src/common/regex.cpp

   1 ///////////////////////////////////////////////////////////////////////////////
   2 // Name:        src/common/regex.cpp
   3 // Purpose:     regular expression matching
   4 // Author:      Karsten Ballüder and Vadim Zeitlin
   5 // Modified by:
   6 // Created:     13.07.01
   7 // RCS-ID:      $Id$
   8 // Copyright:   (c) 2000 Karsten Ballüder <ballueder@gmx.net>
   9 //                  2001 Vadim Zeitlin <vadim@wxwindows.org>
  10 // Licence:     wxWindows licence
  11 ///////////////////////////////////////////////////////////////////////////////
  12
  13 // ============================================================================
  14 // declarations
  15 // ============================================================================
  16
  17 // ----------------------------------------------------------------------------
  18 // headers
  19 // ----------------------------------------------------------------------------
  20
  21 // For compilers that support precompilation, includes "wx.h".
  22 #include "wx/wxprec.h"
  23
  24 #ifdef __BORLANDC__
  25     #pragma hdrstop
  26 #endif
  27
  28 #if wxUSE_REGEX
  29
  30 #ifndef WX_PRECOMP
  31     #include "wx/object.h"
  32     #include "wx/string.h"
  33     #include "wx/log.h"
  34     #include "wx/intl.h"
  35 #endif //WX_PRECOMP
  36
  37 // FreeBSD, Watcom and DMars require this, CW doesn't have nor need it.
  38 // Others also don't seem to need it. If you have an error related to
  39 // (not) including <sys/types.h> please report details to
  40 // wx-dev@lists.wxwindows.org
  41 #if defined(__UNIX__) || defined(__WATCOMC__) || defined(__DIGITALMARS__)
  42 #   include <sys/types.h>
  43 #endif
  44
  45 #include <regex.h>
  46 #include "wx/regex.h"
  47
  48 // WXREGEX_USING_BUILTIN    defined when using the built-in regex lib
  49 // WXREGEX_USING_RE_SEARCH  defined when using re_search in the GNU regex lib
  50 // WXREGEX_IF_NEED_LEN()    wrap the len parameter only used with the built-in
  51 //                          or GNU regex
  52 // WXREGEX_CONVERT_TO_MB    defined when the regex lib is using chars and
  53 //                          wxChar is wide, so conversion must be done
  54 // WXREGEX_CHAR(x)          Convert wxChar to wxRegChar
  55 //
  56 #ifdef __REG_NOFRONT
  57 #   define WXREGEX_USING_BUILTIN
  58 #   define WXREGEX_IF_NEED_LEN(x) ,x
  59 #   if wxUSE_UNICODE
  60 #       define WXREGEX_CHAR(x) x.wc_str()
  61 #   else
  62 #       define WXREGEX_CHAR(x) x.mb_str()
  63 #   endif
  64 #else
  65 #   ifdef HAVE_RE_SEARCH
  66 #       define WXREGEX_IF_NEED_LEN(x) ,x
  67 #       define WXREGEX_USING_RE_SEARCH
  68 #   else
  69 #       define WXREGEX_IF_NEED_LEN(x)
  70 #   endif
  71 #   if wxUSE_UNICODE
  72 #       define WXREGEX_CONVERT_TO_MB
  73 #   endif
  74 #   define WXREGEX_CHAR(x) x.mb_str()
  75 #   define wx_regfree regfree
  76 #   define wx_regerror regerror
  77 #endif
  78
  79 // ----------------------------------------------------------------------------
  80 // private classes
  81 // ----------------------------------------------------------------------------
  82
  83 #ifndef WXREGEX_USING_RE_SEARCH
  84
  85 // the array of offsets for the matches, the usual POSIX regmatch_t array.
