X-Git-Url: https://git.saurik.com/wxWidgets.git/blobdiff_plain/04abe4bc8fbddf68ecc9950e2e743e3746ab26b6..ba86da30cf1a2a8429f3af465cbb6c9f52b307fb:/src/common/string.cpp diff --git a/src/common/string.cpp b/src/common/string.cpp index 6cdcaef003..56af51ae22 100644 --- a/src/common/string.cpp +++ b/src/common/string.cpp @@ -10,13 +10,6 @@ // Licence: wxWindows licence ///////////////////////////////////////////////////////////////////////////// -/* - * About ref counting: - * 1) all empty strings use g_strEmpty, nRefs = -1 (set in Init()) - * 2) AllocBuffer() sets nRefs to 1, Lock() increments it by one - * 3) Unlock() decrements nRefs and frees memory if it goes to 0 - */ - // =========================================================================== // headers, declarations, constants // =========================================================================== @@ -30,6 +23,7 @@ #ifndef WX_PRECOMP #include "wx/string.h" + #include "wx/wxcrtvararg.h" #endif #include @@ -41,10 +35,7 @@ #include #include -#ifdef __SALFORDC__ - #include -#endif - +#include "wx/hashmap.h" // string handling functions used by wxString: #if wxUSE_UNICODE_UTF8 @@ -59,6 +50,18 @@ #define wxStringStrlen wxStrlen #endif +// ---------------------------------------------------------------------------- +// global variables +// ---------------------------------------------------------------------------- + +namespace wxPrivate +{ + +static UntypedBufferData s_untypedNullData(NULL); + +UntypedBufferData * const untypedNullDataPtr = &s_untypedNullData; + +} // namespace wxPrivate // --------------------------------------------------------------------------- // static class variables definition @@ -67,6 +70,105 @@ //According to STL _must_ be a -1 size_t const size_t wxString::npos = (size_t) -1; +#if wxUSE_STRING_POS_CACHE + +#ifdef wxHAS_COMPILER_TLS + +wxTLS_TYPE(wxString::Cache) wxString::ms_cache; + +#else // !wxHAS_COMPILER_TLS + +struct wxStrCacheInitializer +{ + wxStrCacheInitializer() + { + // calling this function triggers s_cache initialization in it, and + // from now on it becomes safe to call from multiple threads + wxString::GetCache(); + } +}; + +/* +wxString::Cache& wxString::GetCache() +{ + static wxTLS_TYPE(Cache) s_cache; + + return wxTLS_VALUE(s_cache); +} +*/ + +static wxStrCacheInitializer gs_stringCacheInit; + +#endif // wxHAS_COMPILER_TLS/!wxHAS_COMPILER_TLS + +// 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 +#ifdef __WXDEBUG__ + +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__ + +#ifdef wxPROFILE_STRING_CACHE + +wxString::CacheStats wxString::ms_cacheStats; + +struct wxStrCacheStatsDumper +{ + ~wxStrCacheStatsDumper() + { + const wxString::CacheStats& stats = wxString::ms_cacheStats; + + 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); + } + } +}; + +static wxStrCacheStatsDumper s_showCacheStats; + +#endif // wxPROFILE_STRING_CACHE + +#endif // wxUSE_STRING_POS_CACHE + // ---------------------------------------------------------------------------- // global functions // ---------------------------------------------------------------------------- @@ -77,11 +179,16 @@ const size_t wxString::npos = (size_t) -1; wxSTD ostream& operator<<(wxSTD ostream& os, const wxCStrData& str) { -// FIXME-UTF8: always, not only if wxUSE_UNICODE -#if wxUSE_UNICODE && !defined(__BORLANDC__) - return os << str.AsWChar(); +#if wxUSE_UNICODE && !wxUSE_UNICODE_UTF8 + const wxCharBuffer buf(str.AsCharBuf()); + if ( !buf ) + os.clear(wxSTD ios_base::failbit); + else + os << buf.data(); + + return os; #else - return os << str.AsChar(); + return os << str.AsInternal(); #endif } @@ -102,24 +209,261 @@ wxSTD ostream& operator<<(wxSTD ostream& os, const wxWCharBuffer& str) } #endif +#if wxUSE_UNICODE && defined(HAVE_WOSTREAM) + +wxSTD wostream& operator<<(wxSTD wostream& wos, const wxString& str) +{ + return wos << str.wc_str(); +} + +wxSTD wostream& operator<<(wxSTD wostream& wos, const wxCStrData& str) +{ + return wos << str.AsWChar(); +} + +wxSTD wostream& operator<<(wxSTD wostream& wos, const wxWCharBuffer& str) +{ + return wos << str.data(); +} + +#endif // wxUSE_UNICODE && defined(HAVE_WOSTREAM) + #endif // wxUSE_STD_IOSTREAM // =========================================================================== // wxString class core // =========================================================================== +#if wxUSE_UNICODE_UTF8 + +void wxString::PosLenToImpl(size_t pos, size_t len, + size_t *implPos, size_t *implLen) const +{ + if ( pos == npos ) + { + *implPos = npos; + } + else // 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(); + } + } +} + +#endif // wxUSE_UNICODE_UTF8 + +// ---------------------------------------------------------------------------- +// wxCStrData converted strings caching +// ---------------------------------------------------------------------------- + +// FIXME-UTF8: temporarily disabled because it doesn't work with global +// string objects; re-enable after fixing this bug and benchmarking +// performance to see if using a hash is a good idea at all +#if 0 + +// For backward compatibility reasons, it must be possible to assign the value +// returned by wxString::c_str() to a char* or wchar_t* variable and work with +// it. Returning wxCharBuffer from (const char*)c_str() wouldn't do the trick, +// because the memory would be freed immediately, but it has to be valid as long +// as the string is not modified, so that code like this still works: +// +// const wxChar *s = str.c_str(); +// while ( s ) { ... } + +// FIXME-UTF8: not thread safe! +// FIXME-UTF8: we currently clear the cached conversion only when the string is +// destroyed, but we should do it when the string is modified, to +// keep memory usage down +// FIXME-UTF8: we do the conversion every time As[W]Char() is called, but if we +// invalidated the cache on every change, we could keep the previous +// conversion +// FIXME-UTF8: add tracing of usage of these two methods - new code is supposed +// to use mb_str() or wc_str() instead of (const [w]char*)c_str() + +template +static inline void DeleteStringFromConversionCache(T& hash, const wxString *s) +{ + typename T::iterator i = hash.find(wxConstCast(s, wxString)); + if ( i != hash.end() ) + { + free(i->second); + hash.erase(i); + } +} + +#if wxUSE_UNICODE +// NB: non-STL implementation doesn't compile with "const wxString*" key type, +// so we have to use wxString* here and const-cast when used +WX_DECLARE_HASH_MAP(wxString*, char*, wxPointerHash, wxPointerEqual, + wxStringCharConversionCache); +static wxStringCharConversionCache gs_stringsCharCache; + +const char* wxCStrData::AsChar() const +{ + // remove previously cache value, if any (see FIXMEs above): + DeleteStringFromConversionCache(gs_stringsCharCache, m_str); + + // convert the string and keep it: + const char *s = gs_stringsCharCache[wxConstCast(m_str, wxString)] = + m_str->mb_str().release(); + + return s + m_offset; +} +#endif // wxUSE_UNICODE + +#if !wxUSE_UNICODE_WCHAR +WX_DECLARE_HASH_MAP(wxString*, wchar_t*, wxPointerHash, wxPointerEqual, + wxStringWCharConversionCache); +static wxStringWCharConversionCache gs_stringsWCharCache; + +const wchar_t* wxCStrData::AsWChar() const +{ + // remove previously cache value, if any (see FIXMEs above): + DeleteStringFromConversionCache(gs_stringsWCharCache, m_str); + + // convert the string and keep it: + const wchar_t *s = gs_stringsWCharCache[wxConstCast(m_str, wxString)] = + m_str->wc_str().release(); + + return s + m_offset; +} +#endif // !wxUSE_UNICODE_WCHAR + +wxString::~wxString() +{ +#if wxUSE_UNICODE + // FIXME-UTF8: do this only if locale is not UTF8 if wxUSE_UNICODE_UTF8 + DeleteStringFromConversionCache(gs_stringsCharCache, this); +#endif +#if !wxUSE_UNICODE_WCHAR + DeleteStringFromConversionCache(gs_stringsWCharCache, this); +#endif +} +#endif + +#if wxUSE_UNICODE && !wxUSE_UTF8_LOCALE_ONLY +const char* wxCStrData::AsChar() const +{ +#if wxUSE_UNICODE_UTF8 + if ( wxLocaleIsUtf8 ) + return AsInternal(); +#endif + // under non-UTF8 locales, we have to convert the internal UTF-8 + // representation using wxConvLibc and cache the result + + wxString *str = wxConstCast(m_str, wxString); + + // convert the string: + // + // FIXME-UTF8: we'd like to do the conversion in the existing buffer (if we + // have it) but it's unfortunately not obvious to implement + // because we don't know how big buffer do we need for the + // given string length (in case of multibyte encodings, e.g. + // ISO-2022-JP or UTF-8 when internal representation is wchar_t) + // + // One idea would be to store more than just m_convertedToChar + // in wxString: then we could record the length of the string + // which was converted the last time and try to reuse the same + // buffer if the current length is not greater than it (this + // could still fail because string could have been modified in + // place but it would work most of the time, so we'd do it and + // only allocate the new