src/common/string.cpp

   1 /////////////////////////////////////////////////////////////////////////////
   2 // Name:        string.cpp
   3 // Purpose:     wxString class
   4 // Author:      Vadim Zeitlin
   5 // Modified by:
   6 // Created:     29/01/98
   7 // RCS-ID:      $Id$
   8 // Copyright:   (c) 1998 Vadim Zeitlin <zeitlin@dptmaths.ens-cachan.fr>
   9 // Licence:     wxWindows license
  10 /////////////////////////////////////////////////////////////////////////////
  11
  12 #ifdef __GNUG__
  13   #pragma implementation "string.h"
  14 #endif
  15
  16 /*
  17  * About ref counting:
  18  *  1) all empty strings use g_strEmpty, nRefs = -1 (set in Init())
  19  *  2) AllocBuffer() sets nRefs to 1, Lock() increments it by one
  20  *  3) Unlock() decrements nRefs and frees memory if it goes to 0
  21  */
  22
  23 // ===========================================================================
  24 // headers, declarations, constants
  25 // ===========================================================================
  26
  27 // For compilers that support precompilation, includes "wx.h".
  28 #include "wx/wxprec.h"
  29
  30 #ifdef __BORLANDC__
  31   #pragma hdrstop
  32 #endif
  33
  34 #ifndef WX_PRECOMP
  35   #include "wx/defs.h"
  36   #include "wx/string.h"
  37   #include "wx/intl.h"
  38   #include "wx/thread.h"
  39 #endif
  40
  41 #include <ctype.h>
  42 #include <string.h>
  43 #include <stdlib.h>
  44
  45 #ifdef __SALFORDC__
  46   #include <clib.h>
  47 #endif
  48
  49 // allocating extra space for each string consumes more memory but speeds up
  50 // the concatenation operations (nLen is the current string's length)
  51 // NB: EXTRA_ALLOC must be >= 0!
  52 #define EXTRA_ALLOC       (19 - nLen % 16)
  53
  54 // ---------------------------------------------------------------------------
  55 // static class variables definition
  56 // ---------------------------------------------------------------------------
  57
  58 #if defined(__VISAGECPP__) && __IBMCPP__ >= 400
  59 // must define this static for VA or else you get multiply defined symbols
  60 // everywhere
  61 const unsigned int wxSTRING_MAXLEN = UINT_MAX - 100;
  62 #endif // Visual Age
  63
  64 #ifdef  wxSTD_STRING_COMPATIBILITY
  65   const size_t wxString::npos = wxSTRING_MAXLEN;
  66 #endif // wxSTD_STRING_COMPATIBILITY
  67
  68 // ----------------------------------------------------------------------------
  69 // static data
  70 // ----------------------------------------------------------------------------
  71
  72 // for an empty string, GetStringData() will return this address: this
  73 // structure has the same layout as wxStringData and it's data() method will
  74 // return the empty string (dummy pointer)
  75 static const struct
  76 {
  77   wxStringData data;
  78   wxChar dummy;
  79 } g_strEmpty = { {-1, 0, 0}, wxT('\0') };
  80
  81 // empty C style string: points to 'string data' byte of g_strEmpty
  82 extern const wxChar WXDLLEXPORT *wxEmptyString = &g_strEmpty.dummy;
  83
  84 // ----------------------------------------------------------------------------
  85 // global functions
  86 // ----------------------------------------------------------------------------
  87
  88 #if defined(wxSTD_STRING_COMPATIBILITY) && wxUSE_STD_IOSTREAM
  89
  90 // MS Visual C++ version 5.0 provides the new STL headers as well as the old
  91 // iostream ones.
  92 //
  93 // ATTN: you can _not_ use both of these in the same program!
  94
  95 wxSTD istream& operator>>(wxSTD istream& is, wxString& WXUNUSED(str))
  96 {
  97 #if 0
  98   int w = is.width(0);
  99   if ( is.ipfx(0) ) {
 100     streambuf *sb = is.rdbuf();
 101     str.erase();
 102     while ( true ) {
 103       int ch = sb->sbumpc ();
 104       if ( ch == EOF ) {
 105         is.setstate(ios::eofbit);
 106         break;
 107       }
 108       else if ( isspace(ch) ) {
 109         sb->sungetc();
 110         break;
 111       }
 112
 113       str += ch;
 114       if ( --w == 1 )
 115         break;
 116     }
 117   }
 118
 119   is.isfx();
 120   if ( str.length() == 0 )
 121     is.setstate(ios::failbit);
 122 #endif
 123   return is;
 124 }
 125
 126 wxSTD ostream& operator<<(wxSTD ostream& os, const wxString& str)
 127 {
 128   os << str.c_str();
 129   return os;
 130 }
 131
 132 #endif  //std::string compatibility
 133
 134 // ----------------------------------------------------------------------------
 135 // private classes
 136 // ----------------------------------------------------------------------------
 137
 138 // this small class is used to gather statistics for performance tuning
 139 //#define WXSTRING_STATISTICS
 140 #ifdef  WXSTRING_STATISTICS
 141   class Averager
 142   {
 143   public:
 144     Averager(const wxChar *sz) { m_sz = sz; m_nTotal = m_nCount = 0; }
 145    ~Averager()
 146    { wxPrintf("wxString: average %s = %f\n", m_sz, ((float)m_nTotal)/m_nCount); }
 147
 148     void Add(size_t n) { m_nTotal += n; m_nCount++; }
 149
 150   private:
 151     size_t m_nCount, m_nTotal;
 152     const wxChar *m_sz;
 153   } g_averageLength("allocation size"),
 154     g_averageSummandLength("summand length"),
 155     g_averageConcatHit("hit probability in concat"),
 156     g_averageInitialLength("initial string length");
 157
 158   #define STATISTICS_ADD(av, val) g_average##av.Add(val)
 159 #else
 160   #define STATISTICS_ADD(av, val)
 161 #endif // WXSTRING_STATISTICS
 162
 163 // ===========================================================================
 164 // wxString class core
 165 // ===========================================================================
 166
 167 // ---------------------------------------------------------------------------
 168 // construction
 169 // ---------------------------------------------------------------------------
 170
 171 // constructs string of <nLength> copies of character <ch>
 172 wxString::wxString(wxChar ch, size_t nLength)
 173 {
 174   Init();
 175
 176   if ( nLength > 0 ) {
 177     if ( !AllocBuffer(nLength) ) {
 178       wxFAIL_MSG( _T("out of memory in wxString::wxString") );
 179       return;
 180     }
 181
 182 #if wxUSE_UNICODE
 183     // memset only works on chars
 184     for ( size_t n = 0; n < nLength; n++ )
 185         m_pchData[n] = ch;
 186 #else
 187     memset(m_pchData, ch, nLength);
 188 #endif
 189   }
 190 }
 191
 192 // takes nLength elements of psz starting at nPos
 193 void wxString::InitWith(const wxChar *psz, size_t nPos, size_t nLength)
 194 {
 195   Init();
 196
 197   // if the length is not given, assume the string to be NUL terminated
 198   if ( nLength == wxSTRING_MAXLEN ) {
 199     wxASSERT_MSG( nPos <= wxStrlen(psz), _T("index out of bounds") );
 200
 201     nLength = wxStrlen(psz + nPos);
 202   }
 203
 204   STATISTICS_ADD(InitialLength, nLength);
 205
 206   if ( nLength > 0 ) {
 207     // trailing '\0' is written in AllocBuffer()
 208     if ( !AllocBuffer(nLength) ) {
 209       wxFAIL_MSG( _T("out of memory in wxString::InitWith") );
 210       return;
 211     }
 212     memcpy(m_pchData, psz + nPos, nLength*sizeof(wxChar));
 213   }
 214 }
 215
 216 #ifdef  wxSTD_STRING_COMPATIBILITY
 217
 218 // poor man's iterators are "void *" pointers
 219 wxString::wxString(const void *pStart, const void *pEnd)
 220 {
 221   InitWith((const wxChar *)pStart, 0,
 222            (const wxChar *)pEnd - (const wxChar *)pStart);
 223 }
 224
 225 #endif  //std::string compatibility
 226
 227 #if wxUSE_UNICODE
 228
 229 // from multibyte string
 230 wxString::wxString(const char *psz, wxMBConv& conv, size_t nLength)
 231 {
 232   // first get necessary size
 233   size_t nLen = psz ? conv.MB2WC((wchar_t *) NULL, psz, 0) : 0;
 234
 235   // nLength is number of *Unicode* characters here!
 236   if ((nLen != (size_t)-1) && (nLen > nLength))
 237     nLen = nLength;
 238
 239   // empty?
