+#if wxUSE_UNICODE_UTF8 && !wxUSE_UTF8_LOCALE_ONLY
+wxArgNormalizerWchar<const wxString&>::wxArgNormalizerWchar(
+ const wxString& s,
+ const wxFormatString *fmt, unsigned index)
+ : wxArgNormalizerWithBuffer<wchar_t>(s.wc_str(), fmt, index)
+{
+}
+
+wxArgNormalizerWchar<const wxCStrData&>::wxArgNormalizerWchar(
+ const wxCStrData& s,
+ const wxFormatString *fmt, unsigned index)
+ : wxArgNormalizerWithBuffer<wchar_t>(s.AsWCharBuf(), fmt, index)
+{
+}
+#endif // wxUSE_UNICODE_UTF8 && !wxUSE_UTF8_LOCALE_ONLY
+
+// ----------------------------------------------------------------------------
+// wxArgNormalizedString
+// ----------------------------------------------------------------------------
+
+wxString wxArgNormalizedString::GetString() const
+{
+ if ( !IsValid() )
+ return wxEmptyString;
+
+#if wxUSE_UTF8_LOCALE_ONLY
+ return wxString(reinterpret_cast<const char*>(m_ptr));
+#else
+ #if wxUSE_UNICODE_UTF8
+ if ( wxLocaleIsUtf8 )
+ return wxString(reinterpret_cast<const char*>(m_ptr));
+ else
+ #endif
+ return wxString(reinterpret_cast<const wxChar*>(m_ptr));
+#endif // !wxUSE_UTF8_LOCALE_ONLY
+}
+
+wxArgNormalizedString::operator wxString() const
+{
+ return GetString();
+}
+
+// ----------------------------------------------------------------------------
+// wxFormatConverter: class doing the "%s" and "%c" normalization
+// ----------------------------------------------------------------------------
+
+/*
+ There are four problems with wxPrintf() etc. format strings:
+
+ 1) The printf vararg macros convert all forms of strings into
+ wxStringCharType* representation. This may make the format string
+ incorrect: for example, if %ls was used together with a wchar_t*
+ variadic argument, this would no longer work, because the templates
+ would change wchar_t* argument to wxStringCharType* and %ls would now
+ be incorrect in e.g. UTF-8 build. We need make sure only one specifier
+ form is used.
+
+ 2) To complicate matters further, the meaning of %s and %c is different
+ under Windows and on Unix. The Windows/MS convention is as follows:
+
+ In ANSI mode:
+
+ format specifier results in
+ -----------------------------------
+ %s, %hs, %hS char*
+ %ls, %S, %lS wchar_t*
+
+ In Unicode mode:
+
+ format specifier results in
+ -----------------------------------
+ %hs, %S, %hS char*
+ %s, %ls, %lS wchar_t*
+
+ (While on POSIX systems we have %C identical to %lc and %c always means
+ char (in any mode) while %lc always means wchar_t.)
+
+ In other words, we should _only_ use %s on Windows and %ls on Unix for
+ wxUSE_UNICODE_WCHAR build.
+
+ 3) To make things even worse, we need two forms in UTF-8 build: one for
+ passing strings to ANSI functions under UTF-8 locales (this one should
+ use %s) and one for widechar functions used under non-UTF-8 locales
+ (this one should use %ls).
+
+ And, of course, the same should be done for %c as well.
+
+
+ wxScanf() family of functions is simpler, because we don't normalize their
+ variadic arguments and we only have to handle 2) above and only for widechar
+ versions.
+*/
+
+template<typename T>
+class wxFormatConverterBase
+{
+public:
+ typedef T CharType;
+
+ wxFormatConverterBase()
+ {
+ m_fmtOrig = NULL;
+ m_fmtLast = NULL;
+ m_nCopied = 0;
+ }
+
+ wxCharTypeBuffer<CharType> Convert(const CharType *format)
+ {
+ // this is reset to NULL if we modify the format string
+ m_fmtOrig = format;
+
+ while ( *format )
+ {
+ if ( CopyFmtChar(*format++) == _T('%') )
+ {
+ // skip any flags
+ while ( IsFlagChar(*format) )
+ CopyFmtChar(*format++);
+
+ // and possible width
+ if ( *format == _T('*') )
+ CopyFmtChar(*format++);
+ else
+ SkipDigits(&format);
+
+ // precision?
