(0x03dd));
1784         addAccents("\\digamma", getutf8(0x03dd));
1785         addAccents("\\hebalef", getutf8(0x05d0));
1786         addAccents("\\aleph", getutf8(0x05d0));
1787         addAccents("\\hebbet", getutf8(0x05d1));
1788         addAccents("\\beth", getutf8(0x05d1));
1789         addAccents("\\hebgimel", getutf8(0x05d2));
1790         addAccents("\\gimel", getutf8(0x05d2));
1791         addAccents("\\hebdalet", getutf8(0x05d3));
1792         addAccents("\\daleth", getutf8(0x05d3));
1793         addAccents("\\hebhe", getutf8(0x05d4));
1794         addAccents("\\hebvav", getutf8(0x05d5));
1795         addAccents("\\hebzayin", getutf8(0x05d6));
1796         addAccents("\\hebhet", getutf8(0x05d7));
1797         addAccents("\\hebtet", getutf8(0x05d8));
1798         addAccents("\\hebyod", getutf8(0x05d9));
1799         addAccents("\\hebfinalkaf", getutf8(0x05da));
1800         addAccents("\\hebkaf", getutf8(0x05db));
1801         addAccents("\\heblamed", getutf8(0x05dc));
1802         addAccents("\\hebfinalmem", getutf8(0x05dd));
1803         addAccents("\\hebmem", getutf8(0x05de));
1804         addAccents("\\hebfinalnun", getutf8(0x05df));
1805         addAccents("\\hebnun", getutf8(0x05e0));
1806         addAccents("\\hebsamekh", getutf8(0x05e1));
1807         addAccents("\\hebayin", getutf8(0x05e2));
1808         addAccents("\\hebfinalpe", getutf8(0x05e3));
1809         addAccents("\\hebpe", getutf8(0x05e4));
1810         addAccents("\\hebfinaltsadi", getutf8(0x05e5));
1811         addAccents("\\hebtsadi", getutf8(0x05e6));
1812         addAccents("\\hebqof", getutf8(0x05e7));
1813         addAccents("\\hebresh", getutf8(0x05e8));
1814         addAccents("\\hebshin", getutf8(0x05e9));
1815         addAccents("\\hebtav", getutf8(0x05ea));
1816
1817         // Thai characters
1818         addAccents("\\thaiKoKai", getutf8(0x0e01));
1819         addAccents("\\thaiKhoKhai", getutf8(0x0e02));
1820         addAccents("\\thaiKhoKhuat", getutf8(0x0e03));
1821         addAccents("\\thaiKhoKhwai", getutf8(0x0e04));
1822         addAccents("\\thaiKhoKhon", getutf8(0x0e05));
1823         addAccents("\\thaiKhoRakhang", getutf8(0x0e06));
1824         addAccents("\\thaiNgoNgu", getutf8(0x0e07));
1825         addAccents("\\thaiChoChan", getutf8(0x0e08));
1826         addAccents("\\thaiChoChing", getutf8(0x0e09));
1827         addAccents("\\thaiChoChang", getutf8(0x0e0a));
1828         addAccents("\\thaiSoSo", getutf8(0x0e0b));
1829         addAccents("\\thaiChoChoe", getutf8(0x0e0c));
1830         addAccents("\\thaiYoYing", getutf8(0x0e0d));
1831         addAccents("\\thaiDoChada", getutf8(0x0e0e));
1832         addAccents("\\thaiToPatak", getutf8(0x0e0f));
1833         addAccents("\\thaiThoThan", getutf8(0x0e10));
1834         addAccents("\\thaiThoNangmontho", getutf8(0x0e11));
1835         addAccents("\\thaiThoPhuthao", getutf8(0x0e12));
1836         addAccents("\\thaiNoNen", getutf8(0x0e13));
1837         addAccents("\\thaiDoDek", getutf8(0x0e14));
1838         addAccents("\\thaiToTao", getutf8(0x0e15));
1839         addAccents("\\thaiThoThung", getutf8(0x0e16));
1840         addAccents("\\thaiThoThahan", getutf8(0x0e17));
1841         addAccents("\\thaiThoThong", getutf8(0x0e18));
1842         addAccents("\\thaiNoNu", getutf8(0x0e19));
1843         addAccents("\\thaiBoBaimai", getutf8(0x0e1a));
1844         addAccents("\\thaiPoPla", getutf8(0x0e1b));
1845         addAccents("\\thaiPhoPhung", getutf8(0x0e1c));
1846         addAccents("\\thaiFoFa", getutf8(0x0e1d));
1847         addAccents("\\thaiPhoPhan", getutf8(0x0e1e));
1848         addAccents("\\thaiFoFan", getutf8(0x0e1f));
1849         addAccents("\\thaiPhoSamphao", getutf8(0x0e20));
1850         addAccents("\\thaiMoMa", getutf8(0x0e21));
1851         addAccents("\\thaiYoYak", getutf8(0x0e22));
1852         addAccents("\\thaiRoRua", getutf8(0x0e23));
1853         addAccents("\\thaiRu", getutf8(0x0e24));
1854         addAccents("\\thaiLoLing", getutf8(0x0e25));
1855         addAccents("\\thaiLu", getutf8(0x0e26));
1856         addAccents("\\thaiWoWaen", getutf8(0x0e27));
1857         addAccents("\\thaiSoSala", getutf8(0x0e28));
1858         addAccents("\\thaiSoRusi", getutf8(0x0e29));
1859         addAccents("\\thaiSoSua", getutf8(0x0e2a));
1860         addAccents("\\thaiHoHip", getutf8(0x0e2b));
1861         addAccents("\\thaiLoChula", getutf8(0x0e2c));
1862         addAccents("\\thaiOAng", getutf8(0x0e2d));
1863         addAccents("\\thaiHoNokhuk", getutf8(0x0e2e));
1864         addAccents("\\thaiPaiyannoi", getutf8(0x0e2f));
1865         addAccents("\\thaiSaraA", getutf8(0x0e30));
1866         addAccents("\\thaiMaiHanakat", getutf8(0x0e31));
1867         addAccents("\\thaiSaraAa", getutf8(0x0e32));
1868         addAccents("\\thaiSaraAm", getutf8(0x0e33));
1869         addAccents("\\thaiSaraI", getutf8(0x0e34));
1870         addAccents("\\thaiSaraIi", getutf8(0x0e35));
1871         addAccents("\\thaiSaraUe", getutf8(0x0e36));
1872         addAccents("\\thaiSaraUee", getutf8(0x0e37));
1873         addAccents("\\thaiSaraU", getutf8(0x0e38));
1874         addAccents("\\thaiSaraUu", getutf8(0x0e39));
1875         addAccents("\\thaiPhinthu", getutf8(0x0e3a));
1876         addAccents("\\thaiSaraE", getutf8(0x0e40));
1877         addAccents("\\thaiSaraAe", getutf8(0x0e41));
1878         addAccents("\\thaiSaraO", getutf8(0x0e42));
1879         addAccents("\\thaiSaraAiMaimuan", getutf8(0x0e43));
1880         addAccents("\\thaiSaraAiMaimalai", getutf8(0x0e44));
1881         addAccents("\\thaiLakkhangyao", getutf8(0x0e45));
1882         addAccents("\\thaiMaiyamok", getutf8(0x0e46));
1883         addAccents("\\thaiMaitaikhu", getutf8(0x0e47));
1884         addAccents("\\thaiMaiEk", getutf8(0x0e48));
1885         addAccents("\\thaiMaiTho", getutf8(0x0e49));
1886         addAccents("\\thaiMaiTri", getutf8(0x0e4a));
1887         addAccents("\\thaiMaiChattawa", getutf8(0x0e4b));
1888         addAccents("\\thaiThanthakhat", getutf8(0x0e4c));
1889         addAccents("\\thaiNikhahit", getutf8(0x0e4d));
1890         addAccents("\\thaiYamakkan", getutf8(0x0e4e));
1891         addAccents("\\thaiFongman", getutf8(0x0e4f));
1892         addAccents("\\thaizero", getutf8(0x0e50));
1893         addAccents("\\thaione", getutf8(0x0e51));
1894         addAccents("\\thaitwo", getutf8(0x0e52));
1895         addAccents("\\thaithree", getutf8(0x0e53));
1896         addAccents("\\thaifour", getutf8(0x0e54));
1897         addAccents("\\thaifive", getutf8(0x0e55));
1898         addAccents("\\thaisix", getutf8(0x0e56));
1899         addAccents("\\thaiseven", getutf8(0x0e57));
1900         addAccents("\\thaieight", getutf8(0x0e58));
1901         addAccents("\\thainine", getutf8(0x0e59));
1902         addAccents("\\thaiAngkhankhu", getutf8(0x0e5a));
1903         addAccents("\\thaiKhomut", getutf8(0x0e5b));
1904         addAccents("\\dag", getutf8(0x2020));
1905         addAccents("\\dagger", getutf8(0x2020));
1906         addAccents("\\textdagger", getutf8(0x2020));
1907         addAccents("\\ddag", getutf8(0x2021));
1908         addAccents("\\ddagger", getutf8(0x2021));
1909         addAccents("\\textdaggerdbl", getutf8(0x2021));
1910         addAccents("\\textbullet", getutf8(0x2022));
1911         addAccents("\\bullet", getutf8(0x2022));
1912         addAccents("\\dots", getutf8(0x2026));
1913         addAccents("\\ldots", getutf8(0x2026));
1914         addAccents("\\textellipsis", getutf8(0x2026));
1915         addAccents("\\textasciiacute", getutf8(0x2032));
1916         addAccents("\\prime", getutf8(0x2032));
1917         addAccents("\\textacutedbl", getutf8(0x2033));
1918         addAccents("\\dprime", getutf8(0x2033));
1919         addAccents("\\textasciigrave", getutf8(0x2035));
1920         addAccents("\\backprime", getutf8(0x2035));
1921         addAccents("\\textsubcircum{ }", getutf8(0x2038));
1922         addAccents("\\caretinsert", getutf8(0x2038));
1923         addAccents("\\textasteriskcentered", getutf8(0x204e));
1924         addAccents("\\ast", getutf8(0x204e));
1925         addAccents("\\textmho", getutf8(0x2127));
1926         addAccents("\\mho", getutf8(0x2127));
1927         addAccents("\\textleftarrow", getutf8(0x2190));
1928         addAccents("\\leftarrow", getutf8(0x2190));
1929         addAccents("\\textuparrow", getutf8(0x2191));
1930         addAccents("\\uparrow", getutf8(0x2191));
1931         addAccents("\\textrightarrow", getutf8(0x2192));
1932         addAccents("\\rightarrow", getutf8(0x2192));
1933         addAccents("\\textdownarrow", getutf8(0x2193));
1934         addAccents("\\downarrow", getutf8(0x2193));
1935         addAccents("\\textglobrise", getutf8(0x2197));
1936         addAccents("\\nearrow", getutf8(0x2197));
1937         addAccents("\\textglobfall", getutf8(0x2198));
1938         addAccents("\\searrow", getutf8(0x2198));
1939         addAccents("\\textsurd", getutf8(0x221a));
1940         addAccents("\\surd", getutf8(0x221a));
1941         addAccents("\\textbigcircle", getutf8(0x25ef));
1942         addAccents("\\bigcirc", getutf8(0x25ef));
1943         addAccents("\\FiveStar", getutf8(0x2605));
1944         addAccents("\\bigstar", getutf8(0x2605));
1945         addAccents("\\FiveStarOpen", getutf8(0x2606));
1946         addAccents("\\bigwhitestar", getutf8(0x2606));
1947         addAccents("\\Checkmark", getutf8(0x2713));
1948         addAccents("\\checkmark", getutf8(0x2713));
1949         addAccents("\\CrossMaltese", getutf8(0x2720));
1950         addAccents("\\maltese", getutf8(0x2720));
1951         addAccents("\\textlangle", getutf8(0x27e8));
1952         addAccents("\\langle", getutf8(0x27e8));
1953         addAccents("\\textrangle", getutf8(0x27e9));
1954         addAccents("\\rangle", getutf8(0x27e9));
1955 }
1956
1957 static void buildAccentsMap()
1958 {
1959         accents["imath"] = "ı";
1960         accents["i"] = "ı";
1961         accents["jmath"] = "ȷ";
1962         accents["cdot"] = "·";
1963         accents["textasciicircum"] = "^";
1964         accents["mathcircumflex"] = "^";
1965         accents["guillemotright"] = "»";
1966         accents["guillemotleft"] = "«";
1967         accents["hairspace"]     = getutf8(0xf0000);    // select from free unicode plane 15
1968         accents["thinspace"]     = getutf8(0xf0002);    // and used _only_ by findadv
1969         accents["negthinspace"]  = getutf8(0xf0003);    // to omit backslashed latex macros
1970         accents["medspace"]      = getutf8(0xf0004);    // See https://en.wikipedia.org/wiki/Private_Use_Areas
1971         accents["negmedspace"]   = getutf8(0xf0005);
1972         accents["thickspace"]    = getutf8(0xf0006);
1973         accents["negthickspace"] = getutf8(0xf0007);
1974         accents["lyx"]           = getutf8(0xf0010);    // Used logos
1975         accents["LyX"]           = getutf8(0xf0010);
1976         accents["tex"]           = getutf8(0xf0011);
1977         accents["TeX"]           = getutf8(0xf0011);
1978         accents["latex"]         = getutf8(0xf0012);
1979         accents["LaTeX"]         = getutf8(0xf0012);
1980         accents["latexe"]        = getutf8(0xf0013);
1981         accents["LaTeXe"]        = getutf8(0xf0013);
1982         accents["lyxarrow"]      = getutf8(0xf0020);
1983         accents["braceleft"]     = getutf8(0xf0030);
1984         accents["braceright"]    = getutf8(0xf0031);
1985         accents["lyxtilde"]      = getutf8(0xf0032);
1986         accents["sim"]           = getutf8(0xf0032);
1987         accents["lyxdollar"]     = getutf8(0xf0033);
1988         accents["backslash lyx"]           = getutf8(0xf0010);  // Used logos inserted with starting \backslash
1989         accents["backslash LyX"]           = getutf8(0xf0010);
1990         accents["backslash tex"]           = getutf8(0xf0011);
1991         accents["backslash TeX"]           = getutf8(0xf0011);
1992         accents["backslash latex"]         = getutf8(0xf0012);
1993         accents["backslash LaTeX"]         = getutf8(0xf0012);
1994         accents["backslash latexe"]        = getutf8(0xf0013);
1995         accents["backslash LaTeXe"]        = getutf8(0xf0013);
1996         accents["backslash lyxarrow"]      = getutf8(0xf0020);
1997         accents["ddot{\\imath}"] = "ï";
1998         buildaccent("ddot", "aAeEhHiIoOtuUwWxXyY",
1999                     "äÄëËḧḦïÏöÖẗüÜẅẄẍẌÿŸ");   // umlaut
2000         buildaccent("dot|.", "aAbBcCdDeEfFGghHIimMnNoOpPrRsStTwWxXyYzZ",
2001                     "ȧȦḃḂċĊḋḊėĖḟḞĠġḣḢİİṁṀṅṄȯȮṗṖṙṘṡṠṫṪẇẆẋẊẏẎżŻ");      // dot{i} can only happen if ignoring case, but there is no lowercase of 'İ'
2002         accents["acute{\\imath}"] = "í";
2003         buildaccent("acute", "aAcCeEgGkKlLmMoOnNpPrRsSuUwWyYzZiI",
2004                     "áÁćĆéÉǵǴḱḰĺĹḿḾóÓńŃṕṔŕŔśŚúÚẃẂýÝźŹíÍ");
2005         buildaccent("dacute|H|h", "oOuU", "őŐűŰ");  // double acute
2006         buildaccent("mathring|r", "aAuUwy",
2007                     "åÅůŮẘẙ");  // ring
2008         accents["check{\\imath}"] = "ǐ";
2009         accents["check{\\jmath}"] = "ǰ";
2010         buildaccent("check|v", "cCdDaAeEiIoOuUgGkKhHlLnNrRsSTtzZ",
2011                     "čČďĎǎǍěĚǐǏǒǑǔǓǧǦǩǨȟȞľĽňŇřŘšŠŤťžŽ");        // caron
2012         accents["hat{\\imath}"] = "î";
2013         accents["hat{\\jmath}"] = "ĵ";
2014         buildaccent("hat|^", "aAcCeEgGhHiIjJoOsSuUwWyYzZ",
2015                     "âÂĉĈêÊĝĜĥĤîÎĵĴôÔŝŜûÛŵŴŷŶẑẐ");  // circ
2016         accents["bar{\\imath}"] = "ī";
2017         buildaccent("bar|=", "aAeEiIoOuUyY",
2018                     "āĀēĒīĪōŌūŪȳȲ");        // macron
2019         accents["tilde{\\imath}"] = "ĩ";
2020         buildaccent("tilde", "aAeEiInNoOuUvVyY",
2021                     "ãÃẽẼĩĨñÑõÕũŨṽṼỹỸ");  // tilde
2022         accents["breve{\\imath}"] = "ĭ";
2023         buildaccent("breve|u", "aAeEgGiIoOuU",
2024                     "ăĂĕĔğĞĭĬŏŎŭŬ");        // breve
2025         accents["grave{\\imath}"] = "ì";
2026         buildaccent("grave|`", "aAeEiIoOuUnNwWyY",
2027                     "àÀèÈìÌòÒùÙǹǸẁẀỳỲ");    // grave
2028         buildaccent("subdot|d", "BbDdHhKkLlMmNnRrSsTtVvWwZzAaEeIiOoUuYy",
2029                     "ḄḅḌḍḤḥḲḳḶḷṂṃṆṇṚṛṢṣṬṭṾṿẈẉẒẓẠạẸẹỊịỌọỤụỴỵ");      // dot below
2030         buildaccent("ogonek|k", "AaEeIiUuOo",
2031                     "ĄąĘęĮįŲųǪǫ");    // ogonek
2032         buildaccent("cedilla|c", "CcGgKkLlNnRrSsTtEeDdHh",
2033                     "ÇçĢģĶķĻļŅņŖŗŞşŢţȨȩḐḑḨḩ");        // cedilla
2034         buildaccent("subring|textsubring", "Aa",
2035                     "Ḁḁ");  // subring
2036         buildaccent("subhat|textsubcircum", "DdEeLlNnTtUu",
2037                     "ḒḓḘḙḼḽṊṋṰṱṶṷ");    // subcircum
2038         buildaccent("subtilde|textsubtilde", "EeIiUu",
2039                     "ḚḛḬḭṴṵ");      // subtilde
2040         accents["dgrave{\\imath}"] = "ȉ";
2041         accents["textdoublegrave{\\i}"] = "ȉ";
2042         buildaccent("dgrave|textdoublegrave", "AaEeIiOoRrUu",
2043                     "ȀȁȄȅȈȉȌȍȐȑȔȕ"); // double grave
2044         accents["rcap{\\imath}"] = "ȋ";
2045         accents["textroundcap{\\i}"] = "ȋ";
2046         buildaccent("rcap|textroundcap", "AaEeIiOoRrUu",
2047                     "ȂȃȆȇȊȋȎȏȒȓȖȗ"); // inverted breve
2048         buildaccent("slashed", "oO",
2049                     "øØ"); // slashed
2050         fillMissingUnicodesymbols(); // Add some still not handled entries contained in 'unicodesynbols'
2051         // LYXERR0("Number of accents " << accents.size());
2052 }
2053
2054 /*
2055  * Created accents in math or regexp environment
2056  * are macros, but we need the utf8 equivalent
2057  */
2058 void Intervall::removeAccents()
2059 {
2060         if (accents.empty())
2061                 buildAccentsMap();
2062         static regex const accre("\\\\("
2063                                  "([\\S]|[A-Za-z]+)\\{[^\\\\\\{\\}]+\\}"
2064                                  "|([\\S]|[A-Za-z]+)\\{\\\\[ij](math)?\\}"
2065                                  "|("
2066                                  "(backslash ([lL]y[xX]|[tT]e[xX]|[lL]a[tT]e[xX]e?|lyxarrow))"
2067                                  "|[A-Za-z]+"
2068                                  ")"
2069                                  "(?![a-zA-Z]))");
2070         smatch sub;
2071         for (sregex_iterator itacc(par.begin(), par.end(), accre), end; itacc != end; ++itacc) {
2072                 sub = *itacc;
2073                 string key = sub.str(1);
2074                 AccentsIterator it_ac = accents.find(key);
2075                 if (it_ac != accents.end()) {
2076                         string val = it_ac->second;
2077                         size_t pos = sub.position(size_t(0));
2078                         for (size_t i = 0; i < val.size(); i++) {
2079                                 par[pos+i] = val[i];
2080                         }
2081                         // Remove possibly following space too
2082                         if (par[pos+sub.str(0).size()] == ' ')
2083                                 addIntervall(pos+val.size(), pos + sub.str(0).size()+1);
2084                         else
2085                                 addIntervall(pos+val.size(), pos + sub.str(0).size());
2086                         for (size_t i = pos+val.size(); i < pos + sub.str(0).size(); i++) {
2087                                 // remove traces of any remaining chars
2088                                 par[i] = ' ';
2089                         }
2090                 }
2091                 else {
2092                         LYXERR(Debug::INFO, "Not added accent for \"" << key << "\"");
2093                 }
2094         }
2095 }
2096
2097 void Intervall::handleOpenP(int i)
2098 {
2099         actualdeptindex++;
2100         if ((size_t) actualdeptindex >= depts.size()) {
2101                 depts.resize(actualdeptindex + 30);
2102                 closes.resize(actualdeptindex + 30);
2103         }
2104         depts[actualdeptindex] = i+1;
2105         closes[actualdeptindex] = -1;
2106         // checkDepthIndex(actualdeptindex);
2107 }
2108
2109 void Intervall::handleCloseP(int i, bool closingAllowed)
2110 {
2111         if (actualdeptindex <= 0) {
2112                 if (! closingAllowed)
2113                         LYXERR(Debug::FINDVERBOSE, "Bad closing parenthesis in latex");  /* should not happen, but the latex input may be wrong */
2114                 // if we are at the very end
2115                 addIntervall(i, i+1);
2116         }
2117         else {
2118                 closes[actualdeptindex] = i+1;
2119                 actualdeptindex--;
2120         }
2121 }
2122
2123 void Intervall::resetOpenedP(int openPos)
2124 {
2125         // Used as initializer for foreignlanguage entry
2126         actualdeptindex = 1;
2127         depts[1] = openPos+1;
2128         closes[1] = -1;
2129 }
2130
2131 int Intervall::previousNotIgnored(int start) const
2132 {
2133         int idx = 0;                          /* int intervalls */
2134         for (idx = ignoreidx; idx >= 0; --idx) {
2135                 if (start > borders[idx].upper)
2136                         return start;
2137                 if (start >= borders[idx].low)
2138                         start = borders[idx].low-1;
2139         }
2140         return start;
2141 }
2142
2143 int Intervall::nextNotIgnored(int start) const
2144 {
2145         int idx = 0;                          /* int intervalls */
2146         for (idx = 0; idx <= ignoreidx; idx++) {
2147                 if (start < borders[idx].low)
2148                         return start;
2149                 if (start < borders[idx].upper)
2150                         start = borders[idx].upper;
2151         }
2152         return start;
2153 }
2154
2155 typedef unordered_map<string, KeyInfo> KeysMap;
2156 typedef unordered_map<string, KeyInfo>::const_iterator KeysIterator;
2157 typedef vector< KeyInfo> Entries;
2158 static KeysMap keys = unordered_map<string, KeyInfo>();
2159
2160 class LatexInfo {
2161 private:
2162         int entidx_;
2163         Entries entries_;
2164         Intervall interval_;
2165         void buildKeys(bool);
2166         void buildEntries(bool);
2167         void makeKey(const string &, KeyInfo, bool isPatternString);
