src/common/string.cpp

   1 /////////////////////////////////////////////////////////////////////////////
   2 // Name:        src/common/string.cpp
   3 // Purpose:     wxString class
   4 // Author:      Vadim Zeitlin, Ryan Norton
   5 // Modified by:
   6 // Created:     29/01/98
   7 // RCS-ID:      $Id$
   8 // Copyright:   (c) 1998 Vadim Zeitlin <zeitlin@dptmaths.ens-cachan.fr>
   9 //              (c) 2004 Ryan Norton <wxprojects@comcast.net>
  10 // Licence:     wxWindows licence
  11 /////////////////////////////////////////////////////////////////////////////
  12
  13 // ===========================================================================
  14 // headers, declarations, constants
  15 // ===========================================================================
  16
  17 // For compilers that support precompilation, includes "wx.h".
  18 #include "wx/wxprec.h"
  19
  20 #ifdef __BORLANDC__
  21     #pragma hdrstop
  22 #endif
  23
  24 #ifndef WX_PRECOMP
  25     #include "wx/string.h"
  26     #include "wx/wxcrtvararg.h"
  27     #include "wx/log.h"
  28 #endif
  29
  30 #include <ctype.h>
  31
  32 #ifndef __WXWINCE__
  33     #include <errno.h>
  34 #endif
  35
  36 #include <string.h>
  37 #include <stdlib.h>
  38
  39 #include "wx/hashmap.h"
  40 #include "wx/vector.h"
  41 #include "wx/xlocale.h"
  42
  43 #ifdef __WXMSW__
  44     #include "wx/msw/wrapwin.h"
  45 #endif // __WXMSW__
  46
  47 // string handling functions used by wxString:
  48 #if wxUSE_UNICODE_UTF8
  49     #define wxStringMemcpy   memcpy
  50     #define wxStringMemcmp   memcmp
  51     #define wxStringMemchr   memchr
  52     #define wxStringStrlen   strlen
  53 #else
  54     #define wxStringMemcpy   wxTmemcpy
  55     #define wxStringMemcmp   wxTmemcmp
  56     #define wxStringMemchr   wxTmemchr
  57     #define wxStringStrlen   wxStrlen
  58 #endif
  59
  60 // ----------------------------------------------------------------------------
  61 // global variables
  62 // ----------------------------------------------------------------------------
  63
  64 namespace wxPrivate
  65 {
  66
  67 static UntypedBufferData s_untypedNullData(NULL, 0);
  68
  69 UntypedBufferData * const untypedNullDataPtr = &s_untypedNullData;
  70
  71 } // namespace wxPrivate
  72
  73 // ---------------------------------------------------------------------------
  74 // static class variables definition
  75 // ---------------------------------------------------------------------------
  76
  77 //According to STL _must_ be a -1 size_t
  78 const size_t wxString::npos = (size_t) -1;
  79
  80 #if wxUSE_STRING_POS_CACHE
  81
  82 #ifdef wxHAS_COMPILER_TLS
  83
  84 wxTLS_TYPE(wxString::Cache) wxString::ms_cache;
  85
  86 #else // !wxHAS_COMPILER_TLS
  87
  88 struct wxStrCacheInitializer
  89 {
  90     wxStrCacheInitializer()
  91     {
  92         // calling this function triggers s_cache initialization in it, and
  93         // from now on it becomes safe to call from multiple threads
  94         wxString::GetCache();
  95     }
  96 };
  97
  98 /*
  99 wxString::Cache& wxString::GetCache()
 100 {
 101     static wxTLS_TYPE(Cache) s_cache;
 102
 103     return wxTLS_VALUE(s_cache);
 104 }
 105 */
 106
 107 static wxStrCacheInitializer gs_stringCacheInit;
 108
 109 #endif // wxHAS_COMPILER_TLS/!wxHAS_COMPILER_TLS
 110
 111 // gdb seems to be unable to display thread-local variables correctly, at least
 112 // not my 6.4.98 version under amd64, so provide this debugging helper to do it
 113 #if wxDEBUG_LEVEL >= 2
 114
 115 struct wxStrCacheDumper
 116 {
 117     static void ShowAll()
 118     {
 119         puts("*** wxString cache dump:");
 120         for ( unsigned n = 0; n < wxString::Cache::SIZE; n++ )
 121         {
 122             const wxString::Cache::Element&
 123                 c = wxString::GetCacheBegin()[n];
 124
 125             printf("\t%u%s\t%p: pos=(%lu, %lu), len=%ld\n",
 126                    n,
 127                    n == wxString::LastUsedCacheElement() ? " [*]" : "",
 128                    c.str,
 129                    (unsigned long)c.pos,
 130                    (unsigned long)c.impl,
 131                    (long)c.len);
 132         }
 133     }
 134 };
 135
 136 void wxDumpStrCache() { wxStrCacheDumper::ShowAll(); }
 137
 138 #endif // wxDEBUG_LEVEL >= 2
 139
 140 #ifdef wxPROFILE_STRING_CACHE
 141
 142 wxString::CacheStats wxString::ms_cacheStats;
 143
 144 struct wxStrCacheStatsDumper
 145 {
 146     ~wxStrCacheStatsDumper()
 147     {
 148         const wxString::CacheStats& stats = wxString::ms_cacheStats;
 149
 150         if ( stats.postot )
 151         {
 152             puts("*** wxString cache statistics:");
 153             printf("\tTotal non-trivial calls to PosToImpl(): %u\n",
 154                    stats.postot);
 155             printf("\tHits %u (of which %u not used) or %.2f%%\n",
 156                    stats.poshits,
 157                    stats.mishits,
 158                    100.*float(stats.poshits - stats.mishits)/stats.postot);
 159             printf("\tAverage position requested: %.2f\n",
 160                    float(stats.sumpos) / stats.postot);
 161             printf("\tAverage offset after cached hint: %.2f\n",
 162                    float(stats.sumofs) / stats.postot);
 163         }
 164
 165         if ( stats.lentot )
 166         {
 167             printf("\tNumber of calls to length(): %u, hits=%.2f%%\n",
 168                    stats.lentot, 100.*float(stats.lenhits)/stats.lentot);
 169         }
 170     }
 171 };
 172
 173 static wxStrCacheStatsDumper s_showCacheStats;
 174
 175 #endif // wxPROFILE_STRING_CACHE
 176
 177 #endif // wxUSE_STRING_POS_CACHE
 178
 179 // ----------------------------------------------------------------------------
 180 // global functions
 181 // ----------------------------------------------------------------------------
 182
 183 #if wxUSE_STD_IOSTREAM
 184
 185 #include <iostream>
 186
 187 wxSTD ostream& operator<<(wxSTD ostream& os, const wxCStrData& str)
 188 {
 189 #if wxUSE_UNICODE && !wxUSE_UNICODE_UTF8
 190     const wxScopedCharBuffer buf(str.AsCharBuf());
 191     if ( !buf )
 192         os.clear(wxSTD ios_base::failbit);
 193     else
 194         os << buf.data();
 195
 196     return os;
 197 #else
 198     return os << str.AsInternal();
 199 #endif
 200 }
 201
 202 wxSTD ostream& operator<<(wxSTD ostream& os, const wxString& str)
 203 {
 204     return os << str.c_str();
 205 }
 206
 207 wxSTD ostream& operator<<(wxSTD ostream& os, const wxScopedCharBuffer& str)
 208 {
 209     return os << str.data();
 210 }
 211
 212 #ifndef __BORLANDC__
 213 wxSTD ostream& operator<<(wxSTD ostream& os, const wxScopedWCharBuffer& str)
 214 {
 215     return os << str.data();
 216 }
 217 #endif
 218
 219 #if wxUSE_UNICODE && defined(HAVE_WOSTREAM)
 220
 221 wxSTD wostream& operator<<(wxSTD wostream& wos, const wxString& str)
 222 {
 223     return wos << str.wc_str();
 224 }
 225
 226 wxSTD wostream& operator<<(wxSTD wostream& wos, const wxCStrData& str)
 227 {
 228     return wos << str.AsWChar();
 229 }
 230
 231 wxSTD wostream& operator<<(wxSTD wostream& wos, const wxScopedWCharBuffer& str)
 232 {
 233     return wos << str.data();
 234 }
 235
 236 #endif  // wxUSE_UNICODE && defined(HAVE_WOSTREAM)
 237
 238 #endif // wxUSE_STD_IOSTREAM
 239
 240 // ===========================================================================
 241 // wxString class core
 242 // ===========================================================================
 243
 244 #if wxUSE_UNICODE_UTF8
 245
 246 void wxString::PosLenToImpl(size_t pos, size_t len,
 247                             size_t *implPos, size_t *implLen) const
 248 {
 249     if ( pos == npos )
 250     {
 251         *implPos = npos;
 252     }
 253     else // have valid start position
 254     {
 255         const const_iterator b = GetIterForNthChar(pos);
 256         *implPos = wxStringImpl::const_iterator(b.impl()) - m_impl.begin();
 257         if ( len == npos )
 258         {
 259             *implLen = npos;
 260         }
 261         else // have valid length too
 262         {
 263             // we need to handle the case of length specifying a substring
 264             // going beyond the end of the string, just as std::string does
 265             const const_iterator e(end());
 266             const_iterator i(b);
 267             while ( len && i <= e )
 268             {
 269                 ++i;
 270                 --len;
 271             }
 272
 273             *implLen = i.impl() - b.impl();
 274         }
 275     }
 276 }
 277
 278 #endif // wxUSE_UNICODE_UTF8
 279
 280 // ----------------------------------------------------------------------------
 281 // wxCStrData converted strings caching
 282 // ----------------------------------------------------------------------------
 283
 284 // FIXME-UTF8: temporarily disabled because it doesn't work with global
 285 //             string objects; re-enable after fixing this bug and benchmarking
 286 //             performance to see if using a hash is a good idea at all
 287 #if 0
 288
 289 // For backward compatibility reasons, it must be possible to assign the value
 290 // returned by wxString::c_str() to a char* or wchar_t* variable and work with
 291 // it. Returning wxCharBuffer from (const char*)c_str() wouldn't do the trick,
 292 // because the memory would be freed immediately, but it has to be valid as long
 293 // as the string is not modified, so that code like this still works:
 294 //
 295 // const wxChar *s = str.c_str();
 296 // while ( s ) { ... }
 297
 298 // FIXME-UTF8: not thread safe!
