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
/////////////////////////////////////////////////////////////////////////////

// ===========================================================================
// 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/wxcrtvararg.h"
#endif

#include <ctype.h>

#ifndef __WXWINCE__
    #include <errno.h>
#endif

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

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

#include "wx/hashmap.h"

// string handling functions used by wxString:
#if wxUSE_UNICODE_UTF8
    #define wxStringMemcpy   memcpy
    #define wxStringMemcmp   memcmp
    #define wxStringMemchr   memchr
    #define wxStringStrlen   strlen
#else
    #define wxStringMemcpy   wxTmemcpy
    #define wxStringMemcmp   wxTmemcmp
    #define wxStringMemchr   wxTmemchr
    #define wxStringStrlen   wxStrlen
#endif


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

//According to STL _must_ be a -1 size_t
const size_t wxString::npos = (size_t) -1;

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

#if wxUSE_STD_IOSTREAM

#include <iostream>

wxSTD ostream& operator<<(wxSTD ostream& os, const wxCStrData& str)
{
#if wxUSE_UNICODE && !wxUSE_UNICODE_UTF8
    return os << (const char *)str.AsCharBuf();
#else
    return os << str.AsInternal();
#endif
}

wxSTD ostream& operator<<(wxSTD ostream& os, const wxString& str)
{
    return os << str.c_str();
}

wxSTD ostream& operator<<(wxSTD ostream& os, const wxCharBuffer& str)
{
    return os << str.data();
}

#ifndef __BORLANDC__
wxSTD ostream& operator<<(wxSTD ostream& os, const wxWCharBuffer& str)
{
    return os << str.data();
}
#endif

#if wxUSE_UNICODE && defined(HAVE_WOSTREAM)

wxSTD wostream& operator<<(wxSTD wostream& wos, const wxString& str)
{
    return wos << str.wc_str();
}

wxSTD wostream& operator<<(wxSTD wostream& wos, const wxCStrData& str)
{
    return wos << str.AsWChar();
}

wxSTD wostream& operator<<(wxSTD wostream& wos, const wxWCharBuffer& str)
{
    return wos << str.data();
}

#endif  // wxUSE_UNICODE && defined(HAVE_WOSTREAM)

#endif // wxUSE_STD_IOSTREAM

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

#if wxUSE_UNICODE_UTF8

void wxString::PosLenToImpl(size_t pos, size_t len,
                            size_t *implPos, size_t *implLen) const
{
    if ( pos == npos )
        *implPos = npos;
    else
    {
        const_iterator i = begin() + pos;
        *implPos = wxStringImpl::const_iterator(i.impl()) - m_impl.begin();
        if ( len == npos )
            *implLen = npos;
        else
        {
            // too large length is interpreted as "to the end of the string"
            // FIXME-UTF8: verify this is the case in std::string, assert
            // otherwise
            if ( pos + len > length() )
                len = length() - pos;

            *implLen = (i + len).impl() - i.impl();
        }
    }
}

#endif // wxUSE_UNICODE_UTF8

// ----------------------------------------------------------------------------
// wxCStrData converted strings caching
// ----------------------------------------------------------------------------

// FIXME-UTF8: temporarily disabled because it doesn't work with global
//             string objects; re-enable after fixing this bug and benchmarking
//             performance to see if using a hash is a good idea at all
#if 0

// For backward compatibility reasons, it must be possible to assign the value
// returned by wxString::c_str() to a char* or wchar_t* variable and work with
// it. Returning wxCharBuffer from (const char*)c_str() wouldn't do the trick,
// because the memory would be freed immediately, but it has to be valid as long
// as the string is not modified, so that code like this still works:
//
// const wxChar *s = str.c_str();
// while ( s ) { ... }

