X-Git-Url: https://git.saurik.com/wxWidgets.git/blobdiff_plain/681e44128670bd58c09ff9b42af52666635b19a6..57d3e26697255a5c5154b4a0dc26d6eab23ff871:/src/common/string.cpp diff --git a/src/common/string.cpp b/src/common/string.cpp index a398b215e5..2978dcdca8 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 +23,9 @@ #ifndef WX_PRECOMP #include "wx/string.h" + #include "wx/wxcrtvararg.h" + #include "wx/intl.h" + #include "wx/log.h" #endif #include @@ -41,11 +37,17 @@ #include #include -#ifdef __SALFORDC__ - #include -#endif - #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 // string handling functions used by wxString: #if wxUSE_UNICODE_UTF8 @@ -60,6 +62,22 @@ #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); + + return &s_untypedNullData; +} + +} // namespace wxPrivate // --------------------------------------------------------------------------- // static class variables definition @@ -68,457 +86,200 @@ //According to STL _must_ be a -1 size_t const size_t wxString::npos = (size_t) -1; -// ---------------------------------------------------------------------------- -// global functions -// ---------------------------------------------------------------------------- +#if wxUSE_STRING_POS_CACHE -#if wxUSE_STD_IOSTREAM +#ifdef wxHAS_COMPILER_TLS -#include +wxTLS_TYPE(wxString::Cache) wxString::ms_cache; -wxSTD ostream& operator<<(wxSTD ostream& os, const wxCStrData& str) -{ -// FIXME-UTF8: always, not only if wxUSE_UNICODE -#if wxUSE_UNICODE && !defined(__BORLANDC__) - return os << (const wchar_t*)str.AsWCharBuf(); -#else - return os << (const char*)str.AsCharBuf(); -#endif -} +#else // !wxHAS_COMPILER_TLS -wxSTD ostream& operator<<(wxSTD ostream& os, const wxString& str) +struct wxStrCacheInitializer { - return os << str.c_str(); -} + wxStrCacheInitializer() + { + // calling this function triggers s_cache initialization in it, and + // from now on it becomes safe to call from multiple threads + wxString::GetCache(); + } +}; -wxSTD ostream& operator<<(wxSTD ostream& os, const wxCharBuffer& str) +/* +wxString::Cache& wxString::GetCache() { - return os << str.data(); -} + static wxTLS_TYPE(Cache) s_cache; -#ifndef __BORLANDC__ -wxSTD ostream& operator<<(wxSTD ostream& os, const wxWCharBuffer& str) -{ - return os << str.data(); + return wxTLS_VALUE(s_cache); } -#endif +*/ -#endif // wxUSE_STD_IOSTREAM +static wxStrCacheInitializer gs_stringCacheInit; -// =========================================================================== -// wxString class core -// =========================================================================== +#endif // wxHAS_COMPILER_TLS/!wxHAS_COMPILER_TLS -#if wxUSE_UNICODE_UTF8 +// 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 -// --------------------------------------------------------------------------- -// UTF-8 operations -// --------------------------------------------------------------------------- - -// -// Table 3.1B from Unicode spec: Legal UTF-8 Byte Sequences -// -// Code Points | 1st Byte | 2nd Byte | 3rd Byte | 4th Byte | -// -------------------+----------+----------+----------+----------+ -// U+0000..U+007F | 00..7F | | | | -// U+0080..U+07FF | C2..DF | 80..BF | | | -// U+0800..U+0FFF | E0 | A0..BF | 80..BF | | -// U+1000..U+FFFF | E1..EF | 80..BF | 80..BF | | -// U+10000..U+3FFFF | F0 | 90..BF | 80..BF | 80..BF | -// U+40000..U+FFFFF | F1..F3 | 80..BF | 80..BF | 80..BF | -// U+100000..U+10FFFF | F4 | 80..8F | 80..BF | 80..BF | -// -------------------+----------+----------+----------+----------+ - -bool wxString::IsValidUtf8String(const char *str) +struct wxStrCacheDumper { - if ( !str ) - return true; // empty string is UTF8 string + 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); + } + } +}; - const unsigned char *c = (const unsigned char*)str; +void wxDumpStrCache() { wxStrCacheDumper::ShowAll(); } - for ( ; *c; ++c ) - { - unsigned char b = *c; +#endif // wxDEBUG_LEVEL >= 2 - if ( b <= 0x7F ) // 00..7F - continue; +#ifdef wxPROFILE_STRING_CACHE - else if ( b < 0xC2 ) // invalid lead bytes: 80..C1 - return false; +wxString::CacheStats wxString::ms_cacheStats; - // two-byte sequences: - else if ( b <= 0xDF ) // C2..DF - { - b = *(++c); - if ( !(b >= 0x80 && b <= 0xBF ) ) - return false; - } +struct wxStrCacheStatsDumper +{ + ~wxStrCacheStatsDumper() + { + const wxString::CacheStats& stats = wxString::ms_cacheStats; - // three-byte sequences: - else if ( b == 0xE0 ) - { - b = *(++c); - if ( !