 240   if ( (nLen != 0) && (nLen != (size_t)-1) ) {
 241     if ( !AllocBuffer(nLen) ) {
 242       wxFAIL_MSG( _T("out of memory in wxString::wxString") );
 243       return;
 244     }
 245     conv.MB2WC(m_pchData, psz, nLen);
 246   }
 247   else {
 248     Init();
 249   }
 250 }
 251
 252 #else // ANSI
 253
 254 #if wxUSE_WCHAR_T
 255 // from wide string
 256 wxString::wxString(const wchar_t *pwz, wxMBConv& conv, size_t nLength)
 257 {
 258   // first get necessary size
 259   size_t nLen = 0;
 260   if (pwz)
 261   {
 262     if (nLength == wxSTRING_MAXLEN)
 263       nLen = conv.WC2MB((char *) NULL, pwz, 0);
 264     else
 265       nLen = nLength;
 266   }
 267
 268   // empty?
 269   if ( (nLen != 0) && (nLen != (size_t)-1) ) {
 270     if ( !AllocBuffer(nLen) ) {
 271       wxFAIL_MSG( _T("out of memory in wxString::wxString") );
 272       return;
 273     }
 274     conv.WC2MB(m_pchData, pwz, nLen);
 275   }
 276   else {
 277     Init();
 278   }
 279 }
 280 #endif // wxUSE_WCHAR_T
 281
 282 #endif // Unicode/ANSI
 283
 284 // ---------------------------------------------------------------------------
 285 // memory allocation
 286 // ---------------------------------------------------------------------------
 287
 288 // allocates memory needed to store a C string of length nLen
 289 bool wxString::AllocBuffer(size_t nLen)
 290 {
 291   // allocating 0 sized buffer doesn't make sense, all empty strings should
 292   // reuse g_strEmpty
 293   wxASSERT( nLen >  0 );
 294
 295   // make sure that we don't overflow
 296   wxASSERT( nLen < (INT_MAX / sizeof(wxChar)) -
 297                    (sizeof(wxStringData) + EXTRA_ALLOC + 1) );
 298
 299   STATISTICS_ADD(Length, nLen);
 300
 301   // allocate memory:
 302   // 1) one extra character for '\0' termination
 303   // 2) sizeof(wxStringData) for housekeeping info
 304   wxStringData* pData = (wxStringData*)
 305     malloc(sizeof(wxStringData) + (nLen + EXTRA_ALLOC + 1)*sizeof(wxChar));
 306
 307   if ( pData == NULL ) {
 308     // allocation failures are handled by the caller
 309     return FALSE;
 310   }
 311
 312   pData->nRefs        = 1;
 313   pData->nDataLength  = nLen;
 314   pData->nAllocLength = nLen + EXTRA_ALLOC;
 315   m_pchData           = pData->data();  // data starts after wxStringData
 316   m_pchData[nLen]     = wxT('\0');
 317   return TRUE;
 318 }
 319
 320 // must be called before changing this string
 321 bool wxString::CopyBeforeWrite()
 322 {
 323   wxStringData* pData = GetStringData();
 324
 325   if ( pData->IsShared() ) {
 326     pData->Unlock();                // memory not freed because shared
 327     size_t nLen = pData->nDataLength;
 328     if ( !AllocBuffer(nLen) ) {
 329       // allocation failures are handled by the caller
 330       return FALSE;
 331     }
 332     memcpy(m_pchData, pData->data(), nLen*sizeof(wxChar));
 333   }
 334
 335   wxASSERT( !GetStringData()->IsShared() );  // we must be the only owner
 336
 337   return TRUE;
 338 }
 339
 340 // must be called before replacing contents of this string
 341 bool wxString::AllocBeforeWrite(size_t nLen)
 342 {
 343   wxASSERT( nLen != 0 );  // doesn't make any sense
 344
 345   // must not share string and must have enough space
 346   wxStringData* pData = GetStringData();
 347   if ( pData->IsShared() || pData->IsEmpty() ) {
 348     // can't work with old buffer, get new one
 349     pData->Unlock();
 350     if ( !AllocBuffer(nLen) ) {
 351       // allocation failures are handled by the caller
 352       return FALSE;
 353     }
 354   }
 355   else {
 356     if ( nLen > pData->nAllocLength ) {
 357       // realloc the buffer instead of calling malloc() again, this is more
 358       // efficient
 359       STATISTICS_ADD(Length, nLen);
 360
 361       nLen += EXTRA_ALLOC;
 362
 363       pData = (wxStringData*)
 364           realloc(pData, sizeof(wxStringData) + (nLen + 1)*sizeof(wxChar));
 365
 366       if ( pData == NULL ) {
 367         // allocation failures are handled by the caller
 368         // keep previous data since reallocation failed
 369         return FALSE;
 370       }
 371
 372       pData->nAllocLength = nLen;
 373       m_pchData = pData->data();
 374     }
 375
 376     // now we have enough space, just update the string length
 377     pData->nDataLength = nLen;
 378   }
 379
 380   wxASSERT( !GetStringData()->IsShared() );  // we must be the only owner
 381
 382   return TRUE;
 383 }
 384
 385 // allocate enough memory for nLen characters
 386 bool wxString::Alloc(size_t nLen)
 387 {
 388   wxStringData *pData = GetStringData();
 389   if ( pData->nAllocLength <= nLen ) {
 390     if ( pData->IsEmpty() ) {
 391       nLen += EXTRA_ALLOC;
 392
 393       wxStringData* pData = (wxStringData*)
 394           malloc(sizeof(wxStringData) + (nLen + 1)*sizeof(wxChar));
 395
 396       if ( pData == NULL ) {
 397         // allocation failure handled by caller
 398         return FALSE;
 399       }
 400
 401       pData->nRefs = 1;
 402       pData->nDataLength = 0;
 403       pData->nAllocLength = nLen;
 404       m_pchData = pData->data();  // data starts after wxStringData
 405       m_pchData[0u] = wxT('\0');
 406     }
 407     else if ( pData->IsShared() ) {
 408       pData->Unlock();                // memory not freed because shared
 409       size_t nOldLen = pData->nDataLength;
 410       if ( !AllocBuffer(nLen) ) {
 411         // allocation failure handled by caller
 412         return FALSE;
 413       }
 414       memcpy(m_pchData, pData->data(), nOldLen*sizeof(wxChar));
 415     }
 416     else {
 417       nLen += EXTRA_ALLOC;
 418
 419       pData = (wxStringData *)
 420         realloc(pData, sizeof(wxStringData) + (nLen + 1)*sizeof(wxChar));
 421
 422       if ( pData == NULL ) {
 423         // allocation failure handled by caller
 424         // keep previous data since reallocation failed
 425         return FALSE;
 426       }
 427
 428       // it's not important if the pointer changed or not (the check for this
 429       // is not faster than assigning to m_pchData in all cases)
 430       pData->nAllocLength = nLen;
 431       m_pchData = pData->data();
 432     }
 433   }
 434   //else: we've already got enough
 435   return TRUE;
 436 }
 437
 438 // shrink to minimal size (releasing extra memory)
 439 bool wxString::Shrink()
 440 {
 441   wxStringData *pData = GetStringData();
 442
 443   size_t nLen = pData->nDataLength;
 444   void *p = realloc(pData, sizeof(wxStringData) + (nLen + 1)*sizeof(wxChar));
 445
 446   if ( p == NULL) {
 447       wxFAIL_MSG( _T("out of memory reallocating wxString data") );
 448       // keep previous data since reallocation failed
 449       return FALSE;
 450   }
 451
 452   if ( p != pData )
 453   {
 454       // contrary to what one might believe, some realloc() implementation do
 455       // move the memory block even when its size is reduced
 456       pData = (wxStringData *)p;
 457
 458       m_pchData = pData->data();
 459   }
 460
 461   pData->nAllocLength = nLen;
 462
 463   return TRUE;
 464 }
 465
 466 // get the pointer to writable buffer of (at least) nLen bytes
 467 wxChar *wxString::GetWriteBuf(size_t nLen)
 468 {
 469   if ( !AllocBeforeWrite(nLen) ) {
 470     // allocation failure handled by caller
 471     return NULL;
 472   }
 473
 474   wxASSERT( GetStringData()->nRefs == 1 );
 475   GetStringData()->Validate(FALSE);
 476
 477   return m_pchData;
 478 }
 479
 480 // put string back in a reasonable state after GetWriteBuf
 481 void wxString::UngetWriteBuf()
 482 {
 483   GetStringData()->nDataLength = wxStrlen(m_pchData);
 484   GetStringData()->Validate(TRUE);
 485 }
 486
 487 void wxString::UngetWriteBuf(size_t nLen)
 488 {
 489   GetStringData()->nDataLength = nLen;
 490   GetStringData()->Validate(TRUE);
 491 }
 492
 493 // ---------------------------------------------------------------------------
 494 // data access
 495 // ---------------------------------------------------------------------------
 496
 497 // all functions are inline in string.h
 498