// FIXME-UTF8: not thread safe!
// FIXME-UTF8: we currently clear the cached conversion only when the string is
//             destroyed, but we should do it when the string is modified, to
//             keep memory usage down
// FIXME-UTF8: we do the conversion every time As[W]Char() is called, but if we
//             invalidated the cache on every change, we could keep the previous
//             conversion
// FIXME-UTF8: add tracing of usage of these two methods - new code is supposed
//             to use mb_str() or wc_str() instead of (const [w]char*)c_str()

template<typename T>
static inline void DeleteStringFromConversionCache(T& hash, const wxString *s)
{
    typename T::iterator i = hash.find(wxConstCast(s, wxString));
    if ( i != hash.end() )
    {
        free(i->second);
        hash.erase(i);
    }
}

#if wxUSE_UNICODE
// NB: non-STL implementation doesn't compile with "const wxString*" key type,
//     so we have to use wxString* here and const-cast when used
WX_DECLARE_HASH_MAP(wxString*, char*, wxPointerHash, wxPointerEqual,
                    wxStringCharConversionCache);
static wxStringCharConversionCache gs_stringsCharCache;

const char* wxCStrData::AsChar() const
{
    // remove previously cache value, if any (see FIXMEs above):
    DeleteStringFromConversionCache(gs_stringsCharCache, m_str);

    // convert the string and keep it:
    const char *s = gs_stringsCharCache[wxConstCast(m_str, wxString)] =
        m_str->mb_str().release();

    return s + m_offset;
}
#endif // wxUSE_UNICODE

#if !wxUSE_UNICODE_WCHAR
WX_DECLARE_HASH_MAP(wxString*, wchar_t*, wxPointerHash, wxPointerEqual,
                    wxStringWCharConversionCache);
static wxStringWCharConversionCache gs_stringsWCharCache;

const wchar_t* wxCStrData::AsWChar() const
{
    // remove previously cache value, if any (see FIXMEs above):
    DeleteStringFromConversionCache(gs_stringsWCharCache, m_str);

    // convert the string and keep it:
    const wchar_t *s = gs_stringsWCharCache[wxConstCast(m_str, wxString)] =
        m_str->wc_str().release();

    return s + m_offset;
}
#endif // !wxUSE_UNICODE_WCHAR

wxString::~wxString()
{
#if wxUSE_UNICODE
    // FIXME-UTF8: do this only if locale is not UTF8 if wxUSE_UNICODE_UTF8
    DeleteStringFromConversionCache(gs_stringsCharCache, this);
#endif
#if !wxUSE_UNICODE_WCHAR
    DeleteStringFromConversionCache(gs_stringsWCharCache, this);
#endif
}
#endif

#if wxUSE_UNICODE && !wxUSE_UTF8_LOCALE_ONLY
const char* wxCStrData::AsChar() const
{
#if wxUSE_UNICODE_UTF8
    if ( wxLocaleIsUtf8 )
        return AsInternal();
#endif
    // under non-UTF8 locales, we have to convert the internal UTF-8
    // representation using wxConvLibc and cache the result

    wxString *str = wxConstCast(m_str, wxString);

    // convert the string:
    //
    // FIXME-UTF8: we'd like to do the conversion in the existing buffer (if we
    //             have it) but it's unfortunately not obvious to implement
    //             because we don't know how big buffer do we need for the
    //             given string length (in case of multibyte encodings, e.g.
    //             ISO-2022-JP or UTF-8 when internal representation is wchar_t)
    //
    //             One idea would be to store more than just m_convertedToChar
    //             in wxString: then we could record the length of the string
    //             which was converted the last time and try to reuse the same
    //             buffer if the current length is not greater than it (this
    //             could still fail because string could have been modified in
    //             place but it would work most of the time, so we'd do it and
    //             only allocate the new buffer if in-place conversion returned
    //             an error). We could also store a bit saying if the string
    //             was modified since the last conversion (and update it in all
    //             operation modifying the string, of course) to avoid unneeded
    //             consequential conversions. But both of these ideas require
    //             adding more fields to wxString and require profiling results
    //             to be sure that we really gain enough from them to justify
    //             doing it.
    wxCharBuffer buf(str->mb_str());