2168         void processRegion(int start, int region_end); /*  remove {} parts */
2169         void removeHead(KeyInfo const &, int count=0);
2170
2171 public:
2172         LatexInfo(string const & par, bool isPatternString)
2173                 : entidx_(-1), interval_(isPatternString, par)
2174         {
2175                 buildKeys(isPatternString);
2176                 entries_ = vector<KeyInfo>();
2177                 buildEntries(isPatternString);
2178         }
2179         int getFirstKey() {
2180                 entidx_ = 0;
2181                 if (entries_.empty()) {
2182                         return -1;
2183                 }
2184                 if (entries_[0].keytype == KeyInfo::isTitle) {
2185                         interval_.hasTitle = true;
2186                         if (! entries_[0].disabled) {
2187                                 interval_.titleValue = entries_[0].head;
2188                         }
2189                         else {
2190                                 interval_.titleValue = "";
2191                         }
2192                         removeHead(entries_[0]);
2193                         if (entries_.size() > 1)
2194                                 return 1;
2195                         else
2196                                 return -1;
2197                 }
2198                 return 0;
2199         }
2200         int getNextKey() {
2201                 entidx_++;
2202                 if (int(entries_.size()) > entidx_) {
2203                         return entidx_;
2204                 }
2205                 else {
2206                         return -1;
2207                 }
2208         }
2209         bool setNextKey(int idx) {
2210                 if ((idx == entidx_) && (entidx_ >= 0)) {
2211                         entidx_--;
2212                         return true;
2213                 }
2214                 else
2215                         return false;
2216         }
2217         int find(int start, KeyInfo::KeyType keytype) const {
2218                 if (start < 0)
2219                         return -1;
2220                 int tmpIdx = start;
2221                 while (tmpIdx < int(entries_.size())) {
2222                         if (entries_[tmpIdx].keytype == keytype)
2223                                 return tmpIdx;
2224                         tmpIdx++;
2225                 }
2226                 return -1;
2227         }
2228         int process(ostringstream & os, KeyInfo const & actual);
2229         int dispatch(ostringstream & os, int previousStart, KeyInfo & actual);
2230         // string show(int lastpos) { return interval.show(lastpos);}
2231         int nextNotIgnored(int start) { return interval_.nextNotIgnored(start);}
2232         KeyInfo &getKeyInfo(int keyinfo) {
2233                 static KeyInfo invalidInfo = KeyInfo();
2234                 if ((keyinfo < 0) || ( keyinfo >= int(entries_.size())))
2235                         return invalidInfo;
2236                 else
2237                         return entries_[keyinfo];
2238         }
2239         void setForDefaultLang(KeyInfo const & defLang) {interval_.setForDefaultLang(defLang);}
2240         void addIntervall(int low, int up) { interval_.addIntervall(low, up); }
2241 };
2242
2243
2244 int Intervall::findclosing(int start, int end, char up = '{', char down = '}', int repeat = 1)
2245 {
2246         int skip = 0;
2247         int depth = 0;
2248         for (int i = start; i < end; i += 1 + skip) {
2249                 char c;
2250                 c = par[i];
2251                 skip = 0;
2252                 if (c == '\\') skip = 1;
2253                 else if (c == up) {
2254                         depth++;
2255                 }
2256                 else if (c == down) {
2257                         if (depth == 0) {
2258                                 repeat--;
2259                                 if ((repeat <= 0) || (par[i+1] != up))
2260                                         return i;
2261                         }
2262                         --depth;
2263                 }
2264         }
2265         return end;
2266 }
2267
2268 void Intervall::removeInvalidClosings(void)
2269 {
2270         // this can happen, if there are deleted parts
2271         int skip = 0;
2272         int depth = 0;
2273         for (unsigned i = 0; i < par.size(); i += 1 + skip) {
2274                 char c = par[i];
2275                 skip = 0;
2276                 if (c == '\\') skip = 1;
2277                 else if (c == '{')
2278                         depth++;
2279                 else if (c == '}') {
2280                         if (depth == 0) {
2281                                 addIntervall(i, i+1);
2282                                 LYXERR(Debug::FINDVERBOSE, "removed invalid closing '}' at " << i);
2283                         }
2284                         else
2285                                 --depth;
2286                 }
2287         }
2288 }
2289 class MathInfo {
2290         class MathEntry {
2291         public:
2292                 string wait;
2293                 size_t mathEnd;
2294                 size_t mathpostfixsize;
2295                 size_t mathStart;
2296                 size_t mathprefixsize;
2297                 size_t mathSize;
2298         };
2299         size_t actualIdx_;
2300         vector<MathEntry> entries_;
2301 public:
2302         MathInfo() {
2303                 actualIdx_ = 0;
2304         }
2305         void insert(string const & wait, size_t start, size_t prefixsize, size_t end, size_t postfixsize) {
2306                 MathEntry m = MathEntry();
2307                 m.wait = wait;
2308                 m.mathStart = start;
2309                 m.mathprefixsize = prefixsize;
2310                 m.mathEnd = end + postfixsize;
2311                 m.mathpostfixsize = postfixsize;
2312                 m.mathSize = m.mathEnd - m.mathStart;
2313                 entries_.push_back(m);
2314         }
2315         bool empty() const { return entries_.empty(); }
2316         size_t getEndPos() const {
2317                 if (entries_.empty() || (actualIdx_ >= entries_.size())) {
2318                         return 0;
2319                 }
2320                 return entries_[actualIdx_].mathEnd;
2321         }
2322         size_t getStartPos() const {
2323                 if (entries_.empty() || (actualIdx_ >= entries_.size())) {
2324                         return 100000;                    /*  definitely enough? */
2325                 }
2326                 return entries_[actualIdx_].mathStart;
2327         }
2328         size_t getPrefixSize() const {
2329                 if (entries_.empty() || (actualIdx_ >= entries_.size())) {
2330                         return 0;
2331                 }
2332                 return entries_[actualIdx_].mathprefixsize;
2333         }
2334         size_t getPostfixSize() const {
2335                 if (entries_.empty() || (actualIdx_ >= entries_.size())) {
2336                         return 0;
2337                 }
2338                 return entries_[actualIdx_].mathpostfixsize;
2339         }
2340         size_t getFirstPos() {
2341                 actualIdx_ = 0;
2342                 return getStartPos();
2343         }
2344         size_t getSize() const {
2345                 if (entries_.empty() || (actualIdx_ >= entries_.size())) {
2346                         return size_t(0);
2347                 }
2348                 return entries_[actualIdx_].mathSize;
2349         }
2350         void incrEntry() { actualIdx_++; }
2351 };
2352
2353 void LatexInfo::buildEntries(bool isPatternString)
2354 {
2355         static regex const rmath("(\\\\)*(\\$|\\\\\\[|\\\\\\]|\\\\(begin|end)\\{((eqnarray|equation|flalign|gather|multline|align|x?x?alignat)\\*?\\})(\\{[0-9]+\\})?)");
2356         static regex const rkeys("(\\\\)*(\\$|\\\\\\[|\\\\\\]|\\\\((([a-zA-Z]+\\*?)(\\{([a-z]+\\*?)\\}|=[0-9]+[a-z]+)?)))");
2357         static bool disableLanguageOverride = false;
2358         smatch sub, submath;
2359         bool evaluatingRegexp = false;
2360         MathInfo mi;
2361         bool evaluatingMath = false;
2362         bool evaluatingCode = false;
2363         size_t codeEnd = 0;
2364         bool evaluatingOptional = false;
2365         size_t optionalEnd = 0;
2366         int codeStart = -1;
2367         KeyInfo found;
2368         bool math_end_waiting = false;
2369         size_t math_pos = 10000;
2370         size_t math_prefix_size = 1;
2371         string math_end;
2372         static vector<string> usedText = vector<string>();
2373         static bool removeMathHull = false;
2374
2375         interval_.removeAccents();
2376         interval_.removeInvalidClosings();
2377
2378         for (sregex_iterator itmath(interval_.par.begin(), interval_.par.end(), rmath), end; itmath != end; ++itmath) {
2379                 submath = *itmath;
2380                 if ((submath.position(2) - submath.position(0)) %2 == 1) {
2381                         // prefixed by odd count of '\\'
2382                         continue;
2383                 }
2384                 if (math_end_waiting) {
2385                         size_t pos = submath.position(size_t(2));
2386                         if ((math_end == "$") &&
2387                                         (submath.str(2) == "$")) {
2388                                 mi.insert("$", math_pos, 1, pos, 1);
2389                                 math_end_waiting = false;
2390                         }
2391                         else if ((math_end == "\\]") &&
2392                                  (submath.str(2) == "\\]")) {
2393                                 mi.insert("\\]", math_pos, 2, pos, 2);
2394                                 math_end_waiting = false;
2395                         }
2396                         else if ((submath.str(3).compare("end") == 0) &&
2397                                  (submath.str(5).compare(math_end) == 0)) {
2398                                 mi.insert(math_end, math_pos, math_prefix_size, pos, submath.str(2).length());
2399                                 math_end_waiting = false;
2400                         }
2401                         else
2402                                 continue;
2403                 }
2404                 else {
2405                         if (submath.str(3).compare("begin") == 0) {
2406                                 math_end_waiting = true;
2407                                 math_end = submath.str(5);
2408                                 math_pos = submath.position(size_t(2));
2409                                 math_prefix_size = submath.str(2).length();
2410                         }
2411                         else if (submath.str(2).compare("\\[") == 0) {
2412                                 math_end_waiting = true;
2413                                 math_end = "\\]";
2414                                 math_pos = submath.position(size_t(2));
2415                         }
2416                         else if (submath.str(2) == "$") {
2417                                 size_t pos = submath.position(size_t(2));
2418                                 math_end_waiting = true;
2419                                 math_end = "$";
2420                                 math_pos = pos;
2421                         }
2422                 }
2423         }
2424         // Ignore language if there is math somewhere in pattern-string
2425         if (isPatternString) {
2426                 for (auto s: usedText) {
2427                         // Remove entries created in previous search runs
2428                         keys.erase(s);
2429                 }
2430                 usedText = vector<string>();
2431                 if (! mi.empty()) {
2432                         // Disable language
2433                         keys["foreignlanguage"].disabled = true;
2434                         disableLanguageOverride = true;
2435                         removeMathHull = false;
2436                 }
2437                 else {
2438                         removeMathHull = true;  // used later if not isPatternString
2439                         disableLanguageOverride = false;
2440                 }
2441         }
2442         else {
2443                 if (disableLanguageOverride) {
2444                         keys["foreignlanguage"].disabled = true;
2445                 }
2446         }
2447         math_pos = mi.getFirstPos();
2448         for (sregex_iterator it(interval_.par.begin(), interval_.par.end(), rkeys), end; it != end; ++it) {
2449                 sub = *it;
2450                 if ((sub.position(2) - sub.position(0)) %2 == 1) {
2451                         // prefixed by odd count of '\\'
2452                         continue;
2453                 }
2454                 string key = sub.str(5);
2455                 if (key == "") {
2456                         if (sub.str(2)[0] == '\\')
2457                                 key = sub.str(2)[1];
2458                         else {
2459                                 key = sub.str(2);
2460                         }
2461                 }
2462                 KeysIterator it_key = keys.find(key);
2463                 if (it_key != keys.end()) {
2464                         if (it_key->second.keytype == KeyInfo::headRemove) {
2465                                 KeyInfo found1 = it_key->second;
2466                                 found1.disabled = true;
2467                                 found1.head = "\\" + key + "{";
2468                                 found1._tokenstart = sub.position(size_t(2));
2469                                 found1._tokensize = found1.head.length();
2470                                 found1._dataStart = found1._tokenstart + found1.head.length();
2471                                 int endpos = interval_.findclosing(found1._dataStart, interval_.par.length(), '{', '}', 1);
2472                                 found1._dataEnd = endpos;
2473                                 removeHead(found1);
2474                                 continue;
2475                         }
2476                 }
2477                 if (evaluatingRegexp) {
2478                         if (sub.str(3).compare("endregexp") == 0) {
2479                                 evaluatingRegexp = false;
2480                                 // found._tokenstart already set
2481                                 found._dataEnd = sub.position(size_t(2)) + 13;
2482                                 found._dataStart = found._dataEnd;
2483                                 found._tokensize = found._dataEnd - found._tokenstart;
2484                                 found.parenthesiscount = 0;
2485                                 found.head = interval_.par.substr(found._tokenstart, found._tokensize);
2486                         }
2487                         else {
2488                                 continue;
2489                         }
2490                 }
2491                 else {
2492                         if (evaluatingMath) {
2493                                 if (size_t(sub.position(size_t(2))) < mi.getEndPos())
2494                                         continue;
2495                                 evaluatingMath = false;
2496                                 mi.incrEntry();
2497                                 math_pos = mi.getStartPos();
2498                         }
2499                         if (it_key == keys.end()) {
2500                                 found = KeyInfo(KeyInfo::isStandard, 0, true);
2501                                 LYXERR(Debug::INFO, "Undefined key " << key << " ==> will be used as text");
2502                                 found = KeyInfo(KeyInfo::isText, 0, false);
2503                                 if (isPatternString) {
2504                                         found.keytype = KeyInfo::isChar;
2505                                         found.disabled = false;
2506                                         found.used = true;
2507                                 }
2508                                 keys[key] = found;
2509                                 usedText.push_back(key);
2510                         }
2511                         else
2512                                 found = keys[key];
2513                         if (key.compare("regexp") == 0) {
2514                                 evaluatingRegexp = true;
2515                                 found._tokenstart = sub.position(size_t(2));
2516                                 found._tokensize = 0;
2517                                 continue;
2518                         }
2519                 }
2520                 // Handle the other params of key
2521                 if (found.keytype == KeyInfo::isIgnored)
2522                         continue;
2523                 else if (found.keytype == KeyInfo::isMath) {
2524                         if (size_t(sub.position(size_t(2))) == math_pos) {
2525                                 found = keys[key];
2526                                 found._tokenstart = sub.position(size_t(2));
2527                                 found._tokensize = mi.getSize();
2528                                 found._dataEnd = found._tokenstart + found._tokensize;
2529                                 found._dataStart = found._dataEnd;
2530                                 found.parenthesiscount = 0;
2531                                 found.head = interval_.par.substr(found._tokenstart, found._tokensize);
2532                                 if (removeMathHull) {
2533                                         interval_.addIntervall(found._tokenstart, found._tokenstart + mi.getPrefixSize());
2534                                         interval_.addIntervall(found._dataEnd - mi.getPostfixSize(), found._dataEnd);
2535                                 }
2536                                 else {
2537                                         // Treate all math constructs as simple math
2538                                         interval_.par[found._tokenstart] = '$';
2539                                         interval_.par[found._dataEnd - mi.getPostfixSize()] = '$';
2540                                         interval_.addIntervall(found._tokenstart + 1, found._tokenstart + mi.getPrefixSize());
2541                                         interval_.addIntervall(found._dataEnd - mi.getPostfixSize() + 1, found._dataEnd);
2542                                 }
2543                                 evaluatingMath = true;
2544                         }
2545                         else {
2546                                 // begin|end of unknown env, discard
2547                                 // First handle tables
2548                                 // longtable|tabular
2549                                 bool discardComment;
2550                                 found = keys[key];
2551                                 found.keytype = KeyInfo::doRemove;
2552                                 if ((sub.str(7).compare("longtable") == 0) ||
2553                                                 (sub.str(7).compare("tabular") == 0)) {
2554                                         discardComment = true;        /* '%' */
2555                                 }
2556                                 else {
2557                                         discardComment = false;
2558                                         static regex const removeArgs("^(multicols|multipar|sectionbox|subsectionbox|tcolorbox)$");
2559                                         smatch sub2;
2560                                         string token = sub.str(7);
2561                                         if (regex_match(token, sub2, removeArgs)) {
2562                                                 found.keytype = KeyInfo::removeWithArg;
2563                                         }
2564                                 }
2565                                 // discard spaces before pos(2)
2566                                 int pos = sub.position(size_t(2));
2567                                 int count;
2568                                 for (count = 0; pos - count > 0; count++) {