 299 // FIXME-UTF8: we currently clear the cached conversion only when the string is
 300 //             destroyed, but we should do it when the string is modified, to
 301 //             keep memory usage down
 302 // FIXME-UTF8: we do the conversion every time As[W]Char() is called, but if we
 303 //             invalidated the cache on every change, we could keep the previous
 304 //             conversion
 305 // FIXME-UTF8: add tracing of usage of these two methods - new code is supposed
 306 //             to use mb_str() or wc_str() instead of (const [w]char*)c_str()
 307
 308 template<typename T>
 309 static inline void DeleteStringFromConversionCache(T& hash, const wxString *s)
 310 {
 311     typename T::iterator i = hash.find(wxConstCast(s, wxString));
 312     if ( i != hash.end() )
 313     {
 314         free(i->second);
 315         hash.erase(i);
 316     }
 317 }
 318
 319 #if wxUSE_UNICODE
 320 // NB: non-STL implementation doesn't compile with "const wxString*" key type,
 321 //     so we have to use wxString* here and const-cast when used
 322 WX_DECLARE_HASH_MAP(wxString*, char*, wxPointerHash, wxPointerEqual,
 323                     wxStringCharConversionCache);
 324 static wxStringCharConversionCache gs_stringsCharCache;
 325
 326 const char* wxCStrData::AsChar() const
 327 {
 328     // remove previously cache value, if any (see FIXMEs above):
 329     DeleteStringFromConversionCache(gs_stringsCharCache, m_str);
 330
 331     // convert the string and keep it:
 332     const char *s = gs_stringsCharCache[wxConstCast(m_str, wxString)] =
 333         m_str->mb_str().release();
 334
 335     return s + m_offset;
 336 }
 337 #endif // wxUSE_UNICODE
 338
 339 #if !wxUSE_UNICODE_WCHAR
 340 WX_DECLARE_HASH_MAP(wxString*, wchar_t*, wxPointerHash, wxPointerEqual,
 341                     wxStringWCharConversionCache);
 342 static wxStringWCharConversionCache gs_stringsWCharCache;
 343
 344 const wchar_t* wxCStrData::AsWChar() const
 345 {
 346     // remove previously cache value, if any (see FIXMEs above):
 347     DeleteStringFromConversionCache(gs_stringsWCharCache, m_str);
 348
 349     // convert the string and keep it:
 350     const wchar_t *s = gs_stringsWCharCache[wxConstCast(m_str, wxString)] =
 351         m_str->wc_str().release();
 352
 353     return s + m_offset;
 354 }
 355 #endif // !wxUSE_UNICODE_WCHAR
 356
 357 wxString::~wxString()
 358 {
 359 #if wxUSE_UNICODE
 360     // FIXME-UTF8: do this only if locale is not UTF8 if wxUSE_UNICODE_UTF8
 361     DeleteStringFromConversionCache(gs_stringsCharCache, this);
 362 #endif
 363 #if !wxUSE_UNICODE_WCHAR
 364     DeleteStringFromConversionCache(gs_stringsWCharCache, this);
 365 #endif
 366 }
 367 #endif
 368
 369 // ===========================================================================
 370 // wxString class core
 371 // ===========================================================================
 372
 373 // ---------------------------------------------------------------------------
 374 // construction and conversion
 375 // ---------------------------------------------------------------------------
 376
 377 #if wxUSE_UNICODE_WCHAR
 378 /* static */
 379 wxString::SubstrBufFromMB wxString::ConvertStr(const char *psz, size_t nLength,
 380                                                const wxMBConv& conv)
 381 {
 382     // anything to do?
 383     if ( !psz || nLength == 0 )
 384         return SubstrBufFromMB(wxWCharBuffer(L""), 0);
 385
 386     if ( nLength == npos )
 387         nLength = wxNO_LEN;
 388
 389     size_t wcLen;
 390     wxScopedWCharBuffer wcBuf(conv.cMB2WC(psz, nLength, &wcLen));
 391     if ( !wcLen )
 392         return SubstrBufFromMB(wxWCharBuffer(L""), 0);
 393     else
 394         return SubstrBufFromMB(wcBuf, wcLen);
 395 }
 396 #endif // wxUSE_UNICODE_WCHAR
 397
 398 #if wxUSE_UNICODE_UTF8
 399 /* static */
 400 wxString::SubstrBufFromMB wxString::ConvertStr(const char *psz, size_t nLength,
 401                                                const wxMBConv& conv)
 402 {
 403     // anything to do?
 404     if ( !psz || nLength == 0 )
 405         return SubstrBufFromMB(wxCharBuffer(""), 0);
 406
 407     // if psz is already in UTF-8, we don't have to do the roundtrip to
 408     // wchar_t* and back:
 409     if ( conv.IsUTF8() )
 410     {
 411         // we need to validate the input because UTF8 iterators assume valid
 412         // UTF-8 sequence and psz may be invalid:
 413         if ( wxStringOperations::IsValidUtf8String(psz, nLength) )
 414         {
 415             // we must pass the real string length to SubstrBufFromMB ctor
 416             if ( nLength == npos )
 417                 nLength = psz ? strlen(psz) : 0;
 418             return SubstrBufFromMB(wxScopedCharBuffer::CreateNonOwned(psz, nLength),
 419                                    nLength);
 420         }
 421         // else: do the roundtrip through wchar_t*
 422     }
 423
 424     if ( nLength == npos )
 425         nLength = wxNO_LEN;
 426
 427     // first convert to wide string:
 428     size_t wcLen;
 429     wxScopedWCharBuffer wcBuf(conv.cMB2WC(psz, nLength, &wcLen));
 430     if ( !wcLen )
 431         return SubstrBufFromMB(wxCharBuffer(""), 0);
 432
 433     // and then to UTF-8:
 434     SubstrBufFromMB buf(ConvertStr(wcBuf, wcLen, wxMBConvStrictUTF8()));
 435     // widechar -> UTF-8 conversion isn't supposed to ever fail:
 436     wxASSERT_MSG( buf.data, wxT("conversion to UTF-8 failed") );
 437
 438     return buf;
 439 }
 440 #endif // wxUSE_UNICODE_UTF8
 441
 442 #if wxUSE_UNICODE_UTF8 || !wxUSE_UNICODE
 443 /* static */
 444 wxString::SubstrBufFromWC wxString::ConvertStr(const wchar_t *pwz, size_t nLength,
 445                                                const wxMBConv& conv)
 446 {
 447     // anything to do?
 448     if ( !pwz || nLength == 0 )
 449         return SubstrBufFromWC(wxCharBuffer(""), 0);
 450
 451     if ( nLength == npos )
 452         nLength = wxNO_LEN;
 453
 454     size_t mbLen;
 455     wxScopedCharBuffer mbBuf(conv.cWC2MB(pwz, nLength, &mbLen));
 456     if ( !mbLen )
 457         return SubstrBufFromWC(wxCharBuffer(""), 0);
 458     else
 459         return SubstrBufFromWC(mbBuf, mbLen);
 460 }
 461 #endif // wxUSE_UNICODE_UTF8 || !wxUSE_UNICODE
 462
 463 // This std::string::c_str()-like method returns a wide char pointer to string
 464 // contents. In wxUSE_UNICODE_WCHAR case it is trivial as it can simply return
 465 // a pointer to the internal representation. Otherwise a conversion is required
 466 // and it returns a temporary buffer.
 467 //
 468 // However for compatibility with c_str() and to avoid breaking existing code
 469 // doing
 470 //
 471 //      for ( const wchar_t *p = s.wc_str(); *p; p++ )
 472 //          ... use *p...
 473 //
 474 // we actually need to ensure that the returned buffer is _not_ temporary and
 475 // so we use wxString::m_convertedToWChar to store the returned data
 476 #if !wxUSE_UNICODE_WCHAR
 477
 478 const wchar_t *wxString::AsWChar(const wxMBConv& conv) const
 479 {
 480     const char * const strMB = m_impl.c_str();
 481     const size_t lenMB = m_impl.length();
 482
 483     // find out the size of the buffer needed
 484     const size_t lenWC = conv.ToWChar(NULL, 0, strMB, lenMB);
 485     if ( lenWC == wxCONV_FAILED )
 486         return NULL;
 487
 488     // keep the same buffer if the string size didn't change: this is not only
 489     // an optimization but also ensure that code which modifies string
 490     // character by character (without changing its length) can continue to use
 491     // the pointer returned by a previous wc_str() call even after changing the
 492     // string
 493
 494     // TODO-UTF8: we could check for ">" instead of "!=" here as this would
 495     //            allow to save on buffer reallocations but at the cost of
 496     //            consuming (even) more memory, we should benchmark this to
 497     //            determine if it's worth doing
 498     if ( !m_convertedToWChar.m_str || lenWC != m_convertedToWChar.m_len )
 499     {
 500         if ( !const_cast<wxString *>(this)->m_convertedToWChar.Extend(lenWC) )
 501             return NULL;
 502     }
 503
 504     // finally do convert
 505     m_convertedToWChar.m_str[lenWC] = L'\0';
 506     if ( conv.ToWChar(m_convertedToWChar.m_str, lenWC,
 507                       strMB, lenMB) == wxCONV_FAILED )
 508         return NULL;
 509
 510     return m_convertedToWChar.m_str;
 511 }
 512
 513 #endif // !wxUSE_UNICODE_WCHAR
 514
 515
 516 // Same thing for mb_str() which returns a normal char pointer to string
 517 // contents: this always requires converting it to the specified encoding in
 518 // non-ANSI build except if we need to convert to UTF-8 and this is what we
 519 // already use internally.
