X-Git-Url: https://git.saurik.com/wxWidgets.git/blobdiff_plain/38444daf663a7e4dfdc4f7ba8387ee5df16789f0..eea4d01c65f9b29baa1193db762b4c6b8144af24:/src/common/string.cpp diff --git a/src/common/string.cpp b/src/common/string.cpp index 2d371b9fa4..2978dcdca8 100644 --- a/src/common/string.cpp +++ b/src/common/string.cpp @@ -1,25 +1,15 @@ ///////////////////////////////////////////////////////////////////////////// -// Name: string.cpp +// Name: src/common/string.cpp // Purpose: wxString class -// Author: Vadim Zeitlin +// Author: Vadim Zeitlin, Ryan Norton // Modified by: // Created: 29/01/98 // RCS-ID: $Id$ // Copyright: (c) 1998 Vadim Zeitlin -// Licence: wxWindows license +// (c) 2004 Ryan Norton +// Licence: wxWindows licence ///////////////////////////////////////////////////////////////////////////// -#ifdef __GNUG__ - #pragma implementation "string.h" -#endif - -/* - * 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 // =========================================================================== @@ -28,530 +18,585 @@ #include "wx/wxprec.h" #ifdef __BORLANDC__ - #pragma hdrstop + #pragma hdrstop #endif #ifndef WX_PRECOMP - #include "wx/defs.h" - #include "wx/string.h" - #include "wx/intl.h" -#if wxUSE_THREADS - #include -#endif + #include "wx/string.h" + #include "wx/wxcrtvararg.h" + #include "wx/intl.h" + #include "wx/log.h" #endif #include + +#ifndef __WXWINCE__ + #include +#endif + #include #include -#ifdef __SALFORDC__ - #include +#include "wx/hashmap.h" +#include "wx/vector.h" +#include "wx/xlocale.h" + +#ifdef __WXMSW__ + #include "wx/msw/wrapwin.h" +#endif // __WXMSW__ + +#if wxUSE_STD_IOSTREAM + #include #endif -#if wxUSE_WCSRTOMBS - #include // for wcsrtombs(), see comments where it's used -#endif // GNU +// 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 + +// define a function declared in wx/buffer.h here as we don't have buffer.cpp +// and don't want to add it just because of this simple function +namespace wxPrivate +{ + +// wxXXXBuffer classes can be (implicitly) used during global statics +// initialization so wrap the status UntypedBufferData variable in a function +// to make it safe to access it even before all global statics are initialized +UntypedBufferData *GetUntypedNullData() +{ + static UntypedBufferData s_untypedNullData(NULL, 0); -#ifdef WXSTRING_IS_WXOBJECT - IMPLEMENT_DYNAMIC_CLASS(wxString, wxObject) -#endif //WXSTRING_IS_WXOBJECT + return &s_untypedNullData; +} -// 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) +} // namespace wxPrivate // --------------------------------------------------------------------------- // static class variables definition // --------------------------------------------------------------------------- -#ifdef wxSTD_STRING_COMPATIBILITY - const size_t wxString::npos = wxSTRING_MAXLEN; -#endif // wxSTD_STRING_COMPATIBILITY +//According to STL _must_ be a -1 size_t +const size_t wxString::npos = (size_t) -1; -// ---------------------------------------------------------------------------- -// static data -// ---------------------------------------------------------------------------- +#if wxUSE_STRING_POS_CACHE + +#ifdef wxHAS_COMPILER_TLS + +wxTLS_TYPE(wxString::Cache) wxString::ms_cache; + +#else // !wxHAS_COMPILER_TLS -// 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 +struct wxStrCacheInitializer { - wxStringData data; - wxChar dummy; -} g_strEmpty = { {-1, 0, 0}, _T('\0') }; + wxStrCacheInitializer() + { + // calling this function triggers s_cache initialization in it, and + // from now on it becomes safe to call from multiple threads + wxString::GetCache(); + } +}; -// empty C style string: points to 'string data' byte of g_strEmpty -extern const wxChar WXDLLEXPORT *g_szNul = &g_strEmpty.dummy; +/* +wxString::Cache& wxString::GetCache() +{ + static wxTLS_TYPE(Cache) s_cache; -// ---------------------------------------------------------------------------- -// conditional compilation -// ---------------------------------------------------------------------------- + return wxTLS_VALUE(s_cache); +} +*/ -// we want to find out if the current platform supports vsnprintf()-like -// function: for Unix this is done with configure, for Windows we test the -// compiler explicitly. -#ifdef __WXMSW__ - #ifdef __VISUALC__ - #define wxVsnprintf _vsnprintf - #endif -#else // !Windows - #ifdef HAVE_VSNPRINTF - #define wxVsnprintf vsnprintf - #endif -#endif // Windows/!Windows - -#ifndef wxVsnprintf - // in this case we'll use vsprintf() (which is ANSI and thus should be - // always available), but it's unsafe because it doesn't check for buffer - // size - so give a warning - #define wxVsnprintf(buffer,len,format,argptr) vsprintf(buffer,format, argptr) - - #if defined(__VISUALC__) - #pragma message("Using sprintf() because no snprintf()-like function defined") - #elif defined(__GNUG__) && !defined(__UNIX__) - #warning "Using sprintf() because no snprintf()-like function defined" - #elif defined(__MWERKS__) - #warning "Using sprintf() because no snprintf()-like function defined" - #endif //compiler -#endif // no vsnprintf - -#ifdef _AIX - // AIX has vsnprintf, but there's no prototype in the system headers. - extern "C" int vsnprintf(char* str, size_t n, const char* format, va_list ap); -#endif +static wxStrCacheInitializer gs_stringCacheInit; -// ---------------------------------------------------------------------------- -// global functions -// ---------------------------------------------------------------------------- +#endif // wxHAS_COMPILER_TLS/!wxHAS_COMPILER_TLS -#ifdef wxSTD_STRING_COMPATIBILITY +// gdb seems to be unable to display thread-local variables correctly, at least +// not my 6.4.98 version under amd64, so provide this debugging helper to do it +#if wxDEBUG_LEVEL >= 2 -// 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! +struct wxStrCacheDumper +{ + static void ShowAll() + { + puts("*** wxString cache dump:"); + for ( unsigned n = 0; n < wxString::Cache::SIZE; n++ ) + { + const wxString::Cache::Element& + c = wxString::GetCacheBegin()[n]; + + printf("\t%u%s\t%p: pos=(%lu, %lu), len=%ld\n", + n, + n == wxString::LastUsedCacheElement() ? " [*]" : "", + c.str, + (unsigned long)c.pos, + (unsigned long)c.impl, + (long)c.len); + } + } +}; + +void wxDumpStrCache() { wxStrCacheDumper::ShowAll(); } + +#endif // wxDEBUG_LEVEL >= 2 -istream& operator>>(istream& is, wxString& WXUNUSED(str)) +#ifdef wxPROFILE_STRING_CACHE + +wxString::CacheStats wxString::ms_cacheStats; + +struct wxStrCacheStatsDumper { -#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; - } + ~wxStrCacheStatsDumper() + { + const wxString::CacheStats& stats = wxString::ms_cacheStats; - str += ch; - if ( --w == 1 ) - break; + if ( stats.postot ) + { + puts("*** wxString cache statistics:"); + printf("\tTotal non-trivial calls to PosToImpl(): %u\n", + stats.postot); + printf("\tHits %u (of which %u not used) or %.2f%%\n", + stats.poshits, + stats.mishits, + 100.*float(stats.poshits - stats.mishits)/stats.postot); + printf("\tAverage position requested: %.2f\n", + float(stats.sumpos) / stats.postot); + printf("\tAverage offset after cached hint: %.2f\n", + float(stats.sumofs) / stats.postot); + } + + if ( stats.lentot ) + { + printf("\tNumber of calls to length(): %u, hits=%.2f%%\n", + stats.lentot, 100.*float(stats.lenhits)/stats.lentot); + } } - } +}; - is.isfx(); - if ( str.length() == 0 ) - is.setstate(ios::failbit); -#endif - return is; -} +static wxStrCacheStatsDumper s_showCacheStats; -#endif //std::string compatibility +#endif // wxPROFILE_STRING_CACHE + +#endif // wxUSE_STRING_POS_CACHE // ---------------------------------------------------------------------------- -// private classes +// global functions // ---------------------------------------------------------------------------- -// this small class is used to gather statistics for performance tuning -//#define WXSTRING_STATISTICS -#ifdef WXSTRING_STATISTICS - class Averager - { - public: - Averager(const char *sz) { m_sz = sz; m_nTotal = m_nCount = 0; } - ~Averager() - { printf("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 char *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 - -// =========================================================================== -// wxString class core -// =========================================================================== +#if wxUSE_STD_IOSTREAM -// --------------------------------------------------------------------------- -// construction -// --------------------------------------------------------------------------- +#include -// constructs string of copies of character -wxString::wxString(wxChar ch, size_t nLength) +wxSTD ostream& operator<<(wxSTD ostream& os, const wxCStrData& str) { - Init(); - - if ( nLength > 0 ) { - AllocBuffer(nLength); +#if wxUSE_UNICODE && !wxUSE_UNICODE_UTF8 + const wxScopedCharBuffer buf(str.AsCharBuf()); + if ( !buf ) + os.clear(wxSTD ios_base::failbit); + else + os << buf.data(); -#if wxUSE_UNICODE - // memset only works on char - for (size_t n=0; n 0 ) { - // trailing '\0' is written in AllocBuffer() - AllocBuffer(nLength); - memcpy(m_pchData, psz + nPos, nLength*sizeof(wxChar)); - } +wxSTD wostream& operator<<(wxSTD wostream& wos, const wxString& str) +{ + return wos << str.wc_str(); } -#ifdef wxSTD_STRING_COMPATIBILITY +wxSTD wostream& operator<<(wxSTD wostream& wos, const wxCStrData& str) +{ + return wos << str.AsWChar(); +} -// poor man's iterators are "void *" pointers -wxString::wxString(const void *pStart, const void *pEnd) +wxSTD wostream& operator<<(wxSTD wostream& wos, const wxScopedWCharBuffer& str) { - InitWith((const wxChar *)pStart, 0, - (const wxChar *)pEnd - (const wxChar *)pStart); + return wos << str.data(); } -#endif //std::string compatibility +#endif // wxUSE_UNICODE && defined(HAVE_WOSTREAM) -#if wxUSE_UNICODE +#endif // wxUSE_STD_IOSTREAM + +// =========================================================================== +// wxString class core +// =========================================================================== + +#if wxUSE_UNICODE_UTF8 -// from multibyte string -wxString::wxString(const char *psz, wxMBConv& conv, size_t nLength) +void wxString::PosLenToImpl(size_t pos, size_t len, + size_t *implPos, size_t *implLen) const { - // first get necessary size - size_t nLen = psz ? conv.MB2WC((wchar_t *) NULL, psz, 0) : 0; + if ( pos == npos ) + { + *implPos = npos; + } + else // have valid start position + { + const const_iterator b = GetIterForNthChar(pos); + *implPos = wxStringImpl::const_iterator(b.impl()) - m_impl.begin(); + if ( len == npos ) + { + *implLen = npos; + } + else // have valid length too + { + // we need to handle the case of length specifying a substring + // going beyond the end of the string, just as std::string does + const const_iterator e(end()); + const_iterator i(b); + while ( len && i <= e ) + { + ++i; + --len; + } + + *implLen = i.impl() - b.impl(); + } + } +} - // nLength is number of *Unicode* characters here! - if (nLen > nLength) - nLen = nLength; +#endif // wxUSE_UNICODE_UTF8 - // empty? - if ( nLen != 0 ) { - AllocBuffer(nLen); - conv.MB2WC(m_pchData, psz, nLen); - } - else { - Init(); - } +// ---------------------------------------------------------------------------- +// 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 +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); + } } -#else +#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; -// from wide string -wxString::wxString(const wchar_t *pwz) +const char* wxCStrData::AsChar() const { - // first get necessary size - size_t nLen = pwz ? wxWC2MB((char *) NULL, pwz, 0) : 0; + // remove previously cache value, if any (see FIXMEs above): + DeleteStringFromConversionCache(gs_stringsCharCache, m_str); - // empty? - if ( nLen != 0 ) { - AllocBuffer(nLen); - wxWC2MB(m_pchData, pwz, nLen); - } - else { - Init(); - } -} + // convert the string and keep it: + const char *s = gs_stringsCharCache[wxConstCast(m_str, wxString)] = + m_str->mb_str().release(); -#endif + return s + m_offset; +} +#endif // wxUSE_UNICODE -// --------------------------------------------------------------------------- -// memory allocation -// --------------------------------------------------------------------------- +#if !wxUSE_UNICODE_WCHAR +WX_DECLARE_HASH_MAP(wxString*, wchar_t*, wxPointerHash, wxPointerEqual, + wxStringWCharConversionCache); +static wxStringWCharConversionCache gs_stringsWCharCache; -// allocates memory needed to store a C string of length nLen -void wxString::AllocBuffer(size_t nLen) +const wchar_t* wxCStrData::AsWChar() const { - wxASSERT( nLen > 0 ); // - wxASSERT( nLen <= INT_MAX-1 ); // max size (enough room for 1 extra) + // remove previously cache value, if any (see FIXMEs above): + DeleteStringFromConversionCache(gs_stringsWCharCache, m_str); - STATISTICS_ADD(Length, nLen); + // convert the string and keep it: + const wchar_t *s = gs_stringsWCharCache[wxConstCast(m_str, wxString)] = + m_str->wc_str().release(); - // 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)); - pData->nRefs = 1; - pData->nDataLength = nLen; - pData->nAllocLength = nLen + EXTRA_ALLOC; - m_pchData = pData->data(); // data starts after wxStringData - m_pchData[nLen] = _T('\0'); + return s + m_offset; } +#endif // !wxUSE_UNICODE_WCHAR -// must be called before changing this string -void wxString::CopyBeforeWrite() +wxString::~wxString() { - wxStringData* pData = GetStringData(); +#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 ( pData->IsShared() ) { - pData->Unlock(); // memory not freed because shared - size_t nLen = pData->nDataLength; - AllocBuffer(nLen); - memcpy(m_pchData, pData->data(), nLen*sizeof(wxChar)); - } +// =========================================================================== +// wxString class core +// =========================================================================== - wxASSERT( !GetStringData()->IsShared() ); // we must be the only owner -} +// --------------------------------------------------------------------------- +// construction and conversion +// --------------------------------------------------------------------------- -// must be called before replacing contents of this string -void wxString::AllocBeforeWrite(size_t nLen) +#if wxUSE_UNICODE_WCHAR +/* static */ +wxString::SubstrBufFromMB wxString::ConvertStr(const char *psz, size_t nLength, + const wxMBConv& conv) { - wxASSERT( nLen != 0 ); // doesn't make any sense + // anything to do? + if ( !psz || nLength == 0 ) + return SubstrBufFromMB(wxWCharBuffer(L""), 0); - // must not share string and must have enough space - wxStringData* pData = GetStringData(); - if ( pData->IsShared() || (nLen > pData->nAllocLength) ) { - // can't work with old buffer, get new one - pData->Unlock(); - AllocBuffer(nLen); - } - else { - // update the string length - pData->nDataLength = nLen; - } + if ( nLength == npos ) + nLength = wxNO_LEN; - wxASSERT( !GetStringData()->IsShared() ); // we must be the only owner + size_t wcLen; + wxScopedWCharBuffer wcBuf(conv.cMB2WC(psz, nLength, &wcLen)); + if ( !wcLen ) + return SubstrBufFromMB(wxWCharBuffer(L""), 0); + else + return SubstrBufFromMB(wcBuf, wcLen); } +#endif // wxUSE_UNICODE_WCHAR -// allocate enough memory for nLen characters -void wxString::Alloc(size_t nLen) +#if wxUSE_UNICODE_UTF8 +/* static */ +wxString::SubstrBufFromMB wxString::ConvertStr(const char *psz, size_t nLength, + const wxMBConv& conv) { - wxStringData *pData = GetStringData(); - if ( pData->nAllocLength <= nLen ) { - if ( pData->IsEmpty() ) { - nLen += EXTRA_ALLOC; + // anything to do? + if ( !psz || nLength == 0 ) + return SubstrBufFromMB(wxCharBuffer(""), 0); - wxStringData* pData = (wxStringData*) - malloc(sizeof(wxStringData) + (nLen + 1)*sizeof(wxChar)); - pData->nRefs = 1; - pData->nDataLength = 0; - pData->nAllocLength = nLen; - m_pchData = pData->data(); // data starts after wxStringData - m_pchData[0u] = _T('\0'); - } - else if ( pData->IsShared() ) { - pData->Unlock(); // memory not freed because shared - size_t nOldLen = pData->nDataLength; - AllocBuffer(nLen); - memcpy(m_pchData, pData->data(), nOldLen*sizeof(wxChar)); + // 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) ) + { + // we must pass the real string length to SubstrBufFromMB ctor + if ( nLength == npos ) + nLength = psz ? strlen(psz) : 0; + return SubstrBufFromMB(wxScopedCharBuffer::CreateNonOwned(psz, nLength), + nLength); + } + // else: do the roundtrip through wchar_t* } - else { - nLen += EXTRA_ALLOC; - wxStringData *p = (wxStringData *) - realloc(pData, sizeof(wxStringData) + (nLen + 1)*sizeof(wxChar)); + if ( nLength == npos ) + nLength = wxNO_LEN; - if ( p == NULL ) { - // @@@ what to do on memory error? - return; - } + // first convert to wide string: + size_t wcLen; + wxScopedWCharBuffer wcBuf(conv.cMB2WC(psz, nLength, &wcLen)); + if ( !wcLen ) + return SubstrBufFromMB(wxCharBuffer(""), 0); - // 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) - p->nAllocLength = nLen; - m_pchData = p->data(); - } - } - //else: we've already got enough + // 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, wxT("conversion to UTF-8 failed") ); + + return buf; } +#endif // wxUSE_UNICODE_UTF8 -// shrink to minimal size (releasing extra memory) -void wxString::Shrink() +#if wxUSE_UNICODE_UTF8 || !wxUSE_UNICODE +/* static */ +wxString::SubstrBufFromWC wxString::ConvertStr(const wchar_t *pwz, size_t nLength, + const wxMBConv& conv) { - wxStringData *pData = GetStringData(); + // anything to do? + if ( !pwz || nLength == 0 ) + return SubstrBufFromWC(wxCharBuffer(""), 0); - // this variable is unused in release build, so avoid the compiler warning by - // just not declaring it -#ifdef __WXDEBUG__ - void *p = -#endif - realloc(pData, sizeof(wxStringData) + (pData->nDataLength + 1)*sizeof(wxChar)); + if ( nLength == npos ) + nLength = wxNO_LEN; - wxASSERT( p != NULL ); // can't free memory? - wxASSERT( p == pData ); // we're decrementing the size - block shouldn't move! + size_t mbLen; + wxScopedCharBuffer mbBuf(conv.cWC2MB(pwz, nLength, &mbLen)); + if ( !mbLen ) + return SubstrBufFromWC(wxCharBuffer(""), 0); + else + return SubstrBufFromWC(mbBuf, mbLen); } +#endif // wxUSE_UNICODE_UTF8 || !wxUSE_UNICODE -// get the pointer to writable buffer of (at least) nLen bytes -wxChar *wxString::GetWriteBuf(size_t nLen) -{ - AllocBeforeWrite(nLen); - - wxASSERT( GetStringData()->nRefs == 1 ); - GetStringData()->Validate(FALSE); +// This std::string::c_str()-like method returns a wide char pointer to string +// contents. In wxUSE_UNICODE_WCHAR case it is trivial as it can simply return +// a pointer to the internal representation. Otherwise a conversion is required +// and it returns a temporary buffer. +// +// However for compatibility with c_str() and to avoid breaking existing code +// doing +// +// for ( const wchar_t *p = s.wc_str(); *p; p++ ) +// ... use *p... +// +// we actually need to ensure that the returned buffer is _not_ temporary and +// so we use wxString::m_convertedToWChar to store the returned data +#if !wxUSE_UNICODE_WCHAR + +const wchar_t *wxString::AsWChar(const wxMBConv& conv) const +{ + const char * const strMB = m_impl.c_str(); + const size_t lenMB = m_impl.length(); + + // find out the size of the buffer needed + const size_t lenWC = conv.ToWChar(NULL, 0, strMB, lenMB); + if ( lenWC == wxCONV_FAILED ) + return NULL; + + // keep the same buffer if the string size didn't change: this is not only + // an optimization but also ensure that code which modifies string + // character by character (without changing its length) can continue to use + // the pointer returned by a previous wc_str() call even after changing the + // string + + // TODO-UTF8: we could check for ">" instead of "!=" here as this would + // allow to save on buffer reallocations but at the cost of + // consuming (even) more memory, we should benchmark this to + // determine if it's worth doing + if ( !m_convertedToWChar.m_str || lenWC != m_convertedToWChar.m_len ) + { + if ( !const_cast(this)->m_convertedToWChar.Extend(lenWC) ) + return NULL; + } - return m_pchData; -} + // finally do convert + m_convertedToWChar.m_str[lenWC] = L'\0'; + if ( conv.ToWChar(m_convertedToWChar.m_str, lenWC, + strMB, lenMB) == wxCONV_FAILED ) + return NULL; -// put string back in a reasonable state after GetWriteBuf -void wxString::UngetWriteBuf() -{ - GetStringData()->nDataLength = wxStrlen(m_pchData); - GetStringData()->Validate(TRUE); + return m_convertedToWChar.m_str; } -// --------------------------------------------------------------------------- -// data access -// --------------------------------------------------------------------------- +#endif // !wxUSE_UNICODE_WCHAR -// all functions are inline in string.h -// --------------------------------------------------------------------------- -// assignment operators -// --------------------------------------------------------------------------- +// Same thing for mb_str() which returns a normal char pointer to string +// contents: this always requires converting it to the specified encoding in +// non-ANSI build except if we need to convert to UTF-8 and this is what we +// already use internally. +#if wxUSE_UNICODE -// helper function: does real copy -void wxString::AssignCopy(size_t nSrcLen, const wxChar *pszSrcData) +const char *wxString::AsChar(const wxMBConv& conv) const { - if ( nSrcLen == 0 ) { - Reinit(); - } - else { - AllocBeforeWrite(nSrcLen); - memcpy(m_pchData, pszSrcData, nSrcLen*sizeof(wxChar)); - GetStringData()->nDataLength = nSrcLen; - m_pchData[nSrcLen] = _T('\0'); - } -} +#if wxUSE_UNICODE_UTF8 + if ( conv.IsUTF8() ) + return m_impl.c_str(); -// assigns one string to another -wxString& wxString::operator=(const wxString& stringSrc) -{ - wxASSERT( stringSrc.GetStringData()->IsValid() ); + const wchar_t * const strWC = AsWChar(wxMBConvStrictUTF8()); + const size_t lenWC = m_convertedToWChar.m_len; +#else // wxUSE_UNICODE_WCHAR + const wchar_t * const strWC = m_impl.c_str(); + const size_t lenWC = m_impl.length(); +#endif // wxUSE_UNICODE_UTF8/wxUSE_UNICODE_WCHAR - // 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(); + const size_t lenMB = conv.FromWChar(NULL, 0, strWC, lenWC); + if ( lenMB == wxCONV_FAILED ) + return NULL; + + if ( !m_convertedToChar.m_str || lenMB != m_convertedToChar.m_len ) + { + if ( !const_cast(this)->m_convertedToChar.Extend(lenMB) ) + return NULL; } - } - return *this; + m_convertedToChar.m_str[lenMB] = '\0'; + if ( conv.FromWChar(m_convertedToChar.m_str, lenMB, + strWC, lenWC) == wxCONV_FAILED ) + return NULL; + + return m_convertedToChar.m_str; } -// assigns a single character -wxString& wxString::operator=(wxChar ch) +#endif // wxUSE_UNICODE + +// shrink to minimal size (releasing extra memory) +bool wxString::Shrink() { - AssignCopy(1, &ch); - return *this; + wxString tmp(begin(), end()); + swap(tmp); + return tmp.length() == length(); } -// assigns C string -wxString& wxString::operator=(const wxChar *psz) +// deprecated compatibility code: +#if WXWIN_COMPATIBILITY_2_8 && !wxUSE_STL_BASED_WXSTRING && !wxUSE_UNICODE_UTF8 +wxStringCharType *wxString::GetWriteBuf(size_t nLen) { - AssignCopy(wxStrlen(psz), psz); - return *this; + return DoGetWriteBuf(nLen); } -#if !wxUSE_UNICODE - -// same as 'signed char' variant -wxString& wxString::operator=(const unsigned char* psz) +void wxString::UngetWriteBuf() { - *this = (const char *)psz; - return *this; + DoUngetWriteBuf(); } -wxString& wxString::operator=(const wchar_t *pwz) +void wxString::UngetWriteBuf(size_t nLen) { - wxString str(pwz); - *this = str; - return *this; + DoUngetWriteBuf(nLen); } +#endif // WXWIN_COMPATIBILITY_2_8 && !wxUSE_STL_BASED_WXSTRING && !wxUSE_UNICODE_UTF8 -#endif // --------------------------------------------------------------------------- -// string concatenation +// data access // --------------------------------------------------------------------------- -// add something to this string -void wxString::ConcatSelf(int nSrcLen, const