src/common/string.cpp

/////////////////////////////////////////////////////////////////////////////
// Name:        src/common/string.cpp
// Purpose:     wxString class
// Author:      Vadim Zeitlin, Ryan Norton
// Modified by:
// Created:     29/01/98
// RCS-ID:      $Id$
// Copyright:   (c) 1998 Vadim Zeitlin <zeitlin@dptmaths.ens-cachan.fr>
//              (c) 2004 Ryan Norton <wxprojects@comcast.net>
// Licence:     wxWindows licence
/////////////////////////////////////////////////////////////////////////////

/*
 * About ref counting:
 *  1) all empty strings use g_strEmpty, nRefs = -1 (set in Init())
 *  2) AllocBuffer() sets nRefs to 1, Lock() increments it by one
 *  3) Unlock() decrements nRefs and frees memory if it goes to 0
 */

// ===========================================================================
// headers, declarations, constants
// ===========================================================================

// For compilers that support precompilation, includes "wx.h".
#include "wx/wxprec.h"

#ifdef __BORLANDC__
    #pragma hdrstop
#endif

#ifndef WX_PRECOMP
    #include "wx/string.h"
    #include "wx/intl.h"
    #include "wx/thread.h"
#endif

#include <ctype.h>
#include <string.h>
#include <stdlib.h>

#ifdef __SALFORDC__
    #include <clib.h>
#endif

// allocating extra space for each string consumes more memory but speeds up
// the concatenation operations (nLen is the current string's length)
// NB: EXTRA_ALLOC must be >= 0!
#define EXTRA_ALLOC       (19 - nLen % 16)

// ---------------------------------------------------------------------------
// static class variables definition
// ---------------------------------------------------------------------------

#if !wxUSE_STL
  //According to STL _must_ be a -1 size_t
  const size_t wxStringBase::npos = (size_t) -1;
#endif

// ----------------------------------------------------------------------------
// static data
// ----------------------------------------------------------------------------

#if wxUSE_STL

extern const wxChar WXDLLIMPEXP_BASE *wxEmptyString = _T("");

#else

// for an empty string, GetStringData() will return this address: this
// structure has the same layout as wxStringData and it's data() method will
// return the empty string (dummy pointer)
static const struct
{
  wxStringData data;
  wxChar dummy;
} g_strEmpty = { {-1, 0, 0}, wxT('\0') };

// empty C style string: points to 'string data' byte of g_strEmpty
extern const wxChar WXDLLIMPEXP_BASE *wxEmptyString = &g_strEmpty.dummy;

#endif

// ----------------------------------------------------------------------------
// global functions
// ----------------------------------------------------------------------------

#if wxUSE_STD_IOSTREAM

// MS Visual C++ version 5.0 provides the new STL headers as well as the old
// iostream ones.
//
// ATTN: you can _not_ use both of these in the same program!

#include <iostream>

wxSTD istream& operator>>(wxSTD istream& is, wxString& WXUNUSED(str))
{
#if 0
  int w = is.width(0);
  if ( is.ipfx(0) ) {
    streambuf *sb = is.rdbuf();
    str.erase();
    while ( true ) {
      int ch = sb->sbumpc ();
      if ( ch == EOF ) {
        is.setstate(ios::eofbit);
        break;
      }
      else if ( isspace(ch) ) {
        sb->sungetc();
        break;
      }

      str += ch;
      if ( --w == 1 )
        break;
    }
  }

  is.isfx();
  if ( str.length() == 0 )
    is.setstate(ios::failbit);
#endif
  return is;
}

wxSTD ostream& operator<<(wxSTD ostream& os, const wxString& str)
{
#ifdef __BORLANDC__
    os << str.mb_str();
#else
    os << str.c_str();
#endif
    return os;
}

#endif // wxUSE_STD_IOSTREAM

// ----------------------------------------------------------------------------
// private classes
// ----------------------------------------------------------------------------

// this small class is used to gather statistics for performance tuning
//#define WXSTRING_STATISTICS
#ifdef  WXSTRING_STATISTICS
  class Averager
  {
  public:
    Averager(const wxChar *sz) { m_sz = sz; m_nTotal = m_nCount = 0; }
   ~Averager()
   { wxPrintf("wxString: average %s = %f\n", m_sz, ((float)m_nTotal)/m_nCount); }

    void Add(size_t n) { m_nTotal += n; m_nCount++; }

  private:
    size_t m_nCount, m_nTotal;
    const wxChar *m_sz;
  } g_averageLength("allocation size"),
    g_averageSummandLength("summand length"),
    g_averageConcatHit("hit probability in concat"),
    g_averageInitialLength("initial string length");

  #define STATISTICS_ADD(av, val) g_average##av.Add(val)
#else
  #define STATISTICS_ADD(av, val)
#endif // WXSTRING_STATISTICS

#if !wxUSE_STL

// ===========================================================================
// wxStringData class deallocation
// ===========================================================================

#if defined(__VISUALC__) && defined(_MT) && !defined(_DLL)
#  pragma message (__FILE__ ": building with Multithreaded non DLL runtime has a performance impact on wxString!")
void wxStringData::Free()
{
    free(this);
}
#endif

// ===========================================================================
// wxStringBase
// ===========================================================================

// takes nLength elements of psz starting at nPos
void wxStringBase::InitWith(const wxChar *psz, size_t nPos, size_t nLength)
{
  Init();

