/////////////////////////////////////////////////////////////////////////////
-// Name: strconv.cpp
+// Name: src/common/strconv.cpp
// Purpose: Unicode conversion classes
// Author: Ove Kaaven, Robert Roebling, Vadim Zeitlin, Vaclav Slavik,
// Ryan Norton, Fredrik Roubert (UTF7)
#define TRACE_STRCONV _T("strconv")
+#if SIZEOF_WCHAR_T == 2
+ #define WC_UTF16
+#endif
+
// ============================================================================
// implementation
// ============================================================================
return buf;
}
-const wxWCharBuffer wxMBConv::cMB2WC(const char *szString, size_t nStringLen, size_t* pOutSize) const
+const wxWCharBuffer
+wxMBConv::cMB2WC(const char *in, size_t inLen, size_t *outLen) const
{
- wxASSERT(pOutSize != NULL);
+ // the currently accumulated wide characters
+ wxWCharBuffer wbuf;
- const char* szEnd = szString + nStringLen + 1;
- const char* szPos = szString;
- const char* szStart = szPos;
+ // the current length of wbuf
+ size_t lenBuf = 0;
- size_t nActualLength = 0;
- size_t nCurrentSize = nStringLen; //try normal size first (should never resize?)
+ // we need to know the representation of L'\0' for this conversion
+ size_t nulLen;
+ const char * const nul = GetMBNul(&nulLen);
+ if ( nulLen == (size_t)-1 || nulLen == 0 )
+ return wxWCharBuffer();
- wxWCharBuffer theBuffer(nCurrentSize);
+ // make a copy of the input string unless it is already properly
+ // NUL-terminated
+ wxCharBuffer bufTmp;
- //Convert the string until the length() is reached, continuing the
- //loop every time a null character is reached
- while(szPos != szEnd)
+ // now we can compute the input size if we were not given it: notice that
+ // in this case the string must be properly NUL-terminated, of course, as
+ // otherwise we have no way of knowing how long it is
+ if ( inLen == (size_t)-1 )
{
- wxASSERT(szPos < szEnd); //something is _really_ screwed up if this rings true
-
- //Get the length of the current (sub)string
- size_t nLen = MB2WC(NULL, szPos, 0);
+ // not the most efficient algorithm but it shouldn't matter as normally
+ // there are not many NULs in the string and so normally memcmp()
+ // should stop on the first character
+ const char *p = in;
+ while ( memcmp(p, nul, nulLen) != 0 )
+ p++;
- //Invalid conversion?
- if( nLen == (size_t)-1 )
+ inLen = p - in + nulLen;
+ }
+ else // we already have the size
+ {
+ // check if it's not already NUL-terminated too to avoid the copy
+ if ( inLen < nulLen || memcmp(in + inLen - nulLen, nul, nulLen) != 0 )
{
- *pOutSize = 0;
- theBuffer.data()[0u] = wxT('\0');
- return theBuffer;
+ // make a copy in order to properly NUL-terminate the string
+ bufTmp = wxCharBuffer(inLen + nulLen - 1 /* 1 will be added */);
+ memcpy(bufTmp.data(), in, inLen);
+ memcpy(bufTmp.data() + inLen, nul, nulLen);
}
+ }
+ if ( bufTmp )
+ in = bufTmp;
- //Increase the actual length (+1 for current null character)
- nActualLength += nLen + 1;
-
- //if buffer too big, realloc the buffer
- if (nActualLength > (nCurrentSize+1))
+ for ( const char * const inEnd = in + inLen;; )