  86 class wxRegExMatches
  87 {
  88 public:
  89     typedef regmatch_t *match_type;
  90
  91     wxRegExMatches(size_t n)        { m_matches = new regmatch_t[n]; }
  92     ~wxRegExMatches()               { delete [] m_matches; }
  93
  94     // we just use casts here because the fields of regmatch_t struct may be 64
  95     // bit but we're limited to size_t in our public API and are not going to
  96     // change it because operating on strings longer than 4GB using it is
  97     // absolutely impractical anyhow
  98     size_t Start(size_t n) const
  99     {
 100         return wx_truncate_cast(size_t, m_matches[n].rm_so);
 101     }
 102
 103     size_t End(size_t n) const
 104     {
 105         return wx_truncate_cast(size_t, m_matches[n].rm_eo);
 106     }
 107
 108     regmatch_t *get() const         { return m_matches; }
 109
 110 private:
 111     regmatch_t *m_matches;
 112 };
 113
 114 #else // WXREGEX_USING_RE_SEARCH
 115
 116 // the array of offsets for the matches, the struct used by the GNU lib
 117 class wxRegExMatches
 118 {
 119 public:
 120     typedef re_registers *match_type;
 121
 122     wxRegExMatches(size_t n)
 123     {
 124         m_matches.num_regs = n;
 125         m_matches.start = new regoff_t[n];
 126         m_matches.end = new regoff_t[n];
 127     }
 128
 129     ~wxRegExMatches()
 130     {
 131         delete [] m_matches.start;
 132         delete [] m_matches.end;
 133     }
 134
 135     size_t Start(size_t n) const    { return m_matches.start[n]; }
 136     size_t End(size_t n) const      { return m_matches.end[n]; }
 137
 138     re_registers *get()             { return &m_matches; }
 139
 140 private:
 141     re_registers m_matches;
 142 };
 143
 144 #endif // WXREGEX_USING_RE_SEARCH
 145
 146 // the character type used by the regular expression engine
 147 #ifndef WXREGEX_CONVERT_TO_MB
 148 typedef wxChar wxRegChar;
 149 #else
 150 typedef char wxRegChar;
 151 #endif
 152
 153 // the real implementation of wxRegEx
 154 class wxRegExImpl
 155 {
 156 public:
 157     // ctor and dtor
 158     wxRegExImpl();
 159     ~wxRegExImpl();
 160
 161     // return true if Compile() had been called successfully
 162     bool IsValid() const { return m_isCompiled; }
 163
 164     // RE operations
 165     bool Compile(const wxString& expr, int flags = 0);
 166     bool Matches(const wxRegChar *str, int flags
 167                  WXREGEX_IF_NEED_LEN(size_t len)) const;
 168     bool GetMatch(size_t *start, size_t *len, size_t index = 0) const;
 169     size_t GetMatchCount() const;
 170     int Replace(wxString *pattern, const wxString& replacement,
 171                 size_t maxMatches = 0) const;
 172
 173 private:
 174     // return the string containing the error message for the given err code
 175     wxString GetErrorMsg(int errorcode, bool badconv) const;
 176
 177     // init the members
 178     void Init()
 179     {
 180         m_isCompiled = false;
 181         m_Matches = NULL;
 182         m_nMatches = 0;
 183     }
 184
 185     // free the RE if compiled
 186     void Free()
 187     {
 188         if ( IsValid() )
 189         {
 190             wx_regfree(&m_RegEx);
 191         }
 192
 193         delete m_Matches;
 194     }
 195
 196     // free the RE if any and reinit the members
 197     void Reinit()
 198     {
 199         Free();
 200         Init();
 201     }
 202
 203     // compiled RE
 204     regex_t         m_RegEx;
 205
 206     // the subexpressions data
 207     wxRegExMatches *m_Matches;
 208     size_t          m_nMatches;
 209
 210     // true if m_RegEx is valid
 211     bool            m_isCompiled;
 212 };
 213
 214
 215 // ============================================================================
 216 // implementation
 217 // ============================================================================
 218
 219 // ----------------------------------------------------------------------------
 220 // wxRegExImpl
 221 // ----------------------------------------------------------------------------
 222
 223 wxRegExImpl::wxRegExImpl()
 224 {
 225     Init();
 226 }
 227
 228 wxRegExImpl::~wxRegExImpl()
 229 {
 230     Free();
 231 }
 232
 233 wxString wxRegExImpl::GetErrorMsg(int errorcode, bool badconv) const
 234 {
 235 #ifdef WXREGEX_CONVERT_TO_MB