wxChar *pszSrcData) -{ - STATISTICS_ADD(SummandLength, nSrcLen); - - // 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(); - AllocBuffer(nNewLen); - memcpy(m_pchData, pOldData->data(), nLen*sizeof(wxChar)); - pOldData->Unlock(); - } - else if ( nNewLen > pData->nAllocLength ) { - STATISTICS_ADD(ConcatHit, 0); - - // we have to grow the buffer - Alloc(nNewLen); - } - 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)); +// all functions are inline in string.h - m_pchData[nNewLen] = _T('\0'); // put terminating '\0' - GetStringData()->nDataLength = nNewLen; // and fix the length - } - //else: the string to append was empty -} +// --------------------------------------------------------------------------- +// concatenation operators +// --------------------------------------------------------------------------- /* * concatenation functions come in 5 flavours: @@ -560,1563 +605,1728 @@ void wxString::ConcatSelf(int nSrcLen, const wxChar *pszSrcData) * C str + string and string + C str */ -wxString operator+(const wxString& string1, const wxString& string2) +wxString operator+(const wxString& str1, const wxString& str2) { - wxASSERT( string1.GetStringData()->IsValid() ); - wxASSERT( string2.GetStringData()->IsValid() ); +#if !wxUSE_STL_BASED_WXSTRING + wxASSERT( str1.IsValid() ); + wxASSERT( str2.IsValid() ); +#endif - wxString s = string1; - s += string2; + wxString s = str1; + s += str2; - return s; + return s; } -wxString operator+(const wxString& string, wxChar ch) +wxString operator+(const wxString& str, wxUniChar ch) { - wxASSERT( string.GetStringData()->IsValid() ); +#if !wxUSE_STL_BASED_WXSTRING + wxASSERT( str.IsValid() ); +#endif - wxString s = string; - s += ch; + wxString s = str; + s += ch; - return s; + return s; } -wxString operator+(wxChar ch, const wxString& string) +wxString operator+(wxUniChar ch, const wxString& str) { - wxASSERT( string.GetStringData()->IsValid() ); +#if !wxUSE_STL_BASED_WXSTRING + wxASSERT( str.IsValid() ); +#endif - wxString s = ch; - s += string; + wxString s = ch; + s += str; - return s; + return s; } -wxString operator+(const wxString& string, const wxChar *psz) +wxString operator+(const wxString& str, const char *psz) { - wxASSERT( string.GetStringData()->IsValid() ); +#if !wxUSE_STL_BASED_WXSTRING + wxASSERT( str.IsValid() ); +#endif - wxString s; - s.Alloc(wxStrlen(psz) + string.Len()); - s = string; - s += psz; + wxString s; + if ( !s.Alloc(strlen(psz) + str.length()) ) { + wxFAIL_MSG( wxT("out of memory in wxString::operator+") ); + } + s += str; + s += psz; - return s; + return s; } -wxString operator+(const wxChar *psz, const wxString& string) +wxString operator+(const wxString& str, const wchar_t *pwz) { - wxASSERT( string.GetStringData()->IsValid() ); +#if !wxUSE_STL_BASED_WXSTRING + wxASSERT( str.IsValid() ); +#endif - wxString s; - s.Alloc(wxStrlen(psz) + string.Len()); - s = psz; - s += string; + wxString s; + if ( !s.Alloc(wxWcslen(pwz) + str.length()) ) { + wxFAIL_MSG( wxT("out of memory in wxString::operator+") ); + } + s += str; + s += pwz; - return s; + return s; } -// =========================================================================== -// other common string functions -// =========================================================================== +wxString operator+(const char *psz, const wxString& str) +{ +#if !wxUSE_STL_BASED_WXSTRING + wxASSERT( str.IsValid() ); +#endif -// --------------------------------------------------------------------------- -// simple sub-string extraction -// --------------------------------------------------------------------------- + wxString s; + if ( !s.Alloc(strlen(psz) + str.length()) ) { + wxFAIL_MSG( wxT("out of memory in wxString::operator+") ); + } + s = psz; + s += str; -// helper function: clone the data attached to this string -void wxString::AllocCopy(wxString& dest, int nCopyLen, int nCopyIndex) const -{ - if ( nCopyLen == 0 ) { - dest.Init(); - } - else { - dest.AllocBuffer(nCopyLen); - memcpy(dest.m_pchData, m_pchData + nCopyIndex, nCopyLen*sizeof(wxChar)); - } + return s; } -// extract string of length nCount starting at nFirst -wxString wxString::Mid(size_t nFirst, size_t nCount) const +wxString operator+(const wchar_t *pwz, const wxString& str) { - wxStringData *pData = GetStringData(); - size_t nLen = pData->nDataLength; - - // default value of nCount is wxSTRING_MAXLEN and means "till the end" - if ( nCount == wxSTRING_MAXLEN ) - { - nCount = nLen - nFirst; - } +#if !wxUSE_STL_BASED_WXSTRING + wxASSERT( str.IsValid() ); +#endif - // out-of-bounds requests return sensible things - if ( nFirst + nCount > nLen ) - { - nCount = nLen - nFirst; - } + wxString s; + if ( !s.Alloc(wxWcslen(pwz) + str.length()) ) { + wxFAIL_MSG( wxT("out of memory in wxString::operator+") ); + } + s = pwz; + s += str; - if ( nFirst > nLen ) - { - // AllocCopy() will return empty string - nCount = 0; - } + return s; +} - wxString dest; - AllocCopy(dest, nCount, nFirst); +// --------------------------------------------------------------------------- +// string comparison +// --------------------------------------------------------------------------- - return dest; +bool wxString::IsSameAs(wxUniChar c, bool compareWithCase) const +{ + return (length() == 1) && (compareWithCase ? GetChar(0u) == c + : wxToupper(GetChar(0u)) == wxToupper(c)); } -// extract nCount last (rightmost) characters -wxString wxString::Right(size_t nCount) const -{ - if ( nCount > (size_t)GetStringData()->nDataLength ) - nCount = GetStringData()->nDataLength; +#ifdef HAVE_STD_STRING_COMPARE - wxString dest; - AllocCopy(dest, nCount, GetStringData()->nDataLength - nCount); - return dest; -} +// 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). -// get all characters after the last occurence of ch -// (returns the whole string if ch not found) -wxString wxString::AfterLast(wxChar ch) const +int wxString::compare(const wxString& str) const { - wxString str; - int iPos = Find(ch, TRUE); - if ( iPos == wxNOT_FOUND ) - str = *this; - else - str = c_str() + iPos + 1; + return m_impl.compare(str.m_impl); +} - return str; +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); } -// extract nCount first (leftmost) characters -wxString wxString::Left(size_t nCount) const +int wxString::compare(size_t nStart, size_t nLen, + const wxString& str, + size_t nStart2, size_t nLen2) const { - if ( nCount > (size_t)GetStringData()->nDataLength ) - nCount = GetStringData()->nDataLength; + size_t pos, len; + PosLenToImpl(nStart, nLen, &pos, &len); - wxString dest; - AllocCopy(dest, nCount, 0); - return dest; + size_t pos2, len2; + str.PosLenToImpl(nStart2, nLen2, &pos2, &len2); + + return m_impl.compare(pos, len, str.m_impl, pos2, len2); } -// get all characters before the first occurence of ch -// (returns the whole string if ch not found) -wxString wxString::BeforeFirst(wxChar ch) const +int wxString::compare(const char* sz) const { - wxString str; - for ( const wxChar *pc = m_pchData; *pc != _T('\0') && *pc != ch; pc++ ) - str += *pc; - - return str; + return m_impl.compare(ImplStr(sz)); } -/// get all characters before the last occurence of ch -/// (returns empty string if ch not found) -wxString wxString::BeforeLast(wxChar ch) const +int wxString::compare(const wchar_t* sz) const { - wxString str; - int iPos = Find(ch, TRUE); - if ( iPos != wxNOT_FOUND && iPos != 0 ) - str = wxString(c_str(), iPos); - - return str; + return m_impl.compare(ImplStr(sz)); } -/// get all characters after the first occurence of ch -/// (returns empty string if ch not found) -wxString wxString::AfterFirst(wxChar ch) const +int wxString::compare(size_t nStart, size_t nLen, + const char* sz, size_t nCount) const { - wxString str; - int iPos = Find(ch); - if ( iPos != wxNOT_FOUND ) - str = c_str() + iPos + 1; + size_t pos, len; + PosLenToImpl(nStart, nLen, &pos, &len); - return str; + SubstrBufFromMB str(ImplStr(sz, nCount)); + + return m_impl.compare(pos, len, str.data, str.len); } -// replace first (or all) occurences of some substring with another one -size_t wxString::Replace(const wxChar *szOld, const wxChar *szNew, bool bReplaceAll) +int wxString::compare(size_t nStart, size_t nLen, + const wchar_t* sz, size_t nCount) const { - size_t uiCount = 0; // count of replacements made + size_t pos, len; + PosLenToImpl(nStart, nLen, &pos, &len); - size_t uiOldLen = wxStrlen(szOld); + SubstrBufFromWC str(ImplStr(sz, nCount)); - wxString strTemp; - const wxChar *pCurrent = m_pchData; - const wxChar *pSubstr; - while ( *pCurrent != _T('\0') ) { - pSubstr = wxStrstr(pCurrent, szOld); - if ( pSubstr == NULL ) { - // strTemp is unused if no replacements were made, so avoid the copy - if ( uiCount == 0 ) - return 0; + return m_impl.compare(pos, len, str.data, str.len); +} - strTemp += pCurrent; // copy the rest - break; // exit the loop +#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 { - // take chars before match - strTemp.ConcatSelf(pSubstr - pCurrent, pCurrent); - strTemp += szNew; - pCurrent = pSubstr + uiOldLen; // restart after match - - uiCount++; - - // stop now? - if ( !bReplaceAll ) { - strTemp += pCurrent; // copy the rest - break; // exit the loop - } + 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()); +} - // only done if there were replacements, otherwise would have returned above - *this = strTemp; +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; - return uiCount; + size_t pos, len; + PosLenToImpl(nStart, nLen, &pos, &len); + + return ::wxDoCmp(m_impl.data() + pos, len, + str.m_impl.data(), str.m_impl.length()); } -bool wxString::IsAscii() const +int wxString::compare(size_t nStart, size_t nLen, + const wxString& str, + size_t nStart2, size_t nLen2) const { - const wxChar *s = (const wxChar*) *this; - while(*s){ - if(!isascii(*s)) return(FALSE); - s++; - } - return(TRUE); + 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); } -bool wxString::IsWord() const +int wxString::compare(const char* sz) const { - const wxChar *s = (const wxChar*) *this; - while(*s){ - if(!wxIsalpha(*s)) return(FALSE); - s++; - } - return(TRUE); + 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); } -bool wxString::IsNumber() const +int wxString::compare(const wchar_t* sz) const { - const wxChar *s = (const wxChar*) *this; - while(*s){ - if(!wxIsdigit(*s)) return(FALSE); - s++; - } - return(TRUE); + 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); } -wxString wxString::Strip(stripType w) const +int wxString::compare(size_t nStart, size_t nLen, + const char* sz, size_t nCount) const { - wxString s = *this; - if ( w & leading ) s.Trim(FALSE); - if ( w & trailing ) s.Trim(TRUE); - return s; + 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 + + // --------------------------------------------------------------------------- -// case conversion +// find_{first,last}_[not]_of functions // --------------------------------------------------------------------------- -wxString& wxString::MakeUpper() +#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 { - CopyBeforeWrite(); + return find_first_of(sz, nStart, wxStrlen(sz)); +} - for ( wxChar *p = m_pchData; *p; p++ ) - *p = (wxChar)wxToupper(*p); +size_t wxString::find_first_not_of(const wxChar* sz, size_t nStart) const +{ + return find_first_not_of(sz, nStart, wxStrlen(sz)); +} - return *this; +size_t wxString::find_first_of(const wxChar* sz, size_t nStart, size_t n) const +{ + wxASSERT_MSG( nStart <= length(), wxT("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; } -wxString& wxString::MakeLower() +size_t wxString::find_first_not_of(const wxChar* sz, size_t nStart, size_t n) const { - CopyBeforeWrite(); + wxASSERT_MSG( nStart <= length(), wxT("invalid index") ); - for ( wxChar *p = m_pchData; *p; p++ ) - *p = (wxChar)wxTolower(*p); + size_t idx = nStart; + for ( const_iterator i = begin() + nStart; i != end(); ++idx, ++i ) + { + if ( !wxTmemchr(sz, *i, n) ) + return idx; + } - return *this; + return npos; } -// --------------------------------------------------------------------------- -// trimming