+ {
+ // try to convert the current chunk if anything left
+ size_t lenChunk = in < inEnd ? MB2WC(NULL, in, 0) : 0;
+ if ( lenChunk == 0 )
{
- wxWCharBuffer theNewBuffer(nCurrentSize << 1);
- memcpy(theNewBuffer.data(), theBuffer.data(), nCurrentSize * sizeof(wchar_t));
- theBuffer = theNewBuffer;
- nCurrentSize <<= 1;
- }
+ // nothing left in the input string, conversion succeeded
+ if ( outLen )
+ {
+ // we shouldn't include the last NUL in the result length
+ *outLen = lenBuf ? lenBuf - 1 : 0;
+ }
- //Convert the current (sub)string
- if ( MB2WC(&theBuffer.data()[szPos - szStart], szPos, nLen + 1) == (size_t)-1 )
- {
- *pOutSize = 0;
- theBuffer.data()[0u] = wxT('\0');
- return theBuffer;
+ return wbuf;
}
- //Increment to next (sub)string
- //Note that we have to use strlen instead of nLen here
- //because XX2XX gives us the size of the output buffer,
- //which is not necessarily the length of the string
- szPos += strlen(szPos) + 1;
+ if ( lenChunk == (size_t)-1 )
+ break;
+
+ const size_t lenBufNew = lenBuf + lenChunk;
+ if ( !wbuf.extend(lenBufNew) )
+ break;
+
+ lenChunk = MB2WC(wbuf.data() + lenBuf, in, lenChunk + 1 /* for NUL */);
+ if ( lenChunk == (size_t)-1 )
+ break;
+
+ // +! for the embedded NUL (if something follows)
+ lenBuf = lenBufNew + 1;
+
+ // advance the input pointer past the end of this chunk
+ while ( memcmp(in, nul, nulLen) != 0 )
+ in++;
+
+ in += nulLen; // skipping over its terminator as well
}
- //success - return actual length and the buffer
- *pOutSize = nActualLength;
- return theBuffer;
+ // conversion failed
+ if ( outLen )
+ *outLen = 0;
+
+ return wxWCharBuffer();
}
-const wxCharBuffer wxMBConv::cWC2MB(const wchar_t *szString, size_t nStringLen, size_t* pOutSize) const
+const wxCharBuffer
+wxMBConv::cWC2MB(const wchar_t *in, size_t inLen, size_t *outLen) const
{
- wxASSERT(pOutSize != NULL);
-
- const wchar_t* szEnd = szString + nStringLen + 1;
- const wchar_t* szPos = szString;
- const wchar_t* szStart = szPos;
+ // the currently accumulated multibyte characters
+ wxCharBuffer buf;
- size_t nActualLength = 0;
- size_t nCurrentSize = nStringLen << 2; //try * 4 first
+ // the current length of buf
+ size_t lenBuf = 0;
- wxCharBuffer theBuffer(nCurrentSize);
-
- //Convert the string until the length() is reached, continuing the
- //loop every time a null character is reached
- while(szPos != szEnd)
+ // make a copy of the input string unless it is already properly
+ // NUL-terminated
+ //
+ // if we don't know its length we have no choice but to assume that it is,
+ // indeed, properly terminated
+ wxWCharBuffer bufTmp;
+ if ( inLen == (size_t)-1 )
{
- wxASSERT(szPos < szEnd); //something is _really_ screwed up if this rings true
+ inLen = wxWcslen(in) + 1;
+ }
+ else if ( inLen != 0 && in[inLen - 1] != L'\0' )
+ {
+ // make a copy in order to properly NUL-terminate the string
+ bufTmp = wxWCharBuffer(inLen);
+ memcpy(bufTmp.data(), in, inLen*sizeof(wchar_t));
+ }
- //Get the length of the current (sub)string
- size_t nLen = WC2MB(NULL, szPos, 0);
+ if ( bufTmp )
+ in = bufTmp;
- //Invalid conversion?