    // if it failed, return empty string and not NULL to avoid crashes in code
    // written with either wxWidgets 2 wxString or std::string behaviour in
    // mind: neither of them ever returns NULL and so we shouldn't neither
    if ( !buf )
        return "";

    if ( str->m_convertedToChar &&
         strlen(buf) == strlen(str->m_convertedToChar) )
    {
        // keep the same buffer for as long as possible, so that several calls
        // to c_str() in a row still work:
        strcpy(str->m_convertedToChar, buf);
    }
    else
    {
        str->m_convertedToChar = buf.release();
    }

    // and keep it:
    return str->m_convertedToChar + m_offset;
}
#endif // wxUSE_UNICODE && !wxUSE_UTF8_LOCALE_ONLY

#if !wxUSE_UNICODE_WCHAR
const wchar_t* wxCStrData::AsWChar() const
{
    wxString *str = wxConstCast(m_str, wxString);

    // convert the string:
    wxWCharBuffer buf(str->wc_str());

    // notice that here, unlike above in AsChar(), conversion can't fail as our
    // internal UTF-8 is always well-formed -- or the string was corrupted and
    // all bets are off anyhow

    // FIXME-UTF8: do the conversion in-place in the existing buffer
    if ( str->m_convertedToWChar &&
         wxWcslen(buf) == wxWcslen(str->m_convertedToWChar) )
    {
        // keep the same buffer for as long as possible, so that several calls
        // to c_str() in a row still work:
        memcpy(str->m_convertedToWChar, buf, sizeof(wchar_t) * wxWcslen(buf));
    }
    else
    {
        str->m_convertedToWChar = buf.release();
    }

    // and keep it:
    return str->m_convertedToWChar + m_offset;
}
#endif // !wxUSE_UNICODE_WCHAR

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

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

#if wxUSE_UNICODE_WCHAR
/* static */
wxString::SubstrBufFromMB wxString::ConvertStr(const char *psz, size_t nLength,
                                               const wxMBConv& conv)
{
    // anything to do?
    if ( !psz || nLength == 0 )
        return SubstrBufFromMB(L"", 0);

    if ( nLength == npos )
        nLength = wxNO_LEN;

    size_t wcLen;
    wxWCharBuffer wcBuf(conv.cMB2WC(psz, nLength, &wcLen));
    if ( !wcLen )
        return SubstrBufFromMB(_T(""), 0);
    else
        return SubstrBufFromMB(wcBuf, wcLen);
}
#endif // wxUSE_UNICODE_WCHAR

#if wxUSE_UNICODE_UTF8
/* static */
wxString::SubstrBufFromMB wxString::ConvertStr(const char *psz, size_t nLength,
                                               const wxMBConv& conv)
{
    // anything to do?
    if ( !psz || nLength == 0 )
        return SubstrBufFromMB("", 0);

    // if psz is already in UTF-8, we don't have to do the roundtrip to
    // wchar_t* and back:
    if ( conv.IsUTF8() )
    {
        // we need to validate the input because UTF8 iterators assume valid
        // UTF-8 sequence and psz may be invalid:
        if ( wxStringOperations::IsValidUtf8String(psz, nLength) )
        {
            return SubstrBufFromMB(wxCharBuffer::CreateNonOwned(psz), nLength);
        }
        // else: do the roundtrip through wchar_t*
    }

    if ( nLength == npos )
        nLength = wxNO_LEN;

    // first convert to wide string:
    size_t wcLen;
    wxWCharBuffer wcBuf(conv.cMB2WC(psz, nLength, &wcLen));
    if ( !wcLen )
        return SubstrBufFromMB("", 0);