 236     // currently only needed when using system library in Unicode mode
 237     if ( badconv )
 238     {
 239         return _("conversion to 8-bit encoding failed");
 240     }
 241 #else
 242     // 'use' badconv to avoid a compiler warning
 243     (void)badconv;
 244 #endif
 245
 246     wxString szError;
 247
 248     // first get the string length needed
 249     int len = wx_regerror(errorcode, &m_RegEx, NULL, 0);
 250     if ( len > 0 )
 251     {
 252         char* szcmbError = new char[++len];
 253
 254         (void)wx_regerror(errorcode, &m_RegEx, szcmbError, len);
 255
 256         szError = wxConvLibc.cMB2WX(szcmbError);
 257         delete [] szcmbError;
 258     }
 259     else // regerror() returned 0
 260     {
 261         szError = _("unknown error");
 262     }
 263
 264     return szError;
 265 }
 266
 267 bool wxRegExImpl::Compile(const wxString& expr, int flags)
 268 {
 269     Reinit();
 270
 271 #ifdef WX_NO_REGEX_ADVANCED
 272 #   define FLAVORS wxRE_BASIC
 273 #else
 274 #   define FLAVORS (wxRE_ADVANCED | wxRE_BASIC)
 275     wxASSERT_MSG( (flags & FLAVORS) != FLAVORS,
 276                   _T("incompatible flags in wxRegEx::Compile") );
 277 #endif
 278     wxASSERT_MSG( !(flags & ~(FLAVORS | wxRE_ICASE | wxRE_NOSUB | wxRE_NEWLINE)),
 279                   _T("unrecognized flags in wxRegEx::Compile") );
 280
 281     // translate our flags to regcomp() ones
 282     int flagsRE = 0;
 283     if ( !(flags & wxRE_BASIC) )
 284 #ifndef WX_NO_REGEX_ADVANCED
 285         if (flags & wxRE_ADVANCED)
 286             flagsRE |= REG_ADVANCED;
 287         else
 288 #endif
 289             flagsRE |= REG_EXTENDED;
 290     if ( flags & wxRE_ICASE )
 291         flagsRE |= REG_ICASE;
 292     if ( flags & wxRE_NOSUB )
 293         flagsRE |= REG_NOSUB;
 294     if ( flags & wxRE_NEWLINE )
 295         flagsRE |= REG_NEWLINE;
 296
 297     // compile it
 298 #ifdef WXREGEX_USING_BUILTIN
 299     bool conv = true;
 300     // FIXME-UTF8: use wc_str() after removing ANSI build
 301     int errorcode = wx_re_comp(&m_RegEx, expr.c_str(), expr.length(), flagsRE);
 302 #else
 303     // FIXME-UTF8: this is potentially broken, we shouldn't even try it
 304     //             and should always use builtin regex library (or PCRE?)
 305     const wxWX2MBbuf conv = expr.mbc_str();
 306     int errorcode = conv ? regcomp(&m_RegEx, conv, flagsRE) : REG_BADPAT;
 307 #endif
 308
 309     if ( errorcode )
 310     {
 311         wxLogError(_("Invalid regular expression '%s': %s"),
 312                    expr.c_str(), GetErrorMsg(errorcode, !conv).c_str());
 313
 314         m_isCompiled = false;
 315     }
 316     else // ok
 317     {
 318         // don't allocate the matches array now, but do it later if necessary
 319         if ( flags & wxRE_NOSUB )
 320         {
 321             // we don't need it at all
 322             m_nMatches = 0;
 323         }
 324         else
 325         {
 326             // we will alloc the array later (only if really needed) but count
 327             // the number of sub-expressions in the regex right now
 328
 329             // there is always one for the whole expression
 330             m_nMatches = 1;
 331
 332             // and some more for bracketed subexperessions
 333             for ( const wxChar *cptr = expr.c_str(); *cptr; cptr++ )
 334             {
 335                 if ( *cptr == _T('\\') )
 336                 {
 337                     // in basic RE syntax groups are inside \(...\)
 338                     if ( *++cptr == _T('(') && (flags & wxRE_BASIC) )
 339                     {
 340                         m_nMatches++;
 341                     }
 342                 }
 343                 else if ( *cptr == _T('(') && !(flags & wxRE_BASIC) )
 344                 {
 345                     // we know that the previous character is not an unquoted
 346                     // backslash because it would have been eaten above, so we
 347                     // have a bare '(' and this indicates a group start for the
 348                     // extended syntax. '(?' is used for extensions by perl-
 349                     // like REs (e.g. advanced), and is not valid for POSIX
 350                     // extended, so ignore them always.