 499 // ---------------------------------------------------------------------------
 500 // assignment operators
 501 // ---------------------------------------------------------------------------
 502
 503 // helper function: does real copy
 504 bool wxString::AssignCopy(size_t nSrcLen, const wxChar *pszSrcData)
 505 {
 506   if ( nSrcLen == 0 ) {
 507     Reinit();
 508   }
 509   else {
 510     if ( !AllocBeforeWrite(nSrcLen) ) {
 511       // allocation failure handled by caller
 512       return FALSE;
 513     }
 514     memcpy(m_pchData, pszSrcData, nSrcLen*sizeof(wxChar));
 515     GetStringData()->nDataLength = nSrcLen;
 516     m_pchData[nSrcLen] = wxT('\0');
 517   }
 518   return TRUE;
 519 }
 520
 521 // assigns one string to another
 522 wxString& wxString::operator=(const wxString& stringSrc)
 523 {
 524   wxASSERT( stringSrc.GetStringData()->IsValid() );
 525
 526   // don't copy string over itself
 527   if ( m_pchData != stringSrc.m_pchData ) {
 528     if ( stringSrc.GetStringData()->IsEmpty() ) {
 529       Reinit();
 530     }
 531     else {
 532       // adjust references
 533       GetStringData()->Unlock();
 534       m_pchData = stringSrc.m_pchData;
 535       GetStringData()->Lock();
 536     }
 537   }
 538
 539   return *this;
 540 }
 541
 542 // assigns a single character
 543 wxString& wxString::operator=(wxChar ch)
 544 {
 545   if ( !AssignCopy(1, &ch) ) {
 546     wxFAIL_MSG( _T("out of memory in wxString::operator=(wxChar)") );
 547   }
 548   return *this;
 549 }
 550
 551
 552 // assigns C string
 553 wxString& wxString::operator=(const wxChar *psz)
 554 {
 555   if ( !AssignCopy(wxStrlen(psz), psz) ) {
 556     wxFAIL_MSG( _T("out of memory in wxString::operator=(const wxChar *)") );
 557   }
 558   return *this;
 559 }
 560
 561 #if !wxUSE_UNICODE
 562
 563 // same as 'signed char' variant
 564 wxString& wxString::operator=(const unsigned char* psz)
 565 {
 566   *this = (const char *)psz;
 567   return *this;
 568 }
 569
 570 #if wxUSE_WCHAR_T
 571 wxString& wxString::operator=(const wchar_t *pwz)
 572 {
 573   wxString str(pwz);
 574   *this = str;
 575   return *this;
 576 }
 577 #endif
 578
 579 #endif
 580
 581 // ---------------------------------------------------------------------------
 582 // string concatenation
 583 // ---------------------------------------------------------------------------
 584
 585 // add something to this string
 586 bool wxString::ConcatSelf(int nSrcLen, const wxChar *pszSrcData)
 587 {
 588   STATISTICS_ADD(SummandLength, nSrcLen);
 589
 590   // concatenating an empty string is a NOP
 591   if ( nSrcLen > 0 ) {
 592     wxStringData *pData = GetStringData();
 593     size_t nLen = pData->nDataLength;
 594     size_t nNewLen = nLen + nSrcLen;
 595
 596     // alloc new buffer if current is too small
 597     if ( pData->IsShared() ) {
 598       STATISTICS_ADD(ConcatHit, 0);
 599
 600       // we have to allocate another buffer
 601       wxStringData* pOldData = GetStringData();
 602       if ( !AllocBuffer(nNewLen) ) {
 603           // allocation failure handled by caller
 604           return FALSE;
 605       }
 606       memcpy(m_pchData, pOldData->data(), nLen*sizeof(wxChar));
 607       pOldData->Unlock();
 608     }
 609     else if ( nNewLen > pData->nAllocLength ) {
 610       STATISTICS_ADD(ConcatHit, 0);
 611
 612       // we have to grow the buffer
 613       if ( !Alloc(nNewLen) ) {
 614           // allocation failure handled by caller
 615           return FALSE;
 616       }
 617     }
 618     else {
 619       STATISTICS_ADD(ConcatHit, 1);
 620
 621       // the buffer is already big enough
 622     }
 623
 624     // should be enough space
 625     wxASSERT( nNewLen <= GetStringData()->nAllocLength );
 626
 627     // fast concatenation - all is done in our buffer
 628     memcpy(m_pchData + nLen, pszSrcData, nSrcLen*sizeof(wxChar));
 629
 630     m_pchData[nNewLen] = wxT('\0');          // put terminating '\0'
 631     GetStringData()->nDataLength = nNewLen; // and fix the length
 632   }
 633   //else: the string to append was empty
 634   return TRUE;
 635 }
 636
 637 /*
 638  * concatenation functions come in 5 flavours:
 639  *  string + string
 640  *  char   + string      and      string + char
 641  *  C str  + string      and      string + C str
 642  */
 643
 644 wxString operator+(const wxString& str1, const wxString& str2)
 645 {
 646   wxASSERT( str1.GetStringData()->IsValid() );
 647   wxASSERT( str2.GetStringData()->IsValid() );
 648
 649   wxString s = str1;
 650   s += str2;
 651
 652   return s;
 653 }
 654
 655 wxString operator+(const wxString& str, wxChar ch)
 656 {
 657   wxASSERT( str.GetStringData()->IsValid() );
 658
 659   wxString s = str;
 660   s += ch;
 661
 662   return s;
 663 }
 664
 665 wxString operator+(wxChar ch, const wxString& str)
 666 {
 667   wxASSERT( str.GetStringData()->IsValid() );
 668
 669   wxString s = ch;
 670   s += str;
 671
 672   return s;
 673 }
 674
 675 wxString operator+(const wxString& str, const wxChar *psz)
 676 {
 677   wxASSERT( str.GetStringData()->IsValid() );
 678
 679   wxString s;
 680   if ( !s.Alloc(wxStrlen(psz) + str.Len()) ) {
 681     wxFAIL_MSG( _T("out of memory in wxString::operator+") );
 682   }
 683   s = str;
 684   s += psz;
 685
 686   return s;
 687 }
 688
 689 wxString operator+(const wxChar *psz, const wxString& str)
 690 {
 691   wxASSERT( str.GetStringData()->IsValid() );
 692
 693   wxString s;
 694   if ( !s.Alloc(wxStrlen(psz) + str.Len()) ) {
 695     wxFAIL_MSG( _T("out of memory in wxString::operator+") );
 696   }
 697   s = psz;
 698   s += str;
 699
 700   return s;
 701 }
 702
 703 // ===========================================================================
 704 // other common string functions
 705 // ===========================================================================
 706
 707 #if wxUSE_UNICODE
 708
 709 wxString wxString::FromAscii(const char *ascii)
 710 {
 711     if (!ascii)
 712        return wxEmptyString;
 713
 714     size_t len = strlen( ascii );
 715     wxString res;
 716
 717     if ( len )
 718     {
 719         wxStringBuffer buf(res, len);
 720
 721         wchar_t *dest = buf;
 722
 723         for ( ;; )
 724         {
 725            if ( (*dest++ = (wchar_t)(unsigned char)*ascii++) == L'\0' )
 726                break;
 727         }
 728     }
 729
 730     return res;
 731 }
 732
 733 const wxCharBuffer wxString::ToAscii() const
 734 {
 735     // this will allocate enough space for the terminating NUL too
 736     wxCharBuffer buffer(length());
 737
 738     signed char *dest = (signed char *)buffer.data();
 739
 740     const wchar_t *pwc = c_str();
 741     for ( ;; )
 742     {
 743         *dest++ = *pwc > SCHAR_MAX ? '_' : *pwc;
 744
 745         // the output string can't have embedded NULs anyhow, so we can safely
 746         // stop at first of them even if we do have any
 747         if ( !*pwc++ )
 748             break;
 749     }
 750
 751     return buffer;
 752 }
 753
 754 #endif // Unicode
 755
 756 // ---------------------------------------------------------------------------
 757 // simple sub-string extraction
 758 // ---------------------------------------------------------------------------
 759
 760 // helper function: clone the data attached to this string
 761 bool wxString::AllocCopy(wxString& dest, int nCopyLen, int nCopyIndex) const
 762 {
 763   if ( nCopyLen == 0 ) {
 764     dest.Init();
 765   }
 766   else {
 767     if ( !dest.AllocBuffer(nCopyLen) ) {
 768       // allocation failure handled by caller
 769       return FALSE;
 770     }
 771     memcpy(dest.m_pchData, m_pchData + nCopyIndex, nCopyLen*sizeof(wxChar));
 772   }
 773   return TRUE;
 774 }
 775
 776 // extract string of length nCount starting at nFirst
 777 wxString wxString::Mid(size_t nFirst, size_t nCount) const
 778 {
 779   wxStringData *pData = GetStringData();
 780   size_t nLen = pData->nDataLength;
 781
 782   // default value of nCount is wxSTRING_MAXLEN and means "till the end"
 783   if ( nCount == wxSTRING_MAXLEN )
 784   {
 785     nCount = nLen - nFirst;
 786   }