  // if the length is not given, assume the string to be NUL terminated
  if ( nLength == npos ) {
    wxASSERT_MSG( nPos <= wxStrlen(psz), _T("index out of bounds") );

    nLength = wxStrlen(psz + nPos);
  }

  STATISTICS_ADD(InitialLength, nLength);

  if ( nLength > 0 ) {
    // trailing '\0' is written in AllocBuffer()
    if ( !AllocBuffer(nLength) ) {
      wxFAIL_MSG( _T("out of memory in wxStringBase::InitWith") );
      return;
    }
    wxTmemcpy(m_pchData, psz + nPos, nLength);
  }
}

// poor man's iterators are "void *" pointers
wxStringBase::wxStringBase(const void *pStart, const void *pEnd)
{
  InitWith((const wxChar *)pStart, 0,
           (const wxChar *)pEnd - (const wxChar *)pStart);
}

wxStringBase::wxStringBase(size_type n, wxChar ch)
{
  Init();
  append(n, ch);
}

// ---------------------------------------------------------------------------
// memory allocation
// ---------------------------------------------------------------------------

// allocates memory needed to store a C string of length nLen
bool wxStringBase::AllocBuffer(size_t nLen)
{
  // allocating 0 sized buffer doesn't make sense, all empty strings should
  // reuse g_strEmpty
  wxASSERT( nLen >  0 );

  // make sure that we don't overflow
  wxASSERT( nLen < (INT_MAX / sizeof(wxChar)) -
                   (sizeof(wxStringData) + EXTRA_ALLOC + 1) );

  STATISTICS_ADD(Length, nLen);

  // allocate memory:
  // 1) one extra character for '\0' termination
  // 2) sizeof(wxStringData) for housekeeping info
  wxStringData* pData = (wxStringData*)
    malloc(sizeof(wxStringData) + (nLen + EXTRA_ALLOC + 1)*sizeof(wxChar));

  if ( pData == NULL ) {
    // allocation failures are handled by the caller
    return false;
  }

  pData->nRefs        = 1;
  pData->nDataLength  = nLen;
  pData->nAllocLength = nLen + EXTRA_ALLOC;
  m_pchData           = pData->data();  // data starts after wxStringData
  m_pchData[nLen]     = wxT('\0');
  return true;
}

// must be called before changing this string
bool wxStringBase::CopyBeforeWrite()
{
  wxStringData* pData = GetStringData();

  if ( pData->IsShared() ) {
    pData->Unlock();                // memory not freed because shared
    size_t nLen = pData->nDataLength;
    if ( !AllocBuffer(nLen) ) {
      // allocation failures are handled by the caller
      return false;
    }
    wxTmemcpy(m_pchData, pData->data(), nLen);
  }

  wxASSERT( !GetStringData()->IsShared() );  // we must be the only owner

  return true;
}

// must be called before replacing contents of this string
bool wxStringBase::AllocBeforeWrite(size_t nLen)
{
  wxASSERT( nLen != 0 );  // doesn't make any sense

  // must not share string and must have enough space
  wxStringData* pData = GetStringData();
  if ( pData->IsShared() || pData->IsEmpty() ) {
    // can't work with old buffer, get new one
    pData->Unlock();
    if ( !AllocBuffer(nLen) ) {
      // allocation failures are handled by the caller
      return false;
    }
  }
  else {
    if ( nLen > pData->nAllocLength ) {
      // realloc the buffer instead of calling malloc() again, this is more
      // efficient
      STATISTICS_ADD(Length, nLen);

      nLen += EXTRA_ALLOC;

      pData = (wxStringData*)
          realloc(pData, sizeof(wxStringData) + (nLen + 1)*sizeof(wxChar));

      if ( pData == NULL ) {
        // allocation failures are handled by the caller
        // keep previous data since reallocation failed
        return false;
      }

      pData->nAllocLength = nLen;
      m_pchData = pData->data();
    }
  }

  wxASSERT( !GetStringData()->IsShared() );  // we must be the only owner

  // it doesn't really matter what the string length is as it's going to be
  // overwritten later but, for extra safety, set it to 0 for now as we may
  // have some junk in m_pchData
  GetStringData()->nDataLength = 0;

  return true;
}

wxStringBase& wxStringBase::append(size_t n, wxChar ch)
{
    size_type len = length();

    if ( !Alloc(len + n) || !CopyBeforeWrite() ) {
      wxFAIL_MSG( _T("out of memory in wxStringBase::append") );
    }
    GetStringData()->nDataLength = len + n;
    m_pchData[len + n] = '\0';
    for ( size_t i = 0; i < n; ++i )
        m_pchData[len + i] = ch;
    return *this;
}

void wxStringBase::resize(size_t nSize, wxChar ch)
{
    size_t len = length();

    if ( nSize < len )
    {
        erase(begin() + nSize, end());
    }
    else if ( nSize > len )
    {
        append(nSize - len, ch);
    }
    //else: we have exactly the specified length, nothing to do
}

