X-Git-Url: https://git.saurik.com/wxWidgets.git/blobdiff_plain/a51fdf81b92e43c7b6167aca131f0a3b9b6724cb..08670ea85abf4b4946a9ce64971b591d7b1ee30b:/src/common/string.cpp diff --git a/src/common/string.cpp b/src/common/string.cpp index 511068e101..2978dcdca8 100644 --- a/src/common/string.cpp +++ b/src/common/string.cpp @@ -24,6 +24,7 @@ #ifndef WX_PRECOMP #include "wx/string.h" #include "wx/wxcrtvararg.h" + #include "wx/intl.h" #include "wx/log.h" #endif @@ -44,6 +45,10 @@ #include "wx/msw/wrapwin.h" #endif // __WXMSW__ +#if wxUSE_STD_IOSTREAM + #include +#endif + // string handling functions used by wxString: #if wxUSE_UNICODE_UTF8 #define wxStringMemcpy memcpy @@ -57,16 +62,20 @@ #define wxStringStrlen wxStrlen #endif -// ---------------------------------------------------------------------------- -// global variables -// ---------------------------------------------------------------------------- - +// 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 { -static UntypedBufferData s_untypedNullData(NULL, 0); +// 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); -UntypedBufferData * const untypedNullDataPtr = &s_untypedNullData; + return &s_untypedNullData; +} } // namespace wxPrivate @@ -433,7 +442,7 @@ wxString::SubstrBufFromMB wxString::ConvertStr(const char *psz, size_t nLength, // 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") ); + wxASSERT_MSG( buf.data, wxT("conversion to UTF-8 failed") ); return buf; } @@ -641,7 +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; @@ -657,7 +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; @@ -673,7 +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; @@ -689,7 +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; @@ -907,7 +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 ) @@ -921,7 +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 ) @@ -954,7 +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; @@ -978,7 +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; @@ -994,7 +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 ) @@ -1016,7 +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; @@ -1074,34 +1083,63 @@ size_t wxString::find_last_not_of(const wxOtherCharType* sz, size_t nStart, int wxString::CmpNoCase(const wxString& s) const { -#if defined(__WXMSW__) && !wxUSE_UNICODE_UTF8 - // prefer to use CompareString() if available as it's more efficient than - // doing it manual or even using wxStricmp() (see #10375) - switch ( ::CompareString(LOCALE_USER_DEFAULT, NORM_IGNORECASE, - m_impl.c_str(), m_impl.length(), - s.m_impl.c_str(), s.m_impl.length()) ) +#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 ( ;; ) { - case CSTR_LESS_THAN: - return -1; + // Compare until the next NUL, if the strings differ this is the final + // result. + rc = wxStricmp(thisCur, thatCur); + if ( rc ) + break; - case CSTR_EQUAL: - return 0; + const size_t lenChunk = wxStrlen(thisCur); + thisCur += lenChunk; + thatCur += lenChunk; - case CSTR_GREATER_THAN: - return 1; + // 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; + } - default: - wxFAIL_MSG( "unexpected CompareString() return value" ); - // fall through + 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; + } - case 0: - wxLogLastError("CompareString"); - // use generic code below + if ( *thatCur ) + { + // Anything non-NUL is greater than NUL. + return -1; + } + } } -#endif // __WXMSW__ && !wxUSE_UNICODE_UTF8 - // do the comparison manually: notice that we can't use wxStricmp() as it - // doesn't handle embedded NULs + 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(); @@ -1125,6 +1163,7 @@ int wxString::CmpNoCase(const wxString& s) const else if ( len1 > len2 ) return 1; return 0; +#endif // !wxUSE_UNICODE_UTF8/wxUSE_UNICODE_UTF8 } @@ -1151,7 +1190,7 @@ wxString wxString::FromAscii(const char *ascii, size_t len) { unsigned char c = (unsigned char)*ascii++; wxASSERT_MSG( c < 0x80, - _T("Non-ASCII value passed to FromAscii().") ); + wxT("Non-ASCII value passed to FromAscii().") ); *dest++ = (wchar_t)c; } @@ -1171,7 +1210,7 @@ wxString wxString::FromAscii(char ascii) unsigned char c = (unsigned char)ascii; - wxASSERT_MSG( c < 0x80, _T("Non-ASCII value passed to FromAscii().") ); + 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)); @@ -1226,7 +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; @@ -1276,7 +1315,7 @@ 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; } @@ -1303,29 +1342,50 @@ 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 