2569                                         char c = interval_.par[pos-count-1];
2570                                         if (discardComment) {
2571                                                 if ((c != ' ') && (c != '%'))
2572                                                         break;
2573                                         }
2574                                         else if (c != ' ')
2575                                                 break;
2576                                 }
2577                                 found._tokenstart = pos - count;
2578                                 if (sub.str(3).compare(0, 5, "begin") == 0) {
2579                                         size_t pos1 = pos + sub.str(2).length();
2580                                         if (sub.str(7).compare("cjk") == 0) {
2581                                                 pos1 = interval_.findclosing(pos1+1, interval_.par.length()) + 1;
2582                                                 if ((interval_.par[pos1] == '{') && (interval_.par[pos1+1] == '}'))
2583                                                         pos1 += 2;
2584                                                 found.keytype = KeyInfo::isMain;
2585                                                 found._dataStart = pos1;
2586                                                 found._dataEnd = interval_.par.length();
2587                                                 found.disabled = keys["foreignlanguage"].disabled;
2588                                                 found.used = keys["foreignlanguage"].used;
2589                                                 found._tokensize = pos1 - found._tokenstart;
2590                                                 found.head = interval_.par.substr(found._tokenstart, found._tokensize);
2591                                         }
2592                                         else {
2593                                                 // Swallow possible optional params
2594                                                 while (interval_.par[pos1] == '[') {
2595                                                         pos1 = interval_.findclosing(pos1+1, interval_.par.length(), '[', ']')+1;
2596                                                 }
2597                                                 // Swallow also the eventual parameter
2598                                                 if (interval_.par[pos1] == '{') {
2599                                                         found._dataEnd = interval_.findclosing(pos1+1, interval_.par.length()) + 1;
2600                                                 }
2601                                                 else {
2602                                                         found._dataEnd = pos1;
2603                                                 }
2604                                                 found._dataStart = found._dataEnd;
2605                                                 found._tokensize = count + found._dataEnd - pos;
2606                                                 found.parenthesiscount = 0;
2607                                                 found.head = interval_.par.substr(found._tokenstart, found._tokensize);
2608                                                 found.disabled = true;
2609                                         }
2610                                 }
2611                                 else {
2612                                         // Handle "\end{...}"
2613                                         found._dataStart = pos + sub.str(2).length();
2614                                         found._dataEnd = found._dataStart;
2615                                         found._tokensize = count + found._dataEnd - pos;
2616                                         found.parenthesiscount = 0;
2617                                         found.head = interval_.par.substr(found._tokenstart, found._tokensize);
2618                                         found.disabled = true;
2619                                 }
2620                         }
2621                 }
2622                 else if (found.keytype != KeyInfo::isRegex) {
2623                         found._tokenstart = sub.position(size_t(2));
2624                         if (found.parenthesiscount == 0) {
2625                                 // Probably to be discarded
2626                                 size_t following_pos = sub.position(size_t(2)) + sub.str(5).length() + 1;
2627                                 char following = interval_.par[following_pos];
2628                                 if (following == ' ')
2629                                         found.head = "\\" + sub.str(5) + " ";
2630                                 else if (following == '=') {
2631                                         // like \uldepth=1000pt
2632                                         found.head = sub.str(2);
2633                                 }
2634                                 else
2635                                         found.head = "\\" + key;
2636                                 found._tokensize = found.head.length();
2637                                 found._dataEnd = found._tokenstart + found._tokensize;
2638                                 found._dataStart = found._dataEnd;
2639                         }
2640                         else {
2641                                 int params = found._tokenstart + key.length() + 1;
2642                                 if (evaluatingOptional) {
2643                                         if (size_t(found._tokenstart) > optionalEnd) {
2644                                                 evaluatingOptional = false;
2645                                         }
2646                                         else {
2647                                                 found.disabled = true;
2648                                         }
2649                                 }
2650                                 int optend = params;
2651                                 while (interval_.par[optend] == '[') {
2652                                         // discard optional parameters
2653                                         optend = interval_.findclosing(optend+1, interval_.par.length(), '[', ']') + 1;
2654                                 }
2655                                 if (optend > params) {
2656                                         key += interval_.par.substr(params, optend-params);
2657                                         evaluatingOptional = true;
2658                                         optionalEnd = optend;
2659                                         if (found.keytype == KeyInfo::isSectioning) {
2660                                                 // Remove optional values (but still keep in header)
2661                                                 interval_.addIntervall(params, optend);
2662                                         }
2663                                 }
2664                                 string token = sub.str(7);
2665                                 int closings;
2666                                 if (interval_.par[optend] != '{') {
2667                                         closings = 0;
2668                                         found.parenthesiscount = 0;
2669                                         found.head = "\\" + key;
2670                                 }
2671                                 else
2672                                         closings = found.parenthesiscount;
2673                                 if (found.parenthesiscount == 1) {
2674                                         found.head = "\\" + key + "{";
2675                                 }
2676                                 else if (found.parenthesiscount > 1) {
2677                                         if (token != "") {
2678                                                 found.head = sub.str(2) + "{";
2679                                                 closings = found.parenthesiscount - 1;
2680                                         }
2681                                         else {
2682                                                 found.head = "\\" + key + "{";
2683                                         }
2684                                 }
2685                                 found._tokensize = found.head.length();
2686                                 found._dataStart = found._tokenstart + found.head.length();
2687                                 if (found.keytype == KeyInfo::doRemove) {
2688                                         if (closings > 0) {
2689                                                 size_t endpar = 2 + interval_.findclosing(found._dataStart, interval_.par.length(), '{', '}', closings);
2690                                                 if (endpar >= interval_.par.length())
2691                                                         found._dataStart = interval_.par.length();
2692                                                 else
2693                                                         found._dataStart = endpar;
2694                                                 found._tokensize = found._dataStart - found._tokenstart;
2695                                         }
2696                                         else {
2697                                                 found._dataStart = found._tokenstart + found._tokensize;
2698                                         }
2699                                         closings = 0;
2700                                 }
2701                                 if (interval_.par.substr(found._dataStart, 15).compare("\\endarguments{}") == 0) {
2702                                         found._dataStart += 15;
2703                                 }
2704                                 size_t endpos;
2705                                 if (closings < 1)
2706                                         endpos = found._dataStart - 1;
2707                                 else
2708                                         endpos = interval_.findclosing(found._dataStart, interval_.par.length(), '{', '}', closings);
2709                                 if (found.keytype == KeyInfo::isList) {
2710                                         // Check if it really is list env
2711                                         static regex const listre("^([a-z]+)$");
2712                                         smatch sub2;
2713                                         if (!regex_match(token, sub2, listre)) {
2714                                                 // Change the key of this entry. It is not in a list/item environment
2715                                                 found.keytype = KeyInfo::endArguments;
2716                                         }
2717                                 }
2718                                 if (found.keytype == KeyInfo::noMain) {
2719                                         evaluatingCode = true;
2720                                         codeEnd = endpos;
2721                                         codeStart = found._dataStart;
2722                                 }
2723                                 else if (evaluatingCode) {
2724                                         if (size_t(found._dataStart) > codeEnd)
2725                                                 evaluatingCode = false;
2726                                         else if (found.keytype == KeyInfo::isMain) {
2727                                                 // Disable this key, treate it as standard
2728                                                 found.keytype = KeyInfo::isStandard;
2729                                                 found.disabled = true;
2730                                                 if ((codeEnd +1 >= interval_.par.length()) &&
2731                                                                 (found._tokenstart == codeStart)) {
2732                                                         // trickery, because the code inset starts
2733                                                         // with \selectlanguage ...
2734                                                         codeEnd = endpos;
2735                                                         if (entries_.size() > 1) {
2736                                                                 entries_[entries_.size()-1]._dataEnd = codeEnd;
2737                                                         }
2738                                                 }
2739                                         }
2740                                 }
2741                                 if ((endpos == interval_.par.length()) &&
2742                                                 (found.keytype == KeyInfo::doRemove)) {
2743                                         // Missing closing => error in latex-input?
2744                                         // therefore do not delete remaining data
2745                                         found._dataStart -= 1;
2746                                         found._dataEnd = found._dataStart;
2747                                 }
2748                                 else
2749                                         found._dataEnd = endpos;
2750                         }
2751                         if (isPatternString) {
2752                                 keys[key].used = true;
2753                         }
2754                 }
2755                 entries_.push_back(found);
2756         }
2757 }
2758
2759 void LatexInfo::makeKey(const string &keysstring, KeyInfo keyI, bool isPatternString)
2760 {
2761         stringstream s(keysstring);
2762         string key;
2763         const char delim = '|';
2764         while (getline(s, key, delim)) {
2765                 KeyInfo keyII(keyI);
2766                 if (isPatternString) {
2767                         keyII.used = false;
2768                 }
2769                 else if ( !keys[key].used)
2770                         keyII.disabled = true;
2771                 keys[key] = keyII;
2772         }
2773 }
2774
2775 void LatexInfo::buildKeys(bool isPatternString)
2776 {
2777
2778         static bool keysBuilt = false;
2779         if (keysBuilt && !isPatternString) return;
2780
2781         // Keys to ignore in any case
2782         makeKey("text|lyxmathsym|ensuremath", KeyInfo(KeyInfo::headRemove, 1, true), true);
2783         makeKey("nonumber|notag", KeyInfo(KeyInfo::headRemove, 0, true), true);
2784         // Known standard keys with 1 parameter.
2785         // Split is done, if not at start of region
2786         makeKey("textsf|textss|texttt", KeyInfo(KeyInfo::isStandard, 1, ignoreFormats.getFamily()), isPatternString);
2787         makeKey("textbf",               KeyInfo(KeyInfo::isStandard, 1, ignoreFormats.getSeries()), isPatternString);
2788         makeKey("textit|textsc|textsl", KeyInfo(KeyInfo::isStandard, 1, ignoreFormats.getShape()), isPatternString);
2789         makeKey("uuline|uline|uwave",   KeyInfo(KeyInfo::isStandard, 1, ignoreFormats.getUnderline()), isPatternString);
2790         makeKey("emph|noun",            KeyInfo(KeyInfo::isStandard, 1, ignoreFormats.getMarkUp()), isPatternString);
2791         makeKey("sout|xout",            KeyInfo(KeyInfo::isStandard, 1, ignoreFormats.getStrikeOut()), isPatternString);
2792
2793         makeKey("section|subsection|subsubsection|paragraph|subparagraph|minisec",
2794                 KeyInfo(KeyInfo::isSectioning, 1, ignoreFormats.getSectioning()), isPatternString);
2795         makeKey("section*|subsection*|subsubsection*|paragraph*",
2796                 KeyInfo(KeyInfo::isSectioning, 1, ignoreFormats.getSectioning()), isPatternString);
2797         makeKey("part|part*|chapter|chapter*", KeyInfo(KeyInfo::isSectioning, 1, ignoreFormats.getSectioning()), isPatternString);
2798         makeKey("title|subtitle|author|subject|publishers|dedication|uppertitleback|lowertitleback|extratitle|lyxaddress|lyxrightaddress", KeyInfo(KeyInfo::isTitle, 1, ignoreFormats.getFrontMatter()), isPatternString);
2799         // Regex
2800         makeKey("regexp", KeyInfo(KeyInfo::isRegex, 1, false), isPatternString);
2801
2802         // Split is done, if not at start of region
2803         makeKey("textcolor", KeyInfo(KeyInfo::isStandard, 2, ignoreFormats.getColor()), isPatternString);
2804         makeKey("latexenvironment", KeyInfo(KeyInfo::isStandard, 2, false), isPatternString);
2805
2806         // Split is done always.
2807         makeKey("foreignlanguage", KeyInfo(KeyInfo::isMain, 2, ignoreFormats.getLanguage()), isPatternString);
2808
2809         // Known charaters
2810         // No split
2811         makeKey("backslash|textbackslash|slash",  KeyInfo(KeyInfo::isChar, 0, false), isPatternString);
2812         makeKey("textasciicircum|textasciitilde", KeyInfo(KeyInfo::isChar, 0, false), isPatternString);
2813         makeKey("textasciiacute|texemdash",       KeyInfo(KeyInfo::isChar, 0, false), isPatternString);
2814         makeKey("dots|ldots",                     KeyInfo(KeyInfo::isChar, 0, false), isPatternString);
2815         // Spaces
2816         makeKey("quad|qquad|hfill|dotfill",               KeyInfo(KeyInfo::isChar, 0, false), isPatternString);
2817         makeKey("textvisiblespace|nobreakspace",          KeyInfo(KeyInfo::isChar, 0, false), isPatternString);
2818         makeKey("negthickspace|negmedspace|negthinspace", KeyInfo(KeyInfo::isChar, 0, false), isPatternString);
2819         makeKey("thickspace|medspace|thinspace",          KeyInfo(KeyInfo::isChar, 0, false), isPatternString);
2820         // Skip
2821         // makeKey("enskip|smallskip|medskip|bigskip|vfill", KeyInfo(KeyInfo::isChar, 0, false), isPatternString);
2822         // Custom space/skip, remove the content (== length value)
2823         makeKey("vspace|vspace*|hspace|hspace*|mspace", KeyInfo(KeyInfo::noContent, 1, false), isPatternString);
2824         // Found in fr/UserGuide.lyx
2825         makeKey("og|fg", KeyInfo(KeyInfo::isChar, 0, false), isPatternString);
2826         // quotes
2827         makeKey("textquotedbl|quotesinglbase|lyxarrow", KeyInfo(KeyInfo::isChar, 0, false), isPatternString);
2828         makeKey("textquotedblleft|textquotedblright", KeyInfo(KeyInfo::isChar, 0, false), isPatternString);
2829         // Known macros to remove (including their parameter)
2830         // No split
2831         makeKey("input|inputencoding|label|ref|index|bibitem", KeyInfo(KeyInfo::doRemove, 1, false), isPatternString);
2832         makeKey("addtocounter|setlength",                 KeyInfo(KeyInfo::noContent, 2, true), isPatternString);
2833         // handle like standard keys with 1 parameter.
2834         makeKey("url|href|vref|thanks", KeyInfo(KeyInfo::isStandard, 1, false), isPatternString);
2835
2836         if (ignoreFormats.getDeleted()) {
2837                 // Ignore deleted text
2838                 makeKey("lyxdeleted", KeyInfo(KeyInfo::doRemove, 3, false), isPatternString);
2839         }
2840         else {
2841                 // but preserve added text
2842                 makeKey("lyxdeleted", KeyInfo(KeyInfo::doRemove, 2, false), isPatternString);
2843         }
2844         makeKey("lyxadded", KeyInfo(KeyInfo::doRemove, 2, false), isPatternString);
2845
2846         // Macros to remove, but let the parameter survive
2847         // No split
2848         makeKey("menuitem|textmd|textrm", KeyInfo(KeyInfo::isStandard, 1, true), isPatternString);
2849
2850         // Remove language spec from content of these insets
2851         makeKey("code", KeyInfo(KeyInfo::noMain, 1, false), isPatternString);
2852
2853         // Same effect as previous, parameter will survive (because there is no one anyway)
2854         // No split
2855         makeKey("noindent|textcompwordmark|maketitle", KeyInfo(KeyInfo::isStandard, 0, true), isPatternString);
2856         // Remove table decorations
2857         makeKey("hline|tabularnewline|toprule|bottomrule|midrule", KeyInfo(KeyInfo::doRemove, 0, true), isPatternString);
2858         // Discard shape-header.