buffer if in-place conversion returned + // an error). We could also store a bit saying if the string + // was modified since the last conversion (and update it in all + // operation modifying the string, of course) to avoid unneeded + // consequential conversions. But both of these ideas require + // adding more fields to wxString and require profiling results + // to be sure that we really gain enough from them to justify + // doing it. + wxCharBuffer buf(str->mb_str()); + + // if it failed, return empty string and not NULL to avoid crashes in code + // written with either wxWidgets 2 wxString or std::string behaviour in + // mind: neither of them ever returns NULL and so we shouldn't neither + if ( !buf ) + return ""; + + if ( str->m_convertedToChar && + strlen(buf) == strlen(str->m_convertedToChar) ) + { + // keep the same buffer for as long as possible, so that several calls + // to c_str() in a row still work: + strcpy(str->m_convertedToChar, buf); + } + else + { + str->m_convertedToChar = buf.release(); + } + + // and keep it: + return str->m_convertedToChar + m_offset; +} +#endif // wxUSE_UNICODE && !wxUSE_UTF8_LOCALE_ONLY + +#if !wxUSE_UNICODE_WCHAR +const wchar_t* wxCStrData::AsWChar() const +{ + wxString *str = wxConstCast(m_str, wxString); + + // convert the string: + wxWCharBuffer buf(str->wc_str()); + + // notice that here, unlike above in AsChar(), conversion can't fail as our + // internal UTF-8 is always well-formed -- or the string was corrupted and + // all bets are off anyhow + + // FIXME-UTF8: do the conversion in-place in the existing buffer + if ( str->m_convertedToWChar && + wxWcslen(buf) == wxWcslen(str->m_convertedToWChar) ) + { + // keep the same buffer for as long as possible, so that several calls + // to c_str() in a row still work: + memcpy(str->m_convertedToWChar, buf, sizeof(wchar_t) * wxWcslen(buf)); + } + else + { + str->m_convertedToWChar = buf.release(); + } + + // and keep it: + return str->m_convertedToWChar + m_offset; +} +#endif // !wxUSE_UNICODE_WCHAR + +// =========================================================================== +// wxString class core +// =========================================================================== + // --------------------------------------------------------------------------- // construction and conversion // --------------------------------------------------------------------------- -#if wxUSE_UNICODE +#if wxUSE_UNICODE_WCHAR /* static */ wxString::SubstrBufFromMB wxString::ConvertStr(const char *psz, size_t nLength, const wxMBConv& conv) { // anything to do? if ( !psz || nLength == 0 ) - return SubstrBufFromMB(); + return SubstrBufFromMB(L"", 0); if ( nLength == npos ) nLength = wxNO_LEN; @@ -127,18 +471,63 @@ wxString::SubstrBufFromMB wxString::ConvertStr(const char *psz, size_t nLength, size_t wcLen; wxWCharBuffer wcBuf(conv.cMB2WC(psz, nLength, &wcLen)); if ( !wcLen ) - return SubstrBufFromMB(); + return SubstrBufFromMB(_T(""), 0); else return SubstrBufFromMB(wcBuf, wcLen); } -#else +#endif // wxUSE_UNICODE_WCHAR + +#if wxUSE_UNICODE_UTF8 +/* static */ +wxString::SubstrBufFromMB wxString::ConvertStr(const char *psz, size_t nLength, + const wxMBConv& conv) +{ + // anything to do? + if ( !psz || nLength == 0 ) + return SubstrBufFromMB("", 0); + + // if psz is already in UTF-8, we don't have to do the roundtrip to + // wchar_t* and back: + if ( conv.IsUTF8() ) + { + // we need to validate the input because UTF8 iterators assume valid + // UTF-8 sequence and psz may be invalid: + if ( wxStringOperations::IsValidUtf8String(psz, nLength) ) + { + // we must pass the real string length to SubstrBufFromMB ctor + if ( nLength == npos ) + nLength = psz ? strlen(psz) : 0; + return SubstrBufFromMB(wxCharBuffer::CreateNonOwned(psz), nLength); + } + // else: do the roundtrip through wchar_t* + } + + if ( nLength == npos ) + nLength = wxNO_LEN; + + // first convert to wide string: + size_t wcLen; + wxWCharBuffer wcBuf(conv.cMB2WC(psz, nLength, &wcLen)); + if ( !wcLen ) + return SubstrBufFromMB("", 0); + + // and then to UTF-8: + SubstrBufFromMB buf(ConvertStr(wcBuf, wcLen, wxMBConvStrictUTF8())); + // widechar -> UTF-8 conversion isn't supposed to ever fail: + wxASSERT_MSG( buf.data, _T("conversion to UTF-8 failed") ); + + return buf; +} +#endif // wxUSE_UNICODE_UTF8 + +#if wxUSE_UNICODE_UTF8 || !wxUSE_UNICODE /* static */ wxString::SubstrBufFromWC wxString::ConvertStr(const wchar_t *pwz, size_t nLength, const wxMBConv& conv) { // anything