    // and then to UTF-8:
    SubstrBufFromMB buf(ConvertStr(wcBuf, wcLen, wxMBConvStrictUTF8()));
    // widechar -> UTF-8 conversion isn't supposed to ever fail:
    wxASSERT_MSG( buf.data, _T("conversion to UTF-8 failed") );

    return buf;
}
#endif // wxUSE_UNICODE_UTF8

#if wxUSE_UNICODE_UTF8 || !wxUSE_UNICODE
/* static */
wxString::SubstrBufFromWC wxString::ConvertStr(const wchar_t *pwz, size_t nLength,
                                               const wxMBConv& conv)
{
    // anything to do?
    if ( !pwz || nLength == 0 )
        return SubstrBufFromWC("", 0);

    if ( nLength == npos )
        nLength = wxNO_LEN;

    size_t mbLen;
    wxCharBuffer mbBuf(conv.cWC2MB(pwz, nLength, &mbLen));
    if ( !mbLen )
        return SubstrBufFromWC("", 0);
    else
        return SubstrBufFromWC(mbBuf, mbLen);
}
#endif // wxUSE_UNICODE_UTF8 || !wxUSE_UNICODE


#if wxUSE_UNICODE_WCHAR

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

#elif wxUSE_UNICODE_UTF8

const wxWCharBuffer wxString::wc_str() const
{
    return wxMBConvStrictUTF8().cMB2WC
                                (
                                    m_impl.c_str(),
                                    m_impl.length() + 1, // size, not length
                                    NULL
                                );
}

const wxCharBuffer wxString::mb_str(const wxMBConv& conv) const
{
    if ( conv.IsUTF8() )
        return wxCharBuffer::CreateNonOwned(m_impl.c_str());

    // FIXME-UTF8: use wc_str() here once we have buffers with length

    size_t wcLen;
    wxWCharBuffer wcBuf(wxMBConvStrictUTF8().cMB2WC
                                             (
                                                m_impl.c_str(),
                                                m_impl.length() + 1, // size
                                                &wcLen
                                             ));
    if ( !wcLen )
        return wxCharBuffer("");

    return conv.cWC2MB(wcBuf, wcLen+1, NULL);
}

#else // ANSI

//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(wx_str(), length() + 1 /* size, not length */, NULL);
}

#endif // Unicode/ANSI

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

// deprecated compatibility code:
#if WXWIN_COMPATIBILITY_2_8 && !wxUSE_STL_BASED_WXSTRING && !wxUSE_UNICODE_UTF8
wxStringCharType *wxString::GetWriteBuf(size_t nLen)
{
    return DoGetWriteBuf(nLen);
}

void wxString::UngetWriteBuf()
{
    DoUngetWriteBuf();
}

void wxString::UngetWriteBuf(size_t nLen)
{
    DoUngetWriteBuf(nLen);
}
#endif // WXWIN_COMPATIBILITY_2_8 && !wxUSE_STL_BASED_WXSTRING && !wxUSE_UNICODE_UTF8


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

// all functions are inline in string.h

// ---------------------------------------------------------------------------
// concatenation operators
// ---------------------------------------------------------------------------

/*
 * 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_BASED_WXSTRING
    wxASSERT( str1.IsValid() );
    wxASSERT( str2.IsValid() );
#endif

    wxString s = str1;
    s += str2;

    return s;
}

wxString operator+(const wxString& str, wxUniChar ch)
{
#if !wxUSE_STL_BASED_WXSTRING
    wxASSERT( str.IsValid() );
#endif

    wxString s = str;
    s += ch;

    return s;
}

wxString operator+(wxUniChar ch, const wxString& str)
{
#if !wxUSE_STL_BASED_WXSTRING
    wxASSERT( str.IsValid() );
#endif

    wxString s = ch;
    s += str;

    return s;
}

wxString operator+(const wxString& str, const char *psz)
{
#if !wxUSE_STL_BASED_WXSTRING
    wxASSERT( str.IsValid() );
#endif

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

    return s;
}

wxString operator+(const wxString& str, const wchar_t *pwz)
{
#if !wxUSE_STL_BASED_WXSTRING
    wxASSERT( str.IsValid() );
#endif

    wxString s;
    if ( !s.Alloc(wxWcslen(pwz) + str.length()) ) {
        wxFAIL_MSG( _T("out of memory in wxString::operator+") );
    }
    s += str;
    s += pwz;

    return s;
}

wxString operator+(const char *psz, const wxString& str)
{
#if !wxUSE_STL_BASED_WXSTRING
    wxASSERT( str.IsValid() );
#endif