// allocate enough memory for nLen characters
bool wxStringBase::Alloc(size_t nLen)
{
  wxStringData *pData = GetStringData();
  if ( pData->nAllocLength <= nLen ) {
    if ( pData->IsEmpty() ) {
      nLen += EXTRA_ALLOC;

      pData = (wxStringData *)
                malloc(sizeof(wxStringData) + (nLen + 1)*sizeof(wxChar));

      if ( pData == NULL ) {
        // allocation failure handled by caller
        return false;
      }

      pData->nRefs = 1;
      pData->nDataLength = 0;
      pData->nAllocLength = nLen;
      m_pchData = pData->data();  // data starts after wxStringData
      m_pchData[0u] = wxT('\0');
    }
    else if ( pData->IsShared() ) {
      pData->Unlock();                // memory not freed because shared
      size_t nOldLen = pData->nDataLength;
      if ( !AllocBuffer(nLen) ) {
        // allocation failure handled by caller
        return false;
      }
      // +1 to copy the terminator, too
      memcpy(m_pchData, pData->data(), (nOldLen+1)*sizeof(wxChar));
      GetStringData()->nDataLength = nOldLen;
    }
    else {
      nLen += EXTRA_ALLOC;

      pData = (wxStringData *)
        realloc(pData, sizeof(wxStringData) + (nLen + 1)*sizeof(wxChar));

      if ( pData == NULL ) {
        // allocation failure handled by caller
        // keep previous data since reallocation failed
        return false;
      }

      // it's not important if the pointer changed or not (the check for this
      // is not faster than assigning to m_pchData in all cases)
      pData->nAllocLength = nLen;
      m_pchData = pData->data();
    }
  }
  //else: we've already got enough
  return true;
}

wxStringBase::iterator wxStringBase::begin()
{
    if (length() > 0)
        CopyBeforeWrite();
    return m_pchData;
}

wxStringBase::iterator wxStringBase::end()
{
    if (length() > 0)
        CopyBeforeWrite();
    return m_pchData + length();
}

wxStringBase::iterator wxStringBase::erase(iterator it)
{
    size_type idx = it - begin();
    erase(idx, 1);
    return begin() + idx;
}

wxStringBase& wxStringBase::erase(size_t nStart, size_t nLen)
{
    wxASSERT(nStart <= length());
    size_t strLen = length() - nStart;
    // delete nLen or up to the end of the string characters
    nLen = strLen < nLen ? strLen : nLen;
    wxString strTmp(c_str(), nStart);
    strTmp.append(c_str() + nStart + nLen, length() - nStart - nLen);

    swap(strTmp);
    return *this;
}

wxStringBase& wxStringBase::insert(size_t nPos, const wxChar *sz, size_t n)
{
    wxASSERT( nPos <= length() );

    if ( n == npos ) n = wxStrlen(sz);
    if ( n == 0 ) return *this;

    if ( !Alloc(length() + n) || !CopyBeforeWrite() ) {
        wxFAIL_MSG( _T("out of memory in wxStringBase::insert") );
    }

    memmove(m_pchData + nPos + n, m_pchData + nPos,
            (length() - nPos) * sizeof(wxChar));
    memcpy(m_pchData + nPos, sz, n * sizeof(wxChar));
    GetStringData()->nDataLength = length() + n;
    m_pchData[length()] = '\0';

    return *this;
}

void wxStringBase::swap(wxStringBase& str)
{
    wxChar* tmp = str.m_pchData;
    str.m_pchData = m_pchData;
    m_pchData = tmp;
}

size_t wxStringBase::find(const wxStringBase& str, size_t nStart) const
{
    wxASSERT( str.GetStringData()->IsValid() );
    wxASSERT( nStart <= length() );

    //anchor
    const wxChar* p = (const wxChar*)wxTmemchr(c_str() + nStart,
                                               str.c_str()[0],
                                               length() - nStart);

    if(!p)
        return npos;

    while(p - c_str() + str.length() <= length() &&
          wxTmemcmp(p, str.c_str(), str.length()) )
    {
        //Previosly passed as the first argument to wxTmemchr,
        //but C/C++ standard does not specify evaluation order
        //of arguments to functions -
        //http://embedded.com/showArticle.jhtml?articleID=9900607
        ++p;

        //anchor again
        p = (const wxChar*)wxTmemchr(p,
                                     str.c_str()[0],
                                     length() - (p - c_str()));

        if(!p)
            return npos;
    }

    return (p - c_str() + str.length() <= length()) ? p - c_str() : npos;
}

size_t wxStringBase::find(const wxChar* sz, size_t nStart, size_t n) const
{
    return find(wxStringBase(sz, n), nStart);
}

size_t wxStringBase::find(wxChar ch, size_t nStart) const
{
    wxASSERT( nStart <= length() );

    const wxChar *p = (const wxChar*)wxTmemchr(c_str() + nStart, ch, length() - nStart);

    return p == NULL ? npos : p - c_str();
}

size_t wxStringBase::rfind(const wxStringBase& str, size_t nStart) const
{
    wxASSERT( str.GetStringData()->IsValid() );
    wxASSERT( nStart == npos || nStart <= length() );

    if ( length() >= str.length() )
    {
        // avoids a corner case later
        if ( length() == 0 && str.length() == 0 )
            return 0;

