/////////////////////////////////////////////////////////////////////////////
-// 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)
// headers
// ----------------------------------------------------------------------------
-#if defined(__GNUG__) && !defined(NO_GCC_PRAGMA)
- #pragma implementation "strconv.h"
-#endif
-
// For compilers that support precompilation, includes "wx.h".
#include "wx/wxprec.h"
#if wxUSE_WCHAR_T
-#ifdef __WXMSW__
- #include "wx/msw/private.h"
-#endif
-
#ifdef __WINDOWS__
+ #include "wx/msw/private.h"
#include "wx/msw/missing.h"
#endif
#define wxHAVE_WIN32_MB2WC
#endif // __WIN32__ but !__WXMICROWIN__
-// ----------------------------------------------------------------------------
-// headers
-// ----------------------------------------------------------------------------
-
#ifdef __SALFORDC__
#include <clib.h>
#endif
#include "wx/utils.h"
#ifdef __WXMAC__
+#ifndef __DARWIN__
#include <ATSUnicode.h>
#include <TextCommon.h>
#include <TextEncodingConverter.h>
+#endif
#include "wx/mac/private.h" // includes mac headers
#endif
-// ----------------------------------------------------------------------------
-// macros
-// ----------------------------------------------------------------------------
-#define BSWAP_UCS4(str, len) { unsigned _c; for (_c=0; _c<len; _c++) str[_c]=wxUINT32_SWAP_ALWAYS(str[_c]); }
-#define BSWAP_UTF16(str, len) { unsigned _c; for (_c=0; _c<len; _c++) str[_c]=wxUINT16_SWAP_ALWAYS(str[_c]); }
+#define TRACE_STRCONV _T("strconv")
-#if SIZEOF_WCHAR_T == 4
- #define WC_NAME "UCS4"
- #define WC_BSWAP BSWAP_UCS4
- #ifdef WORDS_BIGENDIAN
- #define WC_NAME_BEST "UCS-4BE"
- #else
- #define WC_NAME_BEST "UCS-4LE"
- #endif
-#elif SIZEOF_WCHAR_T == 2
- #define WC_NAME "UTF16"
- #define WC_BSWAP BSWAP_UTF16
+#if SIZEOF_WCHAR_T == 2
#define WC_UTF16
- #ifdef WORDS_BIGENDIAN
- #define WC_NAME_BEST "UTF-16BE"
- #else
- #define WC_NAME_BEST "UTF-16LE"
- #endif
-#else // sizeof(wchar_t) != 2 nor 4
- // does this ever happen?
- #error "Unknown sizeof(wchar_t): please report this to wx-dev@lists.wxwindows.org"
#endif
// ============================================================================
output = *input;
return 1;
}
- else if ((input[1]<0xdc00) || (input[1]>=0xdfff))
+ else if ((input[1]<0xdc00) || (input[1]>0xdfff))
{
output = *input;
return (size_t)-1;
}
//Increment to next (sub)string
- //Note that we have to use strlen here instead of nLen
- //here because XX2XX gives us the size of the output buffer,
- //not neccessarly the length of the 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;
}
//success - return actual length and the buffer
*pOutSize = nActualLength;
- return theBuffer;
+ return theBuffer;
}
const wxCharBuffer wxMBConv::cWC2MB(const wchar_t *szString, size_t nStringLen, size_t* pOutSize) const
//Increase the actual length (+1 for current null character)
nActualLength += nLen + 1;
-
+
//if buffer too big, realloc the buffer
if (nActualLength > (nCurrentSize+1))
{
}
//Increment to next (sub)string
- //Note that we have to use wxWcslen here instead of nLen
- //here because XX2XX gives us the size of the output buffer,
- //not neccessarly the length of the 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;
}
//success - return actual length and the buffer
*pOutSize = nActualLength;
- return theBuffer;
+ return theBuffer;
}
// ----------------------------------------------------------------------------
return wxWC2MB(buf, psz, n);
}
+#ifdef __UNIX__
+
// ----------------------------------------------------------------------------
-// wxConvBrokenFileNames is made for GTK2 in Unicode mode when
-// files are accidentally written in an encoding which is not
-// the system encoding. Typically, the system encoding will be
-// UTF8 but there might be files stored in ISO8859-1 in disk.