and padding -// --------------------------------------------------------------------------- -// trims spaces (in the sense of isspace) from left or right side -wxString& wxString::Trim(bool bFromRight) +size_t wxString::find_last_of(const wxChar* sz, size_t nStart) const { - // first check if we're going to modify the string at all - if ( !IsEmpty() && - ( - (bFromRight && wxIsspace(GetChar(Len() - 1))) || - (!bFromRight && wxIsspace(GetChar(0u))) - ) - ) - { - // ok, there is at least one space to trim - CopyBeforeWrite(); + return find_last_of(sz, nStart, wxStrlen(sz)); +} - if ( bFromRight ) +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 ) { - // find last non-space character - wxChar *psz = m_pchData + GetStringData()->nDataLength - 1; - while ( wxIsspace(*psz) && (psz >= m_pchData) ) - psz--; - - // truncate at trailing space start - *++psz = _T('\0'); - GetStringData()->nDataLength = psz - m_pchData; + nStart = len - 1; } else { - // find first non-space character - const wxChar *psz = m_pchData; - while ( wxIsspace(*psz) ) - psz++; - - // fix up data and length - int nDataLength = GetStringData()->nDataLength - (psz - (const wxChar*) m_pchData); - memmove(m_pchData, psz, (nDataLength + 1)*sizeof(wxChar)); - GetStringData()->nDataLength = nDataLength; + wxASSERT_MSG( nStart <= len, wxT("invalid index") ); } - } - return *this; + 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; } -// adds nCount characters chPad to the string from either side -wxString& wxString::Pad(size_t nCount, wxChar chPad, bool bFromRight) +size_t wxString::find_last_not_of(const wxChar* sz, size_t nStart, size_t n) const { - wxString s(chPad, nCount); + size_t len = length(); - if ( bFromRight ) - *this += s; - else - { - s += *this; - *this = s; - } + if ( nStart == npos ) + { + nStart = len - 1; + } + else + { + wxASSERT_MSG( nStart <= len, wxT("invalid index") ); + } - return *this; + 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; } -// truncate the string -wxString& wxString::Truncate(size_t uiLen) +size_t wxString::find_first_not_of(wxUniChar ch, size_t nStart) const { - if ( uiLen < Len() ) { - CopyBeforeWrite(); + wxASSERT_MSG( nStart <= length(), wxT("invalid index") ); - *(m_pchData + uiLen) = _T('\0'); - GetStringData()->nDataLength = uiLen; - } - //else: nothing to do, string is already short enough + size_t idx = nStart; + for ( const_iterator i = begin() + nStart; i != end(); ++idx, ++i ) + { + if ( *i != ch ) + return idx; + } - return *this; + return npos; } -// --------------------------------------------------------------------------- -// finding (return wxNOT_FOUND if not found and index otherwise) -// --------------------------------------------------------------------------- - -// find a character -int wxString::Find(wxChar ch, bool bFromEnd) const +size_t wxString::find_last_not_of(wxUniChar ch, size_t nStart) const { - const wxChar *psz = bFromEnd ? wxStrrchr(m_pchData, ch) : wxStrchr(m_pchData, ch); + size_t len = length(); - return (psz == NULL) ? wxNOT_FOUND : psz - (const wxChar*) m_pchData; + if ( nStart == npos ) + { + nStart = len - 1; + } + else + { + wxASSERT_MSG( nStart <= len, wxT("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; } -// find a sub-string (like strstr) -int wxString::Find(const wxChar *pszSub) const +// 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 { - const wxChar *psz = wxStrstr(m_pchData, pszSub); +#if !wxUSE_UNICODE_UTF8 + // We compare NUL-delimited chunks of the strings inside the loop. We will + // do as many iterations as there are embedded NULs in the string, i.e. + // usually we will run it just once. + + typedef const wxStringImpl::value_type *pchar_type; + const pchar_type thisBegin = m_impl.c_str(); + const pchar_type thatBegin = s.m_impl.c_str(); + + const pchar_type thisEnd = thisBegin + m_impl.length(); + const pchar_type thatEnd = thatBegin + s.m_impl.length(); + + pchar_type thisCur = thisBegin; + pchar_type thatCur = thatBegin; + + int rc; + for ( ;; ) + { + // Compare until the next NUL, if the strings differ this is the final + // result. + rc = wxStricmp(thisCur, thatCur); + if ( rc ) + break; + + const size_t lenChunk = wxStrlen(thisCur); + thisCur += lenChunk; + thatCur += lenChunk; + + // Skip all the NULs as wxStricmp() doesn't handle them. + for ( ; !*thisCur; thisCur++, thatCur++ ) + { + // Check if we exhausted either of the strings. + if ( thisCur == thisEnd ) + { + // This one is exhausted, is the other one too? + return thatCur == thatEnd ? 0 : -1; + } + + if ( thatCur == thatEnd ) + { + // Because of the test above we know that this one is not + // exhausted yet so it's greater than the other one that is. + return 1; + } + + if ( *thatCur ) + { + // Anything non-NUL is greater than NUL. + return -1; + } + } + } + + return rc; +#else // wxUSE_UNICODE_UTF8 + // CRT functions can't be used for case-insensitive comparison of UTF-8 + // strings so do it in the naive, simple and inefficient way. + + // 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(); - return (psz == NULL) ? wxNOT_FOUND : psz - (const wxChar*) m_pchData; + 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; +#endif // !wxUSE_UNICODE_UTF8/wxUSE_UNICODE_UTF8 } -// --------------------------------------------------------------------------- -// stream-like operators -// --------------------------------------------------------------------------- -wxString& wxString::operator<<(int i) + +#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; - res.Printf(_T("%d"), i); - return (*this) << res; + { + wxStringInternalBuffer buf(res, len); + wxStringCharType *dest = buf; + + for ( ; len > 0; --len ) + { + unsigned char c = (unsigned char)*ascii++; + wxASSERT_MSG( c < 0x80, + wxT("Non-ASCII value passed to FromAscii().") ); + + *dest++ = (wchar_t)c; + } + } + + return res; } -wxString& wxString::operator<<(float f) +wxString wxString::FromAscii(const char *ascii) { - wxString res; - res.Printf(_T("%f"), f); + 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, wxT("Non-ASCII value passed to FromAscii().") ); - return (*this) << res; + // NB: the cast to wchar_t causes interpretation of 'ascii' as Latin1 value + return wxString(wxUniChar((wchar_t)c)); } -wxString& wxString::operator<<(double d) +const wxScopedCharBuffer wxString::ToAscii() const { - wxString res; - res.Printf(_T("%g"), d); + // this will allocate enough space for the terminating NUL too + wxCharBuffer buffer(length()); + char *dest = buffer.data(); - return (*this) << res; + 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; } -// --------------------------------------------------------------------------- -// formatted output -// --------------------------------------------------------------------------- -int wxString::Printf(const wxChar *pszFormat, ...) +#endif // wxUSE_UNICODE + +// extract string of length nCount starting at nFirst +wxString wxString::Mid(size_t nFirst, size_t nCount) const { - va_list argptr; - va_start(argptr, pszFormat); + 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; + } - int iLen = PrintfV(pszFormat, argptr); + if ( nFirst > nLen ) + { + // AllocCopy() will return empty string + return wxEmptyString; + } - va_end(argptr); + wxString dest(*this, nFirst, nCount); + if ( dest.length() != nCount ) + { + wxFAIL_MSG( wxT("out of memory in wxString::Mid") ); + } - return iLen; + return dest; } -int wxString::PrintfV(const wxChar* pszFormat, va_list argptr) +// 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 { - // static buffer to avoid dynamic memory allocation each time - static char s_szScratch[1024]; -#if wxUSE_THREADS - // protect the static buffer - static wxCriticalSection critsect; - wxCriticalSectionLocker lock(critsect); -#endif + if ( compare(0, prefix.length(), prefix) != 0 ) + return false; -#if 1 // the new implementation - - Reinit(); - for (size_t n = 0; pszFormat[n]; n++) - if (pszFormat[n] == _T('%')) { - static char s_szFlags[256] = "%"; - size_t flagofs = 1; - bool adj_left = FALSE, in_prec = FALSE, - prec_dot = FALSE, done = FALSE; - int ilen = 0; - size_t min_width = 0, max_width = wxSTRING_MAXLEN; - do { -#define CHECK_PREC if (in_prec && !prec_dot) { s_szFlags[flagofs++] = '.'; prec_dot = TRUE; } - switch (pszFormat[++n]) { - case _T('\0'): - done = TRUE; - break; - case _T('%'): - *this += _T('%'); - done = TRUE; - break; - case _T('#'): - case _T('0'): - case _T(' '): - case _T('+'): - case _T('\''): - CHECK_PREC - s_szFlags[flagofs++] = pszFormat[n]; - break; - case _T('-'): - CHECK_PREC - adj_left = TRUE; - s_szFlags[flagofs++] = pszFormat[n]; - break; - case _T('.'): - CHECK_PREC - in_prec = TRUE; - prec_dot = FALSE; - max_width = 0; - // dot will be auto-added to s_szFlags if non-negative number follows - break; - case _T('h'): - ilen = -1; - CHECK_PREC - s_szFlags[flagofs++] = pszFormat[n]; - break; - case _T('l'): - ilen = 1; - CHECK_PREC - s_szFlags[flagofs++] = pszFormat[n]; - break; - case _T('q'): - case _T('L'): - ilen = 2; - CHECK_PREC - s_szFlags[flagofs++] = pszFormat[n]; - break; - case _T('Z'): - ilen = 3; - CHECK_PREC - s_szFlags[flagofs++] = pszFormat[n]; - break; - case _T('*'): - { - int len = va_arg(argptr, int); - if (in_prec) { - if (len<0) break; - CHECK_PREC - max_width = len; - } else { - if (len<0) { - adj_left = !adj_left; - s_szFlags[flagofs++] = '-'; - len = -len; - } - min_width = len; - } - flagofs += ::sprintf(s_szFlags+flagofs,"%d",len); - } - break; - case _T('1'): case _T('2'): case _T('3'): - case _T('4'): case _T('5'): case _T('6'): - case _T('7'): case _T('8'): case _T('9'): - { - int len = 0; - CHECK_PREC - while ((pszFormat[n]>=_T('0')) && (pszFormat[n]<=_T('9'))) { - s_szFlags[flagofs++] = pszFormat[n]; - len = len*10 + (pszFormat[n] - _T('0')); - n++; - } - if (in_prec) max_width = len; - else min_width = len; - n--; // the main loop pre-increments n again - } - break; - case _T('d'): - case _T('i'): - case _T('o'): - case _T('u'): - case _T('x'): - case _T('X'): - CHECK_PREC - s_szFlags[flagofs++] = pszFormat[n]; - s_szFlags[flagofs] = '\0'; - if (ilen == 0 ) { - int val = va_arg(argptr, int); - ::sprintf(s_szScratch, s_szFlags, val); - } - else if (ilen == -1) { - short int val = va_arg(argptr, short int); - ::sprintf(s_szScratch, s_szFlags, val); - } - else if (ilen == 1) { - long int val = va_arg(argptr, long int); - ::sprintf(s_szScratch, s_szFlags, val); - } - else if (ilen == 2) { -#if SIZEOF_LONG_LONG - long long int val = va_arg(argptr, long long int); - ::sprintf(s_szScratch, s_szFlags, val); -#else - long int val = va_arg(argptr, long int); - ::sprintf(s_szScratch, s_szFlags, val); -#endif - } - else if (ilen == 3) { - size_t val = va_arg(argptr, size_t); - ::sprintf(s_szScratch, s_szFlags, val); - } - *this += wxString(s_szScratch); - done = TRUE; - break; - case _T('e'): - case _T('E'): - case _T('f'): - case _T('g'): - case _T('G'): - CHECK_PREC - s_szFlags[flagofs++] = pszFormat[n]; - s_szFlags[flagofs] = '\0'; - if (ilen == 2) { - long double val = va_arg(argptr, long double); - ::sprintf(s_szScratch, s_szFlags, val); - } else { - double val = va_arg(argptr, double); - ::sprintf(s_szScratch, s_szFlags, val); - } - *this += wxString(s_szScratch); - done = TRUE; - break; - case _T('p'): - { - void *val = va_arg(argptr, void *); - CHECK_PREC - s_szFlags[flagofs++] = pszFormat[n]; - s_szFlags[flagofs] = '\0'; - ::sprintf(s_szScratch, s_szFlags, val); - *this += wxString(s_szScratch); - done = TRUE; - } - break; - case _T('c'): - { - wxChar val = va_arg(argptr, int); - // we don't need to honor padding here, do we? - *this += val; - done = TRUE; - } - break; - case _T('s'): - if (ilen == -1) { - // wx extension: we'll let %hs mean non-Unicode strings - char *val = va_arg(argptr, char *); -#if wxUSE_UNICODE - // ASCII->Unicode constructor handles max_width right - wxString s(val, wxConv_libc, max_width); -#else - size_t len = wxSTRING_MAXLEN; - if (val) { - for (len = 0; val[len] && (len= 1) { - long