2859         // For footnote or shortcut too, because of lang settings
2860         // and wrong handling if used 'KeyInfo::noMain'
2861         makeKey("circlepar|diamondpar|heartpar|nutpar",  KeyInfo(KeyInfo::isStandard, 1, true), isPatternString);
2862         makeKey("trianglerightpar|hexagonpar|starpar",   KeyInfo(KeyInfo::isStandard, 1, true), isPatternString);
2863         makeKey("triangleuppar|triangledownpar|droppar", KeyInfo(KeyInfo::isStandard, 1, true), isPatternString);
2864         makeKey("triangleleftpar|shapepar|dropuppar",    KeyInfo(KeyInfo::isStandard, 1, true), isPatternString);
2865         makeKey("hphantom|vphantom|note|footnote|shortcut|include|includegraphics",     KeyInfo(KeyInfo::isStandard, 1, true), isPatternString);
2866         makeKey("textgreek|textcyrillic", KeyInfo(KeyInfo::isStandard, 1, true), false);
2867         makeKey("parbox", KeyInfo(KeyInfo::doRemove, 1, true), isPatternString);
2868         // like ('tiny{}' or '\tiny ' ... )
2869         makeKey("footnotesize|tiny|scriptsize|small|large|Large|LARGE|huge|Huge", KeyInfo(KeyInfo::isSize, 0, ignoreFormats.getSize()), isPatternString);
2870
2871         // Survives, like known character
2872         // makeKey("lyx|LyX|latex|LaTeX|latexe|LaTeXe|tex|TeX", KeyInfo(KeyInfo::isChar, 0, false), isPatternString);
2873         makeKey("tableofcontents", KeyInfo(KeyInfo::isChar, 0, false), isPatternString);
2874         makeKey("item|listitem", KeyInfo(KeyInfo::isList, 1, false), isPatternString);
2875
2876         makeKey("begin|end", KeyInfo(KeyInfo::isMath, 1, false), isPatternString);
2877         makeKey("[|]", KeyInfo(KeyInfo::isMath, 1, false), isPatternString);
2878         makeKey("$", KeyInfo(KeyInfo::isMath, 1, false), isPatternString);
2879
2880         makeKey("par|uldepth|ULdepth|protect|nobreakdash|medskip|relax", KeyInfo(KeyInfo::isStandard, 0, true), isPatternString);
2881         // Remove RTL/LTR marker
2882         makeKey("l|r|textlr|textfr|textar|beginl|endl", KeyInfo(KeyInfo::isStandard, 0, true), isPatternString);
2883         makeKey("lettrine", KeyInfo(KeyInfo::cleanToStart, 0, true), isPatternString);
2884         makeKey("lyxslide", KeyInfo(KeyInfo::isSectioning, 1, true), isPatternString);
2885         makeKey("endarguments", KeyInfo(KeyInfo::endArguments, 0, true), isPatternString);
2886         makeKey("twocolumn", KeyInfo(KeyInfo::removeWithArg, 2, true), isPatternString);
2887         makeKey("tnotetext|ead|fntext|cortext|address", KeyInfo(KeyInfo::removeWithArg, 0, true), isPatternString);
2888         makeKey("lyxend", KeyInfo(KeyInfo::isStandard, 0, true), isPatternString);
2889         if (isPatternString) {
2890                 // Allow the first searched string to rebuild the keys too
2891                 keysBuilt = false;
2892         }
2893         else {
2894                 // no need to rebuild again
2895                 keysBuilt = true;
2896         }
2897 }
2898
2899 /*
2900  * Keep the list of actual opened parentheses actual
2901  * (e.g. depth == 4 means there are 4 '{' not processed yet)
2902  */
2903 void Intervall::handleParentheses(int lastpos, bool closingAllowed)
2904 {
2905         int skip = 0;
2906         for (int i = depts[actualdeptindex]; i < lastpos; i+= 1 + skip) {
2907                 char c;
2908                 c = par[i];
2909                 skip = 0;
2910                 if (c == '\\') skip = 1;
2911                 else if (c == '{') {
2912                         handleOpenP(i);
2913                 }
2914                 else if (c == '}') {
2915                         handleCloseP(i, closingAllowed);
2916                 }
2917         }
2918 }
2919
2920 #if (0)
2921 string Intervall::show(int lastpos)
2922 {
2923         int idx = 0;                          /* int intervalls */
2924         string s;
2925         int i = 0;
2926         if ((unsigned) lastpos > par.size())
2927                 lastpos = par.size();
2928         for (idx = 0; idx <= ignoreidx; idx++) {
2929                 while (i < lastpos) {
2930                         int printsize;
2931                         if (i <= borders[idx].low) {
2932                                 if (borders[idx].low > lastpos)
2933                                         printsize = lastpos - i;
2934                                 else
2935                                         printsize = borders[idx].low - i;
2936                                 s += par.substr(i, printsize);
2937                                 i += printsize;
2938                                 if (i >= borders[idx].low)
2939                                         i = borders[idx].upper;
2940                         }
2941                         else {
2942                                 i = borders[idx].upper;
2943                                 break;
2944                         }
2945                 }
2946         }
2947         if (lastpos > i) {
2948                 s += par.substr(i, lastpos-i);
2949         }
2950         return s;
2951 }
2952 #endif
2953
2954 void Intervall::output(ostringstream &os, int lastpos)
2955 {
2956         // get number of chars to output
2957         int idx = 0;                          /* int intervalls */
2958         int i = 0;
2959         int printed = 0;
2960         string startTitle = titleValue;
2961         for (idx = 0; idx <= ignoreidx; idx++) {
2962                 if (i < lastpos) {
2963                         if (i <= borders[idx].low) {
2964                                 int printsize;
2965                                 if (borders[idx].low > lastpos)
2966                                         printsize = lastpos - i;
2967                                 else
2968                                         printsize = borders[idx].low - i;
2969                                 if (printsize > 0) {
2970                                         os << startTitle << par.substr(i, printsize);
2971                                         i += printsize;
2972                                         printed += printsize;
2973                                         startTitle = "";
2974                                 }
2975                                 handleParentheses(i, false);
2976                                 if (i >= borders[idx].low)
2977                                         i = borders[idx].upper;
2978                         }
2979                         else {
2980                                 i = borders[idx].upper;
2981                         }
2982                 }
2983                 else
2984                         break;
2985         }
2986         if (lastpos > i) {
2987                 os << startTitle << par.substr(i, lastpos-i);
2988                 printed += lastpos-i;
2989         }
2990         handleParentheses(lastpos, false);
2991         int startindex;
2992         if (keys["foreignlanguage"].disabled)
2993                 startindex = actualdeptindex-langcount;
2994         else
2995                 startindex = actualdeptindex;
2996         for (int i = startindex; i > 0; --i) {
2997                 os << "}";
2998         }
2999         if (hasTitle && (printed > 0))
3000                 os << "}";
3001         if (! isPatternString_)
3002                 os << "\n";
3003         handleParentheses(lastpos, true); /* extra closings '}' allowed here */
3004 }
3005
3006 void LatexInfo::processRegion(int start, int region_end)
3007 {
3008         while (start < region_end) {          /* Let {[} and {]} survive */
3009                 int cnt = interval_.isOpeningPar(start);
3010                 if (cnt == 1) {
3011                         // Closing is allowed past the region
3012                         int closing = interval_.findclosing(start+1, interval_.par.length());
3013                         interval_.addIntervall(start, start+1);
3014                         interval_.addIntervall(closing, closing+1);
3015                 }
3016                 else if (cnt == 3)
3017                         start += 2;
3018                 start = interval_.nextNotIgnored(start+1);
3019         }
3020 }
3021
3022 void LatexInfo::removeHead(KeyInfo const & actual, int count)
3023 {
3024         if (actual.parenthesiscount == 0) {
3025                 // "{\tiny{} ...}" ==> "{{} ...}"
3026                 interval_.addIntervall(actual._tokenstart-count, actual._tokenstart + actual._tokensize);
3027         }
3028         else {
3029                 // Remove header hull, that is "\url{abcd}" ==> "abcd"
3030                 interval_.addIntervall(actual._tokenstart - count, actual._dataStart);
3031                 interval_.addIntervall(actual._dataEnd, actual._dataEnd+1);
3032         }
3033 }
3034
3035 int LatexInfo::dispatch(ostringstream &os, int previousStart, KeyInfo &actual)
3036 {
3037         int nextKeyIdx = 0;
3038         switch (actual.keytype)
3039         {
3040         case KeyInfo::isTitle: {
3041                 removeHead(actual);
3042                 nextKeyIdx = getNextKey();
3043                 break;
3044         }
3045         case KeyInfo::cleanToStart: {
3046                 actual._dataEnd = actual._dataStart;
3047                 nextKeyIdx = getNextKey();
3048                 // Search for end of arguments
3049                 int tmpIdx = find(nextKeyIdx, KeyInfo::endArguments);
3050                 if (tmpIdx > 0) {
3051                         for (int i = nextKeyIdx; i <= tmpIdx; i++) {
3052                                 entries_[i].disabled = true;
3053                         }
3054                         actual._dataEnd = entries_[tmpIdx]._dataEnd;
3055                 }
3056                 while (interval_.par[actual._dataEnd] == ' ')
3057                         actual._dataEnd++;
3058                 interval_.addIntervall(0, actual._dataEnd+1);
3059                 interval_.actualdeptindex = 0;
3060                 interval_.depts[0] = actual._dataEnd+1;
3061                 interval_.closes[0] = -1;
3062                 break;
3063         }
3064         case KeyInfo::isText:
3065                 interval_.par[actual._tokenstart] = '#';
3066                 //interval_.addIntervall(actual._tokenstart, actual._tokenstart+1);
3067                 nextKeyIdx = getNextKey();
3068                 break;
3069         case KeyInfo::noContent: {          /* char like "\hspace{2cm}" */
3070                 if (actual.disabled)
3071                         interval_.addIntervall(actual._tokenstart, actual._dataEnd);
3072                 else
3073                         interval_.addIntervall(actual._dataStart, actual._dataEnd);
3074         }
3075                 // fall through
3076         case KeyInfo::isChar: {
3077                 nextKeyIdx = getNextKey();
3078                 break;
3079         }
3080         case KeyInfo::isSize: {
3081                 if (actual.disabled || (interval_.par[actual._dataStart] != '{') || (interval_.par[actual._dataStart-1] == ' ')) {
3082                         if (actual.parenthesiscount == 0)
3083                                 interval_.addIntervall(actual._tokenstart, actual._dataEnd);
3084                         else {
3085                                 interval_.addIntervall(actual._tokenstart, actual._dataEnd+1);
3086                         }
3087                         nextKeyIdx = getNextKey();
3088                 } else {
3089                         // Here _dataStart points to '{', so correct it
3090                         actual._dataStart += 1;
3091                         actual._tokensize += 1;
3092                         actual.parenthesiscount = 1;
3093                         if (interval_.par[actual._dataStart] == '}') {
3094                                 // Determine the end if used like '{\tiny{}...}'
3095                                 actual._dataEnd = interval_.findclosing(actual._dataStart+1, interval_.par.length()) + 1;
3096                                 interval_.addIntervall(actual._dataStart, actual._dataStart+1);
3097                         }
3098                         else {
3099                                 // Determine the end if used like '\tiny{...}'
3100                                 actual._dataEnd = interval_.findclosing(actual._dataStart, interval_.par.length()) + 1;
3101                         }
3102                         // Split on this key if not at start
3103                         int start = interval_.nextNotIgnored(previousStart);
3104                         if (start < actual._tokenstart) {
3105                                 interval_.output(os, actual._tokenstart);
3106                                 interval_.addIntervall(start, actual._tokenstart);
3107                         }
3108                         // discard entry if at end of actual
3109                         nextKeyIdx = process(os, actual);
3110                 }
3111                 break;
3112         }
3113         case KeyInfo::endArguments: {
3114                 // Remove trailing '{}' too
3115                 actual._dataStart += 1;
3116                 actual._dataEnd += 1;
3117                 interval_.addIntervall(actual._tokenstart, actual._dataEnd+1);
3118                 nextKeyIdx = getNextKey();
3119                 break;
3120         }
3121         case KeyInfo::noMain:
3122                 // fall through
3123         case KeyInfo::isStandard: {
3124                 if (actual.disabled) {
3125                         removeHead(actual);
3126                         processRegion(actual._dataStart, actual._dataStart+1);
3127                         nextKeyIdx = getNextKey();
3128                 } else {
3129                         // Split on this key if not at datastart of calling entry
3130                         int start = interval_.nextNotIgnored(previousStart);
3131                         if (start < actual._tokenstart) {
3132                                 interval_.output(os, actual._tokenstart);
3133                                 interval_.addIntervall(start, actual._tokenstart);
3134                         }
3135                         // discard entry if at end of actual
3136                         nextKeyIdx = process(os, actual);
3137                 }
3138                 break;
3139         }
3140         case KeyInfo::removeWithArg: {
3141                 nextKeyIdx = getNextKey();
3142                 // Search for end of arguments
3143                 int tmpIdx = find(nextKeyIdx, KeyInfo::endArguments);
3144                 if (tmpIdx > 0) {
3145                         for (int i = nextKeyIdx; i <= tmpIdx; i++) {
3146                                 entries_[i].disabled = true;
3147                         }
3148                         actual._dataEnd = entries_[tmpIdx]._dataEnd;
3149                 }
3150                 interval_.addIntervall(actual._tokenstart, actual._dataEnd+1);
3151                 break;
3152         }
3153         case KeyInfo::doRemove: {
3154                 // Remove the key with all parameters and following spaces
3155                 size_t pos;
3156                 size_t start;
3157                 if (interval_.par[actual._dataEnd-1] == ' ' || interval_.par[actual._dataEnd-1] == '}')
3158                         start = actual._dataEnd;
3159                 else
3160                         start = actual._dataEnd+1;
3161                 for (pos = start; pos < interval_.par.length(); pos++) {
3162                         if ((interval_.par[pos] != ' ') && (interval_.par[pos] != '%'))
3163                                 break;
3164                 }
3165                 // Remove also enclosing parentheses [] and {}
3166                 int numpars = 0;
3167                 int spaces = 0;
3168                 while (actual._tokenstart > numpars) {
3169                         if (pos+numpars >= interval_.par.size())
3170                                 break;
3171                         else if (interval_.par[pos+numpars] == ']' && interval_.par[actual._tokenstart-numpars-1] == '[')
3172                                 numpars++;
3173                         else if (interval_.par[pos+numpars] == '}' && interval_.par[actual._tokenstart-numpars-1] == '{')
3174                                 numpars++;
3175                         else
3176                                 break;
3177                 }
3178                 if (numpars > 0) {
3179                         if (interval_.par[pos+numpars] == ' ')
3180                                 spaces++;
3181                 }
3182
3183                 interval_.addIntervall(actual._tokenstart-numpars, pos+numpars+spaces);
3184                 nextKeyIdx = getNextKey();
3185                 break;
3186         }
3187         case KeyInfo::isList: {
3188                 // Discard space before _tokenstart
3189                 int count;
3190                 for (count = 0; count < actual._tokenstart; count++) {
3191                         if (interval_.par[actual._tokenstart-count-1] != ' ')
3192                                 break;
3193                 }
3194                 nextKeyIdx = getNextKey();
3195                 int tmpIdx = find(nextKeyIdx, KeyInfo::endArguments);
3196                 if (tmpIdx > 0) {
3197                         // Special case: \item is not a list, but a command (like in Style Author_Biography in maa-monthly.layout)
3198                         // with arguments
3199                         // How else can we catch this one?
3200                         for (int i = nextKeyIdx; i <= tmpIdx; i++) {
3201                                 entries_[i].disabled = true;
3202                         }
3203                         actual._dataEnd = entries_[tmpIdx]._dataEnd;
3204                 }
3205                 else if (nextKeyIdx > 0) {
3206                         // Ignore any lang entries inside data region
3207                         for (int i = nextKeyIdx; i < int(entries_.size()) && entries_[i]._tokenstart < actual._dataEnd; i++) {
3208                                 if (entries_[i].keytype == KeyInfo::isMain)
3209                                         entries_[i].disabled = true;
3210                         }
3211                 }
3212                 if (actual.disabled) {
3213                         interval_.addIntervall(actual._tokenstart-count, actual._dataEnd+1);
3214                 }
3215                 else {
3216                         interval_.addIntervall(actual._tokenstart-count, actual._tokenstart);
3217                 }
3218                 if (interval_.par[actual._dataEnd+1] == '[') {
3219                         int posdown = interval_.findclosing(actual._dataEnd+2, interval_.par.length(), '[', ']');
3220                         if ((interval_.par[actual._dataEnd+2] == '{') &&
3221                                         (interval_.par[posdown-1] == '}')) {
3222                                 interval_.addIntervall(actual._dataEnd+1,actual._dataEnd+3);
3223                                 interval_.addIntervall(posdown-1, posdown+1);
3224                         }
3225                         else {
3226                                 interval_.addIntervall(actual._dataEnd+1, actual._dataEnd+2);
3227                                 interval_.addIntervall(posdown, posdown+1);
3228                         }
3229                         int blk = interval_.nextNotIgnored(actual._dataEnd+1);
3230                         if (blk > posdown) {
3231                                 // Discard at most 1 space after empty item
3232                                 int count;
3233                                 for (count = 0; count < 1; count++) {
3234                                         if (interval_.par[blk+count] != ' ')
3235                                                 break;
3236                                 }
3237                                 if (count > 0)
3238                                         interval_.addIntervall(blk, blk+count);
3239                         }
3240                 }
3241                 break;
3242         }
3243         case KeyInfo::isSectioning: {
3244                 // Discard spaces before _tokenstart
3245                 int count;
3246                 int val = actual._tokenstart;
3247                 for (count = 0; count < actual._tokenstart;) {
3248                         val = interval_.previousNotIgnored(val-1);
3249                         if (val < 0 || interval_.par[val] != ' ')
3250                                 break;
3251                         else {
3252                                 count = actual._tokenstart - val;
3253                         }
3254                 }
3255                 if (actual.disabled) {
3256                         removeHead(actual, count);
3257                         nextKeyIdx = getNextKey();
3258                 } else {
3259                         interval_.addIntervall(actual._tokenstart-count, actual._tokenstart);
3260                         nextKeyIdx = process(os, actual);
3261                 }
3262                 break;
3263         }
3264         case KeyInfo::isMath: {
3265                 // Same as regex, use the content unchanged
3266                 nextKeyIdx = getNextKey();
3267                 break;
3268         }
3269         case KeyInfo::isRegex: {
3270                 // DO NOT SPLIT ON REGEX
3271                 // Do not disable
3272                 nextKeyIdx = getNextKey();
3273                 break;
3274         }
3275         case KeyInfo::isIgnored: {
3276                 // Treat like a character for now
3277                 nextKeyIdx = getNextKey();
3278                 break;
3279         }
3280         case KeyInfo::isMain: {
3281                 if (interval_.par.substr(actual._dataStart, 2) == "% ")
3282                         interval_.addIntervall(actual._dataStart, actual._dataStart+2);
3283                 if (actual._tokenstart > 0) {
3284                         int prev = interval_.previousNotIgnored(actual._tokenstart - 1);
3285                         if ((prev >= 0) && interval_.par[prev] == '%')
3286                                 interval_.addIntervall(prev, prev+1);
3287                 }
3288                 if (actual.disabled) {
3289                         removeHead(actual);
3290                         interval_.langcount++;
3291                         if ((interval_.par.substr(actual._dataStart, 3) == " \\[") ||
3292                                         (interval_.par.substr(actual._dataStart, 8) == " \\begin{")) {
3293                                 // Discard also the space before math-equation
3294                                 interval_.addIntervall(actual._dataStart, actual._dataStart+1);
3295                         }
3296                         nextKeyIdx = getNextKey();
3297                         // interval.resetOpenedP(actual._dataStart-1);
3298                 }
3299                 else {
3300                         if (actual._tokenstart < 26) {
3301                                 // for the first (and maybe dummy) language
3302                                 interval_.setForDefaultLang(actual);
3303                         }
3304                         interval_.resetOpenedP(actual._dataStart-1);
3305                 }
3306                 break;
3307         }
3308         case KeyInfo::invalid:
3309         case KeyInfo::headRemove:
3310                 // These two cases cannot happen, already handled
3311                 // fall through
3312         default: {
3313                 // LYXERR(Debug::INFO, "Unhandled keytype");
3314                 nextKeyIdx = getNextKey();
3315                 break;
3316         }
3317         }
3318         return nextKeyIdx;
3319 }
3320
3321 int LatexInfo::process(ostringstream & os, KeyInfo const & actual )
3322 {
3323         int end = interval_.nextNotIgnored(actual._dataEnd);
3324         int oldStart = actual._dataStart;
3325         int nextKeyIdx = getNextKey();
3326         while (true) {
3327                 if ((nextKeyIdx < 0) ||
3328                                 (entries_[nextKeyIdx]._tokenstart >= actual._dataEnd) ||
3329                                 (entries_[nextKeyIdx].keytype == KeyInfo::invalid)) {
3330                         if (oldStart <= end) {
3331                                 processRegion(oldStart, end);
3332                                 oldStart = end+1;
3333                         }
3334                         break;
3335                 }
3336                 KeyInfo &nextKey = getKeyInfo(nextKeyIdx);
3337
3338                 if ((nextKey.keytype == KeyInfo::isMain) && !nextKey.disabled) {
3339                         (void) dispatch(os, actual._dataStart, nextKey);
3340                         end = nextKey._tokenstart;
3341                         break;
3342                 }
3343                 processRegion(oldStart, nextKey._tokenstart);
3344                 nextKeyIdx = dispatch(os, actual._dataStart, nextKey);
3345
3346                 oldStart = nextKey._dataEnd+1;
3347         }
3348         // now nextKey is either invalid or is outside of actual._dataEnd
3349         // output the remaining and discard myself
3350         if (oldStart <= end) {
3351                 processRegion(oldStart, end);
3352         }
3353         if (interval_.par.size() > (size_t) end && interval_.par[end] == '}') {
3354                 end += 1;
3355                 // This is the normal case.