- if( nLen == (size_t)-1 )
+ for ( const wchar_t * const inEnd = in + inLen;; )
+ {
+ // try to convert the current chunk, if anything left
+ size_t lenChunk = in < inEnd ? WC2MB(NULL, in, 0) : 0;
+ if ( lenChunk == 0 )
{
- *pOutSize = 0;
- theBuffer.data()[0u] = wxT('\0');
- return theBuffer;
+ // nothing left in the input string, conversion succeeded
+ if ( outLen )
+ *outLen = lenBuf ? lenBuf - 1 : lenBuf;
+
+ return buf;
}
- //Increase the actual length (+1 for current null character)
- nActualLength += nLen + 1;
+ if ( lenChunk == (size_t)-1 )
+ break;
- //if buffer too big, realloc the buffer
- if (nActualLength > (nCurrentSize+1))
- {
- wxCharBuffer theNewBuffer(nCurrentSize << 1);
- memcpy(theNewBuffer.data(), theBuffer.data(), nCurrentSize);
- theBuffer = theNewBuffer;
- nCurrentSize <<= 1;
- }
+ const size_t lenBufNew = lenBuf + lenChunk;
+ if ( !buf.extend(lenBufNew) )
+ break;
- //Convert the current (sub)string
- if(WC2MB(&theBuffer.data()[szPos - szStart], szPos, nLen + 1) == (size_t)-1 )
- {
- *pOutSize = 0;
- theBuffer.data()[0u] = wxT('\0');
- return theBuffer;
- }
+ lenChunk = WC2MB(buf.data() + lenBuf, in, lenChunk + 1 /* for NUL */);
+ if ( lenChunk == (size_t)-1 )
+ break;
- //Increment to next (sub)string
- //Note that we have to use wxWcslen instead of nLen here
- //because XX2XX gives us the size of the output buffer,
- //which is not necessarily the length of the string
- szPos += wxWcslen(szPos) + 1;
+ // chunk successfully converted, go to the next one
+ in += wxWcslen(in) + 1 /* skip NUL too */;
+ lenBuf = lenBufNew + 1;
}
- //success - return actual length and the buffer
- *pOutSize = nActualLength;
- return theBuffer;
+ // conversion failed
+ if ( outLen )
+ *outLen = 0;
+
+ return wxCharBuffer();
}
// ----------------------------------------------------------------------------
return wxWC2MB(buf, psz, n);
}
-#ifdef __UNIX__
-
// ----------------------------------------------------------------------------
// wxConvBrokenFileNames
// ----------------------------------------------------------------------------
+#ifdef __UNIX__
+
wxConvBrokenFileNames::wxConvBrokenFileNames(const wxChar *charset)
{
if ( !charset || wxStricmp(charset, _T("UTF-8")) == 0
m_conv = new wxCSConv(charset);
}
-size_t
-wxConvBrokenFileNames::MB2WC(wchar_t *outputBuf,
- const char *psz,
- size_t outputSize) const
-{
- return m_conv->MB2WC( outputBuf, psz, outputSize );
-}
-
-size_t
-wxConvBrokenFileNames::WC2MB(char *outputBuf,
- const wchar_t *psz,
- size_t outputSize) const
-{
- return m_conv->WC2MB( outputBuf, psz, outputSize );
-}
-
-#endif
+#endif // __UNIX__
// ----------------------------------------------------------------------------
// UTF-7
{
size_t len = 0;
- while (*psz && ((!buf) || (len < n)))
+ while ( *psz && (!buf || (len < n)) )
{
unsigned char cc = *psz++;
if (cc != '+')
len++;
psz++;
}
- else
+ else // start of BASE64 encoded string
{
- // BASE64 encoded string
- bool lsb;
- unsigned char c;
+ bool lsb, ok;
unsigned int d, l;
- for (lsb = false, d = 0, l = 0;
- (cc = utf7unb64[(unsigned char)*psz]) != 0xff; psz++)
+ for ( ok = lsb = false, d = 0, l = 0;