 520 #if wxUSE_UNICODE
 521
 522 const char *wxString::AsChar(const wxMBConv& conv) const
 523 {
 524 #if wxUSE_UNICODE_UTF8
 525     if ( conv.IsUTF8() )
 526         return m_impl.c_str();
 527
 528     const wchar_t * const strWC = AsWChar(wxMBConvStrictUTF8());
 529     const size_t lenWC = m_convertedToWChar.m_len;
 530 #else // wxUSE_UNICODE_WCHAR
 531     const wchar_t * const strWC = m_impl.c_str();
 532     const size_t lenWC = m_impl.length();
 533 #endif // wxUSE_UNICODE_UTF8/wxUSE_UNICODE_WCHAR
 534
 535     const size_t lenMB = conv.FromWChar(NULL, 0, strWC, lenWC);
 536     if ( lenMB == wxCONV_FAILED )
 537         return NULL;
 538
 539     if ( !m_convertedToChar.m_str || lenMB != m_convertedToChar.m_len )
 540     {
 541         if ( !const_cast<wxString *>(this)->m_convertedToChar.Extend(lenMB) )
 542             return NULL;
 543     }
 544
 545     m_convertedToChar.m_str[lenMB] = '\0';
 546     if ( conv.FromWChar(m_convertedToChar.m_str, lenMB,
 547                         strWC, lenWC) == wxCONV_FAILED )
 548         return NULL;
 549
 550     return m_convertedToChar.m_str;
 551 }
 552
 553 #endif // wxUSE_UNICODE
 554
 555 // shrink to minimal size (releasing extra memory)
 556 bool wxString::Shrink()
 557 {
 558   wxString tmp(begin(), end());
 559   swap(tmp);
 560   return tmp.length() == length();
 561 }
 562
 563 // deprecated compatibility code:
 564 #if WXWIN_COMPATIBILITY_2_8 && !wxUSE_STL_BASED_WXSTRING && !wxUSE_UNICODE_UTF8
 565 wxStringCharType *wxString::GetWriteBuf(size_t nLen)
 566 {
 567     return DoGetWriteBuf(nLen);
 568 }
 569
 570 void wxString::UngetWriteBuf()
 571 {
 572     DoUngetWriteBuf();
 573 }
 574
 575 void wxString::UngetWriteBuf(size_t nLen)
 576 {
 577     DoUngetWriteBuf(nLen);
 578 }
 579 #endif // WXWIN_COMPATIBILITY_2_8 && !wxUSE_STL_BASED_WXSTRING && !wxUSE_UNICODE_UTF8
 580
 581
 582 // ---------------------------------------------------------------------------
 583 // data access
 584 // ---------------------------------------------------------------------------
 585
 586 // all functions are inline in string.h
 587
 588 // ---------------------------------------------------------------------------
 589 // concatenation operators
 590 // ---------------------------------------------------------------------------
 591
 592 /*
 593  * concatenation functions come in 5 flavours:
 594  *  string + string
 595  *  char   + string      and      string + char
 596  *  C str  + string      and      string + C str
 597  */
 598
 599 wxString operator+(const wxString& str1, const wxString& str2)
 600 {
 601 #if !wxUSE_STL_BASED_WXSTRING
 602     wxASSERT( str1.IsValid() );
 603     wxASSERT( str2.IsValid() );
 604 #endif
 605
 606     wxString s = str1;
 607     s += str2;
 608
 609     return s;
 610 }
 611
 612 wxString operator+(const wxString& str, wxUniChar ch)
 613 {
 614 #if !wxUSE_STL_BASED_WXSTRING
 615     wxASSERT( str.IsValid() );
 616 #endif
 617
 618     wxString s = str;
 619     s += ch;
 620
 621     return s;
 622 }
 623
 624 wxString operator+(wxUniChar ch, const wxString& str)
 625 {
 626 #if !wxUSE_STL_BASED_WXSTRING
 627     wxASSERT( str.IsValid() );
 628 #endif
 629
 630     wxString s = ch;
 631     s += str;
 632
 633     return s;
 634 }
 635
 636 wxString operator+(const wxString& str, const char *psz)
 637 {
 638 #if !wxUSE_STL_BASED_WXSTRING
 639     wxASSERT( str.IsValid() );
 640 #endif
 641
 642     wxString s;
 643     if ( !s.Alloc(strlen(psz) + str.length()) ) {
 644         wxFAIL_MSG( wxT("out of memory in wxString::operator+") );
 645     }
 646     s += str;
 647     s += psz;
 648
 649     return s;
 650 }
 651
 652 wxString operator+(const wxString& str, const wchar_t *pwz)
 653 {
 654 #if !wxUSE_STL_BASED_WXSTRING
 655     wxASSERT( str.IsValid() );
 656 #endif
 657
 658     wxString s;
 659     if ( !s.Alloc(wxWcslen(pwz) + str.length()) ) {
 660         wxFAIL_MSG( wxT("out of memory in wxString::operator+") );
 661     }
 662     s += str;
 663     s += pwz;
 664
 665     return s;
 666 }
 667
 668 wxString operator+(const char *psz, const wxString& str)
 669 {
 670 #if !wxUSE_STL_BASED_WXSTRING
 671     wxASSERT( str.IsValid() );
 672 #endif
 673
 674     wxString s;
 675     if ( !s.Alloc(strlen(psz) + str.length()) ) {
 676         wxFAIL_MSG( wxT("out of memory in wxString::operator+") );
 677     }
 678     s = psz;
 679     s += str;
 680
 681     return s;
 682 }
 683
 684 wxString operator+(const wchar_t *pwz, const wxString& str)
 685 {
 686 #if !wxUSE_STL_BASED_WXSTRING
 687     wxASSERT( str.IsValid() );
 688 #endif
 689
 690     wxString s;
 691     if ( !s.Alloc(wxWcslen(pwz) + str.length()) ) {
 692         wxFAIL_MSG( wxT("out of memory in wxString::operator+") );
 693     }
 694     s = pwz;
 695     s += str;
 696
 697     return s;
 698 }
 699
 700 // ---------------------------------------------------------------------------
 701 // string comparison
 702 // ---------------------------------------------------------------------------
 703
 704 bool wxString::IsSameAs(wxUniChar c, bool compareWithCase) const
 705 {
 706     return (length() == 1) && (compareWithCase ? GetChar(0u) == c
 707                                : wxToupper(GetChar(0u)) == wxToupper(c));
 708 }
 709
 710 #ifdef HAVE_STD_STRING_COMPARE
 711
 712 // NB: Comparison code (both if HAVE_STD_STRING_COMPARE and if not) works with
 713 //     UTF-8 encoded strings too, thanks to UTF-8's design which allows us to
 714 //     sort strings in characters code point order by sorting the byte sequence
 715 //     in byte values order (i.e. what strcmp() and memcmp() do).