3356                 // But if using the firstlanguage, the closing may be missing
3357         }
3358         // get minimum of 'end' and  'actual._dataEnd' in case that the nextKey.keytype was 'KeyInfo::isMain'
3359         int output_end;
3360         if (actual._dataEnd < end)
3361                 output_end = interval_.nextNotIgnored(actual._dataEnd);
3362         else if (interval_.par.size() > (size_t) end)
3363                 output_end = interval_.nextNotIgnored(end);
3364         else
3365                 output_end = interval_.par.size();
3366         if ((actual.keytype == KeyInfo::isMain) && actual.disabled) {
3367                 interval_.addIntervall(actual._tokenstart, actual._tokenstart+actual._tokensize);
3368         }
3369         // Remove possible empty data
3370         int dstart = interval_.nextNotIgnored(actual._dataStart);
3371         while (interval_.isOpeningPar(dstart) == 1) {
3372                 interval_.addIntervall(dstart, dstart+1);
3373                 int dend = interval_.findclosing(dstart+1, output_end);
3374                 interval_.addIntervall(dend, dend+1);
3375                 dstart = interval_.nextNotIgnored(dstart+1);
3376         }
3377         if (dstart < output_end)
3378                 interval_.output(os, output_end);
3379         if (nextKeyIdx < 0)
3380                 interval_.addIntervall(0, end);
3381         else
3382                 interval_.addIntervall(actual._tokenstart, end);
3383         return nextKeyIdx;
3384 }
3385
3386 string splitOnKnownMacros(string par, bool isPatternString)
3387 {
3388         ostringstream os;
3389         LatexInfo li(par, isPatternString);
3390         // LYXERR(Debug::INFO, "Berfore split: " << par);
3391         KeyInfo DummyKey = KeyInfo(KeyInfo::KeyType::isMain, 2, true);
3392         DummyKey.head = "";
3393         DummyKey._tokensize = 0;
3394         DummyKey._dataStart = 0;
3395         DummyKey._dataEnd = par.length();
3396         DummyKey.disabled = true;
3397         int firstkeyIdx = li.getFirstKey();
3398         string s;
3399         if (firstkeyIdx >= 0) {
3400                 KeyInfo firstKey = li.getKeyInfo(firstkeyIdx);
3401                 DummyKey._tokenstart = firstKey._tokenstart;
3402                 int nextkeyIdx;
3403                 if ((firstKey.keytype != KeyInfo::isMain) || firstKey.disabled) {
3404                         // Use dummy firstKey
3405                         firstKey = DummyKey;
3406                         (void) li.setNextKey(firstkeyIdx);
3407                 }
3408                 else {
3409                         if (par.substr(firstKey._dataStart, 2) == "% ")
3410                                 li.addIntervall(firstKey._dataStart, firstKey._dataStart+2);
3411                 }
3412                 nextkeyIdx = li.process(os, firstKey);
3413                 while (nextkeyIdx >= 0) {
3414                         // Check for a possible gap between the last
3415                         // entry and this one
3416                         int datastart = li.nextNotIgnored(firstKey._dataStart);
3417                         KeyInfo &nextKey = li.getKeyInfo(nextkeyIdx);
3418                         if ((nextKey._tokenstart > datastart)) {
3419                                 // Handle the gap
3420                                 firstKey._dataStart = datastart;
3421                                 firstKey._dataEnd = par.length();
3422                                 (void) li.setNextKey(nextkeyIdx);
3423                                 // Fake the last opened parenthesis
3424                                 li.setForDefaultLang(firstKey);
3425                                 nextkeyIdx = li.process(os, firstKey);
3426                         }
3427                         else {
3428                                 if (nextKey.keytype != KeyInfo::isMain) {
3429                                         firstKey._dataStart = datastart;
3430                                         firstKey._dataEnd = nextKey._dataEnd+1;
3431                                         (void) li.setNextKey(nextkeyIdx);
3432                                         li.setForDefaultLang(firstKey);
3433                                         nextkeyIdx = li.process(os, firstKey);
3434                                 }
3435                                 else {
3436                                         nextkeyIdx = li.process(os, nextKey);
3437                                 }
3438                         }
3439                 }
3440                 // Handle the remaining
3441                 firstKey._dataStart = li.nextNotIgnored(firstKey._dataStart);
3442                 firstKey._dataEnd = par.length();
3443                 // Check if ! empty
3444                 if ((firstKey._dataStart < firstKey._dataEnd) &&
3445                                 (par[firstKey._dataStart] != '}')) {
3446                         li.setForDefaultLang(firstKey);
3447                         (void) li.process(os, firstKey);
3448                 }
3449                 s = os.str();
3450                 // return string definitelly impossible to match, but should be known
3451         }
3452         else
3453                 s = par;                            /* no known macros found */
3454         // LYXERR(Debug::INFO, "After split: " << s);
3455         return s;
3456 }
3457
3458 /*
3459  * Try to unify the language specs in the latexified text.
3460  * Resulting modified string is set to "", if
3461  * the searched tex does not contain all the features in the search pattern
3462  */
3463 static string correctlanguagesetting(string par, bool isPatternString, bool withformat, lyx::Buffer *pbuf = nullptr)
3464 {
3465         static Features regex_f;
3466         static int missed = 0;
3467         static bool regex_with_format = false;
3468
3469         int parlen = par.length();
3470
3471         while ((parlen > 0) && (par[parlen-1] == '\n')) {
3472                 parlen--;
3473         }
3474 #if 0
3475         if (isPatternString && (parlen > 0) && (par[parlen-1] == '~')) {
3476                 // Happens to be there in case of description or labeling environment
3477                 parlen--;
3478         }
3479 #endif
3480         string result;
3481         if (withformat) {
3482                 // Split the latex input into pieces which
3483                 // can be digested by our search engine
3484                 LYXERR(Debug::FINDVERBOSE, "input: \"" << par << "\"");
3485                 if (isPatternString && (pbuf != nullptr)) { // Check if we should disable/enable test for language
3486                         // We check for polyglossia, because in runparams.flavor we use Flavor::XeTeX
3487                         string doclang = pbuf->params().language->polyglossia();
3488                         static regex langre("\\\\(foreignlanguage)\\{([^\\}]+)\\}");
3489                         smatch sub;
3490                         bool toIgnoreLang = true;
3491                         for (sregex_iterator it(par.begin(), par.end(), langre), end; it != end; ++it) {
3492                                 sub = *it;
3493                                 if (sub.str(2) != doclang) {
3494                                         toIgnoreLang = false;
3495                                         break;
3496                                 }
3497                         }
3498                         setIgnoreFormat("language", toIgnoreLang, false);
3499
3500                 }
3501                 result = splitOnKnownMacros(par.substr(0,parlen), isPatternString);
3502                 LYXERR(Debug::FINDVERBOSE, "After splitOnKnownMacros:\n\"" << result << "\"");
3503         }
3504         else
3505                 result = par.substr(0, parlen);
3506         if (isPatternString) {
3507                 missed = 0;
3508                 if (withformat) {
3509                         regex_f = identifyFeatures(result);
3510                         string features = "";
3511                         for (auto it = regex_f.cbegin(); it != regex_f.cend(); ++it) {
3512                                 string a = it->first;
3513                                 regex_with_format = true;
3514                                 features += " " + a;
3515                                 // LYXERR(Debug::INFO, "Identified regex format:" << a);
3516                         }
3517                         LYXERR(Debug::FINDVERBOSE, "Identified Features" << features);
3518
3519                 }
3520         } else if (regex_with_format) {
3521                 Features info = identifyFeatures(result);
3522                 for (auto it = regex_f.cbegin(); it != regex_f.cend(); ++it) {
3523                         string a = it->first;
3524                         bool b = it->second;
3525                         if (b && ! info[a]) {
3526                                 missed++;
3527                                 LYXERR(Debug::FINDVERBOSE, "Missed(" << missed << " " << a <<", srclen = " << parlen );
3528                                 return "";
3529                         }
3530                 }
3531
3532         }
3533         else {
3534                 // LYXERR(Debug::INFO, "No regex formats");
3535         }
3536         return result;
3537 }
3538
3539
3540 // Remove trailing closure of math, macros and environments, so to catch parts of them.
3541 static void identifyClosing(string & t, bool ignoreformat)
3542 {
3543         do {
3544                 LYXERR(Debug::FINDVERBOSE, "identifyClosing(): t now is '" << t << "'");
3545                 if (regex_replace(t, t, "(.*[^\\\\])\\$$", "$1"))
3546                         continue;
3547                 if (regex_replace(t, t, "(.*[^\\\\])\\\\\\]$", "$1"))
3548                         continue;
3549                 if (regex_replace(t, t, "(.*[^\\\\])\\\\end\\{[a-zA-Z_]+\\*?\\}$", "$1"))
3550                         continue;
3551                 if (! ignoreformat) {
3552                         if (regex_replace(t, t, "(.*[^\\\\])\\}$", "$1"))
3553                                 continue;
3554                 }
3555                 break;
3556         } while (true);
3557         return;
3558 }
3559
3560 static int num_replaced = 0;
3561 static bool previous_single_replace = true;
3562
3563 void MatchStringAdv::CreateRegexp(FindAndReplaceOptions const & opt, string regexp_str, string regexp2_str, string par_as_string)
3564 {
3565 #if QTSEARCH
3566         if (regexp_str.empty() || regexp2_str.empty()) {
3567                 regexIsValid = false;
3568                 regexError = "Invalid empty regex";
3569                 return;
3570         }
3571         // Handle \w properly
3572         QRegularExpression::PatternOptions popts = QRegularExpression::UseUnicodePropertiesOption | QRegularExpression::MultilineOption;
3573         if (! opt.casesensitive) {
3574                 popts |= QRegularExpression::CaseInsensitiveOption;
3575         }
3576         regexp = QRegularExpression(QString::fromStdString(regexp_str), popts);
3577         regexp2 = QRegularExpression(QString::fromStdString(regexp2_str), popts);
3578         regexError = "";
3579         if (regexp.isValid() && regexp2.isValid()) {
3580                 regexIsValid = true;
3581                 // Check '{', '}' pairs inside the regex
3582                 int balanced = 0;
3583                 int skip = 1;
3584                 for (unsigned i = 0; i < par_as_string.size(); i+= skip) {
3585                         char c = par_as_string[i];
3586                         if (c == '\\') {
3587                                 skip = 2;
3588                                 continue;
3589                         }
3590                         if (c == '{')
3591                                 balanced++;
3592                         else if (c == '}') {
3593                                 balanced--;
3594                                 if (balanced < 0)
3595                                         break;
3596                         }
3597                         skip = 1;
3598                 }
3599                 if (balanced != 0) {
3600                         regexIsValid = false;
3601                         regexError = "Unbalanced curly brackets in regexp \"" + regexp_str + "\"";
3602                 }
3603         }
3604         else {
3605                 regexIsValid = false;
3606                 if (!regexp.isValid())
3607                         regexError += "Invalid regexp \"" + regexp_str + "\", error = " + regexp.errorString().toStdString();
3608                 else
3609                         regexError += "Invalid regexp2 \"" + regexp2_str + "\", error = " + regexp2.errorString().toStdString();
3610         }
3611 #else
3612         (void)par_as_string;
3613         if (opt.casesensitive) {
3614                 regexp = regex(regexp_str);
3615                 regexp2 = regex(regexp2_str);
3616         }
3617         else {
3618                 regexp = regex(regexp_str, std::regex_constants::icase);
3619                 regexp2 = regex(regexp2_str, std::regex_constants::icase);
3620         }
3621 #endif
3622 }
3623
3624 static void modifyRegexForMatchWord(string &t)
3625 {
3626         string s("");
3627         regex wordre("(\\\\)*((\\.|\\\\b))");
3628         size_t lastpos = 0;
3629         smatch sub;
3630         for (sregex_iterator it(t.begin(), t.end(), wordre), end; it != end; ++it) {
3631                 sub = *it;
3632                 if ((sub.position(2) - sub.position(0)) % 2 == 1) {
3633                         continue;
3634                 }
3635                 else if (sub.str(2) == "\\\\b")
3636                         return;
3637                 if (lastpos < (size_t) sub.position(2))
3638                         s += t.substr(lastpos, sub.position(2) - lastpos);
3639                 s += "\\S";
3640                 lastpos = sub.position(2) + sub.length(2);
3641         }
3642         if (lastpos == 0) {
3643                 s = "\\b" + t + "\\b";
3644                 t = s;
3645                 return;
3646         }
3647         else if (lastpos < t.length())
3648                 s += t.substr(lastpos, t.length() - lastpos);
3649         t = "\\b" + s + "\\b";
3650 }
3651
3652 MatchStringAdv::MatchStringAdv(lyx::Buffer & buf, FindAndReplaceOptions & opt)
3653         : p_buf(&buf), p_first_buf(&buf), opt(opt)
3654 {
3655         Buffer & find_buf = *theBufferList().getBuffer(FileName(to_utf8(opt.find_buf_name)), true);
3656         docstring const & ds = stringifySearchBuffer(find_buf, opt);
3657         if (ds.empty() ) {
3658                 CreateRegexp(opt, "", "", "");
3659                 return;
3660         }
3661         use_regexp = ds.find(from_utf8("\\regexp{")) != std::string::npos;
3662         if (opt.replace_all && previous_single_replace) {
3663                 previous_single_replace = false;
3664                 num_replaced = 0;
3665         }
3666         else if (!opt.replace_all) {
3667                 num_replaced = 0;       // count number of replaced strings
3668                 previous_single_replace = true;
3669         }
3670         // When using regexp, braces are hacked already by escape_for_regex()
3671         par_as_string = normalize(ds, opt.ignoreformat);
3672         open_braces = 0;
3673         close_wildcards = 0;
3674
3675         size_t lead_size = 0;
3676         // correct the language settings
3677         par_as_string = correctlanguagesetting(par_as_string, true, !opt.ignoreformat, &buf);
3678         if (par_as_string.empty()) {
3679                 CreateRegexp(opt, "", "", "");
3680                 return;
3681         }
3682         opt.matchAtStart = false;
3683         if (!use_regexp) {
3684                 identifyClosing(par_as_string, opt.ignoreformat); // Removes math closings ($, ], ...) at end of string
3685                 if (opt.ignoreformat) {
3686                         lead_size = 0;
3687                 }
3688                 else {
3689                         lead_size = identifyLeading(par_as_string);
3690                 }
3691                 lead_as_string = par_as_string.substr(0, lead_size);
3692                 string lead_as_regex_string = string2regex(lead_as_string);
3693                 par_as_string_nolead = par_as_string.substr(lead_size, par_as_string.size() - lead_size);
3694                 string par_as_regex_string_nolead = string2regex(par_as_string_nolead);
3695                 /* Handle whole words too in this case
3696                 */
3697                 if (opt.matchword) {
3698                         par_as_regex_string_nolead = "\\b" + par_as_regex_string_nolead + "\\b";
3699                         opt.matchword = false;
3700                 }
3701                 string regexp_str = "(" + lead_as_regex_string + ")()" + par_as_regex_string_nolead;
3702                 string regexp2_str = "(" + lead_as_regex_string + ")(.*?)" + par_as_regex_string_nolead;
3703                 CreateRegexp(opt, regexp_str, regexp2_str);
3704                 use_regexp = true;
3705                 LYXERR(Debug::FINDVERBOSE|Debug::FIND, "Setting regexp to : '" << regexp_str << "'");
3706                 LYXERR(Debug::FINDVERBOSE|Debug::FIND, "Setting regexp2 to: '" << regexp2_str << "'");
3707                 return;
3708         }
3709
3710         if (!opt.ignoreformat) {
3711                 lead_size = identifyLeading(par_as_string);
3712                 LYXERR(Debug::FINDVERBOSE, "Lead_size: " << lead_size);
3713                 lead_as_string = par_as_string.substr(0, lead_size);
3714                 par_as_string_nolead = par_as_string.substr(lead_size, par_as_string.size() - lead_size);
3715         }
3716
3717         // Here we are using regexp
3718         LASSERT(use_regexp, /**/);
3719         {
3720                 string lead_as_regexp;
3721                 if (lead_size > 0) {
3722                         lead_as_regexp = string2regex(par_as_string.substr(0, lead_size));
3723                         (void)regex_replace(par_as_string_nolead, par_as_string_nolead, "\\$$", "");
3724                         (void)regex_replace(par_as_string_nolead, par_as_string_nolead, "}$", "");
3725                         par_as_string = par_as_string_nolead;
3726                         LYXERR(Debug::FINDVERBOSE, "lead_as_regexp is '" << lead_as_regexp << "'");
3727                         LYXERR(Debug::FINDVERBOSE, "par_as_string now is '" << par_as_string << "'");
3728                 }
3729                 // LYXERR(Debug::FINDVERBOSE, "par_as_string before escape_for_regex() is '" << par_as_string << "'");
3730                 par_as_string = escape_for_regex(par_as_string, !opt.ignoreformat);
3731                 // Insert (.*?) before trailing closure of math, macros and environments, so to catch parts of them.