int *val = va_arg(argptr, long int *); - *val = Len(); - } - done = TRUE; - break; - default: - if (wxIsalpha(pszFormat[n])) - // probably some flag not taken care of here yet - s_szFlags[flagofs++] = pszFormat[n]; - else { - // bad format - *this += _T('%'); // just to pass the glibc tst-printf.c - n--; - done = TRUE; - } - break; - } -#undef CHECK_PREC - } while (!done); - } else *this += pszFormat[n]; + if ( rest ) + { + // put the rest of the string into provided pointer + rest->assign(*this, prefix.length(), npos); + } -#else - // NB: wxVsnprintf() may return either less than the buffer size or -1 if there - // is not enough place depending on implementation - int iLen = wxVsnprintf(s_szScratch, WXSIZEOF(s_szScratch), pszFormat, argptr); - char *buffer; - if ( iLen < (int)WXSIZEOF(s_szScratch) ) { - buffer = s_szScratch; - } - else { - int size = WXSIZEOF(s_szScratch) * 2; - buffer = (char *)malloc(size); - while ( buffer != NULL ) { - iLen = wxVsnprintf(buffer, WXSIZEOF(s_szScratch), pszFormat, argptr); - if ( iLen < size ) { - // ok, there was enough space - break; - } + return true; +} - // still not enough, double it again - buffer = (char *)realloc(buffer, size *= 2); - } - if ( !buffer ) { - // out of memory - return -1; - } - } +// 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(); - wxString s(buffer); - *this = s; + if ( start < 0 || compare(start, npos, suffix) != 0 ) + return false; - if ( buffer != s_szScratch ) - free(buffer); -#endif + if ( rest ) + { + // put the rest of the string into provided pointer + rest->assign(*this, 0, start); + } - return Len(); + return true; } -// ---------------------------------------------------------------------------- -// misc other operations -// ---------------------------------------------------------------------------- -bool wxString::Matches(const wxChar *pszMask) const + +// extract nCount last (rightmost) characters +wxString wxString::Right(size_t nCount) const { - // check char by char - const wxChar *pszTxt; - for ( pszTxt = c_str(); *pszMask != _T('\0'); pszMask++, pszTxt++ ) { - switch ( *pszMask ) { - case _T('?'): - if ( *pszTxt == _T('\0') ) - return FALSE; + if ( nCount > length() ) + nCount = length(); - pszTxt++; - pszMask++; - break; + wxString dest(*this, length() - nCount, nCount); + if ( dest.length() != nCount ) { + wxFAIL_MSG( wxT("out of memory in wxString::Right") ); + } + return dest; +} - case _T('*'): - { - // ignore special chars immediately following this one - while ( *pszMask == _T('*') || *pszMask == _T('?') ) - pszMask++; +// get all characters after the last occurrence 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.assign(*this, iPos + 1, npos); - // if there is nothing more, match - if ( *pszMask == _T('\0') ) - return TRUE; + return str; +} - // are there any other metacharacters in the mask? - size_t uiLenMask; - const wxChar *pEndMask = wxStrpbrk(pszMask, _T("*?")); +// extract nCount first (leftmost) characters +wxString wxString::Left(size_t nCount) const +{ + if ( nCount > length() ) + nCount = length(); - if ( pEndMask != NULL ) { - // we have to match the string between two metachars - uiLenMask = pEndMask - pszMask; - } - else { - // we have to match the remainder of the string - uiLenMask = wxStrlen(pszMask); - } + wxString dest(*this, 0, nCount); + if ( dest.length() != nCount ) { + wxFAIL_MSG( wxT("out of memory in wxString::Left") ); + } + return dest; +} - wxString strToMatch(pszMask, uiLenMask); - const wxChar* pMatch = wxStrstr(pszTxt, strToMatch); - if ( pMatch == NULL ) - return FALSE; +// get all characters before the first occurrence of ch +// (returns the whole string if ch not found) +wxString wxString::BeforeFirst(wxUniChar ch, wxString *rest) const +{ + int iPos = Find(ch); + if ( iPos == wxNOT_FOUND ) + { + iPos = length(); + if ( rest ) + rest->clear(); + } + else + { + if ( rest ) + rest->assign(*this, iPos + 1, npos); + } - // -1 to compensate "++" in the loop - pszTxt = pMatch + uiLenMask - 1; - pszMask += uiLenMask - 1; - } - break; + return wxString(*this, 0, iPos); +} - default: - if ( *pszMask != *pszTxt ) - return FALSE; - break; - } +/// get all characters before the last occurrence of ch +/// (returns empty string if ch not found) +wxString wxString::BeforeLast(wxUniChar ch, wxString *rest) const +{ + wxString str; + int iPos = Find(ch, true); + if ( iPos != wxNOT_FOUND ) + { + if ( iPos != 0 ) + str.assign(*this, 0, iPos); + + if ( rest ) + rest->assign(*this, iPos + 1, npos); + } + else + { + if ( rest ) + *rest = *this; } - // match only if nothing left - return *pszTxt == _T('\0'); + return str; } -// Count the number of chars -int wxString::Freq(wxChar ch) const +/// get all characters after the first occurrence of ch +/// (returns empty string if ch not found) +wxString wxString::AfterFirst(wxUniChar ch) const { - int count = 0; - int len = Len(); - for (int i = 0; i < len; i++) + wxString str; + int iPos = Find(ch); + if ( iPos != wxNOT_FOUND ) + str.assign(*this, iPos + 1, npos); + + return str; +} + +// replace first (or all) occurrences 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, + wxT("wxString::Replace(): invalid parameter") ); + + wxSTRING_INVALIDATE_CACHE(); + + size_t uiCount = 0; // count of replacements made + + // optimize the special common case: replacement of one character by + // another one (in UTF-8 case we can only do this for ASCII characters) + // + // benchmarks show that this special version is around 3 times faster + // (depending on the proportion of matching characters and UTF-8/wchar_t + // build) + if ( strOld.m_impl.length() == 1 && strNew.m_impl.length() == 1 ) { - if (GetChar(i) == ch) - count ++; + const wxStringCharType chOld = strOld.m_impl[0], + chNew = strNew.m_impl[0]; + + // this loop is the simplified version of the one below + for ( size_t pos = 0; ; ) + { + pos = m_impl.find(chOld, pos); + if ( pos == npos ) + break; + + m_impl[pos++] = chNew; + + uiCount++; + + if ( !bReplaceAll ) + break; + } } - return count; -} + else if ( !bReplaceAll) + { + size_t pos = m_impl.find(strOld, 0); + if ( pos != npos ) + { + m_impl.replace(pos, strOld.m_impl.length(), strNew.m_impl); + uiCount = 1; + } + } + else // replace all occurrences + { + const size_t uiOldLen = strOld.m_impl.length(); + const size_t uiNewLen = strNew.m_impl.length(); -// convert to upper case, return the copy of the string -wxString wxString::Upper() const -{ wxString s(*this); return s.MakeUpper(); } + // first scan the string to find all positions at which the replacement + // should be made + wxVector replacePositions; -// convert to lower case, return the copy of the string -wxString wxString::Lower() const { wxString s(*this); return s.MakeLower(); } + size_t pos; + for ( pos = m_impl.find(strOld.m_impl, 0); + pos != npos; + pos = m_impl.find(strOld.m_impl, pos + uiOldLen)) + { + replacePositions.push_back(pos); + ++uiCount; + } -int wxString::sprintf(const wxChar *pszFormat, ...) - { - va_list argptr; - va_start(argptr, pszFormat); - int iLen = PrintfV(pszFormat, argptr); - va_end(argptr); - return iLen; - } + if ( !uiCount ) + return 0; -// --------------------------------------------------------------------------- -// standard C++ library string functions -// --------------------------------------------------------------------------- -#ifdef wxSTD_STRING_COMPATIBILITY + // allocate enough memory for the whole new string + wxString tmp; + tmp.m_impl.reserve(m_impl.length() + uiCount*(uiNewLen - uiOldLen)); + + // copy this string to tmp doing replacements on the fly + size_t replNum = 0; + for ( pos = 0; replNum < uiCount; replNum++ ) + { + const size_t nextReplPos = replacePositions[replNum]; + + if ( pos != nextReplPos ) + { + tmp.m_impl.append(m_impl, pos, nextReplPos - pos); + } + + tmp.m_impl.append(strNew.m_impl); + pos = nextReplPos + uiOldLen; + } + + if ( pos != m_impl.length() ) + { + // append the rest of the string unchanged + tmp.m_impl.append(m_impl, pos, m_impl.length() - pos); + } + + swap(tmp); + } + + return uiCount; +} -wxString& wxString::insert(size_t nPos, const wxString& str) +bool wxString::IsAscii() const { - wxASSERT( str.GetStringData()->IsValid() ); - wxASSERT( nPos <= Len() ); + for ( const_iterator i = begin(); i != end(); ++i ) + { + if ( !(*i).IsAscii() ) + return false; + } - if ( !str.IsEmpty() ) { - wxString strTmp; - wxChar *pc = strTmp.GetWriteBuf(Len() + str.Len()); - wxStrncpy(pc, c_str(), nPos); - wxStrcpy(pc + nPos, str); - wxStrcpy(pc + nPos + str.Len(), c_str() + nPos); - strTmp.UngetWriteBuf(); - *this = strTmp; - } + return true; +} - return *this; +bool wxString::IsWord() const +{ + for ( const_iterator i = begin(); i != end(); ++i ) + { + if ( !wxIsalpha(*i) ) + return false; + } + + return true; } -size_t wxString::find(const wxString& str, size_t nStart) const +bool wxString::IsNumber() const { - wxASSERT( str.GetStringData()->IsValid() ); - wxASSERT( nStart <= Len() ); + if ( empty() ) + return true; + + const_iterator i = begin(); - const wxChar *p = wxStrstr(c_str() + nStart, str); + if ( *i == wxT('-') || *i == wxT('+') ) + ++i; + + for ( ; i != end(); ++i ) + { + if ( !wxIsdigit(*i) ) + return false; + } - return p == NULL ? npos : p - c_str(); + return true; } -// VC++ 1.5 can't cope with the default argument in the header. -#if !defined(__VISUALC__) || defined(__WIN32__) -size_t wxString::find(const wxChar* sz, size_t nStart, size_t n) const +wxString wxString::Strip(stripType w) const { - return find(wxString(sz, n == npos ? 0 : n), nStart); + wxString s = *this; + if ( w & leading ) s.Trim(false); + if ( w & trailing ) s.Trim(true); + return s; } -#endif // VC++ 1.5 -// Gives a duplicate symbol (presumably a case-insensitivity problem) -#if !defined(__BORLANDC__) -size_t wxString::find(wxChar ch, size_t nStart) const -{ - wxASSERT( nStart <= Len() ); +// --------------------------------------------------------------------------- +// case conversion +// --------------------------------------------------------------------------- - const wxChar *p = wxStrchr(c_str() + nStart, ch); +wxString& wxString::MakeUpper() +{ + for ( iterator it = begin(), en = end(); it != en; ++it ) + *it = (wxChar)wxToupper(*it); - return p == NULL ? npos : p - c_str(); + return *this; } -#endif -size_t wxString::rfind(const wxString& str, size_t nStart) const +wxString& wxString::MakeLower() { - wxASSERT( str.GetStringData()->IsValid() ); - wxASSERT( nStart <= Len() ); - - // # could be quicker than that - const wxChar *p = c_str() + (nStart == npos ? Len() : nStart); - while ( p >= c_str() + str.Len() ) { - if ( wxStrncmp(p - str.Len(), str, str.Len()) == 0 ) - return p - str.Len() - c_str(); - p--; - } + for ( iterator it = begin(), en = end(); it != en; ++it ) + *it = (wxChar)wxTolower(*it); - return npos; + return *this; } -// VC++ 1.5 can't cope with the default argument in the header. -#if !defined(__VISUALC__) || defined(__WIN32__) -size_t wxString::rfind(const wxChar* sz, size_t nStart, size_t n) const +wxString& wxString::MakeCapitalized() { - return rfind(wxString(sz, n == npos ? 