to do? if ( !pwz || nLength == 0 ) - return SubstrBufFromWC(); + return SubstrBufFromWC("", 0); if ( nLength == npos ) nLength = wxNO_LEN; @@ -146,34 +535,62 @@ wxString::SubstrBufFromWC wxString::ConvertStr(const wchar_t *pwz, size_t nLengt size_t mbLen; wxCharBuffer mbBuf(conv.cWC2MB(pwz, nLength, &mbLen)); if ( !mbLen ) - return SubstrBufFromWC(); + return SubstrBufFromWC("", 0); else return SubstrBufFromWC(mbBuf, mbLen); } -#endif +#endif // wxUSE_UNICODE_UTF8 || !wxUSE_UNICODE -#if wxUSE_UNICODE +#if wxUSE_UNICODE_WCHAR //Convert wxString in Unicode mode to a multi-byte string const wxCharBuffer wxString::mb_str(const wxMBConv& conv) const { - return conv.cWC2MB(c_str(), length() + 1 /* size, not length */, NULL); + return conv.cWC2MB(wx_str(), length() + 1 /* size, not length */, NULL); } -#else // ANSI +#elif wxUSE_UNICODE_UTF8 + +const wxWCharBuffer wxString::wc_str() const +{ + return wxMBConvStrictUTF8().cMB2WC + ( + m_impl.c_str(), + m_impl.length() + 1, // size, not length + NULL + ); +} + +const wxCharBuffer wxString::mb_str(const wxMBConv& conv) const +{ + if ( conv.IsUTF8() ) + return wxCharBuffer::CreateNonOwned(m_impl.c_str()); + + // FIXME-UTF8: use wc_str() here once we have buffers with length + + size_t wcLen; + wxWCharBuffer wcBuf(wxMBConvStrictUTF8().cMB2WC + ( + m_impl.c_str(), + m_impl.length() + 1, // size + &wcLen + )); + if ( !wcLen ) + return wxCharBuffer(""); + + return conv.cWC2MB(wcBuf, wcLen+1, NULL); +} -#if wxUSE_WCHAR_T +#else // ANSI //Converts this string to a wide character string if unicode //mode is not enabled and wxUSE_WCHAR_T is enabled const wxWCharBuffer wxString::wc_str(const wxMBConv& conv) const { - return conv.cMB2WC(c_str(), length() + 1 /* size, not length */, NULL); + return conv.cMB2WC(wx_str(), length() + 1 /* size, not length */, NULL); } -#endif // wxUSE_WCHAR_T - #endif // Unicode/ANSI // shrink to minimal size (releasing extra memory) @@ -186,7 +603,7 @@ bool wxString::Shrink() // deprecated compatibility code: #if WXWIN_COMPATIBILITY_2_8 && !wxUSE_STL_BASED_WXSTRING && !wxUSE_UNICODE_UTF8 -wxChar *wxString::GetWriteBuf(size_t nLen) +wxStringCharType *wxString::GetWriteBuf(size_t nLen) { return DoGetWriteBuf(nLen); } @@ -210,29 +627,9 @@ void wxString::UngetWriteBuf(size_t nLen) // all functions are inline in string.h // --------------------------------------------------------------------------- -// assignment operators +// concatenation operators // --------------------------------------------------------------------------- -#if !wxUSE_UNICODE - -// same as 'signed char' variant -wxString& wxString::operator=(const unsigned char* psz) -{ - *this = (const char *)psz; - return *this; -} - -#if wxUSE_WCHAR_T -wxString& wxString::operator=(const wchar_t *pwz) -{ - wxString str(pwz); - swap(str); - return *this; -} -#endif - -#endif - /* * concatenation functions come in 5 flavours: * string + string @@ -345,6 +742,12 @@ wxString operator+(const wchar_t *pwz, const wxString& str) // string comparison // --------------------------------------------------------------------------- +bool wxString::IsSameAs(wxUniChar c, bool compareWithCase) const +{ + return (length() == 1) && (compareWithCase ? GetChar(0u) == c + : wxToupper(GetChar(0u)) == wxToupper(c)); +} + #ifdef HAVE_STD_STRING_COMPARE // NB: Comparison code (both if HAVE_STD_STRING_COMPARE and if not) works with @@ -712,15 +1115,15 @@ size_t wxString::find_last_not_of(const wxOtherCharType* sz, size_t nStart, int wxString::CmpNoCase(const wxString& s) const { +#if wxUSE_UNICODE_UTF8 // FIXME-UTF8: use wxUniChar::ToLower/ToUpper once added - size_t idx = 0; const_iterator i1 = begin(); const_iterator end1 = end(); const_iterator i2 = s.begin(); const_iterator end2 = s.end(); - for ( ; i1 != end1 && i2 != end2; ++idx, ++i1, ++i2 ) + for ( ; i1 != end1 && i2 != end2; ++i1, ++i2 ) { wxUniChar lower1 = (wxChar)wxTolower(*i1); wxUniChar lower2 = (wxChar)wxTolower(*i2); @@ -736,6 +1139,9 @@ int wxString::CmpNoCase(const wxString& s) const else if ( len1 > len2 ) return 1; return 0; +#else // wxUSE_UNICODE_WCHAR or ANSI + return wxStricmp(m_impl.c_str(), s.m_impl.c_str()); +#endif } @@ -747,63 +1153,69 @@ int wxString::CmpNoCase(const wxString& s) const #endif #endif -wxString wxString::FromAscii(const char *ascii) +wxString wxString::FromAscii(const char *ascii, size_t len) { - if (!ascii) + if (!ascii || len == 0) return wxEmptyString; - size_t len = strlen( ascii ); wxString res; - if ( len ) { - wxStringBuffer buf(res, len); - - wchar_t *dest = buf; + wxStringInternalBuffer buf(res, len); + wxStringCharType *dest = buf; - for ( ;; ) + for ( ; len > 0; --len ) { - if ( (*dest++ = (wchar_t)(unsigned char)*ascii++) == L'\0' ) - break; + unsigned char c = (unsigned char)*ascii++; + wxASSERT_MSG( c < 0x80, + _T("Non-ASCII value passed to FromAscii().") ); + + *dest++ = (wchar_t)c; } } return res; } -wxString wxString::FromAscii(const char ascii) +wxString wxString::FromAscii(const char *ascii) +{ + return FromAscii(ascii, wxStrlen(ascii)); +} + +wxString wxString::FromAscii(char ascii) { // What do we do with '\0' ? - wxString res; - res += (wchar_t)(unsigned char) ascii; + unsigned char c = (unsigned char)ascii; - return res; + wxASSERT_MSG( c < 0x80, _T("Non-ASCII value passed to FromAscii().") ); + + // NB: the cast to wchar_t causes interpretation of 'ascii' as Latin1 value + return wxString(wxUniChar((wchar_t)c)); } const wxCharBuffer wxString::ToAscii() const { // this will allocate enough space for the terminating NUL too wxCharBuffer buffer(length()); - - char *dest = buffer.data(); - const wchar_t *pwc = c_str(); - for ( ;; ) + for ( const_iterator i = begin(); i != end(); ++i ) { - *dest++ = (char)(*pwc > SCHAR_MAX ? wxT('_') : *pwc); + 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 ( !*pwc++ ) + if ( !c ) break; } return buffer; } -#endif // Unicode +#endif // wxUSE_UNICODE // extract string of length nCount starting at nFirst wxString wxString::Mid(size_t nFirst, size_t nCount) const @@ -839,28 +1251,15 @@ wxString wxString::Mid(size_t nFirst, size_t nCount) const // check that the string starts with prefix and return the rest of the string // in the provided pointer if it is not NULL, otherwise return false -bool wxString::StartsWith(const wxChar *prefix, wxString *rest) const +bool wxString::StartsWith(const wxString& prefix, wxString *rest) const { - wxASSERT_MSG( prefix, _T("invalid parameter in wxString::StartsWith") ); - - // first check if the beginning of the string matches the prefix: note - // that we don't have to check that we don't run out of this string as - // when we reach the terminating NUL, either prefix string ends too (and - // then it's ok) or we break out of the loop because there is no match - const wxChar *p = c_str(); - while ( *prefix ) - { - if ( *prefix++ != *p++ ) - { - // no match - return false; - } - } + if ( compare(0, prefix.length(), prefix) != 0 ) + return false; if ( rest ) { // put the rest of the string into provided pointer - *rest = p; + rest->assign(*this, prefix.length(), npos); } return true; @@ -869,12 +1268,11 @@ bool wxString::StartsWith(const wxChar *prefix, wxString *rest) const // check that the string ends with suffix and return the rest of it in the // provided pointer if it is not NULL, otherwise return false -bool wxString::EndsWith(const wxChar *suffix, wxString *rest) const +bool wxString::EndsWith(const wxString& suffix, wxString *rest) const { - wxASSERT_MSG( suffix, _T("invalid parameter in wxString::EndssWith") ); + int start = length() - suffix.length(); - int start = length() - wxStrlen(suffix); - if ( start < 0 || wxStrcmp(wx_str() + start, suffix) != 0 ) + if ( start < 0 || compare(start, npos, suffix) != 0 ) return false; if ( rest ) @@ -900,7 +1298,7 @@ wxString wxString::Right(size_t nCount) const return dest; } -// get all characters after the last occurence of ch +// get all characters after the last occurrence of ch // (returns the whole string if ch not found) wxString wxString::AfterLast(wxUniChar ch) const { @@ -909,7 +1307,7 @@ wxString wxString::AfterLast(wxUniChar ch) const if ( iPos == wxNOT_FOUND ) str = *this; else - str = wx_str() + iPos + 1; + str.assign(*this, iPos + 1, npos); return str; } @@ -927,16 +1325,17 @@ wxString wxString::Left(size_t nCount) const return dest; } -// get all characters before the first occurence of ch +// get all characters before the first occurrence of ch // (returns the whole string if ch not found) wxString wxString::BeforeFirst(wxUniChar ch) const { int iPos = Find(ch); - if ( iPos == wxNOT_FOUND ) iPos = length(); + if ( iPos == wxNOT_FOUND ) + iPos = length(); return wxString(*this, 0, iPos); } -/// get all characters before the last occurence of ch +/// get all characters before the last occurrence of ch /// (returns empty string if ch not found) wxString wxString::BeforeLast(wxUniChar ch) const { @@ -948,55 +1347,79 @@ wxString wxString::BeforeLast(wxUniChar ch) const