+// wxConvBrokenFileNames
// ----------------------------------------------------------------------------
-class wxConvBrokenFileNames: public wxMBConvLibc
+wxConvBrokenFileNames::wxConvBrokenFileNames(const wxChar *charset)
{
-public:
- virtual size_t MB2WC(wchar_t *outputBuf, const char *psz, size_t outputSize) const;
- virtual size_t WC2MB(char *outputBuf, const wchar_t *psz, size_t outputSize) const;
-};
+ if ( !charset || wxStricmp(charset, _T("UTF-8")) == 0
+ || wxStricmp(charset, _T("UTF8")) == 0 )
+ m_conv = new wxMBConvUTF8(wxMBConvUTF8::MAP_INVALID_UTF8_TO_OCTAL);
+ else
+ m_conv = new wxCSConv(charset);
+}
-size_t wxConvBrokenFileNames::MB2WC(wchar_t *outputBuf, const char *psz, size_t outputSize) const
+size_t
+wxConvBrokenFileNames::MB2WC(wchar_t *outputBuf,
+ const char *psz,
+ size_t outputSize) const
{
-#if 0
- if (we find some invalid characters)
- {
- Convert to Unicode range.
- }
- else
-#endif
- return wxMBConvLibc::MB2WC( outputBuf, psz, outputSize );
+ return m_conv->MB2WC( outputBuf, psz, outputSize );
}
-size_t wxConvBrokenFileNames::WC2MB(char *outputBuf, const wchar_t *psz, size_t outputSize) const
+size_t
+wxConvBrokenFileNames::WC2MB(char *outputBuf,
+ const wchar_t *psz,
+ size_t outputSize) const
{
-#if 0
- Convert back from Unicode range.
-#endif
- return wxMBConvLibc::WC2MB( outputBuf, psz, outputSize );
+ return m_conv->WC2MB( outputBuf, psz, outputSize );
}
+#endif
+
// ----------------------------------------------------------------------------
-// UTF-7
+// UTF-7
// ----------------------------------------------------------------------------
// Implementation (C) 2004 Fredrik Roubert
}
#ifndef WC_UTF16
else if (((wxUint32)cc) > 0xffff)
- {
+ {
// no surrogate pair generation (yet?)
return (size_t)-1;
}
{
// 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 ++)
{
static wxUint32 utf8_max[]=
{ 0x7f, 0x7ff, 0xffff, 0x1fffff, 0x3ffffff, 0x7fffffff, 0xffffffff };
+// boundaries of the private use area we use to (temporarily) remap invalid
+// characters invalid in a UTF-8 encoded string
+const wxUint32 wxUnicodePUA = 0x100000;
+const wxUint32 wxUnicodePUAEnd = wxUnicodePUA + 256;
+
size_t wxMBConvUTF8::MB2WC(wchar_t *buf, const char *psz, size_t n) const
{
size_t len = 0;
while (*psz && ((!buf) || (len < n)))
{
+ const char *opsz = psz;
+ bool invalid = false;
unsigned char cc = *psz++, fc = cc;
unsigned cnt;
for (cnt = 0; fc & 0x80; cnt++)
if (buf)
*buf++ = cc;
len++;
+
+ // escape the escape character for octal escapes
+ if ((m_options & MAP_INVALID_UTF8_TO_OCTAL)
+ && cc == '\\' && (!buf || len < n))
+ {
+ if (buf)
+ *buf++ = cc;
+ len++;
+ }
}
else
{
if (!cnt)
{
// invalid UTF-8 sequence
- return (size_t)-1;
+ invalid = true;
}
else
{
wxUint32 res = cc & (0x3f >> cnt);
while (cnt--)
{
- cc = *psz++;
+ cc = *psz;
if ((cc & 0xC0) != 0x80)
{
// invalid UTF-8 sequence
- return (size_t)-1;
+ invalid = true;
+ break;
}
+ psz++;
res = (res << 6) | (cc & 0x3f);
}
- if (res <= utf8_max[ocnt])
+ if (invalid || res <= utf8_max[ocnt])
{
// illegal UTF-8 encoding
- return (size_t)-1;
+ invalid = true;
+ }
+ else if ((m_options & MAP_INVALID_UTF8_TO_PUA) &&
+ res >= wxUnicodePUA && res < wxUnicodePUAEnd)
+ {
+ // if one of our PUA characters turns up externally
+ // it must also be treated as an illegal sequence
+ // (a bit like you have to escape an escape character)
+ invalid = true;
}
+ else
+ {
#ifdef WC_UTF16
- // cast is ok because wchar_t == wxUuint16 if WC_UTF16