0 : n), nStart); -} + const iterator en = end(); + iterator it = begin(); + if ( it != en ) + { + *it = (wxChar)wxToupper(*it); + for ( ++it; it != en; ++it ) + *it = (wxChar)wxTolower(*it); + } -size_t wxString::rfind(wxChar ch, size_t nStart) const -{ - wxASSERT( nStart <= Len() ); + return *this; +} - const wxChar *p = wxStrrchr(c_str() + nStart, ch); +// --------------------------------------------------------------------------- +// trimming and padding +// --------------------------------------------------------------------------- - return p == NULL ? npos : p - c_str(); -} -#endif // VC++ 1.5 +// some compilers (VC++ 6.0 not to name them) return true for a call to +// isspace('\xEA') in the C locale which seems to be broken to me, but we have +// to live with this by checking that the character is a 7 bit one - even if +// this may fail to detect some spaces (I don't know if Unicode doesn't have +// space-like symbols somewhere except in the first 128 chars), it is arguably +// still better than trimming away accented letters +inline int wxSafeIsspace(wxChar ch) { return (ch < 127) && wxIsspace(ch); } -wxString wxString::substr(size_t nStart, size_t nLen) const +// trims spaces (in the sense of isspace) from left or right side +wxString& wxString::Trim(bool bFromRight) { - // npos means 'take all' - if ( nLen == npos ) - nLen = 0; + // first check if we're going to modify the string at all + if ( !empty() && + ( + (bFromRight && wxSafeIsspace(GetChar(length() - 1))) || + (!bFromRight && wxSafeIsspace(GetChar(0u))) + ) + ) + { + if ( bFromRight ) + { + // find last non-space character + reverse_iterator psz = rbegin(); + while ( (psz != rend()) && wxSafeIsspace(*psz) ) + ++psz; + + // truncate at trailing space start + erase(psz.base(), end()); + } + else + { + // find first non-space character + iterator psz = begin(); + while ( (psz != end()) && wxSafeIsspace(*psz) ) + ++psz; - wxASSERT( nStart + nLen <= Len() ); + // fix up data and length + erase(begin(), psz); + } + } - return wxString(c_str() + nStart, nLen == npos ? 0 : nLen); + return *this; } -wxString& wxString::erase(size_t nStart, size_t nLen) +// adds nCount characters chPad to the string from either side +wxString& wxString::Pad(size_t nCount, wxUniChar chPad, bool bFromRight) { - wxString strTmp(c_str(), nStart); - if ( nLen != npos ) { - wxASSERT( nStart + nLen <= Len() ); + wxString s(chPad, nCount); - strTmp.append(c_str() + nStart + nLen); - } + if ( bFromRight ) + *this += s; + else + { + s += *this; + swap(s); + } - *this = strTmp; - return *this; + return *this; } -wxString& wxString::replace(size_t nStart, size_t nLen, const wxChar *sz) +// truncate the string +wxString& wxString::Truncate(size_t uiLen) { - wxASSERT( nStart + nLen <= wxStrlen(sz) ); - - wxString strTmp; - if ( nStart != 0 ) - strTmp.append(c_str(), nStart); - strTmp += sz; - strTmp.append(c_str() + nStart + nLen); + if ( uiLen < length() ) + { + erase(begin() + uiLen, end()); + } + //else: nothing to do, string is already short enough - *this = strTmp; - return *this; + return *this; } -wxString& wxString::replace(size_t nStart, size_t nLen, size_t nCount, wxChar ch) -{ - return replace(nStart, nLen, wxString(ch, nCount)); -} +// --------------------------------------------------------------------------- +// finding (return wxNOT_FOUND if not found and index otherwise) +// --------------------------------------------------------------------------- -wxString& wxString::replace(size_t nStart, size_t nLen, - const wxString& str, size_t nStart2, size_t nLen2) +// find a character +int wxString::Find(wxUniChar ch, bool bFromEnd) const { - return replace(nStart, nLen, str.substr(nStart2, nLen2)); -} + size_type idx = bFromEnd ? find_last_of(ch) : find_first_of(ch); -wxString& wxString::replace(size_t nStart, size_t nLen, - const wxChar* sz, size_t nCount) -{ - return replace(nStart, nLen, wxString(sz, nCount)); + return (idx == npos) ? wxNOT_FOUND : (int)idx; } -#endif //std::string compatibility +// ---------------------------------------------------------------------------- +// conversion to numbers +// ---------------------------------------------------------------------------- -// ============================================================================ -// ArrayString -// ============================================================================ +// The implementation of all the functions below is exactly the same so factor +// it out. Note that number extraction works correctly on UTF-8 strings, so +// we can use wxStringCharType and wx_str() for maximum efficiency. -// size increment = max(50% of current size, ARRAY_MAXSIZE_INCREMENT) -#define ARRAY_MAXSIZE_INCREMENT 4096 -#ifndef ARRAY_DEFAULT_INITIAL_SIZE // also defined in dynarray.h - #define ARRAY_DEFAULT_INITIAL_SIZE (16) +#ifndef __WXWINCE__ + #define DO_IF_NOT_WINCE(x) x +#else + #define DO_IF_NOT_WINCE(x) #endif -#define STRING(p) ((wxString *)(&(p))) +#define WX_STRING_TO_X_TYPE_START \ + wxCHECK_MSG( pVal, false, wxT("NULL output pointer") ); \ + DO_IF_NOT_WINCE( errno = 0; ) \ + const wxStringCharType *start = wx_str(); \ + wxStringCharType *end; -// ctor -wxArrayString::wxArrayString() -{ - m_nSize = - m_nCount = 0; - m_pItems = (wxChar **) NULL; -} +// notice that we return false without modifying the output parameter at all if +// nothing could be parsed but we do modify it and return false then if we did +// parse something successfully but not the entire string +#define WX_STRING_TO_X_TYPE_END \ + if ( end == start DO_IF_NOT_WINCE(|| errno == ERANGE) ) \ + return false; \ + *pVal = val; \ + return !*end; -// copy ctor -wxArrayString::wxArrayString(const wxArrayString& src) +bool wxString::ToLong(long *pVal, int base) const { - m_nSize = - m_nCount = 0; - m_pItems = (wxChar **) NULL; + wxASSERT_MSG( !base || (base > 1 && base <= 36), wxT("invalid base") ); - *this = src; + WX_STRING_TO_X_TYPE_START + long val = wxStrtol(start, &end, base); + WX_STRING_TO_X_TYPE_END } -// assignment operator -wxArrayString& wxArrayString::operator=(const wxArrayString& src) +bool wxString::ToULong(unsigned long *pVal, int base) const { - if ( m_nSize > 0 ) - Clear(); - - if ( src.m_nCount > ARRAY_DEFAULT_INITIAL_SIZE ) - Alloc(src.m_nCount); + wxASSERT_MSG( !base || (base > 1 && base <= 36), wxT("invalid base") ); - // we can't just copy the pointers here because otherwise we would share - // the strings with another array - for ( size_t n = 0; n < src.m_nCount; n++ ) - Add(src[n]); + WX_STRING_TO_X_TYPE_START + unsigned long val = wxStrtoul(start, &end, base); + WX_STRING_TO_X_TYPE_END +} - if ( m_nCount != 0 ) - memcpy(m_pItems, src.m_pItems, m_nCount*sizeof(wxChar *)); +bool wxString::ToLongLong(wxLongLong_t *pVal, int base) const +{ + wxASSERT_MSG( !base || (base > 1 && base <= 36), wxT("invalid base") ); - return *this; + WX_STRING_TO_X_TYPE_START + wxLongLong_t val = wxStrtoll(start, &end, base); + WX_STRING_TO_X_TYPE_END } -// grow the array -void wxArrayString::Grow() +bool wxString::ToULongLong(wxULongLong_t *pVal, int base) const { - // only do it if no more place - if( m_nCount == m_nSize ) { - if( m_nSize == 0 ) { - // was empty, alloc some memory - m_nSize = ARRAY_DEFAULT_INITIAL_SIZE; - m_pItems = new wxChar *[m_nSize]; - } - else { - // otherwise when it's called for the first time, nIncrement would be 0 - // and the array would never be expanded - wxASSERT( ARRAY_DEFAULT_INITIAL_SIZE != 0 ); - - // add 50% but not too much - size_t nIncrement = m_nSize < ARRAY_DEFAULT_INITIAL_SIZE - ? ARRAY_DEFAULT_INITIAL_SIZE : m_nSize >> 1; - if ( nIncrement > ARRAY_MAXSIZE_INCREMENT ) - nIncrement = ARRAY_MAXSIZE_INCREMENT; - m_nSize += nIncrement; - wxChar **pNew = new wxChar *[m_nSize]; - - // copy data to new location - memcpy(pNew, m_pItems, m_nCount*sizeof(wxChar *)); - - // delete old memory (but do not release the strings!) - wxDELETEA(m_pItems); - - m_pItems = pNew; - } - } + wxASSERT_MSG( !base || (base > 1 && base <= 36), wxT("invalid base") ); + + WX_STRING_TO_X_TYPE_START + wxULongLong_t val = wxStrtoull(start, &end, base); + WX_STRING_TO_X_TYPE_END } -void wxArrayString::Free() +bool wxString::ToDouble(double *pVal) const { - for ( size_t n = 0; n < m_nCount; n++ ) { - STRING(m_pItems[n])->GetStringData()->Unlock(); - } + WX_STRING_TO_X_TYPE_START + double val = wxStrtod(start, &end); + WX_STRING_TO_X_TYPE_END } -// deletes all the strings from the list -void wxArrayString::Empty() +#if wxUSE_XLOCALE + +bool wxString::ToCLong(long *pVal, int base) const { - Free(); + wxASSERT_MSG( !base || (base > 1 && base <= 36), wxT("invalid base") ); - m_nCount = 0; + WX_STRING_TO_X_TYPE_START +#if (wxUSE_UNICODE_UTF8 || !wxUSE_UNICODE) && defined(wxHAS_XLOCALE_SUPPORT) + long val = wxStrtol_lA(start, &end, base, wxCLocale); +#else + long val = wxStrtol_l(start, &end, base, wxCLocale); +#endif + WX_STRING_TO_X_TYPE_END } -// as Empty, but also frees memory -void wxArrayString::Clear() +bool wxString::ToCULong(unsigned long *pVal, int base) const { - Free(); - - m_nSize = - m_nCount = 0; + wxASSERT_MSG( !base || (base > 1 && base <= 36), wxT("invalid base") ); - wxDELETEA(m_pItems); + WX_STRING_TO_X_TYPE_START +#if (wxUSE_UNICODE_UTF8 || !wxUSE_UNICODE) && defined(wxHAS_XLOCALE_SUPPORT) + unsigned long val = wxStrtoul_lA(start, &end, base, wxCLocale); +#else + unsigned long val = wxStrtoul_l(start, &end, base, wxCLocale); +#endif + WX_STRING_TO_X_TYPE_END } -// dtor -wxArrayString::~wxArrayString() +bool wxString::ToCDouble(double *pVal) const { - Free(); - - wxDELETEA(m_pItems); + WX_STRING_TO_X_TYPE_START +#if (wxUSE_UNICODE_UTF8 || !wxUSE_UNICODE) && defined(wxHAS_XLOCALE_SUPPORT) + double val = wxStrtod_lA(start, &end, wxCLocale); +#else + double val = wxStrtod_l(start, &end, wxCLocale); +#endif + WX_STRING_TO_X_TYPE_END } -// pre-allocates memory (frees the previous data!) -void wxArrayString::Alloc(size_t nSize) -{ - wxASSERT( nSize > 0 ); +#else // wxUSE_XLOCALE - // only if old buffer was not big enough - if ( nSize > m_nSize ) { - Free(); - wxDELETEA(m_pItems); - m_pItems = new wxChar *[nSize]; - m_nSize = nSize; - } +// Provide implementation of these functions even when wxUSE_XLOCALE is +// disabled, we still need them in wxWidgets internal code. - m_nCount = 0; +// For integers we just assume the current locale uses the same number +// representation as the C one as there is nothing else we can do. +bool wxString::ToCLong(long *pVal, int base) const +{ + return ToLong(pVal, base); } -// minimizes the memory usage by freeing unused memory -void wxArrayString::Shrink() +bool wxString::ToCULong(unsigned long *pVal, int base) const { - // only do it if we have some memory to free - if( m_nCount < m_nSize ) { - // allocates exactly as much memory as we need - wxChar **pNew = new wxChar *[m_nCount]; - - // copy data to new location - memcpy(pNew, m_pItems, m_nCount*sizeof(wxChar *)); - delete [] m_pItems; - m_pItems = pNew; - } + return ToULong(pVal, base); } -// searches the array for an item (forward or backwards) -int wxArrayString::Index(const wxChar *sz, bool bCase, bool bFromEnd) const +// For floating point numbers we have to handle the problem of the decimal +// point which is different in different locales. +bool wxString::ToCDouble(double *pVal) const { - if ( bFromEnd ) { - if ( m_nCount > 0 ) { - size_t ui = m_nCount; - do { - if ( STRING(m_pItems[--ui])->IsSameAs(sz, bCase) ) - return ui; - } - while ( ui != 0 ); + // Create a copy of this string using the decimal point instead of whatever + // separator the current locale uses. +#if wxUSE_INTL + wxString sep = wxLocale::GetInfo(wxLOCALE_DECIMAL_POINT, + wxLOCALE_CAT_NUMBER); + if ( sep == "." ) + { + // We can avoid an unnecessary string copy in this case. + return ToDouble(pVal); } - } - else { - for( size_t ui = 0; ui < m_nCount; ui++ ) { - if( STRING(m_pItems[ui])->IsSameAs(sz, bCase) ) - return ui; +#else // !wxUSE_INTL + // We don't know what the current separator is so it might even be a point + // already, try to parse the string as a double: + if ( ToDouble(pVal) ) + { + // It must have been the point, nothing else to do. + return true; } - } - return wxNOT_FOUND; -} - -// add item at the end -void wxArrayString::Add(const wxString& str) -{ - wxASSERT( str.GetStringData()->IsValid() ); - - Grow(); + // Try to guess the separator, using the most common alternative value. + wxString sep(","); +#endif // wxUSE_INTL/!wxUSE_INTL + wxString cstr(*this); + cstr.Replace(".", sep); - // the string data must not be deleted! - str.GetStringData()->Lock(); - m_pItems[m_nCount++] = (wxChar *)str.c_str(); + return cstr.ToDouble(pVal); } -// add item at the given position -void wxArrayString::Insert(const wxString& str, size_t nIndex) -{ - wxASSERT( str.GetStringData()->IsValid() ); - - wxCHECK_RET( nIndex <= m_nCount, _("bad index in wxArrayString::Insert") ); - - Grow(); - - memmove(&m_pItems[nIndex + 1], &m_pItems[nIndex], - (m_nCount - nIndex)*sizeof(wxChar *)); +#endif // wxUSE_XLOCALE/!wxUSE_XLOCALE - str.GetStringData()->Lock(); - m_pItems[nIndex] = (wxChar *)str.c_str(); - - m_nCount++; -} +// ---------------------------------------------------------------------------- +// number to string conversion +// ---------------------------------------------------------------------------- -// removes item from array (by index) -void wxArrayString::Remove(size_t nIndex) +/* static */ +wxString wxString::FromDouble(double val, int precision) { - wxCHECK_RET( nIndex <= m_nCount, _("bad index in wxArrayString::Remove") ); + wxCHECK_MSG( precision >= -1, wxString(), "Invalid negative precision" ); - // release our lock - Item(nIndex).GetStringData()->Unlock(); + wxString format; + if ( precision == -1 ) + { + format = "%g"; + } + else // Use fixed precision. + { + format.Printf("%%.