 351                     if ( cptr[1] != _T('?') )
 352                         m_nMatches++;
 353                 }
 354             }
 355         }
 356
 357         m_isCompiled = true;
 358     }
 359
 360     return IsValid();
 361 }
 362
 363 #ifdef WXREGEX_USING_RE_SEARCH
 364
 365 // On GNU, regexec is implemented as a wrapper around re_search. re_search
 366 // requires a length parameter which the POSIX regexec does not have,
 367 // therefore regexec must do a strlen on the search text each time it is
 368 // called. This can drastically affect performance when matching is done in
 369 // a loop along a string, such as during a search and replace. Therefore if
 370 // re_search is detected by configure, it is used directly.
 371 //
 372 static int ReSearch(const regex_t *preg,
 373                     const char *text,
 374                     size_t len,
 375                     re_registers *matches,
 376                     int eflags)
 377 {
 378     regex_t *pattern = wx_const_cast(regex_t*, preg);
 379
 380     pattern->not_bol = (eflags & REG_NOTBOL) != 0;
 381     pattern->not_eol = (eflags & REG_NOTEOL) != 0;
 382     pattern->regs_allocated = REGS_FIXED;
 383
 384     int ret = re_search(pattern, text, len, 0, len, matches);
 385     return ret >= 0 ? 0 : REG_NOMATCH;
 386 }
 387
 388 #endif // WXREGEX_USING_RE_SEARCH
 389
 390 bool wxRegExImpl::Matches(const wxRegChar *str,
 391                           int flags
 392                           WXREGEX_IF_NEED_LEN(size_t len)) const
 393 {
 394     wxCHECK_MSG( IsValid(), false, _T("must successfully Compile() first") );
 395
 396     // translate our flags to regexec() ones
 397     wxASSERT_MSG( !(flags & ~(wxRE_NOTBOL | wxRE_NOTEOL)),
 398                   _T("unrecognized flags in wxRegEx::Matches") );
 399
 400     int flagsRE = 0;
 401     if ( flags & wxRE_NOTBOL )
 402         flagsRE |= REG_NOTBOL;
 403     if ( flags & wxRE_NOTEOL )
 404         flagsRE |= REG_NOTEOL;
 405
 406     // allocate matches array if needed
 407     wxRegExImpl *self = wxConstCast(this, wxRegExImpl);
 408     if ( !m_Matches && m_nMatches )
 409     {
 410         self->m_Matches = new wxRegExMatches(m_nMatches);
 411     }
 412
 413     wxRegExMatches::match_type matches = m_Matches ? m_Matches->get() : NULL;
 414
 415     // do match it
 416 #if defined WXREGEX_USING_BUILTIN
 417     int rc = wx_re_exec(&self->m_RegEx, str, len, NULL, m_nMatches, matches, flagsRE);
 418 #elif defined WXREGEX_USING_RE_SEARCH
 419     int rc = str ? ReSearch(&self->m_RegEx, str, len, matches, flagsRE) : REG_BADPAT;
 420 #else
 421     int rc = str ? regexec(&self->m_RegEx, str, m_nMatches, matches, flagsRE) : REG_BADPAT;
 422 #endif
 423
 424     switch ( rc )
 425     {
 426         case 0:
 427             // matched successfully
 428             return true;
 429
 430         default:
 431             // an error occurred
 432             wxLogError(_("Failed to find match for regular expression: %s"),
 433                        GetErrorMsg(rc, !str).c_str());
 434             // fall through
 435
 436         case REG_NOMATCH:
 437             // no match
 438             return false;
 439     }
 440 }
 441
 442 bool wxRegExImpl::GetMatch(size_t *start, size_t *len, size_t index) const
 443 {
 444     wxCHECK_MSG( IsValid(), false, _T("must successfully Compile() first") );
 445     wxCHECK_MSG( m_nMatches, false, _T("can't use with wxRE_NOSUB") );
 446     wxCHECK_MSG( m_Matches, false, _T("must call Matches() first") );
 447     wxCHECK_MSG( index < m_nMatches, false, _T("invalid match index") );
 448
 449     if ( start )
 450         *start = m_Matches->Start(index);