(b >= 0xA0 && b <= 0xBF ) ) - return false; - b = *(++c); - if ( !(b >= 0x80 && b <= 0xBF ) ) - return false; - } - else if ( b <= 0xEF ) // E1..EF + if ( stats.postot ) { - for ( int i = 0; i < 2; ++i ) - { - b = *(++c); - if ( !(b >= 0x80 && b <= 0xBF ) ) - return false; - } + 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); } - // four-byte sequences: - else if ( b == 0xF0 ) + if ( stats.lentot ) { - b = *(++c); - if ( !(b >= 0x90 && b <= 0xBF ) ) - return false; - for ( int i = 0; i < 2; ++i ) - { - b = *(++c); - if ( !(b >= 0x80 && b <= 0xBF ) ) - return false; - } + printf("\tNumber of calls to length(): %u, hits=%.2f%%\n", + stats.lentot, 100.*float(stats.lenhits)/stats.lentot); } - else if ( b <= 0xF3 ) // F1..F3 - { - for ( int i = 0; i < 3; ++i ) - { - b = *(++c); - if ( !(b >= 0x80 && b <= 0xBF ) ) - return false; - } - } - else if ( b == 0xF4 ) - { - b = *(++c); - if ( !(b >= 0x80 && b <= 0x8F ) ) - return false; - for ( int i = 0; i < 2; ++i ) - { - b = *(++c); - if ( !(b >= 0x80 && b <= 0xBF ) ) - return false; - } - } - else // otherwise, it's invalid lead byte - return false; } +}; - return true; -} +static wxStrCacheStatsDumper s_showCacheStats; -#ifdef __WXDEBUG__ -/* static */ -bool wxString::IsValidUtf8LeadByte(unsigned char c) -{ - return (c <= 0x7F) || (c >= 0xC2 && c <= 0xF4); -} -#endif +#endif // wxPROFILE_STRING_CACHE -unsigned char wxString::ms_utf8IterTable[256] = { - // single-byte sequences (ASCII): - 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, // 00..0F - 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, // 10..1F - 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, // 20..2F - 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, // 30..3F - 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, // 40..4F - 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, // 50..5F - 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, // 60..6F - 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, // 70..7F - - // these are invalid, we use step 1 to skip - // over them (should never happen): - 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, // 80..8F - 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, // 90..9F - 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, // A0..AF - 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, // B0..BF - 1, 1, // C0,C1 - - // two-byte sequences: - 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, // C2..CF - 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, // D0..DF - - // three-byte sequences: - 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, // E0..EF - - // four-byte sequences: - 4, 4, 4, 4, 4, // F0..F4 - - // these are invalid again (5- or 6-byte - // sequences and sequences for code points - // above U+10FFFF, as restricted by RFC 3629): - 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1 // F5..FF -}; +#endif // wxUSE_STRING_POS_CACHE -/* static */ -void wxString::DecIter(wxStringImpl::const_iterator& i) -{ - wxASSERT( IsValidUtf8LeadByte(*i) ); +// ---------------------------------------------------------------------------- +// global functions +// ---------------------------------------------------------------------------- - // Non-lead bytes are all in the 0x80..0xBF range (i.e. 10xxxxxx in - // binary), so we just have to go back until we hit a byte that is either - // < 0x80 (i.e. 0xxxxxxx in binary) or 0xC0..0xFF (11xxxxxx in binary; this - // includes some invalid values, but we can ignore it here, because we - // assume valid UTF-8 input for the purpose of efficient implementation). - --i; - while ( ((*i) & 0xC0) == 0x80 /* 2 highest bits are '10' */ ) - --i; -} +#if wxUSE_STD_IOSTREAM -/* static */ -void wxString::DecIter(wxStringImpl::iterator& i) -{ - // FIXME-UTF8: use template instead - wxASSERT( IsValidUtf8LeadByte(*i) ); - --i; - while ( ((*i) & 0xC0) == 0x80 /* 2 highest bits are '10' */ ) - --i; -} +#include -/* static */ -wxStringImpl::const_iterator -wxString::AddToIter(wxStringImpl::const_iterator i, int n) +wxSTD ostream& operator<<(wxSTD ostream& os, const wxCStrData& str) { - wxStringImpl::const_iterator out(i); - - if ( n > 0 ) - { - for ( int j = 0; j < n; ++j ) - IncIter(out); - } - else if ( n < 0 ) - { - for ( int j = 0; j > n; --j ) - DecIter(out); - } +#if wxUSE_UNICODE && !wxUSE_UNICODE_UTF8 + const wxScopedCharBuffer buf(str.AsCharBuf()); + if ( !buf ) + os.clear(wxSTD ios_base::failbit); + else + os << buf.data(); - return out; + return os; +#else + return os << str.AsInternal(); +#endif } -wxStringImpl::iterator -wxString::AddToIter(wxStringImpl::iterator i, int n) +wxSTD ostream& operator<<(wxSTD ostream& os, const wxString& str) { - // FIXME-UTF8: use template instead - wxStringImpl::iterator out(i); - - if ( n > 0 ) - { - for ( int j = 0; j < n; ++j ) - IncIter(out); - } - else if ( n < 0 ) - { - for ( int j = 0; j > n; --j ) - DecIter(out); - } - - return out; + return os << str.c_str(); } - -/* static */ -int wxString::DiffIters(wxStringImpl::const_iterator i1, - wxStringImpl::const_iterator i2) +wxSTD ostream& operator<<(wxSTD ostream& os, const wxScopedCharBuffer& str) { - int dist = 0; - - if ( i1 < i2 ) - { - while ( i1 != i2 ) - { - IncIter(i1); - dist--; - } - } - else if ( i2 < i1 ) - { - while ( i2 != i1 ) - { - IncIter(i2); - dist++; - } - } - - return dist; + return os << str.data(); } -int wxString::DiffIters(wxStringImpl::iterator i1, wxStringImpl::iterator i2) +#ifndef __BORLANDC__ +wxSTD ostream& operator<<(wxSTD ostream& os, const wxScopedWCharBuffer& str) { - // FIXME-UTF8: use template instead - int dist = 0; + return os << str.data(); +} +#endif - if ( i1 < i2 ) - { - while ( i1 != i2 ) - { - IncIter(i1); - dist--; - } - } - else if ( i2 < i1 ) - { - while ( i2 != i1 ) - { - IncIter(i2); - dist++; - } - } +#if wxUSE_UNICODE && defined(HAVE_WOSTREAM) - return dist; +wxSTD wostream& operator<<(wxSTD wostream& wos, const wxString& str) +{ + return wos << str.wc_str(); } -/* static */ -wxString::Utf8CharBuffer wxString::EncodeChar(wxUniChar ch) +wxSTD wostream& operator<<(wxSTD wostream& wos, const wxCStrData& str) { - Utf8CharBuffer buf; - char *out = buf.data; - - wxUniChar::value_type code = ch.GetValue(); - - // Char. number range | UTF-8 octet sequence - // (hexadecimal) | (binary) - // ----------------------+--------------------------------------------- - // 0000 0000 - 0000 007F | 0xxxxxxx - // 0000 0080 - 0000 07FF | 110xxxxx 10xxxxxx - // 0000 0800 - 0000 FFFF | 1110xxxx 10xxxxxx 10xxxxxx - // 0001 0000 - 0010 FFFF | 11110xxx 10xxxxxx 10xxxxxx 10xxxxxx - // - // Code point value is stored in bits marked with 'x', lowest-order bit - // of the value on the right side in the diagram above. - // (from RFC 3629) - - if ( code <= 0x7F ) - { - out[1] = 0; - out[0] = (char)code; - } - else if ( code <= 0x07FF ) - { - out[2] = 0; - // NB: this line takes 6 least significant bits, encodes them as - // 10xxxxxx and discards them so that the next byte can be encoded: - out[1] = 0x80 | (code & 0x3F); code >>= 6; - out[0] = 0xC0 | code; - } - else if ( code < 0xFFFF ) - { - out[3] = 0; - out[2] = 0x80 | (code & 0x3F); code >>= 6; - out[1] = 0x80 | (code & 0x3F); code >>= 6; - out[0] = 0xE0 | code; - } - else if ( code <= 0x10FFFF ) - { - out[4] = 0; - out[3] = 0x80 | (code & 0x3F); code >>= 6; - out[2] = 0x80 | (code & 0x3F); code >>= 6; - out[1] = 0x80 | (code & 0x3F); code >>= 6; - out[0] = 0xF0 | code; - } - else - { - wxFAIL_MSG( _T("trying to encode undefined Unicode character") ); - out[0] = 0; - } - - return buf; + return wos << str.AsWChar(); } -/* static */ -wxUniChar wxUniCharRef::DecodeChar(wxStringImpl::const_iterator i) +wxSTD wostream& operator<<(wxSTD wostream& wos, const wxScopedWCharBuffer& str) { - wxASSERT( wxString::IsValidUtf8LeadByte(*i) ); // FIXME-UTF8: no "wxString::" - - wxUniChar::value_type code = 0; - size_t len = wxString::GetUtf8CharLength(*i); - wxASSERT_MSG( len <= 4, _T("invalid UTF-8 sequence length") ); - - // Char. number range | UTF-8 octet sequence - // (hexadecimal) | (binary) - // ----------------------+--------------------------------------------- - // 0000 0000 - 0000 007F | 0xxxxxxx - // 0000 0080 - 0000 07FF | 110xxxxx 10xxxxxx - // 0000 0800 - 0000 FFFF | 1110xxxx 10xxxxxx 10xxxxxx - // 0001 0000 - 0010 FFFF | 11110xxx 10xxxxxx 10xxxxxx 10xxxxxx - // - // Code point value is stored in bits marked with 'x', lowest-order bit - // of the value on the right side in the diagram above. - // (from RFC 3629) - - // mask to extract lead byte's value ('x' bits above), by sequence's length: - static const unsigned char s_leadValueMask[4] = { 0x7F, 0x1F, 0x0F, 0x07 }; -#ifdef __WXDEBUG__ - // mask and value of lead byte's most significant bits, by