- size_t pa = encode_utf16(res, (wxUint16 *)buf);
- if (pa == (size_t)-1)
- return (size_t)-1;
- if (buf)
- buf += pa;
- len += pa;
+ // cast is ok because wchar_t == wxUuint16 if WC_UTF16
+ size_t pa = encode_utf16(res, (wxUint16 *)buf);
+ if (pa == (size_t)-1)
+ {
+ invalid = true;
+ }
+ else
+ {
+ if (buf)
+ buf += pa;
+ len += pa;
+ }
#else // !WC_UTF16
- if (buf)
- *buf++ = res;
- len++;
+ if (buf)
+ *buf++ = (wchar_t)res;
+ len++;
#endif // WC_UTF16/!WC_UTF16
+ }
+ }
+ if (invalid)
+ {
+ if (m_options & MAP_INVALID_UTF8_TO_PUA)
+ {
+ while (opsz < psz && (!buf || len < n))
+ {
+#ifdef WC_UTF16
+ // cast is ok because wchar_t == wxUuint16 if WC_UTF16
+ size_t pa = encode_utf16((unsigned char)*opsz + wxUnicodePUA, (wxUint16 *)buf);
+ wxASSERT(pa != (size_t)-1);
+ if (buf)
+ buf += pa;
+ opsz++;
+ len += pa;
+#else
+ if (buf)
+ *buf++ = (wchar_t)(wxUnicodePUA + (unsigned char)*opsz);
+ opsz++;
+ len++;
+#endif
+ }
+ }
+ else if (m_options & MAP_INVALID_UTF8_TO_OCTAL)
+ {
+ while (opsz < psz && (!buf || len < n))
+ {
+ if ( buf && len + 3 < n )
+ {
+ unsigned char on = *opsz;
+ *buf++ = L'\\';
+ *buf++ = (wchar_t)( L'0' + on / 0100 );
+ *buf++ = (wchar_t)( L'0' + (on % 0100) / 010 );
+ *buf++ = (wchar_t)( L'0' + on % 010 );
+ }
+ opsz++;
+ len += 4;
+ }
+ }
+ else // MAP_INVALID_UTF8_NOT
+ {
+ return (size_t)-1;
+ }
}
}
}
return len;
}
+static inline bool isoctal(wchar_t wch)
+{
+ return L'0' <= wch && wch <= L'7';
+}
+
size_t wxMBConvUTF8::WC2MB(char *buf, const wchar_t *psz, size_t n) const
{
size_t len = 0;
#else
cc=(*psz++) & 0x7fffffff;
#endif
- unsigned cnt;
- for (cnt = 0; cc > utf8_max[cnt]; cnt++) {}
- if (!cnt)
+
+ if ( (m_options & MAP_INVALID_UTF8_TO_PUA)
+ && cc >= wxUnicodePUA && cc < wxUnicodePUAEnd )
{
- // plain ASCII char
if (buf)
- *buf++ = (char) cc;
+ *buf++ = (char)(cc - wxUnicodePUA);
len++;
}
+ else if ( (m_options & MAP_INVALID_UTF8_TO_OCTAL)
+ && cc == L'\\' && psz[0] == L'\\' )
+ {
+ if (buf)
+ *buf++ = (char)cc;
+ psz++;
+ len++;
+ }
+ else if ( (m_options & MAP_INVALID_UTF8_TO_OCTAL) &&
+ cc == L'\\' &&
+ isoctal(psz[0]) && isoctal(psz[1]) && isoctal(psz[2]) )
+ {
+ if (buf)
+ {
+ *buf++ = (char) ((psz[0] - L'0')*0100 +
+ (psz[1] - L'0')*010 +
+ (psz[2] - L'0'));
+ }
+ psz += 3;
+ len++;
+ }
else
{
- len += cnt + 1;
- if (buf)
+ unsigned cnt;
+ for (cnt = 0; cc > utf8_max[cnt]; cnt++) {}
+ if (!cnt)
{
- *buf++ = (char) ((-128 >> cnt) | ((cc >> (cnt * 6)) & (0x3f >> cnt)));
- while (cnt--)
- *buf++ = (char) (0x80 | ((cc >> (cnt * 6)) & 0x3f));
+ // plain ASCII char
+ if (buf)
+ *buf++ = (char) cc;
+ len++;
+ }
+
+ else
+ {
+ len += cnt + 1;
+ if (buf)
+ {
+ *buf++ = (char) ((-128 >> cnt) | ((cc >> (cnt * 6)) & (0x3f >> cnt)));
+ while (cnt--)
+ *buf++ = (char) (0x80 | ((cc >> (cnt * 6)) & 0x3f));
+ }
}
}
}
- if (buf && (len<n)) *buf = 0;
+ if (buf && (len<n))
+ *buf = 0;
return len;
}
-
-
-
// ----------------------------------------------------------------------------
// UTF-16
// ----------------------------------------------------------------------------
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);
}
#define ICONV_CHAR_CAST(x) ((ICONV_CONST char **)(x))
+#define ICONV_T_INVALID ((iconv_t)-1)
+
+#if SIZEOF_WCHAR_T == 4
+ #define WC_BSWAP wxUINT32_SWAP_ALWAYS
+ #define WC_ENC wxFONTENCODING_UTF32
+#elif SIZEOF_WCHAR_T == 2
+ #define WC_BSWAP wxUINT16_SWAP_ALWAYS
+ #define WC_ENC wxFONTENCODING_UTF16
+#else // sizeof(wchar_t) != 2 nor 4
+ // does this ever happen?