        // "top" is the point where search starts from
        size_t top = length() - str.length();

        if ( nStart == npos )
            nStart = length() - 1;
        if ( nStart < top )
            top = nStart;

        const wxChar *cursor = c_str() + top;
        do
        {
            if ( wxTmemcmp(cursor, str.c_str(),
                        str.length()) == 0 )
            {
                return cursor - c_str();
            }
        } while ( cursor-- > c_str() );
    }

    return npos;
}

size_t wxStringBase::rfind(const wxChar* sz, size_t nStart, size_t n) const
{
    return rfind(wxStringBase(sz, n), nStart);
}

size_t wxStringBase::rfind(wxChar ch, size_t nStart) const
{
    if ( nStart == npos )
    {
        nStart = length();
    }
    else
    {
        wxASSERT( nStart <= length() );
    }

    const wxChar *actual;
    for ( actual = c_str() + ( nStart == npos ? length() : nStart + 1 );
          actual > c_str(); --actual )
    {
        if ( *(actual - 1) == ch )
            return (actual - 1) - c_str();
    }

    return npos;
}

size_t wxStringBase::find_first_of(const wxChar* sz, size_t nStart) const
{
    wxASSERT(nStart <= length());

    size_t len = wxStrlen(sz);

    size_t i;
    for(i = nStart; i < this->length(); ++i)
    {
        if (wxTmemchr(sz, *(c_str() + i), len))
            break;
    }

    if(i == this->length())
        return npos;
    else
        return i;
}

size_t wxStringBase::find_first_of(const wxChar* sz, size_t nStart,
                                   size_t n) const
{
    return find_first_of(wxStringBase(sz, n), nStart);
}

size_t wxStringBase::find_last_of(const wxChar* sz, size_t nStart) const
{
    if ( nStart == npos )
    {
        nStart = length() - 1;
    }
    else
    {
        wxASSERT_MSG( nStart <= length(),
                        _T("invalid index in find_last_of()") );
    }

    size_t len = wxStrlen(sz);

    for ( const wxChar *p = c_str() + nStart; p >= c_str(); --p )
    {
        if ( wxTmemchr(sz, *p, len) )
            return p - c_str();
    }

    return npos;
}

size_t wxStringBase::find_last_of(const wxChar* sz, size_t nStart,
                                   size_t n) const
{
    return find_last_of(wxStringBase(sz, n), nStart);
}

size_t wxStringBase::find_first_not_of(const wxChar* sz, size_t nStart) const
{
    if ( nStart == npos )
    {
        nStart = length();
    }
    else
    {
        wxASSERT( nStart <= length() );
    }

    size_t len = wxStrlen(sz);

    size_t i;
    for(i = nStart; i < this->length(); ++i)
    {
        if (!wxTmemchr(sz, *(c_str() + i), len))
            break;
    }

    if(i == this->length())
         return npos;
     else
        return i;
}

size_t wxStringBase::find_first_not_of(const wxChar* sz, size_t nStart,
                                       size_t n) const
{
    return find_first_not_of(wxStringBase(sz, n), nStart);
}

size_t wxStringBase::find_first_not_of(wxChar ch, size_t nStart) const
{
    wxASSERT( nStart <= length() );

    for ( const wxChar *p = c_str() + nStart; *p; p++ )
    {
        if ( *p != ch )
            return p - c_str();
    }

    return npos;
}

size_t wxStringBase::find_last_not_of(const wxChar* sz, size_t nStart) const
{
    if ( nStart == npos )
    {
        nStart = length() - 1;
    }
    else
    {
        wxASSERT( nStart <= length() );
    }

    size_t len = wxStrlen(sz);

    for ( const wxChar *p = c_str() + nStart; p >= c_str(); --p )
    {
        if ( !wxTmemchr(sz, *p,len) )
             return p - c_str();
    }

    return npos;
}

size_t wxStringBase::find_last_not_of(const wxChar* sz, size_t nStart,
                                      size_t n) const
{
    return find_last_not_of(wxStringBase(sz, n), nStart);
}

size_t wxStringBase::find_last_not_of(wxChar ch, size_t nStart) const
{
    if ( nStart == npos )
    {
        nStart = length() - 1;
    }
    else
    {
        wxASSERT( nStart <= length() );
    }

    for ( const wxChar *p = c_str() + nStart; p >= c_str(); --p )
    {
        if ( *p != ch )
            return p - c_str();
    }

    return npos;
}

wxStringBase& wxStringBase::replace(size_t nStart, size_t nLen,
                                    const wxChar *sz)
{
  wxASSERT_MSG( nStart <= length(),
                _T("index out of bounds in wxStringBase::replace") );
  size_t strLen = length() - nStart;
  nLen = strLen < nLen ? strLen : nLen;

  wxStringBase strTmp;
  strTmp.reserve(length()); // micro optimisation to avoid multiple mem allocs

  //This is kind of inefficient, but its pretty good considering...
  //we don't want to use character access operators here because on STL
  //it will freeze the reference count of strTmp, which means a deep copy
  //at the end when swap is called
  //
  //Also, we can't use append with the full character pointer and must
  //do it manually because this string can contain null characters
  for(size_t i1 = 0; i1 < nStart; ++i1)
      strTmp.append(1, this->c_str()[i1]);

  //its safe to do the full version here because
  //sz must be a normal c string
  strTmp.append(sz);

  for(size_t i2 = nStart + nLen; i2 < length(); ++i2)
      strTmp.append(1, this->c_str()[i2]);

  swap(strTmp);
  return *this;
}

wxStringBase& wxStringBase::replace(size_t nStart, size_t nLen,
                                    size_t nCount, wxChar ch)
{
  return replace(nStart, nLen, wxStringBase(nCount, ch).c_str());
}

wxStringBase& wxStringBase::replace(size_t nStart, size_t nLen,
                                    const wxStringBase& str,
                                    size_t nStart2, size_t nLen2)
{
  return replace(nStart, nLen, str.substr(nStart2, nLen2));
}

wxStringBase& wxStringBase::replace(size_t nStart, size_t nLen,
                                    const wxChar* sz, size_t nCount)
{
  return replace(nStart, nLen, wxStringBase(sz, nCount).c_str());
}

wxStringBase wxStringBase::substr(size_t nStart, size_t nLen) const
{
  if ( nLen == npos )
    nLen = length() - nStart;
  return wxStringBase(*this, nStart, nLen);
}