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

    return s;
}

wxString operator+(const wchar_t *pwz, const wxString& str)
{
#if !wxUSE_STL_BASED_WXSTRING
    wxASSERT( str.IsValid() );
#endif

    wxString s;
    if ( !s.Alloc(wxWcslen(pwz) + str.length()) ) {
        wxFAIL_MSG( _T("out of memory in wxString::operator+") );
    }
    s = pwz;
    s += str;

    return s;
}

// ---------------------------------------------------------------------------
// string comparison
// ---------------------------------------------------------------------------

bool wxString::IsSameAs(wxUniChar c, bool compareWithCase) const
{
    return (length() == 1) && (compareWithCase ? GetChar(0u) == c
                               : wxToupper(GetChar(0u)) == wxToupper(c));
}

#ifdef HAVE_STD_STRING_COMPARE

// NB: Comparison code (both if HAVE_STD_STRING_COMPARE and if not) works with
//     UTF-8 encoded strings too, thanks to UTF-8's design which allows us to
//     sort strings in characters code point order by sorting the byte sequence
//     in byte values order (i.e. what strcmp() and memcmp() do).

int wxString::compare(const wxString& str) const
{
    return m_impl.compare(str.m_impl);
}

int wxString::compare(size_t nStart, size_t nLen,
                      const wxString& str) const
{
    size_t pos, len;
    PosLenToImpl(nStart, nLen, &pos, &len);
    return m_impl.compare(pos, len, str.m_impl);
}

int wxString::compare(size_t nStart, size_t nLen,
                      const wxString& str,
                      size_t nStart2, size_t nLen2) const
{
    size_t pos, len;
    PosLenToImpl(nStart, nLen, &pos, &len);

    size_t pos2, len2;
    str.PosLenToImpl(nStart2, nLen2, &pos2, &len2);

    return m_impl.compare(pos, len, str.m_impl, pos2, len2);
}

int wxString::compare(const char* sz) const
{
    return m_impl.compare(ImplStr(sz));
}

int wxString::compare(const wchar_t* sz) const
{
    return m_impl.compare(ImplStr(sz));
}

int wxString::compare(size_t nStart, size_t nLen,
                      const char* sz, size_t nCount) const
{
    size_t pos, len;
    PosLenToImpl(nStart, nLen, &pos, &len);

    SubstrBufFromMB str(ImplStr(sz, nCount));

    return m_impl.compare(pos, len, str.data, str.len);
}

int wxString::compare(size_t nStart, size_t nLen,
                      const wchar_t* sz, size_t nCount) const
{
    size_t pos, len;
    PosLenToImpl(nStart, nLen, &pos, &len);

    SubstrBufFromWC str(ImplStr(sz, nCount));

    return m_impl.compare(pos, len, str.data, str.len);
}

#else // !HAVE_STD_STRING_COMPARE

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

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

int wxString::compare(size_t nStart, size_t nLen,
                      const wxString& str) const
{
    wxASSERT(nStart <= length());
    size_type strLen = length() - nStart;
    nLen = strLen < nLen ? strLen : nLen;

    size_t pos, len;
    PosLenToImpl(nStart, nLen, &pos, &len);

    return ::wxDoCmp(m_impl.data() + pos,  len,
                     str.m_impl.data(), str.m_impl.length());
}

int wxString::compare(size_t nStart, size_t nLen,
                      const wxString& 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;

    size_t pos, len;
    PosLenToImpl(nStart, nLen, &pos, &len);
    size_t pos2, len2;
    str.PosLenToImpl(nStart2, nLen2, &pos2, &len2);

    return ::wxDoCmp(m_impl.data() + pos, len,
                     str.m_impl.data() + pos2, len2);
}