+ #error "Unknown sizeof(wchar_t): please report this to wx-dev@lists.wxwindows.org"
+#endif
+
// ----------------------------------------------------------------------------
// wxMBConv_iconv: encapsulates an iconv character set
// ----------------------------------------------------------------------------
virtual size_t WC2MB(char *buf, const wchar_t *psz, size_t n) const;
bool IsOk() const
- { return (m2w != (iconv_t)-1) && (w2m != (iconv_t)-1); }
+ { return (m2w != ICONV_T_INVALID) && (w2m != ICONV_T_INVALID); }
protected:
// the iconv handlers used to translate from multibyte to wide char and in
private:
// the name (for iconv_open()) of a wide char charset -- if none is
// available on this machine, it will remain NULL
- static const char *ms_wcCharsetName;
+ 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;
};
-const char *wxMBConv_iconv::ms_wcCharsetName = NULL;
+// make the constructor available for unit testing
+WXDLLIMPEXP_BASE wxMBConv* new_wxMBConv_iconv( const wxChar* name )
+{
+ wxMBConv_iconv* result = new wxMBConv_iconv( name );
+ if ( !result->IsOk() )
+ {
+ delete result;
+ return 0;
+ }
+ return result;
+}
+
+wxString wxMBConv_iconv::ms_wcCharsetName;
bool wxMBConv_iconv::ms_wcNeedsSwap = false;
wxMBConv_iconv::wxMBConv_iconv(const wxChar *name)
{
- // Do it the hard way
- char cname[100];
- for (size_t i = 0; i < wxStrlen(name)+1; i++)
- cname[i] = (char) name[i];
+ // iconv operates with chars, not wxChars, but luckily it uses only ASCII
+ // names for the charsets
+ const wxCharBuffer cname(wxString(name).ToAscii());
// check for charset that represents wchar_t:
- if (ms_wcCharsetName == NULL)
+ if ( ms_wcCharsetName.empty() )
{
- ms_wcNeedsSwap = false;
+ wxLogTrace(TRACE_STRCONV, _T("Looking for wide char codeset:"));
- // try charset with explicit bytesex info (e.g. "UCS-4LE"):
- ms_wcCharsetName = WC_NAME_BEST;
- m2w = iconv_open(ms_wcCharsetName, cname);
+#if wxUSE_FONTMAP
+ const wxChar **names = wxFontMapperBase::GetAllEncodingNames(WC_ENC);
+#else // !wxUSE_FONTMAP
+ static const wxChar *names[] =
+ {
+#if SIZEOF_WCHAR_T == 4
+ _T("UCS-4"),
+#elif SIZEOF_WCHAR_T = 2
+ _T("UCS-2"),
+#endif
+ NULL
+ };
+#endif // wxUSE_FONTMAP/!wxUSE_FONTMAP
- if (m2w == (iconv_t)-1)
+ for ( ; *names && ms_wcCharsetName.empty(); ++names )
{
- // try charset w/o bytesex info (e.g. "UCS4")
- // and check for bytesex ourselves:
- ms_wcCharsetName = WC_NAME;
- m2w = iconv_open(ms_wcCharsetName, cname);
+ const wxString nameCS(*names);
- // last bet, try if it knows WCHAR_T pseudo-charset
- if (m2w == (iconv_t)-1)
- {
- ms_wcCharsetName = "WCHAR_T";
- m2w = iconv_open(ms_wcCharsetName, cname);
- }
+ // first try charset with explicit bytesex info (e.g. "UCS-4LE"):
+ wxString nameXE(nameCS);
+ #ifdef WORDS_BIGENDIAN
+ nameXE += _T("BE");
+ #else // little endian
+ nameXE += _T("LE");
+ #endif
- if (m2w != (iconv_t)-1)
+ wxLogTrace(TRACE_STRCONV, _T(" trying charset \"%s\""),