3732                 // LYXERR(Debug::FINDVERBOSE, "par_as_string now is '" << par_as_string << "'");
3733                 ++close_wildcards;
3734                 size_t lng = par_as_string.size();
3735                 if (!opt.ignoreformat) {
3736                         // Remove extra '\}' at end if not part of \{\.\}
3737                         while(lng > 2) {
3738                                 if (par_as_string.substr(lng-2, 2).compare("\\}") == 0) {
3739                                         if (lng >= 6) {
3740                                                 if (par_as_string.substr(lng-6,3).compare("\\{\\") == 0)
3741                                                         break;
3742                                         }
3743                                         lng -= 2;
3744                                         open_braces++;
3745                                 }
3746                                 else
3747                                         break;
3748                         }
3749                         if (lng < par_as_string.size())
3750                                 par_as_string.resize(lng);
3751                 }
3752                 LYXERR(Debug::FINDVERBOSE, "par_as_string after correctRegex is '" << par_as_string << "'");
3753                 if ((lng > 0) && (par_as_string[0] == '^')) {
3754                         par_as_string = par_as_string.substr(1);
3755                         --lng;
3756                         opt.matchAtStart = true;
3757                 }
3758                 // LYXERR(Debug::FINDVERBOSE, "par_as_string now is '" << par_as_string << "'");
3759                 // LYXERR(Debug::FINDVERBOSE, "Open braces: " << open_braces);
3760                 // LYXERR(Debug::FINDVERBOSE, "Replaced text (to be used as regex): " << par_as_string);
3761
3762                 // If entered regexp must match at begin of searched string buffer
3763                 // Kornel: Added parentheses to use $1 for size of the leading string
3764                 string regexp_str;
3765                 string regexp2_str;
3766                 {
3767                         // TODO: Adapt '\[12345678]' in par_as_string to acount for the first '()
3768                         // Unfortunately is '\1', '\2', etc not working for strings with extra format
3769                         // so the convert has no effect in that case
3770                         for (int i = 7; i > 0; --i) {
3771                                 string orig = "\\\\" + std::to_string(i);
3772                                 string dest = "\\" + std::to_string(i+2);
3773                                 while (regex_replace(par_as_string, par_as_string, orig, dest));
3774                         }
3775                         if (opt.matchword) {
3776                                 modifyRegexForMatchWord(par_as_string);
3777                                 opt.matchword = false;
3778                         }
3779                         regexp_str = "(" + lead_as_regexp + ")()" + par_as_string;
3780                         regexp2_str = "(" + lead_as_regexp + ")(.*?)" + par_as_string;
3781                 }
3782                 LYXERR(Debug::FINDVERBOSE|Debug::FIND, "Setting regexp to : '" << regexp_str << "'");
3783                 LYXERR(Debug::FINDVERBOSE|Debug::FIND, "Setting regexp2 to: '" << regexp2_str << "'");
3784                 CreateRegexp(opt, regexp_str, regexp2_str, par_as_string);
3785         }
3786 }
3787
3788 MatchResult MatchStringAdv::findAux(DocIterator const & cur, int len, MatchStringAdv::matchType at_begin) const
3789 {
3790         MatchResult mres;
3791
3792         mres.searched_size = len;
3793
3794         docstring docstr = stringifyFromForSearch(opt, cur, len);
3795         string str;
3796         str = normalize(docstr, opt.ignoreformat);
3797         if (!opt.ignoreformat) {
3798                 str = correctlanguagesetting(str, false, !opt.ignoreformat);
3799                 // remove closing '}' and '\n' to allow for use of '$' in regex
3800                 size_t lng = str.size();
3801                 while ((lng > 1) && ((str[lng -1] == '}') || (str[lng -1] == '\n')))
3802                         lng--;
3803                 if (lng != str.size()) {
3804                         str = str.substr(0, lng);
3805                 }
3806                 // Replace occurences of '~' to ' '
3807                 static std::regex specialChars { R"(~)" };
3808                 str = std::regex_replace(str, specialChars,  R"( )" );
3809         }
3810         if (str.empty()) {
3811                 mres.match_len = -1;
3812                 return mres;
3813         }
3814         LYXERR(Debug::FINDVERBOSE|Debug::FIND, "After normalization: Matching against:\n'" << str << "'");
3815
3816         LASSERT(use_regexp, /**/);
3817         {
3818                 // use_regexp always true
3819                 LYXERR(Debug::FINDVERBOSE, "Searching in regexp mode: at_begin=" << matchTypeAsString(at_begin));
3820 #if QTSEARCH
3821                 QString qstr = QString::fromStdString(str);
3822                 QRegularExpression const *p_regexp;
3823                 QRegularExpression::MatchType flags = QRegularExpression::NormalMatch;
3824                 if (at_begin == MatchStringAdv::MatchFromStart) {
3825                         p_regexp = &regexp;
3826                 } else {
3827                         p_regexp = &regexp2;
3828                 }
3829                 QRegularExpressionMatch match = p_regexp->match(qstr, 0, flags);
3830                 if (!match.hasMatch())
3831                         return mres;
3832 #else
3833                 regex const *p_regexp;
3834                 regex_constants::match_flag_type flags;
3835                 if (at_begin == MatchStringAdv::MatchFromStart) {
3836                         flags = regex_constants::match_continuous;
3837                         p_regexp = &regexp;
3838                 } else {
3839                         flags = regex_constants::match_default;
3840                         p_regexp = &regexp2;
3841                 }
3842                 sregex_iterator re_it(str.begin(), str.end(), *p_regexp, flags);
3843                 if (re_it == sregex_iterator())
3844                         return mres;
3845                 match_results<string::const_iterator> const & m = *re_it;
3846 #endif
3847                 // Whole found string, including the leading
3848                 // std: m[0].second - m[0].first
3849                 // Qt: match.capturedEnd(0) - match.capturedStart(0)
3850                 //
3851                 // Size of the leading string
3852                 // std: m[1].second - m[1].first
3853                 // Qt: match.capturedEnd(1) - match.capturedStart(1)
3854                 int leadingsize = 0;
3855 #if QTSEARCH
3856                 if (match.lastCapturedIndex() > 0) {
3857                         leadingsize = match.capturedEnd(1) - match.capturedStart(1);
3858                 }
3859
3860 #else
3861                 if (m.size() > 2) {
3862                         leadingsize = m[1].second - m[1].first;
3863                 }
3864 #endif
3865 #if QTSEARCH
3866                 mres.match_prefix = match.capturedEnd(2) - match.capturedStart(2);
3867                 mres.match_len = match.capturedEnd(0) - match.capturedEnd(2);
3868                 // because of different number of closing at end of string
3869                 // we have to 'unify' the length of the post-match.
3870                 // Done by ignoring closing parenthesis and linefeeds at string end
3871                 int matchend = match.capturedEnd(0);
3872                 size_t strsize = qstr.size();
3873                 if (!opt.ignoreformat) {
3874                         while (mres.match_len > 1) {
3875                                 QChar c = qstr.at(matchend - 1);
3876                                 if ((c == '\n') || (c == '}') || (c == '{')) {
3877                                         mres.match_len--;
3878                                         matchend--;
3879                                 }
3880                                 else
3881                                         break;
3882                         }
3883                         while (strsize > (size_t) match.capturedEnd(0)) {
3884                                 QChar c = qstr.at(strsize-1);
3885                                 if ((c == '\n') || (c == '}')) {
3886                                         --strsize;
3887                                 }
3888                                 else
3889                                         break;
3890                         }
3891                 }
3892                 // LYXERR0(qstr.toStdString());
3893                 mres.match2end = strsize - matchend;
3894                 mres.pos = match.capturedStart(2);
3895 #else
3896                 mres.match_prefix = m[2].second - m[2].first;
3897                 mres.match_len = m[0].second - m[2].second;
3898                 // ignore closing parenthesis and linefeeds at string end
3899                 size_t strend = m[0].second - m[0].first;
3900                 int matchend = strend;
3901                 size_t strsize = str.size();
3902                 if (!opt.ignoreformat) {
3903                         while (mres.match_len > 1) {
3904                                 char c = str.at(matchend - 1);
3905                                 if ((c == '\n') || (c == '}') || (c == '{')) {
3906                                         mres.match_len--;
3907                                         matchend--;
3908                                 }
3909                                 else
3910                                         break;
3911                         }
3912                         while (strsize > strend) {
3913                                 if ((str.at(strsize-1) == '}') || (str.at(strsize-1) == '\n')) {
3914                                         --strsize;
3915                                 }
3916                                 else
3917                                         break;
3918                         }
3919                 }
3920                 // LYXERR0(str);
3921                 mres.match2end = strsize - matchend;
3922                 mres.pos = m[2].first - m[0].first;;
3923 #endif
3924                 if (mres.match2end < 0)
3925                         mres.match_len = 0;
3926                 mres.leadsize = leadingsize;
3927 #if QTSEARCH
3928                 if (mres.match_len > 0) {
3929                         string a0 = match.captured(0).mid(mres.pos + mres.match_prefix, mres.match_len).toStdString();
3930                         mres.result.push_back(a0);
3931                         for (int i = 3; i <= match.lastCapturedIndex(); i++) {
3932                                 mres.result.push_back(match.captured(i).toStdString());
3933                         }
3934                 }
3935 #else
3936                 if (mres.match_len > 0) {
3937                         string a0 = m[0].str().substr(mres.pos + mres.match_prefix, mres.match_len);
3938                         mres.result.push_back(a0);
3939                         for (size_t i = 3; i < m.size(); i++) {
3940                                 mres.result.push_back(m[i]);
3941                         }
3942                 }
3943 #endif
3944                 return mres;
3945         }
3946 }
3947
3948
3949 MatchResult MatchStringAdv::operator()(DocIterator const & cur, int len, MatchStringAdv::matchType at_begin) const
3950 {
3951         MatchResult mres = findAux(cur, len, at_begin);
3952         int res = mres.match_len;
3953         LYXERR(Debug::FINDVERBOSE,
3954                "res=" << res << ", at_begin=" << matchTypeAsString(at_begin)
3955                << ", matchAtStart=" << opt.matchAtStart
3956                << ", inTexted=" << cur.inTexted());
3957         if (opt.matchAtStart) {
3958                 if (cur.pos() != 0)
3959                         mres.match_len = 0;
3960                 else if (mres.match_prefix > 0)
3961                         mres.match_len = 0;
3962                 return mres;
3963         }
3964         else
3965                 return mres;
3966 }
3967
3968 #if 0
3969 static bool simple_replace(string &t, string from, string to)
3970 {
3971         regex repl("(\\\\)*(" + from + ")");
3972         string s("");
3973         size_t lastpos = 0;
3974         smatch sub;
3975         for (sregex_iterator it(t.begin(), t.end(), repl), end; it != end; ++it) {
3976                 sub = *it;
3977                 if ((sub.position(2) - sub.position(0)) % 2 == 1)
3978                         continue;
3979                 if (lastpos < (size_t) sub.position(2))
3980                         s += t.substr(lastpos, sub.position(2) - lastpos);
3981                 s += to;
3982                 lastpos = sub.position(2) + sub.length(2);
3983         }
3984         if (lastpos == 0)
3985                 return false;
3986         else if (lastpos < t.length())
3987                 s += t.substr(lastpos, t.length() - lastpos);
3988         t = s;
3989         return true;
3990 }
3991 #endif
3992
3993 #if 1
3994 static string convertLF2Space(docstring const &s, bool ignore_format)
3995 {
3996         // Using original docstring to handle '\n'
3997
3998         if (s.size() == 0) return "";
3999         stringstream t;
4000         size_t pos;
4001         size_t start = 0;
4002         size_t end = s.size() - 1;
4003         if (!ignore_format) {
4004                 while (s[start] == '\n' && start <= end)
4005                         start++;
4006                 while (end >= start && s[end] == '\n')
4007                         end--;
4008                 if (start >= end + 1)
4009                         return "";
4010         }
4011         do {
4012                 bool dospace = true;
4013                 int skip = -1;
4014                 pos = s.find('\n', start);
4015                 if (pos >= end) {
4016                         t << lyx::to_utf8(s.substr(start, end + 1 - start));
4017                         break;
4018                 }
4019                 if (!ignore_format) {
4020                         if ((pos > start + 1) &&
4021                              s[pos-1] == '\\' &&
4022                              s[pos-2] == '\\') {
4023                                 skip = 2;
4024                                 if ((pos > start + 2) &&
4025                                     (s[pos+1] == '~' || isSpace(s[pos+1]) ||
4026                                      s[pos-3] == '~' || isSpace(s[pos-3]))) {
4027                                         // discard "\\\\\n", do not replace with space
4028                                         dospace = false;
4029                                 }
4030                         }
4031                         else if (pos > start) {
4032                                 if (s[pos-1] == '%') {
4033                                         skip = 1;
4034                                         while ((pos > start+skip) && (s[pos-1-skip] == '%'))
4035                                                 skip++;
4036                                         if ((pos > start+skip) &&
4037                                             (s[pos+1] == '~' || isSpace(s[pos+1]) ||
4038                                              s[pos-1-skip] == '~' || isSpace(s[pos-1-skip]))) {
4039                                                 // discard '%%%%%\n'
4040                                                 dospace = false;
4041                                         }
4042                                 }
4043                                 else if (!isAlnumASCII(s[pos+1]) || !isAlnumASCII(s[pos-1])) {
4044                                         dospace = false;
4045                                         skip = 0;       // remove the '\n' only
4046                                 }
4047                         }
4048                 }
4049                 else {
4050                         dospace = true;
4051                         skip = 0;
4052                 }
4053                 t << lyx::to_utf8(s.substr(start, pos-skip-start));
4054                 if (dospace)
4055                         t << ' ';
4056                 start = pos+1;
4057         } while (start <= end);
4058         return(t.str());
4059 }
4060
4061 #else
4062 static string convertLF2Space(docstring const & s, bool ignore_format)
4063 {
4064         // Using utf8-converted string to handle '\n'
4065
4066         string t;
4067         t = lyx::to_utf8(s);
4068         // Remove \n at begin
4069         while (!t.empty() && t[0] == '\n')
4070                 t = t.substr(1);
4071         // Remove \n* at end
4072         while (!t.empty() && t[t.size() - 1] == '\n') {
4073                 t = t.substr(0, t.size() - 1);
4074         }
4075         size_t pos;
4076         // Handle all other '\n'
4077         while ((pos = t.find("\n")) != string::npos) {
4078                 if (pos > 1 && t[pos-1] == '\\' && t[pos-2] == '\\' ) {
4079                         // Handle '\\\n'
4080                         if (isPrintableNonspace(t[pos+1]) && ((pos < 3) || isPrintableNonspace(t[pos-3]))) {
4081                                 t.replace(pos-2, 3, " ");
4082                         }
4083                         else {
4084                                 // Already a space there
4085                                 t.replace(pos-2, 3, "");
4086                         }
4087                 }
4088                 else {
4089                         if (!isAlnumASCII(t[pos+1]) || !isAlnumASCII(t[pos-1])) {
4090                                 // '\n' adjacent to non-alpha-numerics, discard
4091                                 t.replace(pos, 1, "");
4092                         }
4093                         else {
4094                                 // Replace all other \n with spaces
4095                                 t.replace(pos, 1, " ");
4096                         }
4097                         if (!ignore_format) {
4098                                 size_t count = 0;
4099                                 while ((pos > count + 1) && (t[pos - 1 -count] == '%')) {
4100                                         count++;
4101                                 }
4102                                 if (count > 0) {
4103                                         t.replace(pos - count, count, "");
4104                                 }
4105                         }
4106                 }
4107         }
4108         return(t);
4109
4110 }
4111 #endif
4112
4113 string MatchStringAdv::normalize(docstring const & s, bool ignore_format) const
4114 {
4115         string t = convertLF2Space(s, ignore_format);
4116
4117         // The following replaces are not appropriate in non-format-search mode
4118         if (!ignore_format) {
4119                 // Remove stale empty \emph{}, \textbf{} and similar blocks from latexify
4120                 // Kornel: Added textsl, textsf, textit, texttt and noun
4121                 // + allow to seach for colored text too
4122                 LYXERR(Debug::FINDVERBOSE, "Removing stale empty macros from: " << t);
4123                 while (regex_replace(t, t, "\\\\(emph|noun|text(bf|sl|sf|it|tt)|(u|uu)line|(s|x)out|uwave)(\\{(\\{\\})?\\})+", ""))
4124                         LYXERR(Debug::FINDVERBOSE, "  further removing stale empty \\emph{}, \\textbf{} macros from: " << t);
4125                 while (regex_replace(t, t, "\\\\((sub)?(((sub)?section)|paragraph)|part)\\*?(\\{(\\{\\})?\\})+", ""))
4126                         LYXERR(Debug::FINDVERBOSE, "  further removing stale empty \\section{}, \\part{}, \\paragraph{} macros from: " << t);
4127                 while (regex_replace(t, t, "\\\\(foreignlanguage|textcolor|item)\\{[a-z]+\\}(\\{(\\{\\})?\\})+", ""));
4128         }
4129         return t;
4130 }
4131
4132 docstring stringifyFromCursor(DocIterator const & cur, int len)
4133 {
4134         LYXERR(Debug::FINDVERBOSE, "Stringifying with len=" << len << " from cursor at pos: " << cur);
4135         if (cur.inTexted()) {
4136                 Paragraph const & par = cur.paragraph();
4137                 // TODO what about searching beyond/across paragraph breaks ?
4138                 // TODO Try adding a AS_STR_INSERTS as last arg
4139                 pos_type end = ( len == -1 || cur.pos() + len > int(par.size()) ) ?