+ if ( *format == _T('.') )
+ {
+ CopyFmtChar(*format++);
+ if ( *format == _T('*') )
+ CopyFmtChar(*format++);
+ else
+ SkipDigits(&format);
+ }
+
+ // next we can have a size modifier
+ SizeModifier size;
+
+ switch ( *format )
+ {
+ case 'h':
+ size = Size_Short;
+ format++;
+ break;
+
+ case 'l':
+ // "ll" has a different meaning!
+ if ( format[1] != 'l' )
+ {
+ size = Size_Long;
+ format++;
+ break;
+ }
+ //else: fall through
+
+ default:
+ size = Size_Default;
+ }
+
+ CharType outConv = *format;
+ SizeModifier outSize = size;
+
+ // and finally we should have the type
+ switch ( *format )
+ {
+ case _T('S'):
+ case _T('s'):
+ // all strings were converted into the same form by
+ // wxArgNormalizer<T>, this form depends on the context
+ // in which the value is used (scanf/printf/wprintf):
+ HandleString(*format, size, outConv, outSize);
+ break;
+
+ case _T('C'):
+ case _T('c'):
+ HandleChar(*format, size, outConv, outSize);
+ break;
+
+ default:
+ // nothing special to do
+ break;
+ }
+
+ if ( outConv == *format && outSize == size ) // no change
+ {
+ if ( size != Size_Default )
+ CopyFmtChar(*(format - 1));
+ CopyFmtChar(*format);
+ }
+ else // something changed
+ {
+ switch ( outSize )
+ {
+ case Size_Long:
+ InsertFmtChar(_T('l'));
+ break;
+
+ case Size_Short:
+ InsertFmtChar(_T('h'));
+ break;
+
+ case Size_Default:
+ // nothing to do
+ break;
+ }
+ InsertFmtChar(outConv);
+ }
+
+ format++;
+ }
+ }
+
+ // notice that we only translated the string if m_fmtOrig == NULL (as
+ // set by CopyAllBefore()), otherwise we should simply use the original
+ // format
+ if ( m_fmtOrig )
+ {
+ return wxCharTypeBuffer<CharType>::CreateNonOwned(m_fmtOrig);
+ }
+ else
+ {
+ // NULL-terminate converted format string:
+ *m_fmtLast = 0;
+ return m_fmt;
+ }
+ }
+
+ virtual ~wxFormatConverterBase() {}
+
+protected:
+ enum SizeModifier
+ {
+ Size_Default,
+ Size_Short,
+ Size_Long
+ };
+
+ // called to handle %S or %s; 'conv' is conversion specifier ('S' or 's'
+ // respectively), 'size' is the preceding size modifier; the new values of
+ // conversion and size specifiers must be written to outConv and outSize
+ virtual void HandleString(CharType conv, SizeModifier size,
+ CharType& outConv, SizeModifier& outSize) = 0;
+
+ // ditto for %C or %c
+ virtual void HandleChar(CharType conv, SizeModifier size,
+ CharType& outConv, SizeModifier& outSize) = 0;
+
+private:
+ // copy another character to the translated format: this function does the
+ // copy if we are translating but doesn't do anything at all if we don't,
+ // so we don't create the translated format string at all unless we really
+ // need to (i.e. InsertFmtChar() is called)
+ CharType CopyFmtChar(CharType ch)
+ {
+ if ( !m_fmtOrig )
+ {
+ // we're translating, do copy
+ *(m_fmtLast++) = ch;
+ }
+ else
+ {
+ // simply increase the count which should be copied by
+ // CopyAllBefore() later if needed
+ m_nCopied++;
+ }
+
+ return ch;
+ }
+
+ // insert an extra character
+ void InsertFmtChar(CharType ch)
+ {
+ if ( m_fmtOrig )
+ {
+ // so far we haven't translated anything yet
+ CopyAllBefore();
+ }
+
+ *(m_fmtLast++) = ch;
+ }
+
+ void CopyAllBefore()
+ {
+ wxASSERT_MSG( m_fmtOrig && m_fmt.data() == NULL, "logic error" );
+
+ // the modified format string is guaranteed to be no longer than
+ // 3/2 of the original (worst case: the entire format string consists
+ // of "%s" repeated and is expanded to "%ls" on Unix), so we can
+ // allocate the buffer now and not worry about running out of space if
+ // we over-allocate a bit:
+ size_t fmtLen = wxStrlen(m_fmtOrig);
+ // worst case is of even length, so there's no rounding error in *3/2:
+ m_fmt.extend(fmtLen * 3 / 2);
projects / wxWidgets.git / blobdiff