int wxString::compare(const char* sz) const
{
    SubstrBufFromMB str(ImplStr(sz, npos));
    if ( str.len == npos )
        str.len = wxStringStrlen(str.data);
    return ::wxDoCmp(m_impl.data(), m_impl.length(), str.data, str.len);
}

int wxString::compare(const wchar_t* sz) const
{
    SubstrBufFromWC str(ImplStr(sz, npos));
    if ( str.len == npos )
        str.len = wxStringStrlen(str.data);
    return ::wxDoCmp(m_impl.data(), m_impl.length(), str.data, str.len);
}

int wxString::compare(size_t nStart, size_t nLen,
                      const char* sz, size_t nCount) const
{
    wxASSERT(nStart <= length());
    size_type strLen = length() - nStart;
    nLen = strLen < nLen ? strLen : nLen;

    size_t pos, len;
    PosLenToImpl(nStart, nLen, &pos, &len);

    SubstrBufFromMB str(ImplStr(sz, nCount));
    if ( str.len == npos )
        str.len = wxStringStrlen(str.data);

    return ::wxDoCmp(m_impl.data() + pos, len, str.data, str.len);
}

int wxString::compare(size_t nStart, size_t nLen,
                      const wchar_t* sz, size_t nCount) const
{
    wxASSERT(nStart <= length());
    size_type strLen = length() - nStart;
    nLen = strLen < nLen ? strLen : nLen;

    size_t pos, len;
    PosLenToImpl(nStart, nLen, &pos, &len);

    SubstrBufFromWC str(ImplStr(sz, nCount));
    if ( str.len == npos )
        str.len = wxStringStrlen(str.data);

    return ::wxDoCmp(m_impl.data() + pos, len, str.data, str.len);
}

#endif // HAVE_STD_STRING_COMPARE/!HAVE_STD_STRING_COMPARE


// ---------------------------------------------------------------------------
// find_{first,last}_[not]_of functions
// ---------------------------------------------------------------------------

#if !wxUSE_STL_BASED_WXSTRING || wxUSE_UNICODE_UTF8

// NB: All these functions are implemented  with the argument being wxChar*,
//     i.e. widechar string in any Unicode build, even though native string
//     representation is char* in the UTF-8 build. This is because we couldn't
//     use memchr() to determine if a character is in a set encoded as UTF-8.

size_t wxString::find_first_of(const wxChar* sz, size_t nStart) const
{
    return find_first_of(sz, nStart, wxStrlen(sz));
}

size_t wxString::find_first_not_of(const wxChar* sz, size_t nStart) const
{
    return find_first_not_of(sz, nStart, wxStrlen(sz));
}

size_t wxString::find_first_of(const wxChar* sz, size_t nStart, size_t n) const
{
    wxASSERT_MSG( nStart <= length(),  _T("invalid index") );

    size_t idx = nStart;
    for ( const_iterator i = begin() + nStart; i != end(); ++idx, ++i )
    {
        if ( wxTmemchr(sz, *i, n) )
            return idx;
    }

    return npos;
}

size_t wxString::find_first_not_of(const wxChar* sz, size_t nStart, size_t n) const
{
    wxASSERT_MSG( nStart <= length(),  _T("invalid index") );

    size_t idx = nStart;
    for ( const_iterator i = begin() + nStart; i != end(); ++idx, ++i )
    {
        if ( !wxTmemchr(sz, *i, n) )
            return idx;
    }

    return npos;
}


size_t wxString::find_last_of(const wxChar* sz, size_t nStart) const
{
    return find_last_of(sz, nStart, wxStrlen(sz));
}

size_t wxString::find_last_not_of(const wxChar* sz, size_t nStart) const
{
    return find_last_not_of(sz, nStart, wxStrlen(sz));
}

size_t wxString::find_last_of(const wxChar* sz, size_t nStart, size_t n) const
{
    size_t len = length();

    if ( nStart == npos )
    {
        nStart = len - 1;
    }
    else
    {
        wxASSERT_MSG( nStart <= len, _T("invalid index") );
    }