// assigns one string to another
wxStringBase& wxStringBase::operator=(const wxStringBase& stringSrc)
{
  wxASSERT( stringSrc.GetStringData()->IsValid() );

  // don't copy string over itself
  if ( m_pchData != stringSrc.m_pchData ) {
    if ( stringSrc.GetStringData()->IsEmpty() ) {
      Reinit();
    }
    else {
      // adjust references
      GetStringData()->Unlock();
      m_pchData = stringSrc.m_pchData;
      GetStringData()->Lock();
    }
  }

  return *this;
}

// assigns a single character
wxStringBase& wxStringBase::operator=(wxChar ch)
{
  if ( !AssignCopy(1, &ch) ) {
    wxFAIL_MSG( _T("out of memory in wxStringBase::operator=(wxChar)") );
  }
  return *this;
}

// assigns C string
wxStringBase& wxStringBase::operator=(const wxChar *psz)
{
  if ( !AssignCopy(wxStrlen(psz), psz) ) {
    wxFAIL_MSG( _T("out of memory in wxStringBase::operator=(const wxChar *)") );
  }
  return *this;
}

// helper function: does real copy
bool wxStringBase::AssignCopy(size_t nSrcLen, const wxChar *pszSrcData)
{
  if ( nSrcLen == 0 ) {
    Reinit();
  }
  else {
    if ( !AllocBeforeWrite(nSrcLen) ) {
      // allocation failure handled by caller
      return false;
    }
    memcpy(m_pchData, pszSrcData, nSrcLen*sizeof(wxChar));
    GetStringData()->nDataLength = nSrcLen;
    m_pchData[nSrcLen] = wxT('\0');
  }
  return true;
}

// ---------------------------------------------------------------------------
// string concatenation
// ---------------------------------------------------------------------------

// add something to this string
bool wxStringBase::ConcatSelf(size_t nSrcLen, const wxChar *pszSrcData,
                              size_t nMaxLen)
{
  STATISTICS_ADD(SummandLength, nSrcLen);

  nSrcLen = nSrcLen < nMaxLen ? nSrcLen : nMaxLen;

  // concatenating an empty string is a NOP
  if ( nSrcLen > 0 ) {
    wxStringData *pData = GetStringData();
    size_t nLen = pData->nDataLength;
    size_t nNewLen = nLen + nSrcLen;

    // alloc new buffer if current is too small
    if ( pData->IsShared() ) {
      STATISTICS_ADD(ConcatHit, 0);

      // we have to allocate another buffer
      wxStringData* pOldData = GetStringData();
      if ( !AllocBuffer(nNewLen) ) {
          // allocation failure handled by caller
          return false;
      }
      memcpy(m_pchData, pOldData->data(), nLen*sizeof(wxChar));
      pOldData->Unlock();
    }
    else if ( nNewLen > pData->nAllocLength ) {
      STATISTICS_ADD(ConcatHit, 0);

      reserve(nNewLen);
      // we have to grow the buffer
      if ( capacity() < nNewLen ) {
          // allocation failure handled by caller
          return false;
      }
    }
    else {
      STATISTICS_ADD(ConcatHit, 1);

      // the buffer is already big enough
    }

    // should be enough space
    wxASSERT( nNewLen <= GetStringData()->nAllocLength );

    // fast concatenation - all is done in our buffer
    memcpy(m_pchData + nLen, pszSrcData, nSrcLen*sizeof(wxChar));

    m_pchData[nNewLen] = wxT('\0');          // put terminating '\0'
    GetStringData()->nDataLength = nNewLen; // and fix the length
  }
  //else: the string to append was empty
  return true;
}

// ---------------------------------------------------------------------------
// simple sub-string extraction
// ---------------------------------------------------------------------------

// helper function: clone the data attached to this string
bool wxStringBase::AllocCopy(wxString& dest, int nCopyLen, int nCopyIndex) const
{
  if ( nCopyLen == 0 ) {
    dest.Init();
  }
  else {
    if ( !dest.AllocBuffer(nCopyLen) ) {
      // allocation failure handled by caller
      return false;
    }
    memcpy(dest.m_pchData, m_pchData + nCopyIndex, nCopyLen*sizeof(wxChar));
  }
  return true;
}

#endif // !wxUSE_STL

#if !wxUSE_STL || !defined(HAVE_STD_STRING_COMPARE)