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(); + { + 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 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; } @@ -1348,7 +1408,7 @@ size_t wxString::Replace(const wxString& strOld, { // 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(); @@ -1470,7 +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 ) @@ -1627,23 +1687,23 @@ int wxString::Find(wxUniChar ch, bool bFromEnd) const #endif #define WX_STRING_TO_X_TYPE_START \ - wxCHECK_MSG( pVal, false, _T("NULL output pointer") ); \ + wxCHECK_MSG( pVal, false, wxT("NULL output pointer") ); \ DO_IF_NOT_WINCE( errno = 0; ) \ const wxStringCharType *start = wx_str(); \ wxStringCharType *end; +// 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 \ - /* 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) ) \ + if ( end == start DO_IF_NOT_WINCE(|| errno == ERANGE) ) \ return false; \ *pVal = val; \ - return true; + return !*end; bool wxString::ToLong(long *pVal, int base) const { - wxASSERT_MSG( !base || (base > 1 && base <= 36), _T("invalid base") ); + wxASSERT_MSG( !base || (base > 1 && base <= 36), wxT("invalid base") ); WX_STRING_TO_X_TYPE_START long val = wxStrtol(start, &end, base); @@ -1652,7 +1712,7 @@ bool wxString::ToLong(long *pVal, int base) const bool wxString::ToULong(unsigned long *pVal, int base) const { - wxASSERT_MSG( !base || (base > 1 && base <= 36), _T("invalid base") ); + wxASSERT_MSG( !base || (base > 1 && base <= 36), wxT("invalid base") ); WX_STRING_TO_X_TYPE_START unsigned long val = wxStrtoul(start, &end, base); @@ -1661,7 +1721,7 @@ bool wxString::ToULong(unsigned long *pVal, int base) const bool wxString::ToLongLong(wxLongLong_t *pVal, int base) const { - wxASSERT_MSG( !base || (base > 1 && base <= 36), _T("invalid base") ); + wxASSERT_MSG( !base || (base > 1 && base <= 36), wxT("invalid base") ); WX_STRING_TO_X_TYPE_START wxLongLong_t val = wxStrtoll(start, &end, base); @@ -1670,7 +1730,7 @@ bool wxString::ToLongLong(wxLongLong_t *pVal, int base) const bool wxString::ToULongLong(wxULongLong_t *pVal, int base) const { - wxASSERT_MSG( !base || (base > 1 && base <= 36), _T("invalid base") ); + wxASSERT_MSG( !base || (base > 1 && base <= 36), wxT("invalid base") ); WX_STRING_TO_X_TYPE_START wxULongLong_t val = wxStrtoull(start, &end, base); @@ -1688,7 +1748,7 @@ bool wxString::ToDouble(double *pVal) const bool wxString::ToCLong(long *pVal, int base) const { - wxASSERT_MSG( !base || (base > 1 && base <= 36), _T("invalid base") ); + wxASSERT_MSG( !base || (base > 1 && base <= 36), wxT("invalid base") ); WX_STRING_TO_X_TYPE_START #if (wxUSE_UNICODE_UTF8 || !wxUSE_UNICODE) && defined(wxHAS_XLOCALE_SUPPORT) @@ -1701,7 +1761,7 @@ bool wxString::ToCLong(long *pVal, int base) const bool wxString::ToCULong(unsigned long *pVal, int base) const { - wxASSERT_MSG( !base || (base > 1 && base <= 36), _T("invalid base") ); + wxASSERT_MSG( !base || (base > 1 && base <= 36), wxT("invalid base") ); WX_STRING_TO_X_TYPE_START #if (wxUSE_UNICODE_UTF8 || !wxUSE_UNICODE) && defined(wxHAS_XLOCALE_SUPPORT) @@ -1723,7 +1783,113 @@ bool wxString::ToCDouble(double *pVal) const WX_STRING_TO_X_TYPE_END } -#endif // wxUSE_XLOCALE +#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 @@ -1842,16 +2008,16 @@ int wxString::DoPrintfUtf8(const char *format, ...) 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 + 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 behavior - may be used. The vswprintf function may have vsnprintf-like behavior or - vice-versa. Behavior detected on one release can theoretically change + 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 behavior and try our best to handle every known case, including the + 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. @@ -1862,9 +2028,9 @@ int wxString::DoPrintfUtf8(const char *format, ...) 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. + Darwin (tested on 10.5) follows the C99 behaviour exactly. - Glibc 2.6 almost follows the C99 behavior except vswprintf never sets + 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. */ @@ -1893,11 +2059,6 @@ static int DoStringPrintfV(wxString& str, 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; } @@ -1918,7 +2079,7 @@ static int DoStringPrintfV(wxString& str, // 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'); + 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 @@ -2024,31 +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: @@ -2057,7 +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());