    size_t idx = nStart;
    for ( const_reverse_iterator i = rbegin() + (len - nStart - 1);
          i != rend(); --idx, ++i )
    {
        if ( wxTmemchr(sz, *i, n) )
            return idx;
    }

    return npos;
}

size_t wxString::find_last_not_of(const wxChar* sz, size_t nStart, size_t n) const
{
    size_t len = length();

    if ( nStart == npos )
    {
        nStart = len - 1;
    }
    else
    {
        wxASSERT_MSG( nStart <= len, _T("invalid index") );
    }

    size_t idx = nStart;
    for ( const_reverse_iterator i = rbegin() + (len - nStart - 1);
          i != rend(); --idx, ++i )
    {
        if ( !wxTmemchr(sz, *i, n) )
            return idx;
    }

    return npos;
}

size_t wxString::find_first_not_of(wxUniChar ch, size_t nStart) const
{
    wxASSERT_MSG( nStart <= length(),  _T("invalid index") );

    size_t idx = nStart;
    for ( const_iterator i = begin() + nStart; i != end(); ++idx, ++i )
    {
        if ( *i != ch )
            return idx;
    }

    return npos;
}

size_t wxString::find_last_not_of(wxUniChar ch, size_t nStart) const
{
    size_t len = length();

    if ( nStart == npos )
    {
        nStart = len - 1;
    }
    else
    {
        wxASSERT_MSG( nStart <= len, _T("invalid index") );
    }

    size_t idx = nStart;
    for ( const_reverse_iterator i = rbegin() + (len - nStart - 1);
          i != rend(); --idx, ++i )
    {
        if ( *i != ch )
            return idx;
    }

    return npos;
}

// the functions above were implemented for wchar_t* arguments in Unicode
// build and char* in ANSI build; below are implementations for the other
// version:
#if wxUSE_UNICODE
    #define wxOtherCharType char
    #define STRCONV         (const wxChar*)wxConvLibc.cMB2WC
#else
    #define wxOtherCharType wchar_t
    #define STRCONV         (const wxChar*)wxConvLibc.cWC2MB
#endif

size_t wxString::find_first_of(const wxOtherCharType* sz, size_t nStart) const
    { return find_first_of(STRCONV(sz), nStart); }

size_t wxString::find_first_of(const wxOtherCharType* sz, size_t nStart,
                               size_t n) const
    { return find_first_of(STRCONV(sz, n, NULL), nStart, n); }
size_t wxString::find_last_of(const wxOtherCharType* sz, size_t nStart) const
    { return find_last_of(STRCONV(sz), nStart); }
size_t wxString::find_last_of(const wxOtherCharType* sz, size_t nStart,
                              size_t n) const
    { return find_last_of(STRCONV(sz, n, NULL), nStart, n); }
size_t wxString::find_first_not_of(const wxOtherCharType* sz, size_t nStart) const
    { return find_first_not_of(STRCONV(sz), nStart); }
size_t wxString::find_first_not_of(const wxOtherCharType* sz, size_t nStart,
                                   size_t n) const
    { return find_first_not_of(STRCONV(sz, n, NULL), nStart, n); }
size_t wxString::find_last_not_of(const wxOtherCharType* sz, size_t nStart) const
    { return find_last_not_of(STRCONV(sz), nStart); }
size_t wxString::find_last_not_of(const wxOtherCharType* sz, size_t nStart,
                                  size_t n) const
    { return find_last_not_of(STRCONV(sz, n, NULL), nStart, n); }

#undef wxOtherCharType
#undef STRCONV

#endif // !wxUSE_STL_BASED_WXSTRING || wxUSE_UNICODE_UTF8

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

int wxString::CmpNoCase(const wxString& s) const
{
    // FIXME-UTF8: use wxUniChar::ToLower/ToUpper once added

    const_iterator i1 = begin();
    const_iterator end1 = end();
    const_iterator i2 = s.begin();
    const_iterator end2 = s.end();

    for ( ; i1 != end1 && i2 != end2; ++i1, ++i2 )
    {
        wxUniChar lower1 = (wxChar)wxTolower(*i1);
        wxUniChar lower2 = (wxChar)wxTolower(*i2);
        if ( lower1 != lower2 )
            return lower1 < lower2 ? -1 : 1;
    }

    size_t len1 = length();
    size_t len2 = s.length();

    if ( len1 < len2 )
        return -1;
    else if ( len1 > len2 )
        return 1;
    return 0;
}