 787
 788   // out-of-bounds requests return sensible things
 789   if ( nFirst + nCount > nLen )
 790   {
 791     nCount = nLen - nFirst;
 792   }
 793
 794   if ( nFirst > nLen )
 795   {
 796     // AllocCopy() will return empty string
 797     nCount = 0;
 798   }
 799
 800   wxString dest;
 801   if ( !AllocCopy(dest, nCount, nFirst) ) {
 802       wxFAIL_MSG( _T("out of memory in wxString::Mid") );
 803   }
 804
 805   return dest;
 806 }
 807
 808 // check that the tring starts with prefix and return the rest of the string
 809 // in the provided pointer if it is not NULL, otherwise return FALSE
 810 bool wxString::StartsWith(const wxChar *prefix, wxString *rest) const
 811 {
 812     wxASSERT_MSG( prefix, _T("invalid parameter in wxString::StartsWith") );
 813
 814     // first check if the beginning of the string matches the prefix: note
 815     // that we don't have to check that we don't run out of this string as
 816     // when we reach the terminating NUL, either prefix string ends too (and
 817     // then it's ok) or we break out of the loop because there is no match
 818     const wxChar *p = c_str();
 819     while ( *prefix )
 820     {
 821         if ( *prefix++ != *p++ )
 822         {
 823             // no match
 824             return FALSE;
 825         }
 826     }
 827
 828     if ( rest )
 829     {
 830         // put the rest of the string into provided pointer
 831         *rest = p;
 832     }
 833
 834     return TRUE;
 835 }
 836
 837 // extract nCount last (rightmost) characters
 838 wxString wxString::Right(size_t nCount) const
 839 {
 840   if ( nCount > (size_t)GetStringData()->nDataLength )
 841     nCount = GetStringData()->nDataLength;
 842
 843   wxString dest;
 844   if ( !AllocCopy(dest, nCount, GetStringData()->nDataLength - nCount) ) {
 845     wxFAIL_MSG( _T("out of memory in wxString::Right") );
 846   }
 847   return dest;
 848 }
 849
 850 // get all characters after the last occurence of ch
 851 // (returns the whole string if ch not found)
 852 wxString wxString::AfterLast(wxChar ch) const
 853 {
 854   wxString str;
 855   int iPos = Find(ch, TRUE);
 856   if ( iPos == wxNOT_FOUND )
 857     str = *this;
 858   else
 859     str = c_str() + iPos + 1;
 860
 861   return str;
 862 }
 863
 864 // extract nCount first (leftmost) characters
 865 wxString wxString::Left(size_t nCount) const
 866 {
 867   if ( nCount > (size_t)GetStringData()->nDataLength )
 868     nCount = GetStringData()->nDataLength;
 869
 870   wxString dest;
 871   if ( !AllocCopy(dest, nCount, 0) ) {
 872     wxFAIL_MSG( _T("out of memory in wxString::Left") );
 873   }
 874   return dest;
 875 }
 876
 877 // get all characters before the first occurence of ch
 878 // (returns the whole string if ch not found)
 879 wxString wxString::BeforeFirst(wxChar ch) const
 880 {
 881   wxString str;
 882   for ( const wxChar *pc = m_pchData; *pc != wxT('\0') && *pc != ch; pc++ )
 883     str += *pc;
 884
 885   return str;
 886 }
 887
 888 /// get all characters before the last occurence of ch
 889 /// (returns empty string if ch not found)
 890 wxString wxString::BeforeLast(wxChar ch) const
 891 {
 892   wxString str;
 893   int iPos = Find(ch, TRUE);
 894   if ( iPos != wxNOT_FOUND && iPos != 0 )
 895     str = wxString(c_str(), iPos);
 896
 897   return str;
 898 }
 899
 900 /// get all characters after the first occurence of ch
 901 /// (returns empty string if ch not found)
 902 wxString wxString::AfterFirst(wxChar ch) const
 903 {
 904   wxString str;
 905   int iPos = Find(ch);
 906   if ( iPos != wxNOT_FOUND )
 907     str = c_str() + iPos + 1;
 908
 909   return str;
 910 }
 911
 912 // replace first (or all) occurences of some substring with another one
 913 size_t wxString::Replace(const wxChar *szOld, const wxChar *szNew, bool bReplaceAll)
 914 {
 915   size_t uiCount = 0;   // count of replacements made
 916
 917   size_t uiOldLen = wxStrlen(szOld);
 918
 919   wxString strTemp;
 920   const wxChar *pCurrent = m_pchData;
 921   const wxChar *pSubstr;
 922   while ( *pCurrent != wxT('\0') ) {
 923     pSubstr = wxStrstr(pCurrent, szOld);
 924     if ( pSubstr == NULL ) {
 925       // strTemp is unused if no replacements were made, so avoid the copy
 926       if ( uiCount == 0 )
 927         return 0;
 928
 929       strTemp += pCurrent;    // copy the rest
 930       break;                  // exit the loop
 931     }
 932     else {
 933       // take chars before match
 934       if ( !strTemp.ConcatSelf(pSubstr - pCurrent, pCurrent) ) {
 935         wxFAIL_MSG( _T("out of memory in wxString::Replace") );
 936         return 0;
 937       }
 938       strTemp += szNew;
 939       pCurrent = pSubstr + uiOldLen;  // restart after match
 940
 941       uiCount++;
 942
 943       // stop now?
 944       if ( !bReplaceAll ) {
 945         strTemp += pCurrent;    // copy the rest
 946         break;                  // exit the loop
 947       }
 948     }
 949   }
 950
 951   // only done if there were replacements, otherwise would have returned above
 952   *this = strTemp;
 953
 954   return uiCount;
 955 }
 956
 957 bool wxString::IsAscii() const
 958 {
 959   const wxChar *s = (const wxChar*) *this;
 960   while(*s){
 961     if(!isascii(*s)) return(FALSE);
 962     s++;
 963   }
 964   return(TRUE);
 965 }
 966
 967 bool wxString::IsWord() const
 968 {
 969   const wxChar *s = (const wxChar*) *this;
 970   while(*s){
 971     if(!wxIsalpha(*s)) return(FALSE);
 972     s++;
 973   }
 974   return(TRUE);
 975 }
 976
 977 bool wxString::IsNumber() const
 978 {
 979   const wxChar *s = (const wxChar*) *this;
 980   if (wxStrlen(s))
 981      if ((s[0] == '-') || (s[0] == '+')) s++;
 982   while(*s){
 983     if(!wxIsdigit(*s)) return(FALSE);
 984     s++;
 985   }
 986   return(TRUE);
 987 }
 988
 989 wxString wxString::Strip(stripType w) const
 990 {
 991     wxString s = *this;
 992     if ( w & leading ) s.Trim(FALSE);
 993     if ( w & trailing ) s.Trim(TRUE);
 994     return s;
 995 }
 996
 997 // ---------------------------------------------------------------------------
 998 // case conversion
 999 // ---------------------------------------------------------------------------
1000
1001 wxString& wxString::MakeUpper()
1002 {
1003   if ( !CopyBeforeWrite() ) {
1004     wxFAIL_MSG( _T("out of memory in wxString::MakeUpper") );
1005     return *this;
1006   }
1007
1008   for ( wxChar *p = m_pchData; *p; p++ )
1009     *p = (wxChar)wxToupper(*p);
1010
1011   return *this;
1012 }
1013
1014 wxString& wxString::MakeLower()
1015 {
1016   if ( !CopyBeforeWrite() ) {
1017     wxFAIL_MSG( _T("out of memory in wxString::MakeLower") );
1018     return *this;
1019   }
1020
1021   for ( wxChar *p = m_pchData; *p; p++ )
1022     *p = (wxChar)wxTolower(*p);
1023
1024   return *this;
1025 }
1026
1027 // ---------------------------------------------------------------------------
1028 // trimming and padding
1029 // ---------------------------------------------------------------------------
1030
1031 // some compilers (VC++ 6.0 not to name them) return TRUE for a call to
1032 // isspace('ê') in the C locale which seems to be broken to me, but we have to
1033 // live with this by checking that the character is a 7 bit one - even if this
1034 // may fail to detect some spaces (I don't know if Unicode doesn't have
1035 // space-like symbols somewhere except in the first 128 chars), it is arguably
1036 // still better than trimming away accented letters
1037 inline int wxSafeIsspace(wxChar ch) { return (ch < 127) && wxIsspace(ch); }
1038
1039 // trims spaces (in the sense of isspace) from left or right side
1040 wxString& wxString::Trim(bool bFromRight)
1041 {
1042   // first check if we're going to modify the string at all
1043   if ( !IsEmpty() &&
1044        (
1045         (bFromRight && wxSafeIsspace(GetChar(Len() - 1))) ||
1046         (!bFromRight && wxSafeIsspace(GetChar(0u)))
1047        )
1048      )
1049   {
1050     // ok, there is at least one space to trim
1051     if ( !CopyBeforeWrite() ) {
1052       wxFAIL_MSG( _T("out of memory in wxString::Trim") );