+ (cc = utf7unb64[(unsigned char)*psz]) != 0xff;
+ psz++ )
{
d <<= 6;
d += cc;
for (l += 6; l >= 8; lsb = !lsb)
{
- c = (unsigned char)((d >> (l -= 8)) % 256);
+ unsigned char c = (unsigned char)((d >> (l -= 8)) % 256);
if (lsb)
{
if (buf)
len ++;
}
else
+ {
if (buf)
*buf = (wchar_t)(c << 8);
+ }
+
+ ok = true;
}
}
+
+ if ( !ok )
+ {
+ // in valid UTF7 we should have valid characters after '+'
+ return (size_t)-1;
+ }
+
if (*psz == '-')
psz++;
}
}
- if (buf && (len < n))
- *buf = 0;
+
+ if ( buf && (len < n) )
+ *buf = '\0';
+
return len;
}
size_t wxMBConvUTF7::WC2MB(char *buf, const wchar_t *psz, size_t n) const
{
-
-
size_t len = 0;
while (*psz && ((!buf) || (len < n)))
{
// BASE64 encode string
unsigned int lsb, d, l;
- for (d = 0, l = 0;; psz++)
+ for (d = 0, l = 0; /*nothing*/; psz++)
{
for (lsb = 0; lsb < 2; lsb ++)
{
}
#else // !WC_UTF16
if (buf)
- *buf++ = res;
+ *buf++ = (wchar_t)res;
len++;
#endif // WC_UTF16/!WC_UTF16
}
len += pa;
#else
if (buf)
- *buf++ = wxUnicodePUA + (unsigned char)*opsz;
+ *buf++ = (wchar_t)(wxUnicodePUA + (unsigned char)*opsz);
opsz++;
len++;
#endif
// swap 16bit MB to 16bit String
size_t wxMBConvUTF16swap::MB2WC(wchar_t *buf, const char *psz, size_t n) const
{
- size_t len=0;
+ size_t len = 0;
- while (*(wxUint16*)psz && (!buf || len < n))
+ // UTF16 string must be terminated by 2 NULs as single NULs may occur
+ // inside the string
+ while ( (psz[0] || psz[1]) && (!buf || len < n) )
{
- if (buf)
+ if ( buf )
{
((char *)buf)[0] = psz[1];
((char *)buf)[1] = psz[0];
buf++;
}
len++;
- psz += sizeof(wxUint16);
+ psz += 2;
}
- if (buf && len<n) *buf=0;
+
+ if ( buf && len < n )
+ *buf = L'\0';
return len;
}
// swap 16bit MB to 16bit String
size_t wxMBConvUTF16swap::WC2MB(char *buf, const wchar_t *psz, size_t n) const
{
- size_t len=0;
+ size_t len = 0;
- while (*psz && (!buf || len < n))
+ while ( *psz && (!buf || len < n) )
{
- if (buf)
+ if ( buf )
{
*buf++ = ((char*)psz)[1];
*buf++ = ((char*)psz)[0];
}
- len += sizeof(wxUint16);
+ len += 2;
psz++;
}
- if (buf && len<=n-sizeof(wxUint16)) *(wxUint16*)buf=0;
+
+ if ( buf && len < n )
+ *buf = '\0';
return len;
}
return pa;
if (buf)
- *buf++ = cc;
+ *buf++ = (wchar_t)cc;
len++;
psz += pa * sizeof(wxUint16);
}
return pa;
if (buf)
- *buf++ = cc;
+ *buf++ = (wchar_t)cc;
len++;
psz += pa * sizeof(wxUint16);
while (*(wxUint32*)psz && (!buf || len < n))
{
if (buf)
- *buf++ = *(wxUint32*)psz;
+ *buf++ = (wchar_t)(*(wxUint32*)psz);
len++;
psz += sizeof(wxUint32);
}
#endif
private:
+ virtual const char *GetMBNul(size_t *nulLen) const;
+
// the name (for iconv_open()) of a wide char charset -- if none is
// available on this machine, it will remain NULL
static wxString ms_wcCharsetName;
// true if the wide char encoding we use (i.e. ms_wcCharsetName) has
// different endian-ness than the native one
static bool ms_wcNeedsSwap;
+
+ // NUL representation
+ size_t m_nulLen;
+ char m_nulBuf[8];
};
// make the constructor available for unit testing
wxMBConv_iconv::wxMBConv_iconv(const wxChar *name)
{
+ m_nulLen = (size_t)-2;
+
// iconv operates with chars, not wxChars, but luckily it uses only ASCII