 716
 717 int wxString::compare(const wxString& str) const
 718 {
 719     return m_impl.compare(str.m_impl);
 720 }
 721
 722 int wxString::compare(size_t nStart, size_t nLen,
 723                       const wxString& str) const
 724 {
 725     size_t pos, len;
 726     PosLenToImpl(nStart, nLen, &pos, &len);
 727     return m_impl.compare(pos, len, str.m_impl);
 728 }
 729
 730 int wxString::compare(size_t nStart, size_t nLen,
 731                       const wxString& str,
 732                       size_t nStart2, size_t nLen2) const
 733 {
 734     size_t pos, len;
 735     PosLenToImpl(nStart, nLen, &pos, &len);
 736
 737     size_t pos2, len2;
 738     str.PosLenToImpl(nStart2, nLen2, &pos2, &len2);
 739
 740     return m_impl.compare(pos, len, str.m_impl, pos2, len2);
 741 }
 742
 743 int wxString::compare(const char* sz) const
 744 {
 745     return m_impl.compare(ImplStr(sz));
 746 }
 747
 748 int wxString::compare(const wchar_t* sz) const
 749 {
 750     return m_impl.compare(ImplStr(sz));
 751 }
 752
 753 int wxString::compare(size_t nStart, size_t nLen,
 754                       const char* sz, size_t nCount) const
 755 {
 756     size_t pos, len;
 757     PosLenToImpl(nStart, nLen, &pos, &len);
 758
 759     SubstrBufFromMB str(ImplStr(sz, nCount));
 760
 761     return m_impl.compare(pos, len, str.data, str.len);
 762 }
 763
 764 int wxString::compare(size_t nStart, size_t nLen,
 765                       const wchar_t* sz, size_t nCount) const
 766 {
 767     size_t pos, len;
 768     PosLenToImpl(nStart, nLen, &pos, &len);
 769
 770     SubstrBufFromWC str(ImplStr(sz, nCount));
 771
 772     return m_impl.compare(pos, len, str.data, str.len);
 773 }
 774
 775 #else // !HAVE_STD_STRING_COMPARE
 776
 777 static inline int wxDoCmp(const wxStringCharType* s1, size_t l1,
 778                           const wxStringCharType* s2, size_t l2)
 779 {
 780     if( l1 == l2 )
 781         return wxStringMemcmp(s1, s2, l1);
 782     else if( l1 < l2 )
 783     {
 784         int ret = wxStringMemcmp(s1, s2, l1);
 785         return ret == 0 ? -1 : ret;
 786     }
 787     else
 788     {
 789         int ret = wxStringMemcmp(s1, s2, l2);
 790         return ret == 0 ? +1 : ret;
 791     }
 792 }
 793
 794 int wxString::compare(const wxString& str) const
 795 {
 796     return ::wxDoCmp(m_impl.data(), m_impl.length(),
 797                      str.m_impl.data(), str.m_impl.length());
 798 }
 799
 800 int wxString::compare(size_t nStart, size_t nLen,
 801                       const wxString& str) const
 802 {
 803     wxASSERT(nStart <= length());
 804     size_type strLen = length() - nStart;
 805     nLen = strLen < nLen ? strLen : nLen;
 806
 807     size_t pos, len;
 808     PosLenToImpl(nStart, nLen, &pos, &len);
 809
 810     return ::wxDoCmp(m_impl.data() + pos,  len,
 811                      str.m_impl.data(), str.m_impl.length());
 812 }
 813
 814 int wxString::compare(size_t nStart, size_t nLen,
 815                       const wxString& str,
 816                       size_t nStart2, size_t nLen2) const
 817 {
 818     wxASSERT(nStart <= length());
 819     wxASSERT(nStart2 <= str.length());
 820     size_type strLen  =     length() - nStart,
 821               strLen2 = str.length() - nStart2;
 822     nLen  = strLen  < nLen  ? strLen  : nLen;
 823     nLen2 = strLen2 < nLen2 ? strLen2 : nLen2;
 824
 825     size_t pos, len;
 826     PosLenToImpl(nStart, nLen, &pos, &len);
 827     size_t pos2, len2;
 828     str.PosLenToImpl(nStart2, nLen2, &pos2, &len2);
 829
 830     return ::wxDoCmp(m_impl.data() + pos, len,
 831                      str.m_impl.data() + pos2, len2);
 832 }
 833
 834 int wxString::compare(const char* sz) const
 835 {
 836     SubstrBufFromMB str(ImplStr(sz, npos));
 837     if ( str.len == npos )
 838         str.len = wxStringStrlen(str.data);
 839     return ::wxDoCmp(m_impl.data(), m_impl.length(), str.data, str.len);
 840 }
 841
 842 int wxString::compare(const wchar_t* sz) const
 843 {
 844     SubstrBufFromWC str(ImplStr(sz, npos));
 845     if ( str.len == npos )
 846         str.len = wxStringStrlen(str.data);
 847     return ::wxDoCmp(m_impl.data(), m_impl.length(), str.data, str.len);
 848 }
 849
 850 int wxString::compare(size_t nStart, size_t nLen,
 851                       const char* sz, size_t nCount) const
 852 {
 853     wxASSERT(nStart <= length());
 854     size_type strLen = length() - nStart;
 855     nLen = strLen < nLen ? strLen : nLen;
 856
 857     size_t pos, len;
 858     PosLenToImpl(nStart, nLen, &pos, &len);
 859
 860     SubstrBufFromMB str(ImplStr(sz, nCount));
 861     if ( str.len == npos )
 862         str.len = wxStringStrlen(str.data);
 863
 864     return ::wxDoCmp(m_impl.data() + pos, len, str.data, str.len);
 865 }
 866
 867 int wxString::compare(size_t nStart, size_t nLen,
 868                       const wchar_t* sz, size_t nCount) const
 869 {
 870     wxASSERT(nStart <= length());
 871     size_type strLen = length() - nStart;
 872     nLen = strLen < nLen ? strLen : nLen;
 873
 874     size_t pos, len;
 875     PosLenToImpl(nStart, nLen, &pos, &len);
 876
 877     SubstrBufFromWC str(ImplStr(sz, nCount));
 878     if ( str.len == npos )
 879         str.len = wxStringStrlen(str.data);
 880
 881     return ::wxDoCmp(m_impl.data() + pos, len, str.data, str.len);
 882 }
 883
 884 #endif // HAVE_STD_STRING_COMPARE/!HAVE_STD_STRING_COMPARE
 885
 886
 887 // ---------------------------------------------------------------------------
 888 // find_{first,last}_[not]_of functions
 889 // ---------------------------------------------------------------------------
 890
 891 #if !wxUSE_STL_BASED_WXSTRING || wxUSE_UNICODE_UTF8
 892
 893 // NB: All these functions are implemented  with the argument being wxChar*,
 894 //     i.e. widechar string in any Unicode build, even though native string
 895 //     representation is char* in the UTF-8 build. This is because we couldn't
 896 //     use memchr() to determine if a character is in a set encoded as UTF-8.
 897
 898 size_t wxString::find_first_of(const wxChar* sz, size_t nStart) const
 899 {
 900     return find_first_of(sz, nStart, wxStrlen(sz));
 901 }
 902
 903 size_t wxString::find_first_not_of(const wxChar* sz, size_t nStart) const
 904 {
 905     return find_first_not_of(sz, nStart, wxStrlen(sz));
 906 }
 907
 908 size_t wxString::find_first_of(const wxChar* sz, size_t nStart, size_t n) const
 909 {
 910     wxASSERT_MSG( nStart <= length(),  wxT("invalid index") );
 911
 912     size_t idx = nStart;
 913     for ( const_iterator i = begin() + nStart; i != end(); ++idx, ++i )
 914     {
 915         if ( wxTmemchr(sz, *i, n) )
 916             return idx;
 917     }
 918
 919     return npos;
 920 }
 921
 922 size_t wxString::find_first_not_of(const wxChar* sz, size_t nStart, size_t n) const
 923 {
 924     wxASSERT_MSG( nStart <= length(),  wxT("invalid index") );
 925
 926     size_t idx = nStart;
 927     for ( const_iterator i = begin() + nStart; i != end(); ++idx, ++i )
 928     {
 929         if ( !wxTmemchr(sz, *i, n) )
 930             return idx;
 931     }
 932
 933     return npos;
 934 }
 935
 936
 937 size_t wxString::find_last_of(const wxChar* sz, size_t nStart) const
 938 {
 939     return find_last_of(sz, nStart, wxStrlen(sz));
 940 }
 941
 942 size_t wxString::find_last_not_of(const wxChar* sz, size_t nStart) const
 943 {
 944     return find_last_not_of(sz, nStart, wxStrlen(sz));
 945 }
 946
 947 size_t wxString::find_last_of(const wxChar* sz, size_t nStart, size_t n) const
 948 {
 949     size_t len = length();
 950
 951     if ( nStart == npos )
 952     {
 953         nStart = len - 1;
 954     }
 955     else
 956     {
 957         wxASSERT_MSG( nStart <= len, wxT("invalid index") );
 958     }
 959
 960     size_t idx = nStart;
 961     for ( const_reverse_iterator i = rbegin() + (len - nStart - 1);
 962           i != rend(); --idx, ++i )
 963     {
 964         if ( wxTmemchr(sz, *i, n) )
 965             return idx;
 966     }
 967
 968     return npos;
 969 }
 970
 971 size_t wxString::find_last_not_of(const wxChar* sz, size_t nStart, size_t n) const
 972 {
 973     size_t len = length();
 974
 975     if ( nStart == npos )
 976     {
 977         nStart = len - 1;
 978     }
 979     else
 980     {
 981         wxASSERT_MSG( nStart <= len, wxT("invalid index") );
 982     }
 983
 984     size_t idx = nStart;
 985     for ( const_reverse_iterator i = rbegin() + (len - nStart - 1);
 986           i != rend(); --idx, ++i )
 987     {
 988         if ( !wxTmemchr(sz, *i, n) )
 989             return idx;
 990     }
 991
 992     return npos;
 993 }
 994
 995 size_t wxString::find_first_not_of(wxUniChar ch, size_t nStart) const
 996 {
 997     wxASSERT_MSG( nStart <= length(),  wxT("invalid index") );
 998
 999     size_t idx = nStart;
1000     for ( const_iterator i = begin() + nStart; i != end(); ++idx, ++i )
1001     {
1002         if ( *i != ch )
1003             return idx;
1004     }
1005
1006     return npos;
1007 }
1008
1009 size_t wxString::find_last_not_of(wxUniChar ch, size_t nStart) const
1010 {
1011     size_t len = length();
1012
1013     if ( nStart == npos )
1014     {
1015         nStart = len - 1;
1016     }
1017     else
1018     {
1019         wxASSERT_MSG( nStart <= len, wxT("invalid index") );
1020     }
1021
1022     size_t idx = nStart;
1023     for ( const_reverse_iterator i = rbegin() + (len - nStart - 1);
1024           i != rend(); --idx, ++i )
1025     {