4140                                         int(par.size()) : cur.pos() + len;
4141                 // OutputParams runparams(&cur.buffer()->params().encoding());
4142                 OutputParams runparams(encodings.fromLyXName("utf8"));
4143                 runparams.nice = true;
4144                 setFindParams(runparams);
4145                 int option = AS_STR_INSETS | AS_STR_PLAINTEXT;
4146                 if (ignoreFormats.getDeleted()) {
4147                         option |= AS_STR_SKIPDELETE;
4148                         runparams.find_set_feature(OutputParams::SearchWithoutDeleted);
4149                 }
4150                 else {
4151                         runparams.find_set_feature(OutputParams::SearchWithDeleted);
4152                 }
4153                 if (ignoreFormats.getNonContent()) {
4154                         runparams.find_add_feature(OutputParams::SearchNonOutput);
4155                 }
4156                 LYXERR(Debug::FINDVERBOSE, "Stringifying with cur: "
4157                        << cur << ", from pos: " << cur.pos() << ", end: " << end);
4158                 docstring res = from_utf8(latexNamesToUtf8(par.asString(cur.pos(), end,
4159                                                                         option,
4160                                                                         &runparams), false));
4161                 LYXERR(Debug::FINDVERBOSE|Debug::FIND, "Stringified text from pos(" << cur.pos() << ") len(" << len << "): " << res);
4162                 return res;
4163         } else if (cur.inMathed()) {
4164                 CursorSlice cs = cur.top();
4165                 MathData md = cs.cell();
4166                 MathData::const_iterator it_end =
4167                                 (( len == -1 || cs.pos() + len > int(md.size()))
4168                                  ? md.end()
4169                                  : md.begin() + cs.pos() + len );
4170                 MathData md2;
4171                 for (MathData::const_iterator it = md.begin() + cs.pos(); it != it_end; ++it)
4172                         md2.push_back(*it);
4173                 docstring res = from_utf8(latexNamesToUtf8(asString(md2), false));
4174                 LYXERR(Debug::FINDVERBOSE|Debug::FIND, "Stringified math from pos(" << cur.pos() << ") len(" << len << "): " << res);
4175                 return res;
4176         }
4177         LYXERR(Debug::FINDVERBOSE|Debug::FIND, "Don't know how to stringify from here: " << cur);
4178         return docstring();
4179 }
4180
4181 /** Computes the LaTeX export of buf starting from cur and ending len positions
4182  * after cur, if len is positive, or at the paragraph or innermost inset end
4183  * if len is -1.
4184  */
4185 docstring latexifyFromCursor(DocIterator const & cur, int len)
4186 {
4187         /*
4188         LYXERR(Debug::FINDVERBOSE, "Latexifying with len=" << len << " from cursor at pos: " << cur);
4189         LYXERR(Debug::FINDVERBOSE, "  with cur.lastpost=" << cur.lastpos() << ", cur.lastrow="
4190                << cur.lastrow() << ", cur.lastcol=" << cur.lastcol());
4191         */
4192         Buffer const & buf = *cur.buffer();
4193
4194         odocstringstream ods;
4195         otexstream os(ods);
4196         //OutputParams runparams(&buf.params().encoding());
4197         OutputParams runparams(encodings.fromLyXName("utf8"));
4198         runparams.nice = false;
4199         setFindParams(runparams);
4200         if (ignoreFormats.getDeleted()) {
4201                 runparams.find_set_feature(OutputParams::SearchWithoutDeleted);
4202         }
4203         else {
4204                 runparams.find_set_feature(OutputParams::SearchWithDeleted);
4205         }
4206         if (ignoreFormats.getNonContent()) {
4207                 runparams.find_add_feature(OutputParams::SearchNonOutput);
4208         }
4209
4210         if (cur.inTexted()) {
4211                 // @TODO what about searching beyond/across paragraph breaks ?
4212                 pos_type endpos = cur.paragraph().size();
4213                 if (len != -1 && endpos > cur.pos() + len)
4214                         endpos = cur.pos() + len;
4215                 TeXOnePar(buf, *cur.innerText(), cur.pit(), os, runparams,
4216                           string(), cur.pos(), endpos, true);
4217                 LYXERR(Debug::FINDVERBOSE|Debug::FIND, "Latexified text from pos(" << cur.pos() << ") len(" << len << "): " << ods.str());
4218                 return(ods.str());
4219         } else if (cur.inMathed()) {
4220                 // Retrieve the math environment type, and add '$' or '$[' or others (\begin{equation}) accordingly
4221                 for (int s = cur.depth() - 1; s >= 0; --s) {
4222                         CursorSlice const & cs = cur[s];
4223                         if (cs.asInsetMath() && cs.asInsetMath()->asHullInset()) {
4224                                 TeXMathStream ws(os);
4225                                 cs.asInsetMath()->asHullInset()->header_write(ws);
4226                                 break;
4227                         }
4228                 }
4229
4230                 CursorSlice const & cs = cur.top();
4231                 MathData md = cs.cell();
4232                 MathData::const_iterator it_end =
4233                                 ((len == -1 || cs.pos() + len > int(md.size()))
4234                                  ? md.end()
4235                                  : md.begin() + cs.pos() + len);
4236                 MathData md2;
4237                 for (MathData::const_iterator it = md.begin() + cs.pos();
4238                      it != it_end; ++it)
4239                         md2.push_back(*it);
4240
4241                 ods << asString(md2);
4242                 // Retrieve the math environment type, and add '$' or '$]'
4243                 // or others (\end{equation}) accordingly
4244                 for (int s = cur.depth() - 1; s >= 0; --s) {
4245                         CursorSlice const & cs2 = cur[s];
4246                         InsetMath * inset = cs2.asInsetMath();
4247                         if (inset && inset->asHullInset()) {
4248                                 TeXMathStream ws(os);
4249                                 inset->asHullInset()->footer_write(ws);
4250                                 break;
4251                         }
4252                 }
4253                 LYXERR(Debug::FINDVERBOSE|Debug::FIND, "Latexified math from pos(" << cur.pos() << ") len(" << len << "): " << ods.str());
4254         } else {
4255                 LYXERR(Debug::FINDVERBOSE|Debug::FIND, "Don't know how to stringify from here: " << cur);
4256         }
4257         return ods.str();
4258 }
4259
4260 #if defined(ResultsDebug)
4261 // Debugging output
4262 static void displayMResult(MatchResult &mres, string from, DocIterator & cur)
4263 {
4264         LYXERR0( "from:\t\t\t" << from);
4265         string status;
4266         if (mres.pos_len > 0) {
4267                 // Set in finalize
4268                 status = "FINALSEARCH";
4269         }
4270         else {
4271                 if (mres.match_len > 0) {
4272                         if ((mres.match_prefix == 0) && (mres.pos == mres.leadsize))
4273                                 status = "Good Match";
4274                         else
4275                                 status = "Matched in";
4276                 }
4277                 else
4278                         status = "MissedSearch";
4279         }
4280
4281         LYXERR0( status << "(" << cur.pos() << " ... " << mres.searched_size + cur.pos() << ") cur.lastpos(" << cur.lastpos() << ")");
4282         if ((mres.leadsize > 0) || (mres.match_len > 0) || (mres.match2end > 0))
4283                 LYXERR0( "leadsize(" << mres.leadsize << ") match_len(" << mres.match_len << ") match2end(" << mres.match2end << ")");
4284         if ((mres.pos > 0) || (mres.match_prefix > 0))
4285                 LYXERR0( "pos(" << mres.pos << ") match_prefix(" << mres.match_prefix << ")");
4286         for (size_t i = 0; i < mres.result.size(); i++)
4287                 LYXERR0( "Match " << i << " = \"" << mres.result[i] << "\"");
4288 }
4289 #define displayMres(s, txt, cur) displayMResult(s, txt, cur);
4290 #else
4291 #define displayMres(s, txt, cur)
4292 #endif
4293
4294 /** Finalize an advanced find operation, advancing the cursor to the innermost
4295  ** position that matches, plus computing the length of the matching text to
4296  ** be selected
4297  ** Return the cur.pos() difference between start and end of found match
4298  **/
4299 MatchResult findAdvFinalize(DocIterator & cur, MatchStringAdv const & match, MatchResult const & expected = MatchResult(-1))
4300 {
4301         // Search the foremost position that matches (avoids find of entire math
4302         // inset when match at start of it)
4303         DocIterator old_cur(cur.buffer());
4304         MatchResult mres;
4305         static MatchResult fail = MatchResult();
4306         MatchResult max_match;
4307         // If (prefix_len > 0) means that forwarding 1 position will remove the complete entry
4308         // Happens with e.g. hyperlinks
4309         // either one sees "http://www.bla.bla" or nothing
4310         // so the search for "www" gives prefix_len = 7 (== sizeof("http://")
4311         // and although we search for only 3 chars, we find the whole hyperlink inset
4312         MatchStringAdv::matchType at_begin = (expected.match_prefix == 0) ? MatchStringAdv::MatchFromStart : MatchStringAdv::MatchAnyPlace;
4313         if (!match.opt.forward && match.opt.ignoreformat) {
4314                 if (expected.pos > 0)
4315                         return fail;
4316         }
4317         LASSERT(at_begin == MatchStringAdv::MatchFromStart, /**/);
4318         if (expected.match_len > 0 && at_begin == MatchStringAdv::MatchFromStart) {
4319                 // Search for deepest match
4320                 old_cur = cur;
4321                 max_match = expected;
4322                 do {
4323                         size_t d = cur.depth();
4324                         cur.forwardPos();
4325                         if (!cur)
4326                                 break;
4327                         if (cur.depth() < d)
4328                                 break;
4329                         if (cur.depth() == d)
4330                                 break;
4331                         size_t lastd = d;
4332                         while (cur && cur.depth() > lastd) {
4333                                 lastd = cur.depth();
4334                                 mres = match(cur, -1, at_begin);
4335                                 displayMres(mres, "Checking innermost", cur);
4336                                 if (mres.match_len > 0)
4337                                         break;
4338                                 // maybe deeper?
4339                                 cur.forwardPos();
4340                         }
4341                         if (mres.match_len < expected.match_len)
4342                                 break;
4343                         max_match = mres;
4344                         old_cur = cur;;
4345                 } while(1);
4346                 cur = old_cur;
4347         }
4348         else {
4349                 // (expected.match_len <= 0)
4350                 mres = match(cur, -1, MatchStringAdv::MatchFromStart);      /* match valid only if not searching whole words */
4351                 displayMres(mres, "Start with negative match", cur);
4352                 max_match = mres;
4353         }
4354         // Only now we are really at_begin
4355         if ((max_match.match_len <= 0) ||
4356             (match.opt.restr == FindAndReplaceOptions::R_ONLY_MATHS && !cur.inMathed()))
4357                 return fail;
4358         LYXERR(Debug::FINDVERBOSE, "Ok");
4359
4360         // Compute the match length
4361         int len = 1;
4362         if (cur.pos() + len > cur.lastpos())
4363                 return fail;
4364
4365         LASSERT(match.use_regexp, /**/);
4366         {
4367                 int minl = 1;
4368                 int maxl = cur.lastpos() - cur.pos();
4369                 // Greedy behaviour while matching regexps
4370                 while (maxl > minl) {
4371                         MatchResult mres2;
4372                         mres2 = match(cur, len, at_begin);
4373                         displayMres(mres2, "Finalize loop", cur);
4374                         int actual_match_len = mres2.match_len;
4375                         if (actual_match_len >= max_match.match_len) {
4376                                 // actual_match_len > max_match _can_ happen,
4377                                 // if the search area splits
4378                                 // some following word so that the regex
4379                                 // (e.g. 'r.*r\b' matches 'r' from the middle of the
4380                                 // splitted word)
4381                                 // This means, the len value is too big
4382                                 actual_match_len = max_match.match_len;
4383                                 max_match = mres2;
4384                                 max_match.match_len = actual_match_len;
4385                                 maxl = len;
4386                                 if (maxl - minl < 4)
4387                                         len = (maxl + minl)/2;
4388                                 else
4389                                         len = minl + (maxl - minl + 3)/4;
4390                         }
4391                         else {
4392                                 // (actual_match_len < max_match.match_len)
4393                                 minl = len + 1;
4394                                 len = (maxl + minl)/2;
4395                         }
4396                 }
4397                 len = minl;
4398                 old_cur = cur;
4399                 // Search for real start of matched characters
4400                 while (len > 1) {
4401                         MatchResult actual_match;
4402                         do {
4403                                 cur.forwardPos();
4404                         } while (cur.depth() > old_cur.depth()); /* Skip inner insets */
4405                         if (cur.depth() < old_cur.depth()) {
4406                                 // Outer inset?
4407                                 LYXERR(Debug::INFO, "cur.depth() < old_cur.depth(), this should never happen");
4408                                 break;
4409                         }
4410                         if (cur.pos() != old_cur.pos()) {
4411                                 // OK, forwarded 1 pos in actual inset
4412                                 actual_match = match(cur, len-1, at_begin);
4413                                 if (actual_match.match_len == max_match.match_len) {
4414                                         // Ha, got it! The shorter selection has the same match length
4415                                         len--;
4416                                         old_cur = cur;
4417                                         max_match = actual_match;
4418                                 }
4419                                 else {
4420                                         // OK, the shorter selection matches less chars, revert to previous value
4421                                         cur = old_cur;
4422                                         break;
4423                                 }
4424                         }
4425                         else {
4426                                 LYXERR(Debug::INFO, "cur.pos() == old_cur.pos(), this should never happen");
4427                                 actual_match = match(cur, len, at_begin);
4428                                 if (actual_match.match_len == max_match.match_len) {
4429                                         old_cur = cur;
4430                                         max_match = actual_match;
4431                                 }
4432                         }
4433                 }
4434                 if (len == 0)
4435                         return fail;
4436                 else {
4437                         max_match.pos_len = len;
4438                         displayMres(max_match, "SEARCH RESULT", cur)
4439                                         return max_match;
4440                 }
4441         }
4442 }
4443
4444 /// Finds forward
4445 int findForwardAdv(DocIterator & cur, MatchStringAdv & match)
4446 {
4447         if (!cur)
4448                 return 0;
4449         bool repeat = false;
4450         DocIterator orig_cur;   // to be used if repeat not successful
4451         MatchResult orig_mres;
4452         while (!theApp()->longOperationCancelled() && cur) {
4453                 //(void) findAdvForwardInnermost(cur);
4454                 LYXERR(Debug::FINDVERBOSE, "findForwardAdv() cur: " << cur);
4455                 MatchResult mres = match(cur, -1, MatchStringAdv::MatchAnyPlace);
4456                 string msg = "Starting";
4457                 if (repeat)
4458                         msg = "Repeated";
4459                 displayMres(mres, msg + " findForwardAdv", cur)
4460                                 int match_len = mres.match_len;
4461                 if ((mres.pos > 100000) || (mres.match2end > 100000) || (match_len > 100000)) {
4462                         LYXERR(Debug::INFO, "BIG LENGTHS: " << mres.pos << ", " << match_len << ", " << mres.match2end);
4463                         match_len = 0;
4464                 }
4465                 if (match_len <= 0) {
4466                         // This should exit nested insets, if any, or otherwise undefine the currsor.
4467                         cur.pos() = cur.lastpos();
4468                         LYXERR(Debug::FINDVERBOSE, "Advancing pos: cur=" << cur);
4469                         cur.forwardPos();
4470                 }
4471                 else {  // match_len > 0
4472                         // Try to find the begin of searched string
4473                         int increment;
4474                         int firstInvalid = cur.lastpos() - cur.pos();
4475                         {
4476                                 int incrmatch = (mres.match_prefix + mres.pos - mres.leadsize + 1)*3/4;
4477                                 int incrcur = (firstInvalid + 1 )*3/4;
4478                                 if (incrcur < incrmatch)
4479                                         increment = incrcur;
4480                                 else
4481                                         increment = incrmatch;
4482                                 if (increment < 1)
4483                                         increment = 1;
4484                         }
4485                         LYXERR(Debug::FINDVERBOSE, "Set increment to " << increment);
4486                         while (increment > 0) {
4487                                 DocIterator old_cur = cur;
4488                                 if (cur.pos() + increment >= cur.lastpos()) {
4489                                         increment /= 2;
4490                                         continue;
4491                                 }
4492                                 cur.pos() = cur.pos() + increment;
4493                                 MatchResult mres2 = match(cur, -1, MatchStringAdv::MatchAnyPlace);
4494                                 displayMres(mres2, "findForwardAdv loop", cur)
4495                                                 switch (interpretMatch(mres, mres2)) {
4496                                         case MatchResult::newIsTooFar:
4497                                                 // behind the expected match
4498                                                 firstInvalid = increment;
4499                                                 cur = old_cur;
4500                                                 increment /= 2;
4501                                                 break;
4502                                         case MatchResult::newIsBetter:
4503                                                 // not reached yet, but cur.pos()+increment is bettert
4504                                                 mres = mres2;
4505                                                 firstInvalid -= increment;
4506                                                 if (increment > firstInvalid*3/4)
4507                                                         increment = firstInvalid*3/4;
4508                                                 if ((mres2.pos == mres2.leadsize) && (increment >= mres2.match_prefix)) {
4509                                                         if (increment >= mres2.match_prefix)
4510                                                                 increment = (mres2.match_prefix+1)*3/4;
4511                                                 }
4512                                                 break;
4513                                         default:
4514                                                 // Todo@
4515                                                 // Handle not like MatchResult::newIsTooFar
4516                                                 LYXERR0( "Probably too far: Increment = " << increment << " match_prefix = " << mres.match_prefix);
4517                                                 firstInvalid--;
4518                                                 increment = increment*3/4;
4519                                                 cur = old_cur;
4520                                                 break;
4521                                 }
4522                         }
4523                         if (mres.match_len > 0) {
4524                                 if (mres.match_prefix + mres.pos - mres.leadsize > 0) {
4525                                         // The match seems to indicate some deeper level
4526                                         repeat = true;
4527                                         orig_cur = cur;
4528                                         orig_mres = mres;
4529                                         cur.forwardPos();
4530                                         continue;
4531                                 }
4532                         }
4533                         else if (repeat) {
4534                                 // should never be reached.
4535                                 cur = orig_cur;
4536                                 mres = orig_mres;
4537                         }
4538                         // LYXERR0("Leaving first loop");
4539                         LYXERR(Debug::FINDVERBOSE, "Finalizing 1");
4540                         MatchResult found_match = findAdvFinalize(cur, match, mres);
4541                         if (found_match.match_len > 0) {
4542                                 match.FillResults(found_match);
4543                                 return found_match.pos_len;
4544                         }
4545                         else {
4546                                 // try next possible match
4547                                 cur.forwardPos();
4548                                 repeat = false;
4549                                 continue;
4550                         }
4551                 }
4552         }
4553         return 0;
4554 }
4555
4556
4557 /// Find the most backward consecutive match within same paragraph while searching backwards.