// names for the charsets
const wxCharBuffer cname(wxString(name).ToAscii());
};
#endif // wxUSE_FONTMAP/!wxUSE_FONTMAP
- for ( ; *names; ++names )
+ for ( ; *names && ms_wcCharsetName.empty(); ++names )
{
const wxString nameCS(*names);
return res;
}
+const char *wxMBConv_iconv::GetMBNul(size_t *nulLen) const
+{
+ if ( m_nulLen == (size_t)-2 )
+ {
+ wxMBConv_iconv * const self = wxConstCast(this, wxMBConv_iconv);
+
+#if wxUSE_THREADS
+ // NB: explained in MB2WC
+ wxMutexLocker lock(self->m_iconvMutex);
+#endif
+
+ wchar_t *wnul = L"";
+ size_t inLen = sizeof(wchar_t),
+ outLen = WXSIZEOF(m_nulBuf);
+ char *in = (char *)wnul,
+ *out = self->m_nulBuf;
+ if ( iconv(w2m, &in, &inLen, &out, &outLen) == (size_t)-1 )
+ {
+ self->m_nulLen = (size_t)-1;
+ }
+ else // ok
+ {
+ self->m_nulLen = out - m_nulBuf;
+ }
+ }
+
+ *nulLen = m_nulLen;
+ return m_nulBuf;
+}
+
#endif // HAVE_ICONV
wxMBConv_win32()
{
m_CodePage = CP_ACP;
+ m_nulLen = (size_t)-2;
}
#if wxUSE_FONTMAP
wxMBConv_win32(const wxChar* name)
{
m_CodePage = wxCharsetToCodepage(name);
+ m_nulLen = (size_t)-2;
}
wxMBConv_win32(wxFontEncoding encoding)
{
m_CodePage = wxEncodingToCodepage(encoding);
+ m_nulLen = (size_t)-2;
}
-#endif
+#endif // wxUSE_FONTMAP
size_t MB2WC(wchar_t *buf, const char *psz, size_t n) const
{
// own wxMBConvUTF7 doesn't detect errors (e.g. lone "+" which is
// explicitly ill-formed according to RFC 2152) neither so we don't
// even have any fallback here...
- int flags = m_CodePage == CP_UTF7 ? 0 : MB_ERR_INVALID_CHARS;
+ //
+ // Moreover, MB_ERR_INVALID_CHARS is only supported on Win 2K SP4 or
+ // Win XP or newer and if it is specified on older versions, conversion
+ // from CP_UTF8 (which can have flags only 0 or MB_ERR_INVALID_CHARS)
+ // fails. So we can only use the flag on newer Windows versions.
+ // Additionally, the flag is not supported by UTF7, symbol and CJK
+ // encodings. See here:
+ // http://blogs.msdn.com/michkap/archive/2005/04/19/409566.aspx
+ // http://msdn.microsoft.com/library/en-us/intl/unicode_17si.asp
+ int flags = 0;
+ if ( m_CodePage != CP_UTF7 && m_CodePage != CP_SYMBOL &&
+ m_CodePage < 50000 &&
+ IsAtLeastWin2kSP4() )
+ {
+ flags = MB_ERR_INVALID_CHARS;
+ }
+ else if ( m_CodePage == CP_UTF8 )
+ {
+ // Avoid round-trip in the special case of UTF-8 by using our
+ // own UTF-8 conversion code:
+ return wxMBConvUTF8().MB2WC(buf, psz, n);
+ }
const size_t len = ::MultiByteToWideChar
(
buf, // output string
buf ? n : 0 // size of output buffer
);
+ if ( !len )
+ {
+ // function totally failed
+ return (size_t)-1;
+ }
+
+ // if we were really converting and didn't use MB_ERR_INVALID_CHARS,
+ // check if we succeeded, by doing a double trip:
+ if ( !flags && buf )
+ {
+ const size_t mbLen = strlen(psz);
+ wxCharBuffer mbBuf(mbLen);
+ if ( ::WideCharToMultiByte
+ (
+ m_CodePage,
+ 0,
+ buf,
+ -1,
+ mbBuf.data(),
+ mbLen + 1, // size in bytes, not length
+ NULL,
+ NULL
+ ) == 0 ||
+ strcmp(mbBuf, psz) != 0 )
+ {
+ // we didn't obtain the same thing we started from, hence
+ // the conversion was lossy and we consider that it failed
+ return (size_t)-1;