1026         if ( *i != ch )
1027             return idx;
1028     }
1029
1030     return npos;
1031 }
1032
1033 // the functions above were implemented for wchar_t* arguments in Unicode
1034 // build and char* in ANSI build; below are implementations for the other
1035 // version:
1036 #if wxUSE_UNICODE
1037     #define wxOtherCharType char
1038     #define STRCONV         (const wxChar*)wxConvLibc.cMB2WC
1039 #else
1040     #define wxOtherCharType wchar_t
1041     #define STRCONV         (const wxChar*)wxConvLibc.cWC2MB
1042 #endif
1043
1044 size_t wxString::find_first_of(const wxOtherCharType* sz, size_t nStart) const
1045     { return find_first_of(STRCONV(sz), nStart); }
1046
1047 size_t wxString::find_first_of(const wxOtherCharType* sz, size_t nStart,
1048                                size_t n) const
1049     { return find_first_of(STRCONV(sz, n, NULL), nStart, n); }
1050 size_t wxString::find_last_of(const wxOtherCharType* sz, size_t nStart) const
1051     { return find_last_of(STRCONV(sz), nStart); }
1052 size_t wxString::find_last_of(const wxOtherCharType* sz, size_t nStart,
1053                               size_t n) const
1054     { return find_last_of(STRCONV(sz, n, NULL), nStart, n); }
1055 size_t wxString::find_first_not_of(const wxOtherCharType* sz, size_t nStart) const
1056     { return find_first_not_of(STRCONV(sz), nStart); }
1057 size_t wxString::find_first_not_of(const wxOtherCharType* sz, size_t nStart,
1058                                    size_t n) const
1059     { return find_first_not_of(STRCONV(sz, n, NULL), nStart, n); }
1060 size_t wxString::find_last_not_of(const wxOtherCharType* sz, size_t nStart) const
1061     { return find_last_not_of(STRCONV(sz), nStart); }
1062 size_t wxString::find_last_not_of(const wxOtherCharType* sz, size_t nStart,
1063                                   size_t n) const
1064     { return find_last_not_of(STRCONV(sz, n, NULL), nStart, n); }
1065
1066 #undef wxOtherCharType
1067 #undef STRCONV
1068
1069 #endif // !wxUSE_STL_BASED_WXSTRING || wxUSE_UNICODE_UTF8
1070
1071 // ===========================================================================
1072 // other common string functions
1073 // ===========================================================================
1074
1075 int wxString::CmpNoCase(const wxString& s) const
1076 {
1077 #if defined(__WXMSW__) && !wxUSE_UNICODE_UTF8
1078     // prefer to use CompareString() if available as it's more efficient than
1079     // doing it manual or even using wxStricmp() (see #10375)
1080     switch ( ::CompareString(LOCALE_USER_DEFAULT, NORM_IGNORECASE,
1081                              m_impl.c_str(), m_impl.length(),
1082                              s.m_impl.c_str(), s.m_impl.length()) )
1083     {
1084         case CSTR_LESS_THAN:
1085             return -1;
1086
1087         case CSTR_EQUAL:
1088             return 0;
1089
1090         case CSTR_GREATER_THAN:
1091             return 1;
1092
1093         default:
1094             wxFAIL_MSG( "unexpected CompareString() return value" );
1095             // fall through
1096
1097         case 0:
1098             wxLogLastError("CompareString");
1099             // use generic code below
1100     }
1101 #endif // __WXMSW__ && !wxUSE_UNICODE_UTF8
1102
1103     // do the comparison manually: notice that we can't use wxStricmp() as it
1104     // doesn't handle embedded NULs
1105
1106     // FIXME-UTF8: use wxUniChar::ToLower/ToUpper once added
1107     const_iterator i1 = begin();
1108     const_iterator end1 = end();
1109     const_iterator i2 = s.begin();
1110     const_iterator end2 = s.end();
1111
1112     for ( ; i1 != end1 && i2 != end2; ++i1, ++i2 )
1113     {
1114         wxUniChar lower1 = (wxChar)wxTolower(*i1);
1115         wxUniChar lower2 = (wxChar)wxTolower(*i2);
1116         if ( lower1 != lower2 )
1117             return lower1 < lower2 ? -1 : 1;
1118     }
1119
1120     size_t len1 = length();
1121     size_t len2 = s.length();
1122
1123     if ( len1 < len2 )
1124         return -1;
1125     else if ( len1 > len2 )
1126         return 1;
1127     return 0;
1128 }
1129
1130
1131 #if wxUSE_UNICODE
1132
1133 #ifdef __MWERKS__
1134 #ifndef __SCHAR_MAX__
1135 #define __SCHAR_MAX__ 127
1136 #endif
1137 #endif
1138
1139 wxString wxString::FromAscii(const char *ascii, size_t len)
1140 {
1141     if (!ascii || len == 0)
1142        return wxEmptyString;
1143
1144     wxString res;
1145
1146     {
1147         wxStringInternalBuffer buf(res, len);
1148         wxStringCharType *dest = buf;
1149
1150         for ( ; len > 0; --len )
1151         {
1152             unsigned char c = (unsigned char)*ascii++;
1153             wxASSERT_MSG( c < 0x80,
1154                           wxT("Non-ASCII value passed to FromAscii().") );
1155
1156             *dest++ = (wchar_t)c;
1157         }
1158     }
1159
1160     return res;
1161 }
1162
1163 wxString wxString::FromAscii(const char *ascii)
1164 {
1165     return FromAscii(ascii, wxStrlen(ascii));
1166 }
1167
1168 wxString wxString::FromAscii(char ascii)
1169 {
1170     // What do we do with '\0' ?
1171
1172     unsigned char c = (unsigned char)ascii;
1173
1174     wxASSERT_MSG( c < 0x80, wxT("Non-ASCII value passed to FromAscii().") );
1175
1176     // NB: the cast to wchar_t causes interpretation of 'ascii' as Latin1 value
1177     return wxString(wxUniChar((wchar_t)c));
1178 }
1179
1180 const wxScopedCharBuffer wxString::ToAscii() const
1181 {
1182     // this will allocate enough space for the terminating NUL too
1183     wxCharBuffer buffer(length());
1184     char *dest = buffer.data();
1185
1186     for ( const_iterator i = begin(); i != end(); ++i )
1187     {
1188         wxUniChar c(*i);
1189         // FIXME-UTF8: unify substituted char ('_') with wxUniChar ('?')
1190         *dest++ = c.IsAscii() ? (char)c : '_';
1191
1192         // the output string can't have embedded NULs anyhow, so we can safely
1193         // stop at first of them even if we do have any
1194         if ( !c )
1195             break;
1196     }
1197
1198     return buffer;
1199 }
1200
1201 #endif // wxUSE_UNICODE
1202
1203 // extract string of length nCount starting at nFirst
1204 wxString wxString::Mid(size_t nFirst, size_t nCount) const
1205 {
1206     size_t nLen = length();
1207
1208     // default value of nCount is npos and means "till the end"
1209     if ( nCount == npos )
1210     {
1211         nCount = nLen - nFirst;
1212     }
1213
1214     // out-of-bounds requests return sensible things
1215     if ( nFirst + nCount > nLen )
1216     {
1217         nCount = nLen - nFirst;
1218     }
1219
1220     if ( nFirst > nLen )
1221     {
1222         // AllocCopy() will return empty string
1223         return wxEmptyString;
1224     }
1225
1226     wxString dest(*this, nFirst, nCount);
1227     if ( dest.length() != nCount )
1228     {
1229         wxFAIL_MSG( wxT("out of memory in wxString::Mid") );
1230     }
1231
1232     return dest;
1233 }
1234
1235 // check that the string starts with prefix and return the rest of the string
1236 // in the provided pointer if it is not NULL, otherwise return false
1237 bool wxString::StartsWith(const wxString& prefix, wxString *rest) const
1238 {
1239     if ( compare(0, prefix.length(), prefix) != 0 )
1240         return false;
1241
1242     if ( rest )
1243     {
1244         // put the rest of the string into provided pointer
1245         rest->assign(*this, prefix.length(), npos);
1246     }
1247
1248     return true;
1249 }
1250
1251
1252 // check that the string ends with suffix and return the rest of it in the
1253 // provided pointer if it is not NULL, otherwise return false
1254 bool wxString::EndsWith(const wxString& suffix, wxString *rest) const
1255 {
1256     int start = length() - suffix.length();
1257
1258     if ( start < 0 || compare(start, npos, suffix) != 0 )
1259         return false;
1260
1261     if ( rest )
1262     {
1263         // put the rest of the string into provided pointer
1264         rest->assign(*this, 0, start);
1265     }
1266
1267     return true;
1268 }
1269
1270
1271 // extract nCount last (rightmost) characters
1272 wxString wxString::Right(size_t nCount) const
1273 {
1274   if ( nCount > length() )
1275     nCount = length();
1276
1277   wxString dest(*this, length() - nCount, nCount);
1278   if ( dest.length() != nCount ) {
1279     wxFAIL_MSG( wxT("out of memory in wxString::Right") );
1280   }
1281   return dest;
1282 }
1283
1284 // get all characters after the last occurrence of ch
1285 // (returns the whole string if ch not found)
1286 wxString wxString::AfterLast(wxUniChar ch) const
1287 {
1288   wxString str;
1289   int iPos = Find(ch, true);
1290   if ( iPos == wxNOT_FOUND )
1291     str = *this;
1292   else
1293     str.assign(*this, iPos + 1, npos);
1294
1295   return str;
1296 }
1297
1298 // extract nCount first (leftmost) characters
1299 wxString wxString::Left(size_t nCount) const
1300 {
1301   if ( nCount > length() )
1302     nCount = length();
1303
1304   wxString dest(*this, 0, nCount);
1305   if ( dest.length() != nCount ) {
1306     wxFAIL_MSG( wxT("out of memory in wxString::Left") );
1307   }
1308   return dest;
1309 }
1310
1311 // get all characters before the first occurrence of ch
1312 // (returns the whole string if ch not found)
1313 wxString wxString::BeforeFirst(wxUniChar ch) const
1314 {
1315   int iPos = Find(ch);
1316   if ( iPos == wxNOT_FOUND )
1317       iPos = length();
1318   return wxString(*this, 0, iPos);
1319 }
1320
1321 /// get all characters before the last occurrence of ch