%df", precision); + } - memmove(&m_pItems[nIndex], &m_pItems[nIndex + 1], - (m_nCount - nIndex - 1)*sizeof(wxChar *)); - m_nCount--; + return wxString::Format(format, val); } -// removes item from array (by value) -void wxArrayString::Remove(const wxChar *sz) +/* static */ +wxString wxString::FromCDouble(double val, int precision) { - int iIndex = Index(sz); + wxCHECK_MSG( precision >= -1, wxString(), "Invalid negative precision" ); - wxCHECK_RET( iIndex != wxNOT_FOUND, - _("removing inexistent element in wxArrayString::Remove") ); +#if wxUSE_STD_IOSTREAM && wxUSE_STD_STRING + // We assume that we can use the ostream and not wstream for numbers. + wxSTD ostringstream os; + if ( precision != -1 ) + { + os.precision(precision); + os.setf(std::ios::fixed, std::ios::floatfield); + } - Remove(iIndex); + os << val; + return os.str(); +#else // !wxUSE_STD_IOSTREAM + // Can't use iostream locale support, fall back to the manual method + // instead. + wxString s = FromDouble(val, precision); +#if wxUSE_INTL + wxString sep = wxLocale::GetInfo(wxLOCALE_DECIMAL_POINT, + wxLOCALE_CAT_NUMBER); +#else // !wxUSE_INTL + // As above, this is the most common alternative value. Notice that here it + // doesn't matter if we guess wrongly and the current separator is already + // ".": we'll just waste a call to Replace() in this case. + wxString sep(","); +#endif // wxUSE_INTL/!wxUSE_INTL + + s.Replace(sep, "."); + return s; +#endif // wxUSE_STD_IOSTREAM/!wxUSE_STD_IOSTREAM } -// ---------------------------------------------------------------------------- -// sorting -// ---------------------------------------------------------------------------- - -// we can only sort one array at a time with the quick-sort based -// implementation -#if wxUSE_THREADS - // need a critical section to protect access to gs_compareFunction and - // gs_sortAscending variables - static wxCriticalSection *gs_critsectStringSort = NULL; - - // call this before the value of the global sort vars is changed/after - // you're finished with them - #define START_SORT() wxASSERT( !gs_critsectStringSort ); \ - gs_critsectStringSort = new wxCriticalSection; \ - gs_critsectStringSort->Enter() - #define END_SORT() gs_critsectStringSort->Leave(); \ - delete gs_critsectStringSort; \ - gs_critsectStringSort = NULL -#else // !threads - #define START_SORT() - #define END_SORT() -#endif // wxUSE_THREADS - -// function to use for string comparaison -static wxArrayString::CompareFunction gs_compareFunction = NULL; - -// if we don't use the compare function, this flag tells us if we sort the -// array in ascending or descending order -static bool gs_sortAscending = TRUE; - -// function which is called by quick sort -static int wxStringCompareFunction(const void *first, const void *second) -{ - wxString *strFirst = (wxString *)first; - wxString *strSecond = (wxString *)second; - - if ( gs_compareFunction ) { - return gs_compareFunction(*strFirst, *strSecond); - } - else { - // maybe we should use wxStrcoll - int result = wxStrcmp(strFirst->c_str(), strSecond->c_str()); - - return gs_sortAscending ? result : -result; - } -} +// --------------------------------------------------------------------------- +// formatted output +// --------------------------------------------------------------------------- -// sort array elements using passed comparaison function -void wxArrayString::Sort(CompareFunction compareFunction) +#if !wxUSE_UTF8_LOCALE_ONLY +/* static */ +#ifdef wxNEEDS_WXSTRING_PRINTF_MIXIN +wxString wxStringPrintfMixinBase::DoFormatWchar(const wxChar *format, ...) +#else +wxString wxString::DoFormatWchar(const wxChar *format, ...) +#endif { - START_SORT(); + va_list argptr; + va_start(argptr, format); - wxASSERT( !gs_compareFunction ); // must have been reset to NULL - gs_compareFunction = compareFunction; + wxString s; + s.PrintfV(format, argptr); - DoSort(); + va_end(argptr); - END_SORT(); + return s; } +#endif // !wxUSE_UTF8_LOCALE_ONLY -void wxArrayString::Sort(bool reverseOrder) +#if wxUSE_UNICODE_UTF8 +/* static */ +wxString wxString::DoFormatUtf8(const char *format, ...) { - START_SORT(); + va_list argptr; + va_start(argptr, format); - wxASSERT( !gs_compareFunction ); // must have been reset to NULL - gs_sortAscending = !reverseOrder; + wxString s; + s.PrintfV(format, argptr); - DoSort(); + va_end(argptr); - END_SORT(); + return s; } +#endif // wxUSE_UNICODE_UTF8 -void wxArrayString::DoSort() +/* static */ +wxString wxString::FormatV(const wxString& format, va_list argptr) { - // just sort the pointers using qsort() - of course it only works because - // wxString() *is* a pointer to its data - qsort(m_pItems, m_nCount, sizeof(wxChar *), wxStringCompareFunction); + wxString s; + s.PrintfV(format, argptr); + return s; } -// ============================================================================ -// MBConv -// ============================================================================ +#if !wxUSE_UTF8_LOCALE_ONLY +#ifdef wxNEEDS_WXSTRING_PRINTF_MIXIN +int wxStringPrintfMixinBase::DoPrintfWchar(const wxChar *format, ...) +#else +int wxString::DoPrintfWchar(const wxChar *format, ...) +#endif +{ + va_list argptr; + va_start(argptr, format); -WXDLLEXPORT_DATA(wxMBConv *) wxConv_current = &wxConv_libc; +#ifdef wxNEEDS_WXSTRING_PRINTF_MIXIN + // get a pointer to the wxString instance; we have to use dynamic_cast<> + // because it's the only cast that works safely for downcasting when + // multiple inheritance is used: + wxString *str = static_cast(this); +#else + wxString *str = this; +#endif -// ---------------------------------------------------------------------------- -// standard libc conversion -// ---------------------------------------------------------------------------- + int iLen = str->PrintfV(format, argptr); -WXDLLEXPORT_DATA(wxMBConv) wxConv_libc; + va_end(argptr); -size_t wxMBConv::MB2WC(wchar_t *buf, const char *psz, size_t n) const -{ - return wxMB2WC(buf, psz, n); + return iLen; } +#endif // !wxUSE_UTF8_LOCALE_ONLY -size_t wxMBConv::WC2MB(char *buf, const wchar_t *psz, size_t n) const +#if wxUSE_UNICODE_UTF8 +int wxString::DoPrintfUtf8(const char *format, ...) { - return wxWC2MB(buf, psz, n); -} + va_list argptr; + va_start(argptr, format); -// ---------------------------------------------------------------------------- -// standard file conversion -// ---------------------------------------------------------------------------- + int iLen = PrintfV(format, argptr); -WXDLLEXPORT_DATA(wxMBConv_file) wxConv_file; + va_end(argptr); -// just use the libc conversion for now -size_t wxMBConv_file::MB2WC(wchar_t *buf, const char *psz, size_t n) const -{ - return wxMB2WC(buf, psz, n); + return iLen; } +#endif // wxUSE_UNICODE_UTF8 -size_t wxMBConv_file::WC2MB(char *buf, const wchar_t *psz, size_t n) const +/* + Uses wxVsnprintf and places the result into the this string. + + In ANSI build, wxVsnprintf is effectively vsnprintf but in Unicode build + it is vswprintf. Due to a discrepancy between vsnprintf and vswprintf in + the ISO C99 (and thus SUSv3) standard the return value for the case of + an undersized buffer is inconsistent. For conforming vsnprintf + implementations the function must return the number of characters that + would have been printed had the buffer been large enough. For conforming + vswprintf implementations the function must return a negative number + and set errno. + + What vswprintf sets errno to is undefined but Darwin seems to set it to + EOVERFLOW. The only expected errno are EILSEQ and EINVAL. Both of + those are defined in the standard and backed up by several conformance + statements. Note that ENOMEM mentioned in the manual page does not + apply to swprintf, only wprintf and fwprintf. + + Official manual page: + http://www.opengroup.org/onlinepubs/009695399/functions/swprintf.html + + Some conformance statements (AIX, Solaris): + http://www.opengroup.org/csq/view.mhtml?RID=ibm%2FSD1%2F3 + http://www.theopengroup.org/csq/view.mhtml?norationale=1&noreferences=1&RID=Fujitsu%2FSE2%2F10 + + Since EILSEQ and EINVAL are rather common but EOVERFLOW is not and since + EILSEQ and EINVAL are specifically defined to mean the error is other than + an undersized buffer and no other errno are defined we treat those two + as meaning hard errors and everything else gets the old behaviour which + is to keep looping and increasing buffer size until the function succeeds. + + In practice it's impossible to determine before compilation which behaviour + may be used. The vswprintf function may have vsnprintf-like behaviour or + vice-versa. Behaviour detected on one release can theoretically change + with an updated release. Not to mention that configure testing for it + would require the test to be run on the host system, not the build system + which makes cross compilation difficult. Therefore, we make no assumptions + about behaviour and try our best to handle every known case, including the + case where wxVsnprintf returns a negative number and fails to set errno. + + There is yet one more non-standard implementation and that is our own. + Fortunately, that can be detected at compile-time. + + On top of all that, ISO C99 explicitly defines snprintf to write a null + character to the last position of the specified buffer. That would be at + at the given buffer size minus 1. It is supposed to do this even if it + turns out that the buffer is sized too small. + + Darwin (tested on 10.5) follows the C99 behaviour exactly. + + Glibc 2.6 almost follows the C99 behaviour except vswprintf never sets + errno even when it fails. However, it only seems to ever fail due + to an undersized buffer. +*/ +#if wxUSE_UNICODE_UTF8 +template +#else +// we only need one version in non-UTF8 builds and at least two Windows +// compilers have problems with this function template, so use just one +// normal function here +#endif +static int DoStringPrintfV(wxString& str, + const wxString& format, va_list argptr) { - return wxWC2MB(buf, psz, n); -} + int size = 1024; -// ---------------------------------------------------------------------------- -// standard gdk conversion -// ---------------------------------------------------------------------------- + for ( ;; ) + { +#if wxUSE_UNICODE_UTF8 + BufferType tmp(str, size + 1); + typename BufferType::CharType *buf = tmp; +#else + wxStringBuffer tmp(str, size + 1); + wxChar *buf = tmp; +#endif -#if defined(__WXGTK__) && (GTK_MINOR_VERSION > 0) -WXDLLEXPORT_DATA(wxMBConv_gdk) wxConv_gdk; + if ( !buf ) + { + // out of memory + return -1; + } -#include + // wxVsnprintf() may modify the original arg pointer, so pass it + // only a copy + va_list argptrcopy; + wxVaCopy(argptrcopy, argptr); -size_t wxMBConv_gdk::MB2WC(wchar_t *buf, const char *psz, size_t n) const -{ - if (buf) { - return gdk_mbstowcs((GdkWChar *)buf, psz, n); - } else { - GdkWChar *nbuf = new GdkWChar[n=strlen(psz)]; - size_t len = gdk_mbstowcs(nbuf, psz, n); - delete [] nbuf; - return len; - } +#ifndef __WXWINCE__ + // Set errno to 0 to make it determinate if wxVsnprintf fails to set it. + errno = 0; +#endif + int len = wxVsnprintf(buf, size, format, argptrcopy); + va_end(argptrcopy); + + // some implementations of vsnprintf() don't NUL terminate + // the string if there is not enough space for it so + // always do it manually + // FIXME: This really seems to be the wrong and would be an off-by-one + // bug except the code above allocates an extra character. + buf[size] = wxT('\0'); + + // vsnprintf() may return either -1 (traditional Unix behaviour) or the + // total number of characters which would have been written if the + // buffer were large enough (newer standards such as Unix98) + if ( len < 0 ) + { + // NB: wxVsnprintf() may call either wxCRT_VsnprintfW or + // wxCRT_VsnprintfA in UTF-8 build; wxUSE_WXVSNPRINTF + // is true if *both* of them use our own implementation, + // otherwise we can't be sure +#if wxUSE_WXVSNPRINTF + // we know that our own implementation of wxVsnprintf() returns -1 + // only for a format error - thus there's something wrong with + // the user's format string + buf[0] = '\0'; + return -1; +#else // possibly using system version + // assume it only returns error if there is not enough space, but + // as we don't know how much we need, double the current size of + // the buffer +#ifndef __WXWINCE__ + if( (errno == EILSEQ) || (errno == EINVAL) ) + // If errno was set to one of the two well-known hard errors + // then fail immediately to avoid an infinite loop. + return -1; + else +#endif // __WXWINCE__ + // still not enough, as we don't know how much we need, double the + // current size of the buffer + size *= 2; +#endif // wxUSE_WXVSNPRINTF/!wxUSE_WXVSNPRINTF + } + else if ( len >= size ) + { +#if wxUSE_WXVSNPRINTF + // we know that our own implementation of wxVsnprintf() returns + // size+1 when there's not enough space but that's not the size + // of the required buffer! + size *= 2; // so we just double the current size of the buffer +#else + // some vsnprintf() implementations NUL-terminate the buffer and + // some don't in len == size case, to be safe always add 1 + // FIXME: I don't quite understand this comment. The vsnprintf + // function is specifically defined to return the number of + // characters printed not including the null terminator. + // So OF COURSE you need to add 1 to get the right buffer size. + // The following line is definitely correct, no question. + size = len + 1; +#endif + } + else // ok, there was enough space + { + break; + } + } + + // we could have overshot + str.Shrink(); + + return str.length(); } -size_t wxMBConv_gdk::WC2MB(char *buf, const wchar_t *psz, size_t n) const +int wxString::PrintfV(const wxString& format, va_list argptr) { - char *mbstr = gdk_wcstombs((GdkWChar *)psz); - size_t len = mbstr ? strlen(mbstr) : 0; - if (buf) { - if (len > n) len = n; - memcpy(buf, psz, len); - if (len < n) buf[len] = 0; - } - return len; +#if wxUSE_UNICODE_UTF8 + #if wxUSE_STL_BASED_WXSTRING + typedef wxStringTypeBuffer Utf8Buffer; + #else + typedef wxStringInternalBuffer Utf8Buffer; + #endif +#endif + +#if wxUSE_UTF8_LOCALE_ONLY + return DoStringPrintfV(*this, format, argptr); +#else + #if wxUSE_UNICODE_UTF8 + if ( wxLocaleIsUtf8 ) + return DoStringPrintfV(*this, format, argptr); + else + // wxChar* version + return DoStringPrintfV(*this, format, argptr); + #else + return DoStringPrintfV(*this, format, argptr); + #endif // UTF8/WCHAR +#endif } -#endif // GTK > 1.0 // ---------------------------------------------------------------------------- -// UTF-7 +// misc other operations // ---------------------------------------------------------------------------- -WXDLLEXPORT_DATA(wxMBConv_UTF7) wxConv_UTF7; +// returns true if the string matches the pattern which may contain '*' and +// '?' metacharacters (as usual, '?' matches any character and '*' any number +// of them) +bool wxString::Matches(const wxString& mask) const +{ + // I disable this code as it doesn't seem to be faster (in fact, it seems + // to be much slower) than the old, hand-written code below and using it + // here requires always linking with libregex even if the user code doesn't + // use it +#if 0 // wxUSE_REGEX + // first translate the shell-like mask into a regex + wxString pattern; + pattern.reserve(wxStrlen(pszMask)); + + pattern += wxT('^'); + while ( *pszMask ) + { + switch ( *pszMask ) + { + case wxT('?'): + pattern += wxT('.'); + break; + + case wxT('*'): + pattern += wxT(".*"); + break; + + case wxT('^'): + case wxT('.'): + case wxT('$'): + case wxT('('): + case wxT(')'): + case wxT('|'): + case wxT('+'): + case wxT('\\'): + // these characters are special in a RE, quote them + // (however note that we don't quote '[' and ']' to allow + // using them for Unix shell like matching) + pattern += wxT('\\'); + // fall through + + default: + pattern += *pszMask; + } -// TODO: write actual implementations of UTF-7 here -size_t wxMBConv_UTF7::MB2WC(wchar_t *buf, const char *psz, size_t n) const -{ - return 0; -} + pszMask++; + } + pattern += wxT('$'); + + // and now use it + return wxRegEx(pattern, wxRE_NOSUB | wxRE_EXTENDED).Matches(c_str()); +#else // !wxUSE_REGEX + // TODO: this is, of course, awfully inefficient... + + // FIXME-UTF8: implement using iterators, remove #if +#if wxUSE_UNICODE_UTF8 + const wxScopedWCharBuffer maskBuf = mask.wc_str(); + const wxScopedWCharBuffer txtBuf = wc_str(); + const wxChar *pszMask = maskBuf.data(); + const wxChar *pszTxt = txtBuf.data(); +#else + const wxChar *pszMask = mask.wx_str(); + // the char currently being checked + const wxChar *pszTxt = wx_str(); +#endif -size_t wxMBConv_UTF7::WC2MB(char *buf, const wchar_t *psz, size_t n) const -{ - return 0; -} + // the last location where '*' matched + const wxChar *pszLastStarInText = NULL; + const wxChar *pszLastStarInMask = NULL; -// ---------------------------------------------------------------------------- -// UTF-8 -// ---------------------------------------------------------------------------- +match: + for ( ; *pszMask != wxT('\0'); pszMask++, pszTxt++ ) { + switch ( *pszMask ) { + case wxT('?'): + if ( *pszTxt == wxT('\0') ) + return false; -WXDLLEXPORT_DATA(wxMBConv_UTF8) wxConv_UTF8; + // pszTxt and pszMask will be incremented in the loop statement -// TODO: write actual implementations of UTF-8 here -size_t wxMBConv_UTF8::MB2WC(wchar_t *buf, const char *psz, size_t n) const -{ - return wxMB2WC(buf, psz, n); -} + break; -size_t wxMBConv_UTF8::WC2MB(char *buf, const wchar_t *psz, size_t n) const -{ - return wxWC2MB(buf, psz, n); -} + case wxT('*'): + { + // remember where we started to be able to backtrack later + pszLastStarInText = pszTxt; + pszLastStarInMask = pszMask; -// ---------------------------------------------------------------------------- -// specified character set -// ---------------------------------------------------------------------------- + // ignore special chars immediately following this one + // (should this be an error?) + while ( *pszMask == wxT('*') || *pszMask == wxT('?') ) + pszMask++; -class wxCharacterSet -{ -public: - wxArrayString names; - wchar_t *data; -}; + // if there is nothing more, match + if ( *pszMask == wxT('\0') ) + return true; -#ifndef WX_PRECOMP - #include "wx/dynarray.h" - #include "wx/filefn.h" - #include "wx/textfile.h" - #include "wx/tokenzr.h" - #include "wx/utils.h" -#endif + // are there any other metacharacters in the mask? + size_t uiLenMask; + const wxChar *pEndMask = wxStrpbrk(pszMask, wxT("*?")); + + if ( pEndMask != NULL ) { + // we have to match the string between two metachars + uiLenMask = pEndMask - pszMask; + } + else { + // we have to match the remainder of the string + uiLenMask = wxStrlen(pszMask); + } + + wxString strToMatch(pszMask, uiLenMask); + const wxChar* pMatch = wxStrstr(pszTxt, strToMatch); + if ( pMatch == NULL ) + return false; + + // -1 to compensate "++" in the loop + pszTxt = pMatch + uiLenMask - 1; + pszMask += uiLenMask - 1; + } + break; -WX_DECLARE_OBJARRAY(wxCharacterSet, wxCSArray); -#include "wx/arrimpl.cpp" -WX_DEFINE_OBJARRAY(wxCSArray); - -static wxCSArray wxCharsets; - -static void wxLoadCharacterSets(void) -{ - static bool already_loaded = FALSE; - - if (already_loaded) return; - -#if defined(__UNIX__) - // search through files in /usr/share/i18n/charmaps - wxString fname; - for (fname = ::wxFindFirstFile(_T("/usr/share/i18n/charmaps/*")); - !fname.IsEmpty(); - fname = ::wxFindNextFile()) { - wxTextFile cmap(fname); - if (cmap.Open()) { - wxCharacterSet *cset = new wxCharacterSet; - wxString comchar,escchar; - bool in_charset = FALSE; - - // wxFprintf(stderr,_T("Loaded: %s\n"),fname.c_str()); - - wxString line; - for (line = cmap.GetFirstLine(); - !cmap.Eof(); - line = cmap.GetNextLine()) { - // wxFprintf(stderr,_T("line contents: %s\n"),line.c_str()); - wxStringTokenizer token(line); - wxString cmd = token.GetNextToken(); - if (cmd == comchar) { - if (token.GetNextToken() == _T("alias")) { - wxStringTokenizer names(token.GetNextToken(),_T("/")); - wxString name; - while (!(name = names.GetNextToken()).IsEmpty()) - cset->names.Add(name); - } - } - else if (cmd == _T("")) - cset->names.Add(token.GetNextToken()); - else if (cmd == _T("")) - comchar = token.GetNextToken(); - else if (cmd == _T("")) - escchar = token.GetNextToken(); - else if (cmd == _T("")) { - delete cset; - cset = (wxCharacterSet *) NULL; - break; // we don't support multibyte charsets ourselves (yet) - } - else if (cmd == _T("CHARMAP")) { - cset->data = (wchar_t *)calloc(256, sizeof(wchar_t)); - in_charset = TRUE; - } - else if (cmd == _T("END")) { - if (token.GetNextToken() == _T("CHARMAP")) - in_charset = FALSE; - } - else if (in_charset) { - // format: /x00 NULL (NUL) - // /x41 LATIN CAPITAL LETTER A - wxString hex = token.GetNextToken(); - // skip whitespace (why doesn't wxStringTokenizer do this?) - while (wxIsEmpty(hex) && token.HasMoreTokens()) hex = token.GetNextToken(); - wxString uni = token.GetNextToken(); - // skip whitespace again - while (wxIsEmpty(uni) && token.HasMoreTokens()) uni = token.GetNextToken(); - - if ((hex.GetChar(0) == escchar) && (hex.GetChar(1) == _T('x')) && - (uni.Left(2) == _T("=0) { - unsigned long uni1 = ::wxHexToDec(uni.Mid(2,2)); - unsigned long uni2 = ::wxHexToDec(uni.Mid(4,2)); - cset->data[pos] = (uni1 << 16) | uni2; - // wxFprintf(stderr,_T("char %02x mapped to %04x (%c)\n"),pos,cset->data[pos],cset->data[pos]); - } - } - } - } - if (cset) { - cset->names.Shrink(); - wxCharsets.Add(cset); - } + default: + if ( *pszMask != *pszTxt ) + return false; + break; } } -#endif - wxCharsets.Shrink(); - already_loaded = TRUE; -} -static wxCharacterSet *wxFindCharacterSet(const wxChar *charset) -{ - if (!charset) return (wxCharacterSet *)NULL; - wxLoadCharacterSets(); - for (size_t n=0; n4) { - if (wxString(charset,4) == _T("8859")) { - codeset << _T("8859-"); - if (*charset == _T('-')) charset++; - } - } - } - codeset << charset; - codeset.MakeUpper(); - m_name = wxStrdup(codeset.c_str()); - m_deferred = TRUE; -#endif + goto match; } -} -void wxCSConv::LoadNow() -{ -// wxPrintf(_T("Conversion request\n")); - if (m_deferred) { - if (!m_name) { -#ifdef __UNIX__ - wxChar *lang = wxGetenv(_T("LANG")); - wxChar *dot = lang ? wxStrchr(lang, _T('.')) : (wxChar *)NULL; - if (dot) SetName(dot+1); -#endif - } - m_cset = wxFindCharacterSet(m_name); - m_deferred = FALSE; - } + return false; +#endif // wxUSE_REGEX/!wxUSE_REGEX } -size_t wxCSConv::MB2WC(wchar_t *buf, const char *psz, size_t n) const +// Count the number of chars +int wxString::Freq(wxUniChar ch) const { - ((wxCSConv *)this)->LoadNow(); // discard constness - if (buf) { - if (m_cset) { - for (size_t c=0; cdata[psz[c]]; - } else { - // latin-1 (direct) - for (size_t c=0; cLoadNow(); // discard constness - if (buf) { - if (m_cset) { - for (size_t c=0; cdata[n] != psz[c]); n++); - buf[c] = (n>0xff) ? '?' : n; - } - } else { - // latin-1 (direct) - for (size_t c=0; c0xff) ? '?' : psz[c]; + int count = 0; + for ( const_iterator i = begin(); i != end(); ++i ) + { + if ( *i == ch ) + count ++; } - return n; - } - return wcslen(psz); + return count; } +