length: - static const unsigned char s_leadMarkerMask[4] = { 0x80, 0xE0, 0xF0, 0xF8 }; - static const unsigned char s_leadMarkerVal[4] = { 0x00, 0xC0, 0xE0, 0xF0 }; -#endif - - // extract the lead byte's value bits: - wxASSERT_MSG( ((unsigned char)*i & s_leadMarkerMask[len-1]) == - s_leadMarkerVal[len-1], - _T("invalid UTF-8 lead byte") ); - code = (unsigned char)*i & s_leadValueMask[len-1]; - - // all remaining bytes, if any, are handled in the same way regardless of - // sequence's length: - for ( ++i ; len > 1; --len, ++i ) - { - wxASSERT_MSG( ((unsigned char)*i & 0xC0) == 0x80, - _T("invalid UTF-8 byte") ); - - code <<= 6; - code |= (unsigned char)*i & 0x3F; - } - - return wxUniChar(code); + return wos << str.data(); } -/* static */ -wxCharBuffer wxString::EncodeNChars(size_t n, wxUniChar ch) -{ - Utf8CharBuffer once(EncodeChar(ch)); - // the IncIter() table can be used to determine the length of ch's encoding: - size_t len = ms_utf8IterTable[(unsigned char)once.data[0]]; +#endif // wxUSE_UNICODE && defined(HAVE_WOSTREAM) - wxCharBuffer buf(n * len); - char *ptr = buf.data(); - for ( size_t i = 0; i < n; i++, ptr += len ) - { - memcpy(ptr, once.data, len); - } +#endif // wxUSE_STD_IOSTREAM - return buf; -} +// =========================================================================== +// 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 + } + else // have valid start position { - const_iterator i = begin() + pos; - *implPos = wxStringImpl::const_iterator(i.impl()) - m_impl.begin(); + const const_iterator b = GetIterForNthChar(pos); + *implPos = wxStringImpl::const_iterator(b.impl()) - m_impl.begin(); if ( len == npos ) + { *implLen = npos; - else + } + else // have valid length too { - // too large length is interpreted as "to the end of the string" - // FIXME-UTF8: verify this is the case in std::string, assert - // otherwise - if ( pos + len > length() ) - len = length() - pos; + // 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 + len).impl() - i.impl(); + *implLen = i.impl() - b.impl(); } } } @@ -614,58 +375,6 @@ wxString::~wxString() } #endif -#if wxUSE_UNICODE -const char* wxCStrData::AsChar() const -{ - wxString *str = wxConstCast(m_str, wxString); - - // convert the string: - wxCharBuffer buf(str->mb_str()); - - // FIXME-UTF8: do the conversion in-place in the existing buffer - 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 - -#if !wxUSE_UNICODE_WCHAR -const wchar_t* wxCStrData::AsWChar() const -{ - wxString *str = wxConstCast(m_str, wxString); - - // convert the string: - wxWCharBuffer buf(str->wc_str()); - - // 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 // =========================================================================== @@ -681,15 +390,15 @@ wxString::SubstrBufFromMB wxString::ConvertStr(const char *psz, size_t nLength, { // anything to do? if ( !psz || nLength == 0 ) - return SubstrBufFromMB(L"", 0); + return SubstrBufFromMB(wxWCharBuffer(L""), 0); if ( nLength == npos ) nLength = wxNO_LEN; size_t wcLen; - wxWCharBuffer wcBuf(conv.cMB2WC(psz, nLength, &wcLen)); + wxScopedWCharBuffer wcBuf(conv.cMB2WC(psz, nLength, &wcLen)); if ( !wcLen ) - return SubstrBufFromMB(_T(""), 0); + return SubstrBufFromMB(wxWCharBuffer(L""), 0); else return SubstrBufFromMB(wcBuf, wcLen); } @@ -700,27 +409,40 @@ wxString::SubstrBufFromMB wxString::ConvertStr(const char *psz, size_t nLength, wxString::SubstrBufFromMB wxString::ConvertStr(const char *psz, size_t nLength, const wxMBConv& conv) { - // FIXME-UTF8: return as-is without copying under UTF8 locale, return - // converted string under other locales - needs wxCharBuffer - // changes - // anything to do? if ( !psz || nLength == 0 ) - return SubstrBufFromMB("", 0); + return SubstrBufFromMB(wxCharBuffer(""), 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(wxScopedCharBuffer::CreateNonOwned(psz, nLength), + 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)); + wxScopedWCharBuffer wcBuf(conv.cMB2WC(psz, nLength, &wcLen)); if ( !wcLen ) - return SubstrBufFromMB("", 0); + return SubstrBufFromMB(wxCharBuffer(""), 0); // and then to UTF-8: - SubstrBufFromMB buf(ConvertStr(wcBuf, wcLen, wxConvUTF8)); + 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") ); + wxASSERT_MSG( buf.data, wxT("conversion to UTF-8 failed") ); return buf; } @@ -733,65 +455,111 @@ wxString::SubstrBufFromWC wxString::ConvertStr(const wchar_t *pwz, size_t nLengt { // anything to do? if ( !pwz || nLength == 0 ) - return SubstrBufFromWC("", 0); + return SubstrBufFromWC(wxCharBuffer(""), 0); if ( nLength == npos ) nLength = wxNO_LEN; size_t mbLen; - wxCharBuffer mbBuf(conv.cWC2MB(pwz, nLength, &mbLen)); + wxScopedCharBuffer mbBuf(conv.cWC2MB(pwz, nLength, &mbLen)); if ( !mbLen ) - return SubstrBufFromWC("", 0); + return SubstrBufFromWC(wxCharBuffer(""), 0); else return SubstrBufFromWC(mbBuf, mbLen); } #endif // wxUSE_UNICODE_UTF8 || !wxUSE_UNICODE +// 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 -#if wxUSE_UNICODE_WCHAR - -//Convert wxString in Unicode mode to a multi-byte string -const wxCharBuffer wxString::mb_str(const wxMBConv& conv) const +const wchar_t *wxString::AsWChar(const wxMBConv& conv) const { - return conv.cWC2MB(wx_str(), length() + 1 /* size, not length */, NULL); -} + const char * const strMB = m_impl.c_str(); + const size_t lenMB = m_impl.length(); -#elif wxUSE_UNICODE_UTF8 + // find out the size of the buffer needed + const size_t lenWC = conv.ToWChar(NULL, 0, strMB, lenMB); + if ( lenWC == wxCONV_FAILED ) + return NULL; -const wxWCharBuffer wxString::wc_str() const -{ - return wxConvUTF8.cMB2WC(m_impl.c_str(), - m_impl.length() + 1 /* size, not length */, - 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 -const wxCharBuffer wxString::mb_str(const wxMBConv& conv) const -{ - // FIXME-UTF8: optimize the case when conv==wxConvUTF8 or wxConvLibc - // under UTF8 locale - // FIXME-UTF8: use wc_str() here once we have buffers with length + // 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; + } - size_t wcLen; - wxWCharBuffer wcBuf( - wxConvUTF8.cMB2WC(m_impl.c_str(), - m_impl.length() + 1 /* size, not length */, - &wcLen)); - if ( !wcLen ) - return wxCharBuffer(""); + // finally do convert + m_convertedToWChar.m_str[lenWC] = L'\0'; + if ( conv.ToWChar(m_convertedToWChar.m_str, lenWC, + strMB, lenMB) == wxCONV_FAILED ) + return NULL; - return conv.cWC2MB(wcBuf, wcLen, NULL); + return m_convertedToWChar.m_str; } -#else // ANSI +#endif // !wxUSE_UNICODE_WCHAR + + +// 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 -//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 +const char *wxString::AsChar(const wxMBConv& conv) const { - return conv.cMB2WC(wx_str(), length() + 1 /* size, not length */, NULL); +#if wxUSE_UNICODE_UTF8 + if ( conv.IsUTF8() ) + return m_impl.c_str(); + + 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 + + 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; + } + + 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; } -#endif // Unicode/ANSI +#endif // wxUSE_UNICODE // shrink to minimal size (releasing extra memory) bool wxString::Shrink() @@ -803,7 +571,7 @@ 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); } @@ -882,7 +650,7 @@ wxString operator+(const wxString& str, const char *psz) wxString s; if ( !s.Alloc(strlen(psz) + str.length()) ) { - wxFAIL_MSG( _T("out of memory in wxString::operator+") ); + wxFAIL_MSG( wxT("out of memory in wxString::operator+") ); } s += str; s += psz; @@ -898,7 +666,7 @@ wxString operator+(const wxString& str, const wchar_t *pwz) wxString s; if ( !s.Alloc(wxWcslen(pwz) + str.length()) ) { - wxFAIL_MSG( _T("out of memory in wxString::operator+") ); + wxFAIL_MSG( wxT("out of memory in wxString::operator+") ); } s += str; s += pwz; @@ -914,7 +682,7 @@ wxString operator+(const char *psz, const wxString& str) wxString s; if ( !s.Alloc(strlen(psz) + str.length()) ) { - wxFAIL_MSG( _T("out of memory in wxString::operator+") ); + wxFAIL_MSG( wxT("out of memory in wxString::operator+") ); } s = psz; s += str; @@ -930,7 +698,7 @@ wxString operator+(const wchar_t *pwz, const wxString& str) wxString s; if ( !s.Alloc(wxWcslen(pwz) + str.length()) ) { - wxFAIL_MSG( _T("out of memory in wxString::operator+") ); + wxFAIL_MSG( wxT("out of memory in wxString::operator+") ); } s = pwz; s += str; @@ -942,6 +710,12 @@ 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 @@ -1142,7 +916,7 @@ size_t wxString::find_first_not_of(const wxChar* sz, size_t nStart) const size_t wxString::find_first_of(const wxChar* sz, size_t nStart, size_t n) const { - wxASSERT_MSG( nStart <= length(), _T("invalid index") ); + wxASSERT_MSG( nStart <= length(), wxT("invalid index") ); size_t idx = nStart; for ( const_iterator i = begin() + nStart; i != end(); ++idx, ++i ) @@ -1156,7 +930,7 @@ size_t wxString::find_first_of(const wxChar* sz, size_t nStart, size_t n) const size_t wxString::find_first_not_of(const wxChar* sz, size_t nStart, size_t n) const { - wxASSERT_MSG( nStart <= length(), _T("invalid index") ); + wxASSERT_MSG( nStart <= length(), wxT("invalid index") ); size_t idx = nStart; for ( const_iterator i = begin() + nStart; i != end(); ++idx, ++i ) @@ -1189,7 +963,7 @@ size_t wxString::find_last_of(const wxChar* sz, size_t nStart, size_t n) const } else { - wxASSERT_MSG( nStart <= len, _T("invalid index") ); + wxASSERT_MSG( nStart <= len, wxT("invalid index") ); } size_t idx = nStart; @@ -1213,7 +987,7 @@ size_t wxString::find_last_not_of(const wxChar* sz, size_t nStart, size_t n) con } else { - wxASSERT_MSG( nStart <= len, _T("invalid index") ); + wxASSERT_MSG( nStart <= len, wxT("invalid index") ); } size_t idx = nStart; @@ -1229,7 +1003,7 @@ size_t wxString::find_last_not_of(const wxChar* sz, size_t nStart, size_t n) con size_t wxString::find_first_not_of(wxUniChar ch, size_t nStart) const { - wxASSERT_MSG( nStart <= length(), _T("invalid index") ); + wxASSERT_MSG( nStart <= length(), wxT("invalid index") ); size_t idx = nStart; for ( const_iterator i = begin() + nStart; i != end(); ++idx, ++i ) @@ -1251,7 +1025,7 @@ size_t wxString::find_last_not_of(wxUniChar ch, size_t nStart) const } else { - wxASSERT_MSG( nStart <= len, _T("invalid index") ); + wxASSERT_MSG( nStart <= len, wxT("invalid index") ); } size_t idx = nStart; @@ -1309,15 +1083,71 @@ size_t wxString::find_last_not_of(const wxOtherCharType* sz, size_t nStart, int wxString::CmpNoCase(const wxString& s) const { - // FIXME-UTF8: use wxUniChar::ToLower/ToUpper once added +#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; - size_t idx = 0; + // 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(); - 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); @@ -1333,6 +1163,7 @@ int wxString::CmpNoCase(const wxString& s) const else if ( len1 > len2 ) return 1; return 0; +#endif // !wxUSE_UNICODE_UTF8/wxUSE_UNICODE_UTF8 } @@ -1344,63 +1175,69 @@ 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); + wxStringInternalBuffer buf(res, len); + wxStringCharType *dest = buf; - wchar_t *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, + wxT("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, wxT("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 +const wxScopedCharBuffer 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 @@ -1428,7 +1265,7 @@ wxString wxString::Mid(size_t nFirst, size_t nCount) const wxString dest(*this, nFirst, nCount); if ( dest.length() != nCount ) { - wxFAIL_MSG( _T("out of memory in wxString::Mid") ); + wxFAIL_MSG( wxT("out of memory in wxString::Mid") ); } return dest; @@ -1436,28 +1273,15 @@ 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; @@ -1466,11 +1290,9 @@ 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() - wxStrlen(suffix); + int start = length() - suffix.length(); if ( start < 0 || compare(start, npos, suffix) != 0 ) return false; @@ -1493,12 +1315,12 @@ wxString wxString::Right(size_t nCount) const wxString dest(*this, length() - nCount, nCount); if ( dest.length() != nCount ) { - wxFAIL_MSG( _T("out of memory in wxString::Right") ); + wxFAIL_MSG( wxT("out of memory in wxString::Right") ); } 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 { @@ -1507,7 +1329,7 @@ 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; } @@ -1520,82 +1342,160 @@ wxString wxString::Left(size_t nCount) const wxString dest(*this, 0, nCount); if ( dest.length() != nCount ) { - wxFAIL_MSG( _T("out of memory in wxString::Left") ); + wxFAIL_MSG( wxT("out of memory in wxString::Left") ); } 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 +wxString wxString::BeforeFirst(wxUniChar ch, wxString *rest) const { int iPos = Find(ch); - if ( iPos == wxNOT_FOUND ) iPos = length(); + if ( iPos == wxNOT_FOUND ) + { + iPos = length(); + if ( rest ) + rest->clear(); + } + else + { + if ( rest ) + rest->assign(*this, iPos + 1, npos); + } + 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 +wxString wxString::BeforeLast(wxUniChar ch, wxString *rest) const { wxString str; int iPos = Find(ch, true); - if ( iPos != wxNOT_FOUND && iPos != 0 ) - str = wxString(c_str(), iPos); + 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; + } 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 +// 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, - _T("wxString::Replace(): invalid parameter") ); + wxT("wxString::Replace(): invalid parameter") ); + + wxSTRING_INVALIDATE_CACHE(); size_t uiCount = 0; // count of replacements made - size_t uiOldLen = strOld.length(); - size_t uiNewLen = strNew.length(); + // 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]; + + // 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; - size_t dwPos = 0; + uiCount++; - while ( (*this)[dwPos] != wxT('\0') ) + if ( !bReplaceAll ) + break; + } + } + else if ( !bReplaceAll) { - //DO NOT USE STRSTR HERE - //this string can contain embedded null characters, - //so strstr will function incorrectly - dwPos = find(strOld, dwPos); - if ( dwPos == npos ) - break; // exit the loop - else + size_t pos = m_impl.find(strOld, 0); + if ( pos != npos ) { - //replace this occurance of the old string with the new one - replace(dwPos, uiOldLen, strNew, uiNewLen); + 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(); - //move up pos past the string that was replaced - dwPos += uiNewLen; + // first scan the string to find all positions at which the replacement + // should be made + wxVector replacePositions; - //increase replace count + 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; + } - // stop now? - if ( !bReplaceAll ) - break; // exit the loop + if ( !uiCount ) + return 0; + + // 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; @@ -1630,7 +1530,7 @@ bool wxString::IsNumber() const const_iterator i = begin(); - if ( *i == _T('-') || *i == _T('+') ) + if ( *i == wxT('-') || *i == wxT('+') ) ++i; for ( ; i != end(); ++i ) @@ -1670,14 +1570,28 @@ 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); } @@ -1698,7 +1612,7 @@ 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()); @@ -1708,7 +1622,7 @@ 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); @@ -1762,84 +1676,231 @@ int wxString::Find(wxUniChar ch, bool bFromEnd) const // 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); +#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; - // 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 - ; +// 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; + +bool wxString::ToLong(long *pVal, int base) const +{ + wxASSERT_MSG( !base || (base > 1 && base <= 36), wxT("invalid base") ); + + WX_STRING_TO_X_TYPE_START + long val = wxStrtol(start, &end, base); + WX_STRING_TO_X_TYPE_END } -bool wxString::ToLong(long *val, int base) const +bool wxString::ToULong(unsigned long *pVal, int base) const { - return wxStringToIntType((const wxChar*)c_str(), val, base, wxStrtol); + wxASSERT_MSG( !base || (base > 1 && base <= 36), wxT("invalid base") ); + + WX_STRING_TO_X_TYPE_START + unsigned long val = wxStrtoul(start, &end, base); + WX_STRING_TO_X_TYPE_END } -bool wxString::ToULong(unsigned long *val, int base) const +bool wxString::ToLongLong(wxLongLong_t *pVal, int base) const { - return wxStringToIntType((const wxChar*)c_str(), val, base, wxStrtoul); + wxASSERT_MSG( !base || (base > 1 && base <= 36), wxT("invalid base") ); + + WX_STRING_TO_X_TYPE_START + wxLongLong_t val = wxStrtoll(start, &end, base); + WX_STRING_TO_X_TYPE_END } -bool wxString::ToLongLong(wxLongLong_t *val, int base) const +bool wxString::ToULongLong(wxULongLong_t *pVal, int base) const { - return wxStringToIntType((const wxChar*)c_str(), val, base, wxStrtoll); + 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 } -bool wxString::ToULongLong(wxULongLong_t *val, int base) const +bool wxString::ToDouble(double *pVal) const { - return wxStringToIntType((const wxChar*)c_str(), val, base, wxStrtoull); + WX_STRING_TO_X_TYPE_START + double val = wxStrtod(start, &end); + WX_STRING_TO_X_TYPE_END } -bool wxString::ToDouble(double *val) const +#if wxUSE_XLOCALE + +bool wxString::ToCLong(long *pVal, int base) const { - wxCHECK_MSG( val, false, _T("NULL pointer in wxString::ToDouble") ); + wxASSERT_MSG( !base || (base > 1 && base <= 36), wxT("invalid base") ); -#ifndef __WXWINCE__ - errno = 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 +} - const wxChar *start = c_str(); - wxChar *end; - *val = wxStrtod(start, &end); +bool wxString::ToCULong(unsigned long *pVal, int base) const +{ + wxASSERT_MSG( !base || (base > 1 && base <= 36), wxT("invalid 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) + 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 +} + +bool wxString::ToCDouble(double *pVal) const +{ + 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 +} + +#else // wxUSE_XLOCALE + +// Provide implementation of these functions even when wxUSE_XLOCALE is +// disabled, we still need them in wxWidgets internal code. + +// 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); +} + +bool wxString::ToCULong(unsigned long *pVal, int base) const +{ + return ToULong(pVal, base); +} + +// 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 +{ + // 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 // !