1322 /// (returns empty string if ch not found)
1323 wxString wxString::BeforeLast(wxUniChar ch) const
1324 {
1325   wxString str;
1326   int iPos = Find(ch, true);
1327   if ( iPos != wxNOT_FOUND && iPos != 0 )
1328     str = wxString(c_str(), iPos);
1329
1330   return str;
1331 }
1332
1333 /// get all characters after the first occurrence of ch
1334 /// (returns empty string if ch not found)
1335 wxString wxString::AfterFirst(wxUniChar ch) const
1336 {
1337   wxString str;
1338   int iPos = Find(ch);
1339   if ( iPos != wxNOT_FOUND )
1340       str.assign(*this, iPos + 1, npos);
1341
1342   return str;
1343 }
1344
1345 // replace first (or all) occurrences of some substring with another one
1346 size_t wxString::Replace(const wxString& strOld,
1347                          const wxString& strNew, bool bReplaceAll)
1348 {
1349     // if we tried to replace an empty string we'd enter an infinite loop below
1350     wxCHECK_MSG( !strOld.empty(), 0,
1351                  wxT("wxString::Replace(): invalid parameter") );
1352
1353     wxSTRING_INVALIDATE_CACHE();
1354
1355     size_t uiCount = 0;   // count of replacements made
1356
1357     // optimize the special common case: replacement of one character by
1358     // another one (in UTF-8 case we can only do this for ASCII characters)
1359     //
1360     // benchmarks show that this special version is around 3 times faster
1361     // (depending on the proportion of matching characters and UTF-8/wchar_t
1362     // build)
1363     if ( strOld.m_impl.length() == 1 && strNew.m_impl.length() == 1 )
1364     {
1365         const wxStringCharType chOld = strOld.m_impl[0],
1366                                chNew = strNew.m_impl[0];
1367
1368         // this loop is the simplified version of the one below
1369         for ( size_t pos = 0; ; )
1370         {
1371             pos = m_impl.find(chOld, pos);
1372             if ( pos == npos )
1373                 break;
1374
1375             m_impl[pos++] = chNew;
1376
1377             uiCount++;
1378
1379             if ( !bReplaceAll )
1380                 break;
1381         }
1382     }
1383     else if ( !bReplaceAll)
1384     {
1385         size_t pos = m_impl.find(strOld, 0);
1386         if ( pos != npos )
1387         {
1388             m_impl.replace(pos, strOld.m_impl.length(), strNew.m_impl);
1389             uiCount = 1;
1390         }
1391     }
1392     else // replace all occurrences
1393     {
1394         const size_t uiOldLen = strOld.m_impl.length();
1395         const size_t uiNewLen = strNew.m_impl.length();
1396
1397         // first scan the string to find all positions at which the replacement
1398         // should be made
1399         wxVector<size_t> replacePositions;
1400
1401         size_t pos;
1402         for ( pos = m_impl.find(strOld.m_impl, 0);
1403               pos != npos;
1404               pos = m_impl.find(strOld.m_impl, pos + uiOldLen))
1405         {
1406             replacePositions.push_back(pos);
1407             ++uiCount;
1408         }
1409
1410         if ( !uiCount )
1411             return 0;
1412
1413         // allocate enough memory for the whole new string
1414         wxString tmp;
1415         tmp.m_impl.reserve(m_impl.length() + uiCount*(uiNewLen - uiOldLen));
1416
1417         // copy this string to tmp doing replacements on the fly
1418         size_t replNum = 0;
1419         for ( pos = 0; replNum < uiCount; replNum++ )
1420         {
1421             const size_t nextReplPos = replacePositions[replNum];
1422
1423             if ( pos != nextReplPos )
1424             {
1425                 tmp.m_impl.append(m_impl, pos, nextReplPos - pos);
1426             }
1427
1428             tmp.m_impl.append(strNew.m_impl);
1429             pos = nextReplPos + uiOldLen;
1430         }
1431
1432         if ( pos != m_impl.length() )
1433         {
1434             // append the rest of the string unchanged
1435             tmp.m_impl.append(m_impl, pos, m_impl.length() - pos);
1436         }
1437
1438         swap(tmp);
1439     }
1440
1441     return uiCount;
1442 }
1443
1444 bool wxString::IsAscii() const
1445 {
1446     for ( const_iterator i = begin(); i != end(); ++i )
1447     {
1448         if ( !(*i).IsAscii() )
1449             return false;
1450     }
1451
1452     return true;
1453 }
1454
1455 bool wxString::IsWord() const
1456 {
1457     for ( const_iterator i = begin(); i != end(); ++i )
1458     {
1459         if ( !wxIsalpha(*i) )
1460             return false;
1461     }
1462
1463     return true;
1464 }
1465
1466 bool wxString::IsNumber() const
1467 {
1468     if ( empty() )
1469         return true;
1470
1471     const_iterator i = begin();
1472
1473     if ( *i == wxT('-') || *i == wxT('+') )
1474         ++i;
1475
1476     for ( ; i != end(); ++i )
1477     {
1478         if ( !wxIsdigit(*i) )
1479             return false;
1480     }
1481
1482     return true;
1483 }
1484
1485 wxString wxString::Strip(stripType w) const
1486 {
1487     wxString s = *this;
1488     if ( w & leading ) s.Trim(false);
1489     if ( w & trailing ) s.Trim(true);
1490     return s;
1491 }
1492
1493 // ---------------------------------------------------------------------------
1494 // case conversion
1495 // ---------------------------------------------------------------------------
1496
1497 wxString& wxString::MakeUpper()
1498 {
1499   for ( iterator it = begin(), en = end(); it != en; ++it )
1500     *it = (wxChar)wxToupper(*it);
1501
1502   return *this;
1503 }
1504
1505 wxString& wxString::MakeLower()
1506 {
1507   for ( iterator it = begin(), en = end(); it != en; ++it )
1508     *it = (wxChar)wxTolower(*it);
1509
1510   return *this;
1511 }
1512
1513 wxString& wxString::MakeCapitalized()
1514 {
1515     const iterator en = end();
1516     iterator it = begin();
1517     if ( it != en )
1518     {
1519         *it = (wxChar)wxToupper(*it);
1520         for ( ++it; it != en; ++it )
1521             *it = (wxChar)wxTolower(*it);
1522     }
1523
1524     return *this;
1525 }
1526
1527 // ---------------------------------------------------------------------------
1528 // trimming and padding
1529 // ---------------------------------------------------------------------------
1530
1531 // some compilers (VC++ 6.0 not to name them) return true for a call to
1532 // isspace('\xEA') in the C locale which seems to be broken to me, but we have
1533 // to live with this by checking that the character is a 7 bit one - even if
1534 // this may fail to detect some spaces (I don't know if Unicode doesn't have
1535 // space-like symbols somewhere except in the first 128 chars), it is arguably
1536 // still better than trimming away accented letters
1537 inline int wxSafeIsspace(wxChar ch) { return (ch < 127) && wxIsspace(ch); }
1538
1539 // trims spaces (in the sense of isspace) from left or right side
1540 wxString& wxString::Trim(bool bFromRight)
1541 {
1542     // first check if we're going to modify the string at all
1543     if ( !empty() &&
1544          (
1545           (bFromRight && wxSafeIsspace(GetChar(length() - 1))) ||
1546           (!bFromRight && wxSafeIsspace(GetChar(0u)))
1547          )
1548        )
1549     {
1550         if ( bFromRight )
1551         {
1552             // find last non-space character
1553             reverse_iterator psz = rbegin();
1554             while ( (psz != rend()) && wxSafeIsspace(*psz) )
1555                 ++psz;
1556
1557             // truncate at trailing space start
1558             erase(psz.base(), end());
1559         }
1560         else
1561         {
1562             // find first non-space character
1563             iterator psz = begin();
1564             while ( (psz != end()) && wxSafeIsspace(*psz) )
1565                 ++psz;
1566
1567             // fix up data and length
1568             erase(begin(), psz);
1569         }
1570     }
1571
1572     return *this;
1573 }
1574
1575 // adds nCount characters chPad to the string from either side
1576 wxString& wxString::Pad(size_t nCount, wxUniChar chPad, bool bFromRight)
1577 {
1578     wxString s(chPad, nCount);
1579
1580     if ( bFromRight )
1581         *this += s;
1582     else
1583     {
1584         s += *this;
1585         swap(s);
1586     }
1587
1588     return *this;
1589 }
1590
1591 // truncate the string
1592 wxString& wxString::Truncate(size_t uiLen)
1593 {
1594     if ( uiLen < length() )
1595     {
1596         erase(begin() + uiLen, end());
1597     }
1598     //else: nothing to do, string is already short enough
1599
1600     return *this;
1601 }
1602
1603 // ---------------------------------------------------------------------------
1604 // finding (return wxNOT_FOUND if not found and index otherwise)
1605 // ---------------------------------------------------------------------------
1606
1607 // find a character
1608 int wxString::Find(wxUniChar ch, bool bFromEnd) const
1609 {
1610     size_type idx = bFromEnd ? find_last_of(ch) : find_first_of(ch);
1611
1612     return (idx == npos) ? wxNOT_FOUND : (int)idx;
1613 }
1614
1615 // ----------------------------------------------------------------------------
1616 // conversion to numbers
1617 // ----------------------------------------------------------------------------
1618
1619 // The implementation of all the functions below is exactly the same so factor
1620 // it out. Note that number extraction works correctly on UTF-8 strings, so
1621 // we can use wxStringCharType and wx_str() for maximum efficiency.