1053       return *this;
1054     }
1055
1056     if ( bFromRight )
1057     {
1058       // find last non-space character
1059       wxChar *psz = m_pchData + GetStringData()->nDataLength - 1;
1060       while ( wxSafeIsspace(*psz) && (psz >= m_pchData) )
1061         psz--;
1062
1063       // truncate at trailing space start
1064       *++psz = wxT('\0');
1065       GetStringData()->nDataLength = psz - m_pchData;
1066     }
1067     else
1068     {
1069       // find first non-space character
1070       const wxChar *psz = m_pchData;
1071       while ( wxSafeIsspace(*psz) )
1072         psz++;
1073
1074       // fix up data and length
1075       int nDataLength = GetStringData()->nDataLength - (psz - (const wxChar*) m_pchData);
1076       memmove(m_pchData, psz, (nDataLength + 1)*sizeof(wxChar));
1077       GetStringData()->nDataLength = nDataLength;
1078     }
1079   }
1080
1081   return *this;
1082 }
1083
1084 // adds nCount characters chPad to the string from either side
1085 wxString& wxString::Pad(size_t nCount, wxChar chPad, bool bFromRight)
1086 {
1087   wxString s(chPad, nCount);
1088
1089   if ( bFromRight )
1090     *this += s;
1091   else
1092   {
1093     s += *this;
1094     *this = s;
1095   }
1096
1097   return *this;
1098 }
1099
1100 // truncate the string
1101 wxString& wxString::Truncate(size_t uiLen)
1102 {
1103   if ( uiLen < Len() ) {
1104     if ( !CopyBeforeWrite() ) {
1105       wxFAIL_MSG( _T("out of memory in wxString::Truncate") );
1106       return *this;
1107     }
1108
1109     *(m_pchData + uiLen) = wxT('\0');
1110     GetStringData()->nDataLength = uiLen;
1111   }
1112   //else: nothing to do, string is already short enough
1113
1114   return *this;
1115 }
1116
1117 // ---------------------------------------------------------------------------
1118 // finding (return wxNOT_FOUND if not found and index otherwise)
1119 // ---------------------------------------------------------------------------
1120
1121 // find a character
1122 int wxString::Find(wxChar ch, bool bFromEnd) const
1123 {
1124   const wxChar *psz = bFromEnd ? wxStrrchr(m_pchData, ch) : wxStrchr(m_pchData, ch);
1125
1126   return (psz == NULL) ? wxNOT_FOUND : psz - (const wxChar*) m_pchData;
1127 }
1128
1129 // find a sub-string (like strstr)
1130 int wxString::Find(const wxChar *pszSub) const
1131 {
1132   const wxChar *psz = wxStrstr(m_pchData, pszSub);
1133
1134   return (psz == NULL) ? wxNOT_FOUND : psz - (const wxChar*) m_pchData;
1135 }
1136
1137 // ----------------------------------------------------------------------------
1138 // conversion to numbers
1139 // ----------------------------------------------------------------------------
1140
1141 bool wxString::ToLong(long *val, int base) const
1142 {
1143     wxCHECK_MSG( val, FALSE, _T("NULL pointer in wxString::ToLong") );
1144     wxASSERT_MSG( !base || (base > 1 && base <= 36), _T("invalid base") );
1145
1146     const wxChar *start = c_str();
1147     wxChar *end;
1148     *val = wxStrtol(start, &end, base);
1149
1150     // return TRUE only if scan was stopped by the terminating NUL and if the
1151     // string was not empty to start with
1152     return !*end && (end != start);
1153 }
1154
1155 bool wxString::ToULong(unsigned long *val, int base) const
1156 {
1157     wxCHECK_MSG( val, FALSE, _T("NULL pointer in wxString::ToULong") );
1158     wxASSERT_MSG( !base || (base > 1 && base <= 36), _T("invalid base") );
1159
1160     const wxChar *start = c_str();
1161     wxChar *end;
1162     *val = wxStrtoul(start, &end, base);
1163
1164     // return TRUE only if scan was stopped by the terminating NUL and if the
1165     // string was not empty to start with
1166     return !*end && (end != start);
1167 }
1168
1169 bool wxString::ToDouble(double *val) const
1170 {
1171     wxCHECK_MSG( val, FALSE, _T("NULL pointer in wxString::ToDouble") );
1172
1173     const wxChar *start = c_str();
1174     wxChar *end;
1175     *val = wxStrtod(start, &end);
1176
1177     // return TRUE only if scan was stopped by the terminating NUL and if the
1178     // string was not empty to start with
1179     return !*end && (end != start);
1180 }
1181
1182 // ---------------------------------------------------------------------------
1183 // formatted output
1184 // ---------------------------------------------------------------------------
1185
1186 /* static */
1187 wxString wxString::Format(const wxChar *pszFormat, ...)
1188 {
1189     va_list argptr;
1190     va_start(argptr, pszFormat);
1191
1192     wxString s;
1193     s.PrintfV(pszFormat, argptr);
1194
1195     va_end(argptr);
1196
1197     return s;
1198 }
1199
1200 /* static */
1201 wxString wxString::FormatV(const wxChar *pszFormat, va_list argptr)
1202 {
1203     wxString s;
1204     s.PrintfV(pszFormat, argptr);
1205     return s;
1206 }
1207
1208 int wxString::Printf(const wxChar *pszFormat, ...)
1209 {
1210   va_list argptr;
1211   va_start(argptr, pszFormat);
1212
1213   int iLen = PrintfV(pszFormat, argptr);
1214
1215   va_end(argptr);
1216
1217   return iLen;
1218 }
1219
1220 int wxString::PrintfV(const wxChar* pszFormat, va_list argptr)
1221 {
1222     int size = 1024;
1223     for ( ;; )
1224     {
1225         wxChar *buf = GetWriteBuf(size + 1);
1226         if ( !buf )
1227         {
1228             // out of memory
1229             return -1;
1230         }
1231
1232         int len = wxVsnprintf(buf, size, pszFormat, argptr);
1233
1234         // some implementations of vsnprintf() don't NUL terminate the string
1235         // if there is not enough space for it so always do it manually
1236         buf[size] = _T('\0');
1237
1238         UngetWriteBuf();
1239
1240         if ( len >= 0 )
1241         {
1242             // ok, there was enough space
1243             break;
1244         }
1245
1246         // still not enough, double it again
1247         size *= 2;
1248     }
1249
1250     // we could have overshot
1251     Shrink();
1252
1253     return Len();
1254 }
1255
1256 // ----------------------------------------------------------------------------
1257 // misc other operations
1258 // ----------------------------------------------------------------------------
1259
1260 // returns TRUE if the string matches the pattern which may contain '*' and
1261 // '?' metacharacters (as usual, '?' matches any character and '*' any number
1262 // of them)
1263 bool wxString::Matches(const wxChar *pszMask) const
1264 {
1265     // I disable this code as it doesn't seem to be faster (in fact, it seems
1266     // to be much slower) than the old, hand-written code below and using it
1267     // here requires always linking with libregex even if the user code doesn't
1268     // use it
1269 #if 0 // wxUSE_REGEX
1270     // first translate the shell-like mask into a regex
1271     wxString pattern;
1272     pattern.reserve(wxStrlen(pszMask));
1273
1274     pattern += _T('^');
1275     while ( *pszMask )
1276     {
1277         switch ( *pszMask )
1278         {
1279             case _T('?'):
1280                 pattern += _T('.');
1281                 break;
1282
1283             case _T('*'):
1284                 pattern += _T(".*");
1285                 break;
1286
1287             case _T('^'):
1288             case _T('.'):
1289             case _T('$'):
1290             case _T('('):
1291             case _T(')'):
1292             case _T('|'):
1293             case _T('+'):
1294             case _T('\\'):
1295                 // these characters are special in a RE, quote them
1296                 // (however note that we don't quote '[' and ']' to allow
1297                 // using them for Unix shell like matching)
1298                 pattern += _T('\\');
1299                 // fall through
1300
1301             default:
1302                 pattern += *pszMask;
1303         }
1304
1305         pszMask++;
1306     }
1307     pattern += _T('$');
1308
1309     // and now use it
1310     return wxRegEx(pattern, wxRE_NOSUB | wxRE_EXTENDED).Matches(c_str());
1311 #else // !wxUSE_REGEX
1312   // TODO: this is, of course, awfully inefficient...