return str; } -/// get all characters after the first occurence of ch +/// get all characters after the first occurrence of ch /// (returns empty string if ch not found) wxString wxString::AfterFirst(wxUniChar ch) const { wxString str; int iPos = Find(ch); if ( iPos != wxNOT_FOUND ) - str = wx_str() + iPos + 1; + str.assign(*this, iPos + 1, npos); return str; } -// replace first (or all) occurences of some substring with another one -size_t wxString::Replace(const wxChar *szOld, - const wxChar *szNew, bool bReplaceAll) +// 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( szOld && *szOld && szNew, 0, + wxCHECK_MSG( !strOld.empty(), 0, _T("wxString::Replace(): invalid parameter") ); + wxSTRING_INVALIDATE_CACHE(); + size_t uiCount = 0; // count of replacements made - size_t uiOldLen = wxStrlen(szOld); - size_t uiNewLen = wxStrlen(szNew); + // 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 ) + { + const wxStringCharType chOld = strOld.m_impl[0], + chNew = strNew.m_impl[0]; - size_t dwPos = 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; - while ( this->c_str()[dwPos] != wxT('\0') ) + m_impl[pos++] = chNew; + + uiCount++; + + if ( !bReplaceAll ) + break; + } + } + else // general case { - //DO NOT USE STRSTR HERE - //this string can contain embedded null characters, - //so strstr will function incorrectly - dwPos = find(szOld, dwPos); - if ( dwPos == npos ) - break; // exit the loop - else + const size_t uiOldLen = strOld.m_impl.length(); + const size_t uiNewLen = strNew.m_impl.length(); + + for ( size_t pos = 0; ; ) { - //replace this occurance of the old string with the new one - replace(dwPos, uiOldLen, szNew, uiNewLen); + pos = m_impl.find(strOld.m_impl, pos); + if ( pos == npos ) + break; + + // replace this occurrence of the old string with the new one + m_impl.replace(pos, uiOldLen, strNew.m_impl); - //move up pos past the string that was replaced - dwPos += uiNewLen; + // move up pos past the string that was replaced + pos += uiNewLen; - //increase replace count - ++uiCount; + // increase replace count + uiCount++; - // stop now? + // stop after the first one? if ( !bReplaceAll ) - break; // exit the loop + break; } } @@ -1005,34 +1428,43 @@ size_t wxString::Replace(const wxChar *szOld, bool wxString::IsAscii() const { - const wxChar *s = (const wxChar*) *this; - while(*s){ - if(!isascii(*s)) return(false); - s++; - } - return(true); + for ( const_iterator i = begin(); i != end(); ++i ) + { + if ( !(*i).IsAscii() ) + return false; + } + + return true; } bool wxString::IsWord() const { - const wxChar *s = (const wxChar*) *this; - while(*s){ - if(!wxIsalpha(*s)) return(false); - s++; - } - return(true); + for ( const_iterator i = begin(); i != end(); ++i ) + { + if ( !wxIsalpha(*i) ) + return false; + } + + return true; } bool wxString::IsNumber() const { - const wxChar *s = (const wxChar*) *this; - if (wxStrlen(s)) - if ((s[0] == wxT('-')) || (s[0] == wxT('+'))) s++; - while(*s){ - if(!wxIsdigit(*s)) return(false); - s++; - } - return(true); + if ( empty() ) + return true; + + const_iterator i = begin(); + + if ( *i == _T('-') || *i == _T('+') ) + ++i; + + for ( ; i != end(); ++i ) + { + if ( !wxIsdigit(*i) ) + return false; + } + + return true; } wxString wxString::Strip(stripType w) const @@ -1063,14 +1495,28 @@ wxString& wxString::MakeLower() return *this; } +wxString& wxString::MakeCapitalized() +{ + const iterator en = end(); + iterator it = begin(); + if ( it != en ) + { + *it = (wxChar)wxToupper(*it); + for ( ++it; it != en; ++it ) + *it = (wxChar)wxTolower(*it); + } + + return *this; +} + // --------------------------------------------------------------------------- // trimming and padding // --------------------------------------------------------------------------- // some compilers (VC++ 6.0 not to name them) return true for a call to -// isspace('ê') 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 +// 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); } @@ -1091,7 +1537,7 @@ wxString& wxString::Trim(bool bFromRight) // find last non-space character reverse_iterator psz = rbegin(); while ( (psz != rend()) && wxSafeIsspace(*psz) ) - psz++; + ++psz; // truncate at trailing space start erase(psz.base(), end()); @@ -1101,7 +1547,7 @@ wxString& wxString::Trim(bool bFromRight) // find first non-space character iterator psz = begin(); while ( (psz != end()) && wxSafeIsspace(*psz) ) - psz++; + ++psz; // fix up data and length erase(begin(), psz); @@ -1151,110 +1597,88 @@ int wxString::Find(wxUniChar ch, bool bFromEnd) const return (idx == npos) ? wxNOT_FOUND : (int)idx; } -// find a sub-string (like strstr) -int wxString::Find(const wxChar *pszSub) const -{ - size_type idx = find(pszSub); - - return (idx == npos) ? wxNOT_FOUND : (int)idx; -} - // ---------------------------------------------------------------------------- // conversion to numbers // ---------------------------------------------------------------------------- -// the implementation of all the functions below is exactly the same so factor -// it out - -template -bool wxStringToIntType(const wxChar *start, - T *val, - int base, - F func) -{ - wxCHECK_MSG( val, false, _T("NULL output pointer") ); - wxASSERT_MSG( !base || (base > 1 && base <= 36), _T("invalid base") ); +// 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. #ifndef __WXWINCE__ - errno = 0; + #define DO_IF_NOT_WINCE(x) x +#else + #define DO_IF_NOT_WINCE(x) #endif - wxChar *end; - *val = (*func)(start, &end, base); - - // return true only if scan was stopped by the terminating NUL and if the - // string was not empty to start with and no under/overflow occurred - return !*end && (end != start) -#ifndef __WXWINCE__ - && (errno != ERANGE) -#endif - ; -} +#define WX_STRING_TO_INT_TYPE(out, base, func, T) \ + wxCHECK_MSG( out, false, _T("NULL output pointer") ); \ + wxASSERT_MSG( !base || (base > 1 && base <= 36), _T("invalid base") ); \ + \ + DO_IF_NOT_WINCE( errno = 0; ) \ + \ + const wxStringCharType *start = wx_str(); \ + wxStringCharType *end; \ + T val = func(start, &end, base); \ + \ + /* return true only if scan was stopped by the terminating NUL and */ \ + /* if the string was not empty to start with and no under/overflow */ \ + /* occurred: */ \ + if ( *end || end == start DO_IF_NOT_WINCE(|| errno == ERANGE) ) \ + return false; \ + *out = val; \ + return true -bool wxString::ToLong(long *val, int base) const +bool wxString::ToLong(long *pVal, int base) const { - return wxStringToIntType((const wxChar*)c_str(), val, base, wxStrtol); + WX_STRING_TO_INT_TYPE(pVal, base, wxStrtol, long); } -bool wxString::ToULong(unsigned long *val, int base) const +bool wxString::ToULong(unsigned long *pVal, int base) const { - return wxStringToIntType((const wxChar*)c_str(), val, base, wxStrtoul); + WX_STRING_TO_INT_TYPE(pVal, base, wxStrtoul, unsigned long); } -bool wxString::ToLongLong(wxLongLong_t *val, int base) const +bool wxString::ToLongLong(wxLongLong_t *pVal, int base) const { -#ifdef wxHAS_STRTOLL - return wxStringToIntType((const wxChar*)c_str(), val, base, wxStrtoll); -#else - // TODO: implement this ourselves - wxUnusedVar(val); - wxUnusedVar(base); - return false; -#endif // wxHAS_STRTOLL + WX_STRING_TO_INT_TYPE(pVal, base, wxStrtoll, wxLongLong_t); } -bool wxString::ToULongLong(wxULongLong_t *val, int base) const +bool wxString::ToULongLong(wxULongLong_t *pVal, int base) const { -#ifdef wxHAS_STRTOLL - return wxStringToIntType((const wxChar*)c_str(), val, base, wxStrtoull); -#else - // TODO: implement this ourselves - wxUnusedVar(val); - wxUnusedVar(base); - return false; -#endif + WX_STRING_TO_INT_TYPE(pVal, base, wxStrtoull, wxULongLong_t); } -bool wxString::ToDouble(double *val) const +bool wxString::ToDouble(double *pVal) const { - wxCHECK_MSG( val, false, _T("NULL pointer in wxString::ToDouble") ); + wxCHECK_MSG( pVal, false, _T("NULL output pointer") ); -#ifndef __WXWINCE__ - errno = 0; -#endif + DO_IF_NOT_WINCE( errno = 0; ) const wxChar *start = c_str(); wxChar *end; - *val = wxStrtod(start, &end); + double val = wxStrtod(start, &end); // return true only if scan was stopped by the terminating NUL and if the // string was not empty to start with and no under/overflow occurred - return !*end && (end != start) -#ifndef __WXWINCE__ - && (errno != ERANGE) -#endif - ; + if ( *end || end == start DO_IF_NOT_WINCE(|| errno == ERANGE) ) + return false; + + *pVal = val; + + return true; } // --------------------------------------------------------------------------- // formatted output // --------------------------------------------------------------------------- +#if !wxUSE_UTF8_LOCALE_ONLY /* static */ #ifdef wxNEEDS_WXSTRING_PRINTF_MIXIN -wxString wxStringPrintfMixinBase::DoFormat(const wxChar *format, ...) +wxString wxStringPrintfMixinBase::DoFormatWchar(const wxChar *format, ...) #else -wxString wxString::DoFormat(const wxChar *format, ...) +wxString wxString::DoFormatWchar(const wxChar *format, ...) #endif { va_list argptr; @@ -1267,6 +1691,23 @@ wxString wxString::DoFormat(const wxChar *format, ...) return s; } +#endif // !wxUSE_UTF8_LOCALE_ONLY + +#if wxUSE_UNICODE_UTF8 +/* static */ +wxString wxString::DoFormatUtf8(const char *format, ...) +{ + va_list argptr; + va_start(argptr, format); + + wxString s; + s.PrintfV(format, argptr); + + va_end(argptr); + + return s; +} +#endif // wxUSE_UNICODE_UTF8 /* static */ wxString wxString::FormatV(const wxString& format, va_list argptr) @@ -1276,10 +1717,11 @@ wxString wxString::FormatV(const wxString& format, va_list argptr) return s; } +#if !wxUSE_UTF8_LOCALE_ONLY #ifdef wxNEEDS_WXSTRING_PRINTF_MIXIN -int wxStringPrintfMixinBase::DoPrintf(const wxChar *format, ...) +int wxStringPrintfMixinBase::DoPrintfWchar(const wxChar *format, ...) #else -int wxString::DoPrintf(const wxChar *format, ...) +int wxString::DoPrintfWchar(const wxChar *format, ...) #endif { va_list argptr; @@ -1300,19 +1742,106 @@ int wxString::DoPrintf(const wxChar *format, ...) return iLen; } +#endif // !wxUSE_UTF8_LOCALE_ONLY -int wxString::PrintfV(const wxString& format, va_list argptr) +#if wxUSE_UNICODE_UTF8 +int wxString::DoPrintfUtf8(const char *format, ...) +{ + va_list argptr; + va_start(argptr, format); + + int iLen = PrintfV(format, argptr); + + va_end(argptr); + + return iLen; +} +#endif // wxUSE_UNICODE_UTF8 + +/* + 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 behavior which + is to keep looping and increasing buffer size until the function succeeds. + + In practice it's impossible to determine before compilation which behavior + may be used. The vswprintf function may have vsnprintf-like behavior or + vice-versa. Behavior 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 behavior 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 behavior exactly. + + Glibc 2.6 almost follows the C99 behavior 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) { int size = 1024; for ( ;; ) { - wxStringBuffer tmp(*this, size + 1); +#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 ( !buf ) { // out of memory + + // in UTF-8 build, leaving uninitialized junk in the buffer + // could result in invalid non-empty UTF-8 string, so just + // reset the string to empty on failure: + buf[0] = '\0'; return -1; } @@ -1320,12 +1849,19 @@ int wxString::PrintfV(const wxString& format, va_list argptr) // only a copy va_list argptrcopy; wxVaCopy(argptrcopy, argptr); + +#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] = _T('\0'); // vsnprintf() may return either -1 (traditional Unix behaviour) or the @@ -1333,27 +1869,47 @@ int wxString::PrintfV(const wxString& format, va_list argptr) // 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 // assume that system version only returns error if not enough space +#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; + size *= 2; #endif // wxUSE_WXVSNPRINTF/!wxUSE_WXVSNPRINTF } else if ( len >= size ) { #if wxUSE_WXVSNPRINTF - // we know that our own implementation of wxVsnprintf() returns + // 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 } @@ -1364,9 +1920,34 @@ int wxString::PrintfV(const wxString& format, va_list argptr) } // we could have overshot - Shrink(); + str.Shrink(); - return length(); + return str.length(); +} + +int wxString::PrintfV(const wxString& format, va_list argptr) +{ +#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 } // ---------------------------------------------------------------------------- @@ -1376,7 +1957,7 @@ int wxString::PrintfV(const wxString& format, va_list argptr) // 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 wxChar *pszMask) const +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 @@ -1427,8 +2008,17 @@ bool wxString::Matches(const wxChar *pszMask) const #else // !wxUSE_REGEX // TODO: this is, of course, awfully inefficient... + // FIXME-UTF8: implement using iterators, remove #if +#if wxUSE_UNICODE_UTF8 + wxWCharBuffer maskBuf = mask.wc_str(); + wxWCharBuffer 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 = c_str(); + const wxChar *pszTxt = wx_str(); +#endif // the last location where '*' matched const wxChar *pszLastStarInText = NULL; @@ -1523,10 +2113,3 @@ int wxString::Freq(wxUniChar ch) const return count; } -// convert to upper case, return the copy of the string -wxString wxString::Upper() const -{ wxString s(*this); return s.MakeUpper(); } - -// convert to lower case, return the copy of the string -wxString wxString::Lower() const { wxString s(*this); return s.MakeLower(); } -