1622
1623 #ifndef __WXWINCE__
1624     #define DO_IF_NOT_WINCE(x) x
1625 #else
1626     #define DO_IF_NOT_WINCE(x)
1627 #endif
1628
1629 #define WX_STRING_TO_X_TYPE_START                                           \
1630     wxCHECK_MSG( pVal, false, wxT("NULL output pointer") );                  \
1631     DO_IF_NOT_WINCE( errno = 0; )                                           \
1632     const wxStringCharType *start = wx_str();                               \
1633     wxStringCharType *end;
1634
1635 // notice that we return false without modifying the output parameter at all if
1636 // nothing could be parsed but we do modify it and return false then if we did
1637 // parse something successfully but not the entire string
1638 #define WX_STRING_TO_X_TYPE_END                                             \
1639     if ( end == start DO_IF_NOT_WINCE(|| errno == ERANGE) )                 \
1640         return false;                                                       \
1641     *pVal = val;                                                            \
1642     return !*end;
1643
1644 bool wxString::ToLong(long *pVal, int base) const
1645 {
1646     wxASSERT_MSG( !base || (base > 1 && base <= 36), wxT("invalid base") );
1647
1648     WX_STRING_TO_X_TYPE_START
1649     long val = wxStrtol(start, &end, base);
1650     WX_STRING_TO_X_TYPE_END
1651 }
1652
1653 bool wxString::ToULong(unsigned long *pVal, int base) const
1654 {
1655     wxASSERT_MSG( !base || (base > 1 && base <= 36), wxT("invalid base") );
1656
1657     WX_STRING_TO_X_TYPE_START
1658     unsigned long val = wxStrtoul(start, &end, base);
1659     WX_STRING_TO_X_TYPE_END
1660 }
1661
1662 bool wxString::ToLongLong(wxLongLong_t *pVal, int base) const
1663 {
1664     wxASSERT_MSG( !base || (base > 1 && base <= 36), wxT("invalid base") );
1665
1666     WX_STRING_TO_X_TYPE_START
1667     wxLongLong_t val = wxStrtoll(start, &end, base);
1668     WX_STRING_TO_X_TYPE_END
1669 }
1670
1671 bool wxString::ToULongLong(wxULongLong_t *pVal, int base) const
1672 {
1673     wxASSERT_MSG( !base || (base > 1 && base <= 36), wxT("invalid base") );
1674
1675     WX_STRING_TO_X_TYPE_START
1676     wxULongLong_t val = wxStrtoull(start, &end, base);
1677     WX_STRING_TO_X_TYPE_END
1678 }
1679
1680 bool wxString::ToDouble(double *pVal) const
1681 {
1682     WX_STRING_TO_X_TYPE_START
1683     double val = wxStrtod(start, &end);
1684     WX_STRING_TO_X_TYPE_END
1685 }
1686
1687 #if wxUSE_XLOCALE
1688
1689 bool wxString::ToCLong(long *pVal, int base) const
1690 {
1691     wxASSERT_MSG( !base || (base > 1 && base <= 36), wxT("invalid base") );
1692
1693     WX_STRING_TO_X_TYPE_START
1694 #if (wxUSE_UNICODE_UTF8 || !wxUSE_UNICODE) && defined(wxHAS_XLOCALE_SUPPORT)
1695     long val = wxStrtol_lA(start, &end, base, wxCLocale);
1696 #else
1697     long val = wxStrtol_l(start, &end, base, wxCLocale);
1698 #endif
1699     WX_STRING_TO_X_TYPE_END
1700 }
1701
1702 bool wxString::ToCULong(unsigned long *pVal, int base) const
1703 {
1704     wxASSERT_MSG( !base || (base > 1 && base <= 36), wxT("invalid base") );
1705
1706     WX_STRING_TO_X_TYPE_START
1707 #if (wxUSE_UNICODE_UTF8 || !wxUSE_UNICODE) && defined(wxHAS_XLOCALE_SUPPORT)
1708     unsigned long val = wxStrtoul_lA(start, &end, base, wxCLocale);
1709 #else
1710     unsigned long val = wxStrtoul_l(start, &end, base, wxCLocale);
1711 #endif
1712     WX_STRING_TO_X_TYPE_END
1713 }
1714
1715 bool wxString::ToCDouble(double *pVal) const
1716 {
1717     WX_STRING_TO_X_TYPE_START
1718 #if (wxUSE_UNICODE_UTF8 || !wxUSE_UNICODE) && defined(wxHAS_XLOCALE_SUPPORT)
1719     double val = wxStrtod_lA(start, &end, wxCLocale);
1720 #else
1721     double val = wxStrtod_l(start, &end, wxCLocale);
1722 #endif
1723     WX_STRING_TO_X_TYPE_END
1724 }
1725
1726 #endif  // wxUSE_XLOCALE
1727
1728 // ---------------------------------------------------------------------------
1729 // formatted output
1730 // ---------------------------------------------------------------------------
1731
1732 #if !wxUSE_UTF8_LOCALE_ONLY
1733 /* static */
1734 #ifdef wxNEEDS_WXSTRING_PRINTF_MIXIN
1735 wxString wxStringPrintfMixinBase::DoFormatWchar(const wxChar *format, ...)
1736 #else
1737 wxString wxString::DoFormatWchar(const wxChar *format, ...)
1738 #endif
1739 {
1740     va_list argptr;
1741     va_start(argptr, format);
1742
1743     wxString s;
1744     s.PrintfV(format, argptr);
1745
1746     va_end(argptr);
1747
1748     return s;
1749 }
1750 #endif // !wxUSE_UTF8_LOCALE_ONLY
1751
1752 #if wxUSE_UNICODE_UTF8
1753 /* static */
1754 wxString wxString::DoFormatUtf8(const char *format, ...)
1755 {
1756     va_list argptr;
1757     va_start(argptr, format);
1758
1759     wxString s;
1760     s.PrintfV(format, argptr);
1761
1762     va_end(argptr);
1763
1764     return s;
1765 }
1766 #endif // wxUSE_UNICODE_UTF8
1767
1768 /* static */
1769 wxString wxString::FormatV(const wxString& format, va_list argptr)
1770 {
1771     wxString s;
1772     s.PrintfV(format, argptr);
1773     return s;
1774 }
1775
1776 #if !wxUSE_UTF8_LOCALE_ONLY
1777 #ifdef wxNEEDS_WXSTRING_PRINTF_MIXIN
1778 int wxStringPrintfMixinBase::DoPrintfWchar(const wxChar *format, ...)
1779 #else
1780 int wxString::DoPrintfWchar(const wxChar *format, ...)
1781 #endif
1782 {
1783     va_list argptr;
1784     va_start(argptr, format);
1785
1786 #ifdef wxNEEDS_WXSTRING_PRINTF_MIXIN
1787     // get a pointer to the wxString instance; we have to use dynamic_cast<>
1788     // because it's the only cast that works safely for downcasting when
1789     // multiple inheritance is used:
1790     wxString *str = static_cast<wxString*>(this);
1791 #else
1792     wxString *str = this;
1793 #endif
1794
1795     int iLen = str->PrintfV(format, argptr);
1796
1797     va_end(argptr);
1798
1799     return iLen;
1800 }
1801 #endif // !wxUSE_UTF8_LOCALE_ONLY
1802
1803 #if wxUSE_UNICODE_UTF8
1804 int wxString::DoPrintfUtf8(const char *format, ...)
1805 {
1806     va_list argptr;
1807     va_start(argptr, format);
1808
1809     int iLen = PrintfV(format, argptr);
1810
1811     va_end(argptr);
1812
1813     return iLen;
1814 }
1815 #endif // wxUSE_UNICODE_UTF8
1816
1817 /*
1818     Uses wxVsnprintf and places the result into the this string.
1819
1820     In ANSI build, wxVsnprintf is effectively vsnprintf but in Unicode build
1821     it is vswprintf.  Due to a discrepancy between vsnprintf and vswprintf in
1822     the ISO C99 (and thus SUSv3) standard the return value for the case of
1823     an undersized buffer is inconsistent.  For conforming vsnprintf
1824     implementations the function must return the number of characters that
1825     would have been printed had the buffer been large enough.  For conforming
1826     vswprintf implementations the function must return a negative number
1827     and set errno.
1828
1829     What vswprintf sets errno to is undefined but Darwin seems to set it to
1830     EOVERFLOW.  The only expected errno are EILSEQ and EINVAL.  Both of
1831     those are defined in the standard and backed up by several conformance
1832     statements.  Note that ENOMEM mentioned in the manual page does not
1833     apply to swprintf, only wprintf and fwprintf.
1834
1835     Official manual page:
1836     http://www.opengroup.org/onlinepubs/009695399/functions/swprintf.html
1837
1838     Some conformance statements (AIX, Solaris):
1839     http://www.opengroup.org/csq/view.mhtml?RID=ibm%2FSD1%2F3
1840     http://www.theopengroup.org/csq/view.mhtml?norationale=1&noreferences=1&RID=Fujitsu%2FSE2%2F10
1841
1842     Since EILSEQ and EINVAL are rather common but EOVERFLOW is not and since
1843     EILSEQ and EINVAL are specifically defined to mean the error is other than
1844     an undersized buffer and no other errno are defined we treat those two
1845     as meaning hard errors and everything else gets the old behavior which
1846     is to keep looping and increasing buffer size until the function succeeds.
1847
1848     In practice it's impossible to determine before compilation which behavior
1849     may be used.  The vswprintf function may have vsnprintf-like behavior or
1850     vice-versa.  Behavior detected on one release can theoretically change
1851     with an updated release.  Not to mention that configure testing for it
1852     would require the test to be run on the host system, not the build system
1853     which makes cross compilation difficult. Therefore, we make no assumptions
1854     about behavior and try our best to handle every known case, including the
1855     case where wxVsnprintf returns a negative number and fails to set errno.
1856
1857     There is yet one more non-standard implementation and that is our own.
1858     Fortunately, that can be detected at compile-time.
1859
1860     On top of all that, ISO C99 explicitly defines snprintf to write a null
1861     character to the last position of the specified buffer.  That would be at
1862     at the given buffer size minus 1.  It is supposed to do this even if it
1863     turns out that the buffer is sized too small.
1864
1865     Darwin (tested on 10.5) follows the C99 behavior exactly.
1866
1867     Glibc 2.6 almost follows the C99 behavior except vswprintf never sets
1868     errno even when it fails.  However, it only seems to ever fail due
1869     to an undersized buffer.
1870 */
1871 #if wxUSE_UNICODE_UTF8
1872 template<typename BufferType>
1873 #else
1874 // we only need one version in non-UTF8 builds and at least two Windows
1875 // compilers have problems with this function template, so use just one
1876 // normal function here
1877 #endif
1878 static int DoStringPrintfV(wxString& str,
1879                            const wxString& format, va_list argptr)
1880 {
1881     int size = 1024;
1882
1883     for ( ;; )
1884     {
1885 #if wxUSE_UNICODE_UTF8
1886         BufferType tmp(str, size + 1);
1887         typename BufferType::CharType *buf = tmp;
1888 #else
1889         wxStringBuffer tmp(str, size + 1);
1890         wxChar *buf = tmp;
1891 #endif
1892
1893         if ( !buf )
1894         {
1895             // out of memory
1896
1897             // in UTF-8 build, leaving uninitialized junk in the buffer
1898             // could result in invalid non-empty UTF-8 string, so just
1899             // reset the string to empty on failure:
1900             buf[0] = '\0';
1901             return -1;
1902         }
1903
1904         // wxVsnprintf() may modify the original arg pointer, so pass it
1905         // only a copy
1906         va_list argptrcopy;
1907         wxVaCopy(argptrcopy, argptr);
1908
1909 #ifndef __WXWINCE__
1910         // Set errno to 0 to make it determinate if wxVsnprintf fails to set it.