+ }
+ }
// note that it returns count of written chars for buf != NULL and size
// of the needed buffer for buf == NULL so in either case the length of
// the string (which never includes the terminating NUL) is one less
- return len ? len - 1 : (size_t)-1;
+ return len - 1;
}
size_t WC2MB(char *buf, const wchar_t *pwz, size_t n) const
return s_isWin98Or2k == 1;
}
+ static bool IsAtLeastWin2kSP4()
+ {
+#ifdef __WXWINCE__
+ return false;
+#else
+ static int s_isAtLeastWin2kSP4 = -1;
+
+ if ( s_isAtLeastWin2kSP4 == -1 )
+ {
+ OSVERSIONINFOEX ver;
+
+ memset(&ver, 0, sizeof(ver));
+ ver.dwOSVersionInfoSize = sizeof(ver);
+ GetVersionEx((OSVERSIONINFO*)&ver);
+
+ s_isAtLeastWin2kSP4 =
+ ((ver.dwMajorVersion > 5) || // Vista+
+ (ver.dwMajorVersion == 5 && ver.dwMinorVersion > 0) || // XP/2003
+ (ver.dwMajorVersion == 5 && ver.dwMinorVersion == 0 &&
+ ver.wServicePackMajor >= 4)) // 2000 SP4+
+ ? 1 : 0;
+ }
+
+ return s_isAtLeastWin2kSP4 == 1;
+#endif
+ }
+
+ virtual const char *GetMBNul(size_t *nulLen) const
+ {
+ if ( m_nulLen == (size_t)-2 )
+ {
+ wxMBConv_win32 * const self = wxConstCast(this, wxMBConv_win32);
+
+ self->m_nulLen = ::WideCharToMultiByte
+ (
+ m_CodePage, // code page
+ 0, // no flags
+ L"", // input string
+ 1, // translate just NUL
+ self->m_nulBuf, // output buffer
+ WXSIZEOF(m_nulBuf), // and its size
+ NULL, // "replacement" char
+ NULL // [out] was it used?
+ );
+
+ if ( m_nulLen == 0 )
+ self->m_nulLen = (size_t)-1;
+ }
+
+ *nulLen = m_nulLen;
+ return m_nulBuf;
+ }
+
long m_CodePage;
+ size_t m_nulLen;
+ char m_nulBuf[8];
};
#endif // wxHAVE_WIN32_MB2WC
wxFontEncoding m_enc;
wxEncodingConverter m2w, w2m;
+private:
+ virtual const char *GetMBNul(size_t *nulLen) const
+ {
+ switch ( m_enc )
+ {
+ case wxFONTENCODING_UTF16BE:
+ case wxFONTENCODING_UTF16LE:
+ *nulLen = 2;
+ return "\0";
+
+ case wxFONTENCODING_UTF32BE:
+ case wxFONTENCODING_UTF32LE:
+ *nulLen = 4;
+ return "\0\0\0";
+
+ default:
+ *nulLen = 1;
+ return "";
+ }
+ }
+
// were we initialized successfully?
bool m_ok;
SetName(charset);
}
+#if wxUSE_FONTMAP
+ m_encoding = wxFontMapperBase::GetEncodingFromName(charset);
+#else
m_encoding = wxFONTENCODING_SYSTEM;
+#endif
}
wxCSConv::wxCSConv(wxFontEncoding encoding)
// check for the special case of ASCII or ISO8859-1 charset: as we have
// special knowledge of it anyhow, we don't need to create a special
// conversion object
- if ( m_encoding == wxFONTENCODING_ISO8859_1 )
+ if ( m_encoding == wxFONTENCODING_ISO8859_1 ||
+ m_encoding == wxFONTENCODING_DEFAULT )
{
// don't convert at all
return NULL;
return len;
}
+const char *wxCSConv::GetMBNul(size_t *nulLen) const
+{
+ CreateConvIfNeeded();
+
+ if ( m_convReal )
+ {
+ // cast needed just to call private function of m_convReal
+ return ((wxCSConv *)m_convReal)->GetMBNul(nulLen);
+ }
+
+ *nulLen = 1;
+ return "";
+}
+
// ----------------------------------------------------------------------------
// globals
// ----------------------------------------------------------------------------
wxConvUTF8;
#endif // wxUSE_WCHAR_T/!wxUSE_WCHAR_T
-
-
projects / wxWidgets.git / blobdiff