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; + } + + // Try to guess the separator, using the most common alternative value. + wxString sep(","); +#endif // wxUSE_INTL/!wxUSE_INTL + wxString cstr(*this); + cstr.Replace(".", sep); + + return cstr.ToDouble(pVal); +} + +#endif // wxUSE_XLOCALE/!wxUSE_XLOCALE + +// ---------------------------------------------------------------------------- +// number to string conversion +// ---------------------------------------------------------------------------- + +/* static */ +wxString wxString::FromDouble(double val, int precision) +{ + wxCHECK_MSG( precision >= -1, wxString(), "Invalid negative precision" ); + + wxString format; + if ( precision == -1 ) + { + format = "%g"; + } + else // Use fixed precision. + { + format.Printf("%%.%df", precision); + } + + return wxString::Format(format, val); +} + +/* static */ +wxString wxString::FromCDouble(double val, int precision) +{ + wxCHECK_MSG( precision >= -1, wxString(), "Invalid negative precision" ); + +#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); + } + + 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 } // --------------------------------------------------------------------------- // 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; @@ -1852,6 +1913,23 @@ 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) @@ -1861,10 +1939,11 @@ 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; @@ -1885,15 +1964,97 @@ 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 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) { 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 ) { @@ -1905,40 +2066,67 @@ int wxString::PrintfV(const wxString& format, va_list argptr) // only a copy va_list argptrcopy; wxVaCopy(argptrcopy, argptr); - int len = wxVsnprintf(buf, size, (const wxChar*)/*FIXME-UTF8*/format, argptrcopy); + +#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 - buf[size] = _T('\0'); + // 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 // 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 } @@ -1949,9 +2137,34 @@ int wxString::PrintfV(const wxString& format, va_list argptr) } // we could have overshot - Shrink(); + str.Shrink(); + + 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 - return length(); +#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 } // ---------------------------------------------------------------------------- @@ -1972,31 +2185,31 @@ bool wxString::Matches(const wxString& mask) const wxString pattern; pattern.reserve(wxStrlen(pszMask)); - pattern += _T('^'); + pattern += wxT('^'); while ( *pszMask ) { switch ( *pszMask ) { - case _T('?'): - pattern += _T('.'); + case wxT('?'): + pattern += wxT('.'); break; - case _T('*'): - pattern += _T(".*"); + case wxT('*'): + pattern += wxT(".*"); break; - case _T('^'): - case _T('.'): - case _T('$'): - case _T('('): - case _T(')'): - case _T('|'): - case _T('+'): - case _T('\\'): + case wxT('^'): + case wxT('.'): + case wxT('$'): + 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 += _T('\\'); + pattern += wxT('\\'); // fall through default: @@ -2005,7 +2218,7 @@ bool wxString::Matches(const wxString& mask) const pszMask++; } - pattern += _T('$'); + pattern += wxT('$'); // and now use it return wxRegEx(pattern, wxRE_NOSUB | wxRE_EXTENDED).Matches(c_str()); @@ -2014,8 +2227,8 @@ bool wxString::Matches(const wxString& mask) const // FIXME-UTF8: implement using iterators, remove #if #if wxUSE_UNICODE_UTF8 - wxWCharBuffer maskBuf = mask.wc_str(); - wxWCharBuffer txtBuf = wc_str(); + const wxScopedWCharBuffer maskBuf = mask.wc_str(); + const wxScopedWCharBuffer txtBuf = wc_str(); const wxChar *pszMask = maskBuf.data(); const wxChar *pszTxt = txtBuf.data(); #else @@ -2117,9 +2330,3 @@ 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(); }