 451     if ( len )
 452         *len = m_Matches->End(index) - m_Matches->Start(index);
 453
 454     return true;
 455 }
 456
 457 size_t wxRegExImpl::GetMatchCount() const
 458 {
 459     wxCHECK_MSG( IsValid(), 0, _T("must successfully Compile() first") );
 460     wxCHECK_MSG( m_nMatches, 0, _T("can't use with wxRE_NOSUB") );
 461
 462     return m_nMatches;
 463 }
 464
 465 int wxRegExImpl::Replace(wxString *text,
 466                          const wxString& replacement,
 467                          size_t maxMatches) const
 468 {
 469     wxCHECK_MSG( text, wxNOT_FOUND, _T("NULL text in wxRegEx::Replace") );
 470     wxCHECK_MSG( IsValid(), wxNOT_FOUND, _T("must successfully Compile() first") );
 471
 472     // the input string
 473 #ifndef WXREGEX_CONVERT_TO_MB
 474     const wxChar *textstr = text->c_str();
 475     size_t textlen = text->length();
 476 #else
 477     const wxWX2MBbuf textstr = WXREGEX_CHAR(*text);
 478     if (!textstr)
 479     {
 480         wxLogError(_("Failed to find match for regular expression: %s"),
 481                    GetErrorMsg(0, true).c_str());
 482         return 0;
 483     }
 484     size_t textlen = strlen(textstr);
 485     text->clear();
 486 #endif
 487
 488     // the replacement text
 489     wxString textNew;
 490
 491     // the result, allow 25% extra
 492     wxString result;
 493     result.reserve(5 * textlen / 4);
 494
 495     // attempt at optimization: don't iterate over the string if it doesn't
 496     // contain back references at all
 497     bool mayHaveBackrefs =
 498         replacement.find_first_of(_T("\\&")) != wxString::npos;
 499
 500     if ( !mayHaveBackrefs )
 501     {
 502         textNew = replacement;
 503     }
 504
 505     // the position where we start looking for the match
 506     size_t matchStart = 0;
 507
 508     // number of replacement made: we won't make more than maxMatches of them
 509     // (unless maxMatches is 0 which doesn't limit the number of replacements)
 510     size_t countRepl = 0;
 511
 512     // note that "^" shouldn't match after the first call to Matches() so we
 513     // use wxRE_NOTBOL to prevent it from happening
 514     while ( (!maxMatches || countRepl < maxMatches) &&
 515             Matches(textstr + matchStart,
 516                     countRepl ? wxRE_NOTBOL : 0
 517                     WXREGEX_IF_NEED_LEN(textlen - matchStart)) )
 518     {
 519         // the string possibly contains back references: we need to calculate
 520         // the replacement text anew after each match
 521         if ( mayHaveBackrefs )
 522         {
 523             mayHaveBackrefs = false;
 524             textNew.clear();
 525             textNew.reserve(replacement.length());
 526
 527             for ( const wxChar *p = replacement.c_str(); *p; p++ )
 528             {
 529                 size_t index = (size_t)-1;
 530
 531                 if ( *p == _T('\\') )
 532                 {
 533                     if ( wxIsdigit(*++p) )
 534                     {
 535                         // back reference
 536                         wxChar *end;
 537                         index = (size_t)wxStrtoul(p, &end, 10);
 538                         p = end - 1; // -1 to compensate for p++ in the loop
 539                     }
 540                     //else: backslash used as escape character
 541                 }
 542                 else if ( *p == _T('&') )
 543                 {
 544                     // treat this as "\0" for compatbility with ed and such
 545                     index = 0;
 546                 }
 547
 548                 // do we have a back reference?