1313
1314   // the char currently being checked
1315   const wxChar *pszTxt = c_str();
1316
1317   // the last location where '*' matched
1318   const wxChar *pszLastStarInText = NULL;
1319   const wxChar *pszLastStarInMask = NULL;
1320
1321 match:
1322   for ( ; *pszMask != wxT('\0'); pszMask++, pszTxt++ ) {
1323     switch ( *pszMask ) {
1324       case wxT('?'):
1325         if ( *pszTxt == wxT('\0') )
1326           return FALSE;
1327
1328         // pszTxt and pszMask will be incremented in the loop statement
1329
1330         break;
1331
1332       case wxT('*'):
1333         {
1334           // remember where we started to be able to backtrack later
1335           pszLastStarInText = pszTxt;
1336           pszLastStarInMask = pszMask;
1337
1338           // ignore special chars immediately following this one
1339           // (should this be an error?)
1340           while ( *pszMask == wxT('*') || *pszMask == wxT('?') )
1341             pszMask++;
1342
1343           // if there is nothing more, match
1344           if ( *pszMask == wxT('\0') )
1345             return TRUE;
1346
1347           // are there any other metacharacters in the mask?
1348           size_t uiLenMask;
1349           const wxChar *pEndMask = wxStrpbrk(pszMask, wxT("*?"));
1350
1351           if ( pEndMask != NULL ) {
1352             // we have to match the string between two metachars
1353             uiLenMask = pEndMask - pszMask;
1354           }
1355           else {
1356             // we have to match the remainder of the string
1357             uiLenMask = wxStrlen(pszMask);
1358           }
1359
1360           wxString strToMatch(pszMask, uiLenMask);
1361           const wxChar* pMatch = wxStrstr(pszTxt, strToMatch);
1362           if ( pMatch == NULL )
1363             return FALSE;
1364
1365           // -1 to compensate "++" in the loop
1366           pszTxt = pMatch + uiLenMask - 1;
1367           pszMask += uiLenMask - 1;
1368         }
1369         break;
1370
1371       default:
1372         if ( *pszMask != *pszTxt )
1373           return FALSE;
1374         break;
1375     }
1376   }
1377
1378   // match only if nothing left
1379   if ( *pszTxt == wxT('\0') )
1380     return TRUE;
1381
1382   // if we failed to match, backtrack if we can
1383   if ( pszLastStarInText ) {
1384     pszTxt = pszLastStarInText + 1;
1385     pszMask = pszLastStarInMask;
1386
1387     pszLastStarInText = NULL;
1388
1389     // don't bother resetting pszLastStarInMask, it's unnecessary
1390
1391     goto match;
1392   }
1393
1394   return FALSE;
1395 #endif // wxUSE_REGEX/!wxUSE_REGEX
1396 }
1397
1398 // Count the number of chars
1399 int wxString::Freq(wxChar ch) const
1400 {
1401     int count = 0;
1402     int len = Len();
1403     for (int i = 0; i < len; i++)
1404     {
1405         if (GetChar(i) == ch)
1406             count ++;
1407     }
1408     return count;
1409 }
1410
1411 // convert to upper case, return the copy of the string
1412 wxString wxString::Upper() const
1413 { wxString s(*this); return s.MakeUpper(); }
1414
1415 // convert to lower case, return the copy of the string
1416 wxString wxString::Lower() const { wxString s(*this); return s.MakeLower(); }
1417
1418 int wxString::sprintf(const wxChar *pszFormat, ...)
1419   {
1420     va_list argptr;
1421     va_start(argptr, pszFormat);
1422     int iLen = PrintfV(pszFormat, argptr);
1423     va_end(argptr);
1424     return iLen;
1425   }
1426
1427 // ---------------------------------------------------------------------------
1428 // standard C++ library string functions
1429 // ---------------------------------------------------------------------------
1430
1431 #ifdef  wxSTD_STRING_COMPATIBILITY
1432
1433 void wxString::resize(size_t nSize, wxChar ch)
1434 {
1435     size_t len = length();
1436
1437     if ( nSize < len )
1438     {
1439         Truncate(nSize);
1440     }
1441     else if ( nSize > len )
1442     {
1443         *this += wxString(ch, nSize - len);
1444     }
1445     //else: we have exactly the specified length, nothing to do
1446 }
1447
1448 void wxString::swap(wxString& str)
1449 {
1450     // this is slightly less efficient than fiddling with m_pchData directly,
1451     // but it is still quite efficient as we don't copy the string here because
1452     // ref count always stays positive
1453     wxString tmp = str;
1454     str = *this;
1455     *this = tmp;
1456 }
1457
1458 wxString& wxString::insert(size_t nPos, const wxString& str)
1459 {
1460   wxASSERT( str.GetStringData()->IsValid() );
1461   wxASSERT( nPos <= Len() );
1462
1463   if ( !str.IsEmpty() ) {
1464     wxString strTmp;
1465     wxChar *pc = strTmp.GetWriteBuf(Len() + str.Len());
1466     wxStrncpy(pc, c_str(), nPos);
1467     wxStrcpy(pc + nPos, str);
1468     wxStrcpy(pc + nPos + str.Len(), c_str() + nPos);
1469     strTmp.UngetWriteBuf();
1470     *this = strTmp;
1471   }
1472
1473   return *this;
1474 }
1475
1476 size_t wxString::find(const wxString& str, size_t nStart) const
1477 {
1478   wxASSERT( str.GetStringData()->IsValid() );
1479   wxASSERT( nStart <= Len() );
1480
1481   const wxChar *p = wxStrstr(c_str() + nStart, str);
1482
1483   return p == NULL ? npos : p - c_str();
1484 }
1485
1486 // VC++ 1.5 can't cope with the default argument in the header.
1487 #if !defined(__VISUALC__) || defined(__WIN32__)
1488 size_t wxString::find(const wxChar* sz, size_t nStart, size_t n) const
1489 {
1490   return find(wxString(sz, n), nStart);
1491 }
1492 #endif // VC++ 1.5
1493
1494 // Gives a duplicate symbol (presumably a case-insensitivity problem)
1495 #if !defined(__BORLANDC__)
1496 size_t wxString::find(wxChar ch, size_t nStart) const
1497 {
1498   wxASSERT( nStart <= Len() );
1499
1500   const wxChar *p = wxStrchr(c_str() + nStart, ch);
1501
1502   return p == NULL ? npos : p - c_str();
1503 }
1504 #endif
1505
1506 size_t wxString::rfind(const wxString& str, size_t nStart) const
1507 {
1508   wxASSERT( str.GetStringData()->IsValid() );
1509   wxASSERT( nStart == npos || nStart <= Len() );
1510
1511   // TODO could be made much quicker than that
1512   const wxChar *p = c_str() + (nStart == npos ? Len() : nStart);
1513   while ( p >= c_str() + str.Len() ) {
1514     if ( wxStrncmp(p - str.Len(), str, str.Len()) == 0 )
1515       return p - str.Len() - c_str();
1516     p--;
1517   }
1518
1519   return npos;
1520 }
1521
1522 // VC++ 1.5 can't cope with the default argument in the header.