#if !wxUSE_STL
    #define STRINGCLASS wxStringBase
#else
    #define STRINGCLASS wxString
#endif

static inline int wxDoCmp(const wxChar* s1, size_t l1,
                          const wxChar* s2, size_t l2)
{
    if( l1 == l2 )
        return wxTmemcmp(s1, s2, l1);
    else if( l1 < l2 )
    {
        int ret = wxTmemcmp(s1, s2, l1);
        return ret == 0 ? -1 : ret;
    }
    else
    {
        int ret = wxTmemcmp(s1, s2, l2);
        return ret == 0 ? +1 : ret;
    }
}

int STRINGCLASS::compare(const wxStringBase& str) const
{
    return ::wxDoCmp(data(), length(), str.data(), str.length());
}

int STRINGCLASS::compare(size_t nStart, size_t nLen,
                         const wxStringBase& str) const
{
    wxASSERT(nStart <= length());
    size_type strLen = length() - nStart;
    nLen = strLen < nLen ? strLen : nLen;
    return ::wxDoCmp(data() + nStart, nLen, str.data(), str.length());
}

int STRINGCLASS::compare(size_t nStart, size_t nLen,
                         const wxStringBase& str,
                         size_t nStart2, size_t nLen2) const
{
    wxASSERT(nStart <= length());
    wxASSERT(nStart2 <= str.length());
    size_type strLen  =     length() - nStart,
              strLen2 = str.length() - nStart2;
    nLen  = strLen  < nLen  ? strLen  : nLen;
    nLen2 = strLen2 < nLen2 ? strLen2 : nLen2;
    return ::wxDoCmp(data() + nStart, nLen, str.data() + nStart2, nLen2);
}

int STRINGCLASS::compare(const wxChar* sz) const
{
    size_t nLen = wxStrlen(sz);
    return ::wxDoCmp(data(), length(), sz, nLen);
}

int STRINGCLASS::compare(size_t nStart, size_t nLen,
                         const wxChar* sz, size_t nCount) const
{
    wxASSERT(nStart <= length());
    size_type strLen = length() - nStart;
    nLen = strLen < nLen ? strLen : nLen;
    if( nCount == npos )
        nCount = wxStrlen(sz);

    return ::wxDoCmp(data() + nStart, nLen, sz, nCount);
}

#undef STRINGCLASS

#endif // !wxUSE_STL || !defined(HAVE_STD_STRING_COMPARE)

// ===========================================================================
// wxString class core
// ===========================================================================

// ---------------------------------------------------------------------------
// construction and conversion
// ---------------------------------------------------------------------------

#if wxUSE_UNICODE

// from multibyte string
wxString::wxString(const char *psz, const wxMBConv& conv, size_t nLength)
{
    // anything to do?
    if ( psz && nLength != 0 )
    {
        if ( nLength == npos )
        {
            nLength = wxNO_LEN;
        }

        size_t nLenWide;
        wxWCharBuffer wbuf = conv.cMB2WC(psz, nLength, &nLenWide);

        if ( nLenWide )
            assign(wbuf, nLenWide);
    }
}

//Convert wxString in Unicode mode to a multi-byte string
const wxCharBuffer wxString::mb_str(const wxMBConv& conv) const
{
    return conv.cWC2MB(c_str(), length() + 1 /* size, not length */, NULL);
}

#else // ANSI

#if wxUSE_WCHAR_T

// from wide string
wxString::wxString(const wchar_t *pwz, const wxMBConv& conv, size_t nLength)
{
    // anything to do?
    if ( pwz && nLength != 0 )
    {
        if ( nLength == npos )
        {
            nLength = wxNO_LEN;
        }

        size_t nLenMB;
        wxCharBuffer buf = conv.cWC2MB(pwz, nLength, &nLenMB);

        if ( nLenMB )
            assign(buf, nLenMB);
    }
}

//Converts this string to a wide character string if unicode
//mode is not enabled and wxUSE_WCHAR_T is enabled
const wxWCharBuffer wxString::wc_str(const wxMBConv& conv) const
{
    return conv.cMB2WC(c_str(), length() + 1 /* size, not length */, NULL);
}

#endif // wxUSE_WCHAR_T

#endif // Unicode/ANSI

// shrink to minimal size (releasing extra memory)
bool wxString::Shrink()
{
  wxString tmp(begin(), end());
  swap(tmp);
  return tmp.length() == length();
}

#if !wxUSE_STL
// get the pointer to writable buffer of (at least) nLen bytes
wxChar *wxString::GetWriteBuf(size_t nLen)
{
  if ( !AllocBeforeWrite(nLen) ) {
    // allocation failure handled by caller
    return NULL;
  }

  wxASSERT( GetStringData()->nRefs == 1 );
  GetStringData()->Validate(false);

  return m_pchData;
}

// put string back in a reasonable state after GetWriteBuf
void wxString::UngetWriteBuf()
{
  GetStringData()->nDataLength = wxStrlen(m_pchData);
  GetStringData()->Validate(true);
}

void wxString::UngetWriteBuf(size_t nLen)
{
  GetStringData()->nDataLength = nLen;
  GetStringData()->Validate(true);
}
#endif

// ---------------------------------------------------------------------------
// data access
// ---------------------------------------------------------------------------

// all functions are inline in string.h

// ---------------------------------------------------------------------------
// assignment operators
// ---------------------------------------------------------------------------

#if !wxUSE_UNICODE

// same as 'signed char' variant
wxString& wxString::operator=(const unsigned char* psz)
{
  *this = (const char *)psz;
  return *this;
}

#if wxUSE_WCHAR_T
wxString& wxString::operator=(const wchar_t *pwz)
{
  wxString str(pwz);
  swap(str);
  return *this;
}
#endif