4558 MatchResult findMostBackwards(DocIterator & cur, MatchStringAdv const & match, MatchResult &expected)
4559 {
4560         DocIterator cur_begin = cur;
4561         cur_begin.pos() = 0;
4562         DocIterator tmp_cur = cur;
4563         MatchResult mr = findAdvFinalize(tmp_cur, match, expected);
4564         Inset & inset = cur.inset();
4565         for (; cur != cur_begin; cur.backwardPos()) {
4566                 LYXERR(Debug::FINDVERBOSE, "findMostBackwards(): cur=" << cur);
4567                 DocIterator new_cur = cur;
4568                 new_cur.backwardPos();
4569                 if (new_cur == cur || &new_cur.inset() != &inset
4570                     || match(new_cur, -1, MatchStringAdv::MatchFromStart).match_len <= 0)
4571                         break;
4572                 MatchResult new_mr = findAdvFinalize(new_cur, match, expected);
4573                 if (new_mr.match_len == mr.match_len)
4574                         break;
4575                 mr = new_mr;
4576         }
4577         LYXERR(Debug::FINDVERBOSE, "findMostBackwards(): exiting with cur=" << cur);
4578         return mr;
4579 }
4580
4581
4582 /// Finds backwards
4583 int findBackwardsAdv(DocIterator & cur, MatchStringAdv & match)
4584 {
4585         if (! cur)
4586                 return 0;
4587         // Backup of original position
4588         DocIterator cur_begin = doc_iterator_begin(cur.buffer());
4589         if (cur == cur_begin)
4590                 return 0;
4591         cur.backwardPos();
4592         DocIterator cur_orig(cur);
4593         bool pit_changed = false;
4594         do {
4595                 cur.pos() = 0;
4596                 MatchResult found_match = match(cur, -1, MatchStringAdv::MatchAnyPlace);
4597
4598                 if (found_match.match_len > 0) {
4599                         if (pit_changed)
4600                                 cur.pos() = cur.lastpos();
4601                         else
4602                                 cur.pos() = cur_orig.pos();
4603                         LYXERR(Debug::FINDVERBOSE, "findBackAdv2: cur: " << cur);
4604                         DocIterator cur_prev_iter;
4605                         do {
4606                                 found_match = match(cur, -1, MatchStringAdv::MatchFromStart);
4607                                 LYXERR(Debug::FINDVERBOSE, "findBackAdv3: found_match="
4608                                        << (found_match.match_len > 0) << ", cur: " << cur);
4609                                 if (found_match.match_len > 0) {
4610                                         MatchResult found_mr = findMostBackwards(cur, match, found_match);
4611                                         if (found_mr.pos_len > 0) {
4612                                                 match.FillResults(found_mr);
4613                                                 return found_mr.pos_len;
4614                                         }
4615                                 }
4616
4617                                 // Stop if begin of document reached
4618                                 if (cur == cur_begin)
4619                                         break;
4620                                 cur_prev_iter = cur;
4621                                 cur.backwardPos();
4622                         } while (true);
4623                 }
4624                 if (cur == cur_begin)
4625                         break;
4626                 if (cur.pit() > 0)
4627                         --cur.pit();
4628                 else
4629                         cur.backwardPos();
4630                 pit_changed = true;
4631         } while (!theApp()->longOperationCancelled());
4632         return 0;
4633 }
4634
4635
4636 } // namespace
4637
4638
4639 docstring stringifyFromForSearch(FindAndReplaceOptions const & opt,
4640                                  DocIterator const & cur, int len)
4641 {
4642         if (cur.pos() < 0 || cur.pos() > cur.lastpos())
4643                 return docstring();
4644         if (!opt.ignoreformat)
4645                 return latexifyFromCursor(cur, len);
4646         else
4647                 return stringifyFromCursor(cur, len);
4648 }
4649
4650
4651 FindAndReplaceOptions::FindAndReplaceOptions(
4652                 docstring const & _find_buf_name, bool _casesensitive,
4653                 bool _matchword, bool _forward, bool _expandmacros, bool _ignoreformat,
4654                 docstring const & _repl_buf_name, bool _keep_case,
4655                 SearchScope _scope, SearchRestriction _restr, bool _replace_all)
4656         : find_buf_name(_find_buf_name), casesensitive(_casesensitive), matchword(_matchword),
4657           forward(_forward), expandmacros(_expandmacros), ignoreformat(_ignoreformat),
4658           repl_buf_name(_repl_buf_name), keep_case(_keep_case), scope(_scope), restr(_restr), replace_all(_replace_all)
4659 {
4660 }
4661
4662
4663 namespace {
4664
4665
4666 /** Check if 'len' letters following cursor are all non-lowercase */
4667 static bool allNonLowercase(Cursor const & cur, int len)
4668 {
4669         pos_type beg_pos = cur.selectionBegin().pos();
4670         pos_type end_pos = cur.selectionBegin().pos() + len;
4671         if (len > cur.lastpos() + 1 - beg_pos) {
4672                 LYXERR(Debug::FINDVERBOSE, "This should not happen, more debug needed");
4673                 len = cur.lastpos() + 1 - beg_pos;
4674                 end_pos = beg_pos + len;
4675         }
4676         for (pos_type pos = beg_pos; pos != end_pos; ++pos)
4677                 if (isLowerCase(cur.paragraph().getChar(pos)))
4678                         return false;
4679         return true;
4680 }
4681
4682
4683 /** Check if first letter is upper case and second one is lower case */
4684 static bool firstUppercase(Cursor const & cur)
4685 {
4686         char_type ch1, ch2;
4687         pos_type pos = cur.selectionBegin().pos();
4688         if (pos >= cur.lastpos() - 1) {
4689                 LYXERR(Debug::FINDVERBOSE, "No upper-case at cur: " << cur);
4690                 return false;
4691         }
4692         ch1 = cur.paragraph().getChar(pos);
4693         ch2 = cur.paragraph().getChar(pos + 1);
4694         bool result = isUpperCase(ch1) && isLowerCase(ch2);
4695         LYXERR(Debug::FINDVERBOSE, "firstUppercase(): "
4696                << "ch1=" << ch1 << "(" << char(ch1) << "), ch2="
4697                << ch2 << "(" << char(ch2) << ")"
4698                << ", result=" << result << ", cur=" << cur);
4699         return result;
4700 }
4701
4702
4703 /** Make first letter of supplied buffer upper-case, and the rest lower-case.
4704  **
4705  ** \fixme What to do with possible further paragraphs in replace buffer ?
4706  **/
4707 static void changeFirstCase(Buffer & buffer, TextCase first_case, TextCase others_case)
4708 {
4709         ParagraphList::iterator pit = buffer.paragraphs().begin();
4710         LASSERT(!pit->empty(), /**/);
4711         pos_type right = pos_type(1);
4712         pit->changeCase(buffer.params(), pos_type(0), right, first_case);
4713         right = pit->size();
4714         pit->changeCase(buffer.params(), pos_type(1), right, others_case);
4715 }
4716 } // namespace
4717
4718 static bool replaceMatches(string &t, int maxmatchnum, vector <string> const & replacements)
4719 {
4720         // Should replace the string "$" + std::to_string(matchnum) with replacement
4721         // if the char '$' is not prefixed with odd number of char '\\'
4722         static regex const rematch("(\\\\)*(\\$\\$([0-9]))");
4723         string s;
4724         size_t lastpos = 0;
4725         smatch sub;
4726         for (sregex_iterator it(t.begin(), t.end(), rematch), end; it != end; ++it) {
4727                 sub = *it;
4728                 if ((sub.position(2) - sub.position(0)) % 2 == 1)
4729                         continue;
4730                 int num = stoi(sub.str(3), nullptr, 10);
4731                 if (num >= maxmatchnum)
4732                         continue;
4733                 if (lastpos < (size_t) sub.position(2))
4734                         s += t.substr(lastpos, sub.position(2) - lastpos);
4735                 s += replacements[num];
4736                 lastpos = sub.position(2) + sub.length(2);
4737         }
4738         if (lastpos == 0)
4739                 return false;
4740         else if (lastpos < t.length())
4741                 s += t.substr(lastpos, t.length() - lastpos);
4742         t = s;
4743         return true;
4744 }
4745
4746 ///
4747 static int findAdvReplace(BufferView * bv, FindAndReplaceOptions const & opt, MatchStringAdv & matchAdv)
4748 {
4749         Cursor & cur = bv->cursor();
4750         if (opt.repl_buf_name.empty()
4751                         || theBufferList().getBuffer(FileName(to_utf8(opt.repl_buf_name)), true) == 0
4752                         || theBufferList().getBuffer(FileName(to_utf8(opt.find_buf_name)), true) == 0)
4753                 return 0;
4754
4755         DocIterator sel_beg = cur.selectionBegin();
4756         DocIterator sel_end = cur.selectionEnd();
4757         if (&sel_beg.inset() != &sel_end.inset()
4758                         || sel_beg.pit() != sel_end.pit()
4759                         || sel_beg.idx() != sel_end.idx())
4760                 return 0;
4761         int sel_len = sel_end.pos() - sel_beg.pos();
4762         LYXERR(Debug::FINDVERBOSE, "sel_beg: " << sel_beg << ", sel_end: " << sel_end
4763                << ", sel_len: " << sel_len << endl);
4764         if (sel_len == 0)
4765                 return 0;
4766         LASSERT(sel_len > 0, return 0);
4767
4768         if (matchAdv(sel_beg, sel_len, MatchStringAdv::MatchFromStart).match_len <= 0)
4769                 return 0;
4770
4771         // Build a copy of the replace buffer, adapted to the KeepCase option
4772         Buffer const & repl_buffer_orig = *theBufferList().getBuffer(FileName(to_utf8(opt.repl_buf_name)), true);
4773         ostringstream oss;
4774         repl_buffer_orig.write(oss);
4775         string lyx = oss.str();
4776         if (matchAdv.valid_matches > 0)
4777                 replaceMatches(lyx, matchAdv.valid_matches, matchAdv.matches);
4778         Buffer repl_buffer(string(), false);
4779         repl_buffer.setInternal(true);
4780         repl_buffer.setUnnamed(true);
4781         LASSERT(repl_buffer.readString(lyx), return 0);
4782         if (opt.keep_case && sel_len >= 2) {
4783                 LYXERR(Debug::FINDVERBOSE, "keep_case true: cur.pos()=" << cur.pos() << ", sel_len=" << sel_len);
4784                 if (cur.inTexted()) {
4785                         if (firstUppercase(cur))
4786                                 changeFirstCase(repl_buffer, text_uppercase, text_lowercase);
4787                         else if (allNonLowercase(cur, sel_len))
4788                                 changeFirstCase(repl_buffer, text_uppercase, text_uppercase);
4789                 }
4790         }
4791         cap::cutSelection(cur, false);
4792         if (cur.inTexted()) {
4793                 repl_buffer.changeLanguage(
4794                                         repl_buffer.language(),
4795                                         cur.getFont().language());
4796                 LYXERR(Debug::FINDVERBOSE, "Replacing by pasteParagraphList()ing repl_buffer");
4797                 LYXERR(Debug::FINDVERBOSE, "Before pasteParagraphList() cur=" << cur << endl);
4798                 cap::pasteParagraphList(cur, repl_buffer.paragraphs(),
4799                                         repl_buffer.params().documentClassPtr(),
4800                                         repl_buffer.params().authors(),
4801                                         bv->buffer().errorList("Paste"));
4802                 LYXERR(Debug::FINDVERBOSE, "After pasteParagraphList() cur=" << cur << endl);
4803                 sel_len = repl_buffer.paragraphs().begin()->size();
4804         } else if (cur.inMathed()) {
4805                 odocstringstream ods;
4806                 otexstream os(ods);
4807                 // OutputParams runparams(&repl_buffer.params().encoding());
4808                 OutputParams runparams(encodings.fromLyXName("utf8"));
4809                 runparams.nice = false;
4810                 setFindParams(runparams);
4811                 TeXOnePar(repl_buffer, repl_buffer.text(), 0, os, runparams, string(), -1, -1, true);
4812                 //repl_buffer.getSourceCode(ods, 0, repl_buffer.paragraphs().size(), false);
4813                 docstring repl_latex = ods.str();
4814                 LYXERR(Debug::FINDVERBOSE, "Latexified replace_buffer: '" << repl_latex << "'");
4815                 string s;
4816                 (void)regex_replace(to_utf8(repl_latex), s, "\\$(.*)\\$", "$1");
4817                 (void)regex_replace(s, s, "\\\\\\[(.*)\\\\\\]", "$1");
4818                 repl_latex = from_utf8(s);
4819                 LYXERR(Debug::FINDVERBOSE, "Replacing by insert()ing latex: '" << repl_latex << "' cur=" << cur << " with depth=" << cur.depth());
4820                 MathData ar(cur.buffer());
4821                 asArray(repl_latex, ar, Parse::NORMAL);
4822                 cur.insert(ar);
4823                 sel_len = ar.size();
4824                 LYXERR(Debug::FINDVERBOSE, "After insert() cur=" << cur << " with depth: " << cur.depth() << " and len: " << sel_len);
4825         }
4826         if (cur.pos() >= sel_len)
4827                 cur.pos() -= sel_len;
4828         else
4829                 cur.pos() = 0;
4830         LYXERR(Debug::FINDVERBOSE, "After pos adj cur=" << cur << " with depth: " << cur.depth() << " and len: " << sel_len);
4831         bv->putSelectionAt(DocIterator(cur), sel_len, !opt.forward);
4832         bv->processUpdateFlags(Update::Force);
4833         return 1;
4834 }
4835
4836
4837 /// Perform a FindAdv operation.
4838 bool findAdv(BufferView * bv, FindAndReplaceOptions & opt)
4839 {
4840         DocIterator cur;
4841         int pos_len = 0;
4842
4843         // e.g., when invoking word-findadv from mini-buffer wither with
4844         //       wrong options syntax or before ever opening advanced F&R pane
4845         if (theBufferList().getBuffer(FileName(to_utf8(opt.find_buf_name)), true) == 0)
4846                 return false;
4847
4848         try {
4849                 MatchStringAdv matchAdv(bv->buffer(), opt);
4850 #if QTSEARCH
4851                 if (!matchAdv.regexIsValid) {
4852                         bv->message(lyx::from_utf8(matchAdv.regexError));
4853                         return(false);
4854                 }
4855 #endif
4856                 int length = bv->cursor().selectionEnd().pos() - bv->cursor().selectionBegin().pos();
4857                 if (length > 0)
4858                         bv->putSelectionAt(bv->cursor().selectionBegin(), length, !opt.forward);
4859                 num_replaced += findAdvReplace(bv, opt, matchAdv);
4860                 cur = bv->cursor();
4861                 if (opt.forward)
4862                         pos_len = findForwardAdv(cur, matchAdv);
4863                 else
4864                         pos_len = findBackwardsAdv(cur, matchAdv);
4865         } catch (exception & ex) {
4866                 bv->message(from_utf8(ex.what()));
4867                 return false;
4868         }
4869
4870         if (pos_len == 0) {
4871                 if (num_replaced > 0) {
4872                         switch (num_replaced)
4873                         {
4874                         case 1:
4875                                 bv->message(_("One match has been replaced."));
4876                                 break;
4877                         case 2:
4878                                 bv->message(_("Two matches have been replaced."));
4879                                 break;
4880                         default:
4881                                 bv->message(bformat(_("%1$d matches have been replaced."), num_replaced));
4882                                 break;
4883                         }
4884                         num_replaced = 0;
4885                 }
4886                 else {
4887                         bv->message(_("Match not found."));
4888                 }
4889                 return false;
4890         }
4891
4892         if (num_replaced > 0)
4893                 bv->message(_("Match has been replaced."));
4894         else
4895                 bv->message(_("Match found."));
4896
4897         if (cur.pos() + pos_len > cur.lastpos()) {
4898                 // Prevent crash in bv->putSelectionAt()
4899                 // Should never happen, maybe LASSERT() here?
4900                 pos_len = cur.lastpos() - cur.pos();
4901         }
4902         LYXERR(Debug::FINDVERBOSE|Debug::FIND, "Putting selection at cur=" << cur << " with len: " << pos_len);
4903         bv->putSelectionAt(cur, pos_len, !opt.forward);
4904
4905         return true;
4906 }
4907
4908
4909 ostringstream & operator<<(ostringstream & os, FindAndReplaceOptions const & opt)
4910 {
4911         os << to_utf8(opt.find_buf_name) << "\nEOSS\n"
4912            << opt.casesensitive << ' '
4913            << opt.matchword << ' '
4914            << opt.forward << ' '
4915            << opt.expandmacros << ' '
4916            << opt.ignoreformat << ' '
4917            << opt.replace_all << ' '
4918            << to_utf8(opt.repl_buf_name) << "\nEOSS\n"
4919            << opt.keep_case << ' '
4920            << int(opt.scope) << ' '
4921            << int(opt.restr);
4922
4923         LYXERR(Debug::FINDVERBOSE, "built: " << os.str());
4924
4925         return os;
4926 }
4927
4928
4929 istringstream & operator>>(istringstream & is, FindAndReplaceOptions & opt)
4930 {
4931         // LYXERR(Debug::FINDVERBOSE, "parsing");
4932         string s;
4933         string line;
4934         getline(is, line);
4935         while (line != "EOSS") {
4936                 if (! s.empty())
4937                         s = s + "\n";
4938                 s = s + line;
4939                 if (is.eof())   // Tolerate malformed request
4940                         break;
4941                 getline(is, line);
4942         }
4943         // LYXERR(Debug::FINDVERBOSE, "file_buf_name: '" << s << "'");
4944         opt.find_buf_name = from_utf8(s);
4945         is >> opt.casesensitive >> opt.matchword >> opt.forward >> opt.expandmacros >> opt.ignoreformat >> opt.replace_all;
4946         is.get();       // Waste space before replace string
4947         s = "";
4948         getline(is, line);
4949         while (line != "EOSS") {
4950                 if (! s.empty())
4951                         s = s + "\n";
4952                 s = s + line;
4953                 if (is.eof())   // Tolerate malformed request
4954                         break;
4955                 getline(is, line);
4956         }
4957         // LYXERR(Debug::FINDVERBOSE, "repl_buf_name: '" << s << "'");
4958         opt.repl_buf_name = from_utf8(s);
4959         is >> opt.keep_case;
4960         int i;
4961         is >> i;
4962         opt.scope = FindAndReplaceOptions::SearchScope(i);
4963         is >> i;
4964         opt.restr = FindAndReplaceOptions::SearchRestriction(i);
4965
4966         /*
4967         LYXERR(Debug::FINDVERBOSE, "parsed: " << opt.casesensitive << ' ' << opt.matchword << ' ' << opt.forward << ' '
4968                << opt.expandmacros << ' ' << opt.ignoreformat << ' ' << opt.keep_case << ' '
4969                << opt.scope << ' ' << opt.restr);
4970         */
4971         return is;
4972 }
4973
4974 } // namespace lyx