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