#endif

/*
 * concatenation functions come in 5 flavours:
 *  string + string
 *  char   + string      and      string + char
 *  C str  + string      and      string + C str
 */

wxString operator+(const wxString& str1, const wxString& str2)
{
#if !wxUSE_STL
  wxASSERT( str1.GetStringData()->IsValid() );
  wxASSERT( str2.GetStringData()->IsValid() );
#endif

  wxString s = str1;
  s += str2;

  return s;
}

wxString operator+(const wxString& str, wxChar ch)
{
#if !wxUSE_STL
  wxASSERT( str.GetStringData()->IsValid() );
#endif

  wxString s = str;
  s += ch;

  return s;
}

wxString operator+(wxChar ch, const wxString& str)
{
#if !wxUSE_STL
  wxASSERT( str.GetStringData()->IsValid() );
#endif

  wxString s = ch;
  s += str;

  return s;
}

wxString operator+(const wxString& str, const wxChar *psz)
{
#if !wxUSE_STL
  wxASSERT( str.GetStringData()->IsValid() );
#endif

  wxString s;
  if ( !s.Alloc(wxStrlen(psz) + str.Len()) ) {
    wxFAIL_MSG( _T("out of memory in wxString::operator+") );
  }
  s += str;
  s += psz;

  return s;
}

wxString operator+(const wxChar *psz, const wxString& str)
{
#if !wxUSE_STL
  wxASSERT( str.GetStringData()->IsValid() );
#endif

  wxString s;
  if ( !s.Alloc(wxStrlen(psz) + str.Len()) ) {
    wxFAIL_MSG( _T("out of memory in wxString::operator+") );
  }
  s = psz;
  s += str;

  return s;
}

// ===========================================================================
// other common string functions
// ===========================================================================

int wxString::Cmp(const wxString& s) const
{
    return compare(s);
}

int wxString::Cmp(const wxChar* psz) const
{
    return compare(psz);
}

static inline int wxDoCmpNoCase(const wxChar* s1, size_t l1,
                                const wxChar* s2, size_t l2)
{
    size_t i;

    if( l1 == l2 )
    {
        for(i = 0; i < l1; ++i)
        {
            if(wxTolower(s1[i]) != wxTolower(s2[i]))
                break;
        }
        return i == l1 ? 0 : wxTolower(s1[i]) < wxTolower(s2[i]) ? -1 : 1;
    }
    else if( l1 < l2 )
    {
        for(i = 0; i < l1; ++i)
        {
            if(wxTolower(s1[i]) != wxTolower(s2[i]))
                break;
        }
        return i == l1 ? -1 : wxTolower(s1[i]) < wxTolower(s2[i]) ? -1 : 1;
    }
    else
    {
        for(i = 0; i < l2; ++i)
        {
            if(wxTolower(s1[i]) != wxTolower(s2[i]))
                break;
        }
        return i == l2 ? 1 : wxTolower(s1[i]) < wxTolower(s2[i]) ? -1 : 1;
    }
}

int wxString::CmpNoCase(const wxString& s) const
{
    return wxDoCmpNoCase(data(), length(), s.data(), s.length());
}

int wxString::CmpNoCase(const wxChar* psz) const
{
    int nLen = wxStrlen(psz);

    return wxDoCmpNoCase(data(), length(), psz, nLen);
}


#if wxUSE_UNICODE

#ifdef __MWERKS__
#ifndef __SCHAR_MAX__
#define __SCHAR_MAX__ 127
#endif
#endif

wxString wxString::FromAscii(const char *ascii)
{
    if (!ascii)
       return wxEmptyString;

    size_t len = strlen( ascii );
    wxString res;

    if ( len )
    {
        wxStringBuffer buf(res, len);

        wchar_t *dest = buf;

        for ( ;; )
        {
           if ( (*dest++ = (wchar_t)(unsigned char)*ascii++) == L'\0' )
               break;
        }
    }

    return res;
}

wxString wxString::FromAscii(const char ascii)
{
    // What do we do with '\0' ?

    wxString res;
    res += (wchar_t)(unsigned char) ascii;

    return res;
}

const wxCharBuffer wxString::ToAscii() const
{
    // this will allocate enough space for the terminating NUL too
    wxCharBuffer buffer(length());


    char *dest = buffer.data();

    const wchar_t *pwc = c_str();
    for ( ;; )
    {
        *dest++ = (char)(*pwc > SCHAR_MAX ? wxT('_') : *pwc);

        // the output string can't have embedded NULs anyhow, so we can safely
        // stop at first of them even if we do have any
        if ( !*pwc++ )
            break;
    }

    return buffer;
}

#endif // Unicode

// extract string of length nCount starting at nFirst
wxString wxString::Mid(size_t nFirst, size_t nCount) const
{
    size_t nLen = length();

    // default value of nCount is npos and means "till the end"
    if ( nCount == npos )
    {
        nCount = nLen - nFirst;
    }

    // out-of-bounds requests return sensible things
    if ( nFirst + nCount > nLen )
    {
        nCount = nLen - nFirst;
    }

    if ( nFirst > nLen )
    {
        // AllocCopy() will return empty string
        return wxEmptyString;
    }

    wxString dest(*this, nFirst, nCount);
    if ( dest.length() != nCount )
    {
        wxFAIL_MSG( _T("out of memory in wxString::Mid") );
    }

    return dest;
}

// check that the string starts with prefix and return the rest of the string
// in the provided pointer if it is not NULL, otherwise return false
bool wxString::StartsWith(const wxChar *prefix, wxString *rest) const
{
    wxASSERT_MSG( prefix, _T("invalid parameter in wxString::StartsWith") );