#if wxUSE_UNICODE

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

wxString wxString::FromAscii(const char *ascii, size_t len)
{
    if (!ascii || len == 0)
       return wxEmptyString;

    wxString res;

    {
        wxStringInternalBuffer buf(res, len);
        wxStringCharType *dest = buf;

        for ( ; len > 0; --len )
        {
            unsigned char c = (unsigned char)*ascii++;
            wxASSERT_MSG( c < 0x80,
                          _T("Non-ASCII value passed to FromAscii().") );

            *dest++ = (wchar_t)c;
        }
    }

    return res;
}

wxString wxString::FromAscii(const char *ascii)
{
    return FromAscii(ascii, wxStrlen(ascii));
}

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

    unsigned char c = (unsigned char)ascii;

    wxASSERT_MSG( c < 0x80, _T("Non-ASCII value passed to FromAscii().") );

    // NB: the cast to wchar_t causes interpretation of 'ascii' as Latin1 value
    return wxString(wxUniChar((wchar_t)c));
}

const wxCharBuffer wxString::ToAscii() const
{
    // this will allocate enough space for the terminating NUL too
    wxCharBuffer buffer(length());
    char *dest = buffer.data();

    for ( const_iterator i = begin(); i != end(); ++i )
    {
        wxUniChar c(*i);
        // FIXME-UTF8: unify substituted char ('_') with wxUniChar ('?')
        *dest++ = c.IsAscii() ? (char)c : '_';

        // 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 ( !c )
            break;
    }

    return buffer;
}

#endif // wxUSE_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 wxString& prefix, wxString *rest) const
{
    if ( compare(0, prefix.length(), prefix) != 0 )
        return false;

    if ( rest )
    {
        // put the rest of the string into provided pointer
        rest->assign(*this, prefix.length(), npos);
    }

    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 wxString& suffix, wxString *rest) const
{
    int start = length() - suffix.length();

    if ( start < 0 || compare(start, npos, 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(wxUniChar ch) const
{
  wxString str;
  int iPos = Find(ch, true);
  if ( iPos == wxNOT_FOUND )
    str = *this;
  else
    str = wx_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(wxUniChar 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(wxUniChar 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(wxUniChar ch) const
{
  wxString str;
  int iPos = Find(ch);
  if ( iPos != wxNOT_FOUND )
    str = wx_str() + iPos + 1;

  return str;
}

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

    size_t uiCount = 0;   // count of replacements made

    size_t uiOldLen = strOld.length();
    size_t uiNewLen = strNew.length();

    size_t dwPos = 0;

    while ( (*this)[dwPos] != wxT('\0') )
    {
        //DO NOT USE STRSTR HERE
        //this string can contain embedded null characters,
        //so strstr will function incorrectly
        dwPos = find(strOld, dwPos);
        if ( dwPos == npos )
            break;                  // exit the loop
        else
        {
            //replace this occurance of the old string with the new one
            replace(dwPos, uiOldLen, strNew, 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
{
    for ( const_iterator i = begin(); i != end(); ++i )
    {
        if ( !(*i).IsAscii() )
            return false;
    }

    return true;
}

bool wxString::IsWord() const
{
    for ( const_iterator i = begin(); i != end(); ++i )
    {
        if ( !wxIsalpha(*i) )
            return false;
    }

    return true;
}

bool wxString::IsNumber() const
{
    if ( empty() )
        return true;

    const_iterator i = begin();

    if ( *i == _T('-') || *i == _T('+') )
        ++i;

    for ( ; i != end(); ++i )
    {
        if ( !wxIsdigit(*i) )
            return false;
    }

    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