1911         errno = 0;
1912 #endif
1913         int len = wxVsnprintf(buf, size, format, argptrcopy);
1914         va_end(argptrcopy);
1915
1916         // some implementations of vsnprintf() don't NUL terminate
1917         // the string if there is not enough space for it so
1918         // always do it manually
1919         // FIXME: This really seems to be the wrong and would be an off-by-one
1920         // bug except the code above allocates an extra character.
1921         buf[size] = wxT('\0');
1922
1923         // vsnprintf() may return either -1 (traditional Unix behaviour) or the
1924         // total number of characters which would have been written if the
1925         // buffer were large enough (newer standards such as Unix98)
1926         if ( len < 0 )
1927         {
1928             // NB: wxVsnprintf() may call either wxCRT_VsnprintfW or
1929             //     wxCRT_VsnprintfA in UTF-8 build; wxUSE_WXVSNPRINTF
1930             //     is true if *both* of them use our own implementation,
1931             //     otherwise we can't be sure
1932 #if wxUSE_WXVSNPRINTF
1933             // we know that our own implementation of wxVsnprintf() returns -1
1934             // only for a format error - thus there's something wrong with
1935             // the user's format string
1936             buf[0] = '\0';
1937             return -1;
1938 #else // possibly using system version
1939             // assume it only returns error if there is not enough space, but
1940             // as we don't know how much we need, double the current size of
1941             // the buffer
1942 #ifndef __WXWINCE__
1943             if( (errno == EILSEQ) || (errno == EINVAL) )
1944             // If errno was set to one of the two well-known hard errors
1945             // then fail immediately to avoid an infinite loop.
1946                 return -1;
1947             else
1948 #endif // __WXWINCE__
1949             // still not enough, as we don't know how much we need, double the
1950             // current size of the buffer
1951                 size *= 2;
1952 #endif // wxUSE_WXVSNPRINTF/!wxUSE_WXVSNPRINTF
1953         }
1954         else if ( len >= size )
1955         {
1956 #if wxUSE_WXVSNPRINTF
1957             // we know that our own implementation of wxVsnprintf() returns
1958             // size+1 when there's not enough space but that's not the size
1959             // of the required buffer!
1960             size *= 2;      // so we just double the current size of the buffer
1961 #else
1962             // some vsnprintf() implementations NUL-terminate the buffer and
1963             // some don't in len == size case, to be safe always add 1
1964             // FIXME: I don't quite understand this comment.  The vsnprintf
1965             // function is specifically defined to return the number of
1966             // characters printed not including the null terminator.
1967             // So OF COURSE you need to add 1 to get the right buffer size.
1968             // The following line is definitely correct, no question.
1969             size = len + 1;
1970 #endif
1971         }
1972         else // ok, there was enough space
1973         {
1974             break;
1975         }
1976     }
1977
1978     // we could have overshot
1979     str.Shrink();
1980
1981     return str.length();
1982 }
1983
1984 int wxString::PrintfV(const wxString& format, va_list argptr)
1985 {
1986 #if wxUSE_UNICODE_UTF8
1987     #if wxUSE_STL_BASED_WXSTRING
1988         typedef wxStringTypeBuffer<char> Utf8Buffer;
1989     #else
1990         typedef wxStringInternalBuffer Utf8Buffer;
1991     #endif
1992 #endif
1993
1994 #if wxUSE_UTF8_LOCALE_ONLY
1995     return DoStringPrintfV<Utf8Buffer>(*this, format, argptr);
1996 #else
1997     #if wxUSE_UNICODE_UTF8
1998     if ( wxLocaleIsUtf8 )
1999         return DoStringPrintfV<Utf8Buffer>(*this, format, argptr);
2000     else
2001         // wxChar* version
2002         return DoStringPrintfV<wxStringBuffer>(*this, format, argptr);
2003     #else
2004         return DoStringPrintfV(*this, format, argptr);
2005     #endif // UTF8/WCHAR
2006 #endif
2007 }
2008
2009 // ----------------------------------------------------------------------------
2010 // misc other operations
2011 // ----------------------------------------------------------------------------
2012
2013 // returns true if the string matches the pattern which may contain '*' and
2014 // '?' metacharacters (as usual, '?' matches any character and '*' any number
2015 // of them)
2016 bool wxString::Matches(const wxString& mask) const
2017 {
2018     // I disable this code as it doesn't seem to be faster (in fact, it seems
2019     // to be much slower) than the old, hand-written code below and using it
2020     // here requires always linking with libregex even if the user code doesn't
2021     // use it
2022 #if 0 // wxUSE_REGEX
2023     // first translate the shell-like mask into a regex
2024     wxString pattern;
2025     pattern.reserve(wxStrlen(pszMask));
2026
2027     pattern += wxT('^');
2028     while ( *pszMask )
2029     {
2030         switch ( *pszMask )
2031         {
2032             case wxT('?'):
2033                 pattern += wxT('.');
2034                 break;
2035
2036             case wxT('*'):
2037                 pattern += wxT(".*");
2038                 break;
2039
2040             case wxT('^'):
2041             case wxT('.'):
2042             case wxT('$'):
2043             case wxT('('):
2044             case wxT(')'):
2045             case wxT('|'):
2046             case wxT('+'):
2047             case wxT('\\'):
2048                 // these characters are special in a RE, quote them
2049                 // (however note that we don't quote '[' and ']' to allow
2050                 // using them for Unix shell like matching)
2051                 pattern += wxT('\\');
2052                 // fall through
2053
2054             default:
2055                 pattern += *pszMask;
2056         }
2057
2058         pszMask++;
2059     }
2060     pattern += wxT('$');
2061
2062     // and now use it
2063     return wxRegEx(pattern, wxRE_NOSUB | wxRE_EXTENDED).Matches(c_str());
2064 #else // !wxUSE_REGEX
2065   // TODO: this is, of course, awfully inefficient...
2066
2067   // FIXME-UTF8: implement using iterators, remove #if
2068 #if wxUSE_UNICODE_UTF8
2069   const wxScopedWCharBuffer maskBuf = mask.wc_str();
2070   const wxScopedWCharBuffer txtBuf = wc_str();
2071   const wxChar *pszMask = maskBuf.data();
2072   const wxChar *pszTxt = txtBuf.data();
2073 #else
2074   const wxChar *pszMask = mask.wx_str();
2075   // the char currently being checked
2076   const wxChar *pszTxt = wx_str();
2077 #endif
2078
2079   // the last location where '*' matched
2080   const wxChar *pszLastStarInText = NULL;
2081   const wxChar *pszLastStarInMask = NULL;
2082
2083 match:
2084   for ( ; *pszMask != wxT('\0'); pszMask++, pszTxt++ ) {
2085     switch ( *pszMask ) {
2086       case wxT('?'):
2087         if ( *pszTxt == wxT('\0') )
2088           return false;
2089
2090         // pszTxt and pszMask will be incremented in the loop statement
2091
2092         break;
2093
2094       case wxT('*'):
2095         {
2096           // remember where we started to be able to backtrack later
2097           pszLastStarInText = pszTxt;
2098           pszLastStarInMask = pszMask;
2099
2100           // ignore special chars immediately following this one
2101           // (should this be an error?)
2102           while ( *pszMask == wxT('*') || *pszMask == wxT('?') )
2103             pszMask++;
2104
2105           // if there is nothing more, match
2106           if ( *pszMask == wxT('\0') )
2107             return true;
2108
2109           // are there any other metacharacters in the mask?
2110           size_t uiLenMask;
2111           const wxChar *pEndMask = wxStrpbrk(pszMask, wxT("*?"));
2112
2113           if ( pEndMask != NULL ) {
2114             // we have to match the string between two metachars
2115             uiLenMask = pEndMask - pszMask;
2116           }
2117           else {
2118             // we have to match the remainder of the string
2119             uiLenMask = wxStrlen(pszMask);
2120           }
2121
2122           wxString strToMatch(pszMask, uiLenMask);
2123           const wxChar* pMatch = wxStrstr(pszTxt, strToMatch);
2124           if ( pMatch == NULL )
2125             return false;
2126
2127           // -1 to compensate "++" in the loop
2128           pszTxt = pMatch + uiLenMask - 1;
2129           pszMask += uiLenMask - 1;
2130         }
2131         break;
2132
2133       default:
2134         if ( *pszMask != *pszTxt )
2135           return false;
2136         break;
2137     }
2138   }
2139
2140   // match only if nothing left
2141   if ( *pszTxt == wxT('\0') )
2142     return true;
2143
2144   // if we failed to match, backtrack if we can
2145   if ( pszLastStarInText ) {
2146     pszTxt = pszLastStarInText + 1;
2147     pszMask = pszLastStarInMask;
2148
2149     pszLastStarInText = NULL;
2150
2151     // don't bother resetting pszLastStarInMask, it's unnecessary
2152
2153     goto match;
2154   }
2155
2156   return false;
2157 #endif // wxUSE_REGEX/!wxUSE_REGEX
2158 }
2159
2160 // Count the number of chars
2161 int wxString::Freq(wxUniChar ch) const
2162 {
2163     int count = 0;
2164     for ( const_iterator i = begin(); i != end(); ++i )
2165     {
2166         if ( *i == ch )
2167             count ++;
2168     }
2169     return count;
2170 }
2171