    // first check if the beginning of the string matches the prefix: note
    // that we don't have to check that we don't run out of this string as
    // when we reach the terminating NUL, either prefix string ends too (and
    // then it's ok) or we break out of the loop because there is no match
    const wxChar *p = c_str();
    while ( *prefix )
    {
        if ( *prefix++ != *p++ )
        {
            // no match
            return false;
        }
    }

    if ( rest )
    {
        // put the rest of the string into provided pointer
        *rest = p;
    }

    return true;
}


// check that the string ends with suffix and return the rest of it in the
// provided pointer if it is not NULL, otherwise return false
bool wxString::EndsWith(const wxChar *suffix, wxString *rest) const
{
    wxASSERT_MSG( suffix, _T("invalid parameter in wxString::EndssWith") );

    int start = length() - wxStrlen(suffix);
    if ( start < 0 || wxStrcmp(c_str() + start, suffix) != 0 )
        return false;

    if ( rest )
    {
        // put the rest of the string into provided pointer
        rest->assign(*this, 0, start);
    }

    return true;
}


// extract nCount last (rightmost) characters
wxString wxString::Right(size_t nCount) const
{
  if ( nCount > length() )
    nCount = length();

  wxString dest(*this, length() - nCount, nCount);
  if ( dest.length() != nCount ) {
    wxFAIL_MSG( _T("out of memory in wxString::Right") );
  }
  return dest;
}

// get all characters after the last occurence of ch
// (returns the whole string if ch not found)
wxString wxString::AfterLast(wxChar ch) const
{
  wxString str;
  int iPos = Find(ch, true);
  if ( iPos == wxNOT_FOUND )
    str = *this;
  else
    str = c_str() + iPos + 1;

  return str;
}

// extract nCount first (leftmost) characters
wxString wxString::Left(size_t nCount) const
{
  if ( nCount > length() )
    nCount = length();

  wxString dest(*this, 0, nCount);
  if ( dest.length() != nCount ) {
    wxFAIL_MSG( _T("out of memory in wxString::Left") );
  }
  return dest;
}

// get all characters before the first occurence of ch
// (returns the whole string if ch not found)
wxString wxString::BeforeFirst(wxChar ch) const
{
  int iPos = Find(ch);
  if ( iPos == wxNOT_FOUND ) iPos = length();
  return wxString(*this, 0, iPos);
}

/// get all characters before the last occurence of ch
/// (returns empty string if ch not found)
wxString wxString::BeforeLast(wxChar ch) const
{
  wxString str;
  int iPos = Find(ch, true);
  if ( iPos != wxNOT_FOUND && iPos != 0 )
    str = wxString(c_str(), iPos);

  return str;
}

/// get all characters after the first occurence of ch
/// (returns empty string if ch not found)
wxString wxString::AfterFirst(wxChar ch) const
{
  wxString str;
  int iPos = Find(ch);
  if ( iPos != wxNOT_FOUND )
    str = c_str() + iPos + 1;

  return str;
}

// replace first (or all) occurences of some substring with another one
size_t wxString::Replace(const wxChar *szOld,
                  const wxChar *szNew, bool bReplaceAll)
{
    // if we tried to replace an empty string we'd enter an infinite loop below
    wxCHECK_MSG( szOld && *szOld && szNew, 0,
                 _T("wxString::Replace(): invalid parameter") );

    size_t uiCount = 0;   // count of replacements made

    size_t uiOldLen = wxStrlen(szOld);
    size_t uiNewLen = wxStrlen(szNew);

    size_t dwPos = 0;

    while ( this->c_str()[dwPos] != wxT('\0') )
    {
        //DO NOT USE STRSTR HERE
        //this string can contain embedded null characters,
        //so strstr will function incorrectly
        dwPos = find(szOld, dwPos);
        if ( dwPos == npos )
            break;                  // exit the loop
        else
        {
            //replace this occurance of the old string with the new one
            replace(dwPos, uiOldLen, szNew, uiNewLen);

            //move up pos past the string that was replaced
            dwPos += uiNewLen;

            //increase replace count
            ++uiCount;

            // stop now?
            if ( !bReplaceAll )
                break;                  // exit the loop
        }
    }

    return uiCount;
}

bool wxString::IsAscii() const
{
  const wxChar *s = (const wxChar*) *this;
  while(*s){
    if(!isascii(*s)) return(false);
    s++;
  }
  return(true);
}

bool wxString::IsWord() const
{
  const wxChar *s = (const wxChar*) *this;
  while(*s){
    if(!wxIsalpha(*s)) return(false);
    s++;
  }
  return(true);
}

bool wxString::IsNumber() const
{
  const wxChar *s = (const wxChar*) *this;
  if (wxStrlen(s))
     if ((s[0] == wxT('-')) || (s[0] == wxT('+'))) s++;
  while(*s){
    if(!wxIsdigit(*s)) return(false);
    s++;
  }
  return(true);
}

wxString wxString::Strip(stripType w) const
{
    wxString s = *this;
    if ( w & leading ) s.Trim(false);
    if ( w & trailing ) s.Trim(true);
    return s;
}

// ---------------------------------------------------------------------------
// case conversion
// ---------------------------------------------------------------------------

wxString& wxString::MakeUpper()
{
  for ( iterator it = begin(), en = end(); it != en; ++it )
    *it = (wxChar)wxToupper(*it);

  return *this;
}

wxString& wxString::MakeLower()
{
  for ( iterator it = begin(), en = end(); it != en; ++it )
    *it = (wxChar)wxTolower(*it);

  return *this;
}

// ---------------------------------------------------------------------------
// trimming and padding
// ---------------------------------------------------------------------------

// some compilers (VC++ 6.0 not to name them) return true for a call to