1523 #if !defined(__VISUALC__) || defined(__WIN32__)
1524 size_t wxString::rfind(const wxChar* sz, size_t nStart, size_t n) const
1525 {
1526     return rfind(wxString(sz, n == npos ? wxSTRING_MAXLEN : n), nStart);
1527 }
1528
1529 size_t wxString::rfind(wxChar ch, size_t nStart) const
1530 {
1531     if ( nStart == npos )
1532     {
1533         nStart = Len();
1534     }
1535     else
1536     {
1537         wxASSERT( nStart <= Len() );
1538     }
1539
1540     const wxChar *p = wxStrrchr(c_str(), ch);
1541
1542     if ( p == NULL )
1543         return npos;
1544
1545     size_t result = p - c_str();
1546     return ( result > nStart ) ? npos : result;
1547 }
1548 #endif // VC++ 1.5
1549
1550 size_t wxString::find_first_of(const wxChar* sz, size_t nStart) const
1551 {
1552     const wxChar *start = c_str() + nStart;
1553     const wxChar *firstOf = wxStrpbrk(start, sz);
1554     if ( firstOf )
1555         return firstOf - c_str();
1556     else
1557         return npos;
1558 }
1559
1560 size_t wxString::find_last_of(const wxChar* sz, size_t nStart) const
1561 {
1562     if ( nStart == npos )
1563     {
1564         nStart = Len();
1565     }
1566     else
1567     {
1568         wxASSERT( nStart <= Len() );
1569     }
1570
1571     for ( const wxChar *p = c_str() + length() - 1; p >= c_str(); p-- )
1572     {
1573         if ( wxStrchr(sz, *p) )
1574             return p - c_str();
1575     }
1576
1577     return npos;
1578 }
1579
1580 size_t wxString::find_first_not_of(const wxChar* sz, size_t nStart) const
1581 {
1582     if ( nStart == npos )
1583     {
1584         nStart = Len();
1585     }
1586     else
1587     {
1588         wxASSERT( nStart <= Len() );
1589     }
1590
1591     size_t nAccept = wxStrspn(c_str() + nStart, sz);
1592     if ( nAccept >= length() - nStart )
1593         return npos;
1594     else
1595         return nAccept;
1596 }
1597
1598 size_t wxString::find_first_not_of(wxChar ch, size_t nStart) const
1599 {
1600     wxASSERT( nStart <= Len() );
1601
1602     for ( const wxChar *p = c_str() + nStart; *p; p++ )
1603     {
1604         if ( *p != ch )
1605             return p - c_str();
1606     }
1607
1608     return npos;
1609 }
1610
1611 size_t wxString::find_last_not_of(const wxChar* sz, size_t nStart) const
1612 {
1613     if ( nStart == npos )
1614     {
1615         nStart = Len();
1616     }
1617     else
1618     {
1619         wxASSERT( nStart <= Len() );
1620     }
1621
1622     for ( const wxChar *p = c_str() + nStart - 1; p >= c_str(); p-- )
1623     {
1624         if ( !wxStrchr(sz, *p) )
1625             return p - c_str();
1626     }
1627
1628     return npos;
1629 }
1630
1631 size_t wxString::find_last_not_of(wxChar ch, size_t nStart) const
1632 {
1633     if ( nStart == npos )
1634     {
1635         nStart = Len();
1636     }
1637     else
1638     {
1639         wxASSERT( nStart <= Len() );
1640     }
1641
1642     for ( const wxChar *p = c_str() + nStart - 1; p >= c_str(); p-- )
1643     {
1644         if ( *p != ch )
1645             return p - c_str();
1646     }
1647
1648     return npos;
1649 }
1650
1651 wxString& wxString::erase(size_t nStart, size_t nLen)
1652 {
1653   wxString strTmp(c_str(), nStart);
1654   if ( nLen != npos ) {
1655     wxASSERT( nStart + nLen <= Len() );
1656
1657     strTmp.append(c_str() + nStart + nLen);
1658   }
1659
1660   *this = strTmp;
1661   return *this;
1662 }
1663
1664 wxString& wxString::replace(size_t nStart, size_t nLen, const wxChar *sz)
1665 {
1666   wxASSERT_MSG( nStart + nLen <= Len(),
1667                 _T("index out of bounds in wxString::replace") );
1668
1669   wxString strTmp;
1670   strTmp.Alloc(Len());      // micro optimisation to avoid multiple mem allocs
1671
1672   if ( nStart != 0 )
1673     strTmp.append(c_str(), nStart);
1674   strTmp << sz << c_str() + nStart + nLen;
1675
1676   *this = strTmp;
1677   return *this;
1678 }
1679
1680 wxString& wxString::replace(size_t nStart, size_t nLen, size_t nCount, wxChar ch)
1681 {
1682   return replace(nStart, nLen, wxString(ch, nCount));
1683 }
1684
1685 wxString& wxString::replace(size_t nStart, size_t nLen,
1686                             const wxString& str, size_t nStart2, size_t nLen2)
1687 {
1688   return replace(nStart, nLen, str.substr(nStart2, nLen2));
1689 }
1690
1691 wxString& wxString::replace(size_t nStart, size_t nLen,
1692                         const wxChar* sz, size_t nCount)
1693 {
1694   return replace(nStart, nLen, wxString(sz, nCount));
1695 }
1696
1697 #endif  //std::string compatibility
1698
1699 // ============================================================================
1700 // ArrayString
1701 // ============================================================================
1702
1703 // size increment = min(50% of current size, ARRAY_MAXSIZE_INCREMENT)
1704 #define   ARRAY_MAXSIZE_INCREMENT       4096
1705
1706 #ifndef   ARRAY_DEFAULT_INITIAL_SIZE    // also defined in dynarray.h
1707 #define   ARRAY_DEFAULT_INITIAL_SIZE    (16)
1708 #endif
1709
1710 #define   STRING(p)   ((wxString *)(&(p)))
1711
1712 // ctor
1713 void wxArrayString::Init(bool autoSort)
1714 {
1715   m_nSize  =
1716   m_nCount = 0;
1717   m_pItems = (wxChar **) NULL;
1718   m_autoSort = autoSort;
1719 }
1720
1721 // copy ctor
1722 wxArrayString::wxArrayString(const wxArrayString& src)
1723 {
1724   Init(src.m_autoSort);
1725
1726   *this = src;
1727 }
1728
1729 // assignment operator
1730 wxArrayString& wxArrayString::operator=(const wxArrayString& src)
1731 {
1732   if ( m_nSize > 0 )
1733     Clear();
1734
1735   Copy(src);
1736
1737   m_autoSort = src.m_autoSort;
1738
1739   return *this;
1740 }
1741
1742 void wxArrayString::Copy(const wxArrayString& src)
1743 {
1744   if ( src.m_nCount > ARRAY_DEFAULT_INITIAL_SIZE )
1745     Alloc(src.m_nCount);
1746
1747   for ( size_t n = 0; n < src.m_nCount; n++ )
1748     Add(src[n]);
1749 }
1750
1751 // grow the array
1752 void wxArrayString::Grow(size_t nIncrement)
1753 {
1754   // only do it if no more place
1755   if ( m_nCount == m_nSize ) {
1756     // if ARRAY_DEFAULT_INITIAL_SIZE were set to 0, the initially empty would
1757     // be never resized!
1758     #if ARRAY_DEFAULT_INITIAL_SIZE == 0
1759       #error "ARRAY_DEFAULT_INITIAL_SIZE must be > 0!"
1760     #endif
1761
1762     if ( m_nSize == 0 ) {
1763       // was empty, alloc some memory
1764       m_nSize = ARRAY_DEFAULT_INITIAL_SIZE;
1765       if (m_nSize < nIncrement)
1766           m_nSize = nIncrement;
1767       m_pItems = new wxChar *[m_nSize];
1768     }
1769     else {
1770       // otherwise when it's called for the first time, nIncrement would be 0
1771       // and the array would never be expanded
1772       // add 50% but not too much
1773       size_t ndefIncrement = m_nSize < ARRAY_DEFAULT_INITIAL_SIZE
1774                           ? ARRAY_DEFAULT_INITIAL_SIZE : m_nSize >> 1;
1775       if ( ndefIncrement > ARRAY_MAXSIZE_INCREMENT )
1776         ndefIncrement = ARRAY_MAXSIZE_INCREMENT;
1777       if ( nIncrement < ndefIncrement )
1778         nIncrement = ndefIncrement;
1779       m_nSize += nIncrement;
1780       wxChar **pNew = new wxChar *[m_nSize];
1781
1782       // copy data to new location
1783       memcpy(pNew, m_pItems, m_nCount*sizeof(wxChar *));
1784
1785       // delete old memory (but do not release the strings!)
1786       wxDELETEA(m_pItems);
1787
1788       m_pItems = pNew;
1789     }
1790   }
1791 }
1792
1793 void wxArrayString::Free()
1794 {
1795   for ( size_t n = 0; n < m_nCount; n++ ) {
1796     STRING(m_pItems[n])->GetStringData()->Unlock();
1797   }
1798 }
1799
1800 // deletes all the strings from the list
1801 void wxArrayString::Empty()
1802 {
1803   Free();
1804
1805   m_nCount = 0;
1806 }
1807
1808 // as Empty, but also frees memory
1809 void wxArrayString::Clear()
1810 {
1811   Free();
1812
1813   m_nSize  =
1814   m_nCount = 0;
1815
1816   wxDELETEA(m_pItems);
1817 }
1818
1819 // dtor
1820 wxArrayString::~wxArrayString()
1821 {
1822   Free();
1823
1824   wxDELETEA(m_pItems);
1825 }
1826
1827 // pre-allocates memory (frees the previous data!)
1828 void wxArrayString::Alloc(size_t nSize)
1829 {
1830   // only if old buffer was not big enough
1831   if ( nSize > m_nSize ) {
1832     Free();
1833     wxDELETEA(m_pItems);
1834     m_pItems = new wxChar *[nSize];
1835     m_nSize  = nSize;
1836   }
1837
1838   m_nCount = 0;
1839 }
1840
1841 // minimizes the memory usage by freeing unused memory
1842 void wxArrayString::Shrink()
1843 {
1844   // only do it if we have some memory to free
1845   if( m_nCount < m_nSize ) {
1846     // allocates exactly as much memory as we need
1847     wxChar **pNew = new wxChar *[m_nCount];
1848
1849     // copy data to new location
1850     memcpy(pNew, m_pItems, m_nCount*sizeof(wxChar *));
1851     delete [] m_pItems;
1852     m_pItems = pNew;
1853   }
1854 }
1855
1856 // return a wxString[] as required for some control ctors.