 549                 if ( index != (size_t)-1 )
 550                 {
 551                     // yes, get its text
 552                     size_t start, len;
 553                     if ( !GetMatch(&start, &len, index) )
 554                     {
 555                         wxFAIL_MSG( _T("invalid back reference") );
 556
 557                         // just eat it...
 558                     }
 559                     else
 560                     {
 561                         textNew += wxString(textstr + matchStart + start,
 562                                             *wxConvCurrent, len);
 563
 564                         mayHaveBackrefs = true;
 565                     }
 566                 }
 567                 else // ordinary character
 568                 {
 569                     textNew += *p;
 570                 }
 571             }
 572         }
 573
 574         size_t start, len;
 575         if ( !GetMatch(&start, &len) )
 576         {
 577             // we did have match as Matches() returned true above!
 578             wxFAIL_MSG( _T("internal logic error in wxRegEx::Replace") );
 579
 580             return wxNOT_FOUND;
 581         }
 582
 583         // an insurance against implementations that don't grow exponentially
 584         // to ensure building the result takes linear time
 585         if (result.capacity() < result.length() + start + textNew.length())
 586             result.reserve(2 * result.length());
 587
 588 #ifndef WXREGEX_CONVERT_TO_MB
 589         result.append(*text, matchStart, start);
 590 #else
 591         result.append(wxString(textstr + matchStart, *wxConvCurrent, start));
 592 #endif
 593         matchStart += start;
 594         result.append(textNew);
 595
 596         countRepl++;
 597
 598         matchStart += len;
 599     }
 600
 601 #ifndef WXREGEX_CONVERT_TO_MB
 602     result.append(*text, matchStart, wxString::npos);
 603 #else
 604     result.append(wxString(textstr + matchStart, *wxConvCurrent));
 605 #endif
 606     *text = result;
 607
 608     return countRepl;
 609 }
 610
 611 // ----------------------------------------------------------------------------
 612 // wxRegEx: all methods are mostly forwarded to wxRegExImpl
 613 // ----------------------------------------------------------------------------
 614
 615 void wxRegEx::Init()
 616 {
 617     m_impl = NULL;
 618 }
 619
 620 wxRegEx::~wxRegEx()
 621 {
 622     delete m_impl;
 623 }
 624
 625 bool wxRegEx::Compile(const wxString& expr, int flags)
 626 {
 627     if ( !m_impl )
 628     {
 629         m_impl = new wxRegExImpl;
 630     }
 631
 632     if ( !m_impl->Compile(expr, flags) )
 633     {
 634         // error message already given in wxRegExImpl::Compile
 635         delete m_impl;
 636         m_impl = NULL;
 637
 638         return false;
 639     }
 640
 641     return true;
 642 }
 643
 644 bool wxRegEx::Matches(const wxString& str, int flags) const
 645 {
 646     wxCHECK_MSG( IsValid(), false, _T("must successfully Compile() first") );
 647
 648     return m_impl->Matches(WXREGEX_CHAR(str), flags
 649                             WXREGEX_IF_NEED_LEN(str.length()));
 650 }
 651
 652 bool wxRegEx::GetMatch(size_t *start, size_t *len, size_t index) const
 653 {
 654     wxCHECK_MSG( IsValid(), false, _T("must successfully Compile() first") );
 655
 656     return m_impl->GetMatch(start, len, index);
 657 }
 658
 659 wxString wxRegEx::GetMatch(const wxString& text, size_t index) const
 660 {
 661     size_t start, len;
 662     if ( !GetMatch(&start, &len, index) )
 663         return wxEmptyString;
 664
 665     return text.Mid(start, len);
 666 }
 667
 668 size_t wxRegEx::GetMatchCount() const
 669 {
 670     wxCHECK_MSG( IsValid(), 0, _T("must successfully Compile() first") );
 671
 672     return m_impl->GetMatchCount();
 673 }
 674
 675 int wxRegEx::Replace(wxString *pattern,
 676                      const wxString& replacement,
 677                      size_t maxMatches) const
 678 {
 679     wxCHECK_MSG( IsValid(), wxNOT_FOUND, _T("must successfully Compile() first") );
 680
 681     return m_impl->Replace(pattern, replacement, maxMatches);
 682 }
 683
 684 #endif // wxUSE_REGEX