1857 wxString* wxArrayString::GetStringArray() const
1858 {
1859     wxString *array = 0;
1860
1861     if( m_nCount > 0 )
1862     {
1863         array = new wxString[m_nCount];
1864         for( size_t i = 0; i < m_nCount; i++ )
1865             array[i] = m_pItems[i];
1866     }
1867
1868     return array;
1869 }
1870
1871 // searches the array for an item (forward or backwards)
1872 int wxArrayString::Index(const wxChar *sz, bool bCase, bool bFromEnd) const
1873 {
1874   if ( m_autoSort ) {
1875     // use binary search in the sorted array
1876     wxASSERT_MSG( bCase && !bFromEnd,
1877                   wxT("search parameters ignored for auto sorted array") );
1878
1879     size_t i,
1880            lo = 0,
1881            hi = m_nCount;
1882     int res;
1883     while ( lo < hi ) {
1884       i = (lo + hi)/2;
1885
1886       res = wxStrcmp(sz, m_pItems[i]);
1887       if ( res < 0 )
1888         hi = i;
1889       else if ( res > 0 )
1890         lo = i + 1;
1891       else
1892         return i;
1893     }
1894
1895     return wxNOT_FOUND;
1896   }
1897   else {
1898     // use linear search in unsorted array
1899     if ( bFromEnd ) {
1900       if ( m_nCount > 0 ) {
1901         size_t ui = m_nCount;
1902         do {
1903           if ( STRING(m_pItems[--ui])->IsSameAs(sz, bCase) )
1904             return ui;
1905         }
1906         while ( ui != 0 );
1907       }
1908     }
1909     else {
1910       for( size_t ui = 0; ui < m_nCount; ui++ ) {
1911         if( STRING(m_pItems[ui])->IsSameAs(sz, bCase) )
1912           return ui;
1913       }
1914     }
1915   }
1916
1917   return wxNOT_FOUND;
1918 }
1919
1920 // add item at the end
1921 size_t wxArrayString::Add(const wxString& str, size_t nInsert)
1922 {
1923   if ( m_autoSort ) {
1924     // insert the string at the correct position to keep the array sorted
1925     size_t i,
1926            lo = 0,
1927            hi = m_nCount;
1928     int res;
1929     while ( lo < hi ) {
1930       i = (lo + hi)/2;
1931
1932       res = wxStrcmp(str, m_pItems[i]);
1933       if ( res < 0 )
1934         hi = i;
1935       else if ( res > 0 )
1936         lo = i + 1;
1937       else {
1938         lo = hi = i;
1939         break;
1940       }
1941     }
1942
1943     wxASSERT_MSG( lo == hi, wxT("binary search broken") );
1944
1945     Insert(str, lo, nInsert);
1946
1947     return (size_t)lo;
1948   }
1949   else {
1950     wxASSERT( str.GetStringData()->IsValid() );
1951
1952     Grow(nInsert);
1953
1954     for (size_t i = 0; i < nInsert; i++)
1955     {
1956         // the string data must not be deleted!
1957         str.GetStringData()->Lock();
1958
1959         // just append
1960         m_pItems[m_nCount + i] = (wxChar *)str.c_str(); // const_cast
1961     }
1962     size_t ret = m_nCount;
1963     m_nCount += nInsert;
1964     return ret;
1965   }
1966 }
1967
1968 // add item at the given position
1969 void wxArrayString::Insert(const wxString& str, size_t nIndex, size_t nInsert)
1970 {
1971   wxASSERT( str.GetStringData()->IsValid() );
1972
1973   wxCHECK_RET( nIndex <= m_nCount, wxT("bad index in wxArrayString::Insert") );
1974   wxCHECK_RET( m_nCount <= m_nCount + nInsert,
1975                wxT("array size overflow in wxArrayString::Insert") );
1976
1977   Grow(nInsert);
1978
1979   memmove(&m_pItems[nIndex + nInsert], &m_pItems[nIndex],
1980           (m_nCount - nIndex)*sizeof(wxChar *));
1981
1982   for (size_t i = 0; i < nInsert; i++)
1983   {
1984       str.GetStringData()->Lock();
1985       m_pItems[nIndex + i] = (wxChar *)str.c_str();
1986   }
1987   m_nCount += nInsert;
1988 }
1989
1990 // expand the array
1991 void wxArrayString::SetCount(size_t count)
1992 {
1993     Alloc(count);
1994
1995     wxString s;
1996     while ( m_nCount < count )
1997         m_pItems[m_nCount++] = (wxChar *)s.c_str();
1998 }
1999
2000 // removes item from array (by index)
2001 void wxArrayString::Remove(size_t nIndex, size_t nRemove)
2002 {
2003   wxCHECK_RET( nIndex < m_nCount, wxT("bad index in wxArrayString::Remove") );
2004   wxCHECK_RET( nIndex + nRemove <= m_nCount,
2005                wxT("removing too many elements in wxArrayString::Remove") );
2006
2007   // release our lock
2008   for (size_t i = 0; i < nRemove; i++)
2009       Item(nIndex + i).GetStringData()->Unlock();
2010
2011   memmove(&m_pItems[nIndex], &m_pItems[nIndex + nRemove],
2012           (m_nCount - nIndex - nRemove)*sizeof(wxChar *));
2013   m_nCount -= nRemove;
2014 }
2015
2016 // removes item from array (by value)
2017 void wxArrayString::Remove(const wxChar *sz)
2018 {
2019   int iIndex = Index(sz);
2020
2021   wxCHECK_RET( iIndex != wxNOT_FOUND,
2022                wxT("removing inexistent element in wxArrayString::Remove") );
2023
2024   Remove(iIndex);
2025 }
2026
2027 // ----------------------------------------------------------------------------
2028 // sorting
2029 // ----------------------------------------------------------------------------
2030
2031 // we can only sort one array at a time with the quick-sort based
2032 // implementation
2033 #if wxUSE_THREADS
2034   // need a critical section to protect access to gs_compareFunction and
2035   // gs_sortAscending variables
2036   static wxCriticalSection *gs_critsectStringSort = NULL;
2037
2038   // call this before the value of the global sort vars is changed/after
2039   // you're finished with them
2040   #define START_SORT()     wxASSERT( !gs_critsectStringSort );                \
2041                            gs_critsectStringSort = new wxCriticalSection;     \
2042                            gs_critsectStringSort->Enter()
2043   #define END_SORT()       gs_critsectStringSort->Leave();                    \
2044                            delete gs_critsectStringSort;                      \
2045                            gs_critsectStringSort = NULL
2046 #else // !threads
2047   #define START_SORT()
2048   #define END_SORT()
2049 #endif // wxUSE_THREADS
2050
2051 // function to use for string comparaison
2052 static wxArrayString::CompareFunction gs_compareFunction = NULL;
2053
2054 // if we don't use the compare function, this flag tells us if we sort the
2055 // array in ascending or descending order
2056 static bool gs_sortAscending = TRUE;
2057
2058 // function which is called by quick sort
2059 extern "C" int LINKAGEMODE
2060 wxStringCompareFunction(const void *first, const void *second)
2061 {
2062   wxString *strFirst = (wxString *)first;
2063   wxString *strSecond = (wxString *)second;
2064
2065   if ( gs_compareFunction ) {
2066     return gs_compareFunction(*strFirst, *strSecond);
2067   }
2068   else {
2069     // maybe we should use wxStrcoll
2070     int result = wxStrcmp(strFirst->c_str(), strSecond->c_str());
2071
2072     return gs_sortAscending ? result : -result;
2073   }
2074 }
2075
2076 // sort array elements using passed comparaison function
2077 void wxArrayString::Sort(CompareFunction compareFunction)
2078 {
2079   START_SORT();
2080
2081   wxASSERT( !gs_compareFunction );  // must have been reset to NULL
2082   gs_compareFunction = compareFunction;
2083
2084   DoSort();
2085
2086   // reset it to NULL so that Sort(bool) will work the next time
2087   gs_compareFunction = NULL;
2088
2089   END_SORT();
2090 }
2091
2092 void wxArrayString::Sort(bool reverseOrder)
2093 {
2094   START_SORT();
2095
2096   wxASSERT( !gs_compareFunction );  // must have been reset to NULL
2097   gs_sortAscending = !reverseOrder;
2098
2099   DoSort();
2100
2101   END_SORT();
2102 }
2103
2104 void wxArrayString::DoSort()
2105 {
2106   wxCHECK_RET( !m_autoSort, wxT("can't use this method with sorted arrays") );
2107
2108   // just sort the pointers using qsort() - of course it only works because
2109   // wxString() *is* a pointer to its data
2110   qsort(m_pItems, m_nCount, sizeof(wxChar *), wxStringCompareFunction);
2111 }
2112
2113 bool wxArrayString::operator==(const wxArrayString& a) const
2114 {
2115     if ( m_nCount != a.m_nCount )
2116         return FALSE;
2117
2118     for ( size_t n = 0; n < m_nCount; n++ )
2119     {
2120         if ( Item(n) != a[n] )
2121             return FALSE;
2122     }
2123
2124     return TRUE;
2125 }
2126