+ nameXE.c_str());
+
+ m2w = iconv_open(nameXE.ToAscii(), cname);
+ if ( m2w == ICONV_T_INVALID )
{
- char buf[2], *bufPtr;
- wchar_t wbuf[2], *wbufPtr;
- size_t insz, outsz;
- size_t res;
-
- buf[0] = 'A';
- buf[1] = 0;
- wbuf[0] = 0;
- insz = 2;
- outsz = SIZEOF_WCHAR_T * 2;
- wbufPtr = wbuf;
- bufPtr = buf;
-
- res = iconv(m2w, ICONV_CHAR_CAST(&bufPtr), &insz,
- (char**)&wbufPtr, &outsz);
-
- if (ICONV_FAILED(res, insz))
- {
- ms_wcCharsetName = NULL;
- wxLogLastError(wxT("iconv"));
- wxLogError(_("Conversion to charset '%s' doesn't work."), name);
- }
- else
+ // try charset w/o bytesex info (e.g. "UCS4")
+ wxLogTrace(TRACE_STRCONV, _T(" trying charset \"%s\""),
+ nameCS.c_str());
+ m2w = iconv_open(nameCS.ToAscii(), cname);
+
+ // and check for bytesex ourselves:
+ if ( m2w != ICONV_T_INVALID )
{
- ms_wcNeedsSwap = wbuf[0] != (wchar_t)buf[0];
+ char buf[2], *bufPtr;
+ wchar_t wbuf[2], *wbufPtr;
+ size_t insz, outsz;
+ size_t res;
+
+ buf[0] = 'A';
+ buf[1] = 0;
+ wbuf[0] = 0;
+ insz = 2;
+ outsz = SIZEOF_WCHAR_T * 2;
+ wbufPtr = wbuf;
+ bufPtr = buf;
+
+ res = iconv(m2w, ICONV_CHAR_CAST(&bufPtr), &insz,
+ (char**)&wbufPtr, &outsz);
+
+ if (ICONV_FAILED(res, insz))
+ {
+ wxLogLastError(wxT("iconv"));
+ wxLogError(_("Conversion to charset '%s' doesn't work."),
+ nameCS.c_str());
+ }
+ else // ok, can convert to this encoding, remember it
+ {
+ ms_wcCharsetName = nameCS;
+ ms_wcNeedsSwap = wbuf[0] != (wchar_t)buf[0];
+ }
}
}
- else
+ else // use charset not requiring byte swapping
{
- ms_wcCharsetName = NULL;
-
- // VS: we must not output an error here, since wxWidgets will safely
- // fall back to using wxEncodingConverter.
- wxLogTrace(wxT("strconv"), wxT("Impossible to convert to/from charset '%s' with iconv, falling back to wxEncodingConverter."), name);
- //wxLogError(
+ ms_wcCharsetName = nameXE;
}
}
- wxLogTrace(wxT("strconv"), wxT("wchar_t charset is '%s', needs swap: %i"), ms_wcCharsetName, ms_wcNeedsSwap);
+
+ wxLogTrace(TRACE_STRCONV,
+ wxT("iconv wchar_t charset is \"%s\"%s"),
+ ms_wcCharsetName.empty() ? _T("<none>")
+ : ms_wcCharsetName.c_str(),
+ ms_wcNeedsSwap ? _T(" (needs swap)")
+ : _T(""));
}
else // we already have ms_wcCharsetName
{
- m2w = iconv_open(ms_wcCharsetName, cname);
+ m2w = iconv_open(ms_wcCharsetName.ToAscii(), cname);
}
- // NB: don't ever pass NULL to iconv_open(), it may crash!
- if ( ms_wcCharsetName )
+ if ( ms_wcCharsetName.empty() )
{
- w2m = iconv_open( cname, ms_wcCharsetName);
+ w2m = ICONV_T_INVALID;
}
else
{
- w2m = (iconv_t)-1;
+ w2m = iconv_open(cname, ms_wcCharsetName.ToAscii());
+ if ( w2m == ICONV_T_INVALID )
+ {
+ wxLogTrace(TRACE_STRCONV,
+ wxT("\"%s\" -> \"%s\" works but not the converse!?"),
+ ms_wcCharsetName.c_str(), cname.data());
+ }
}
}
wxMBConv_iconv::~wxMBConv_iconv()
{
- if ( m2w != (iconv_t)-1 )
+ if ( m2w != ICONV_T_INVALID )
iconv_close(m2w);
- if ( w2m != (iconv_t)-1 )
+ if ( w2m != ICONV_T_INVALID )
iconv_close(w2m);
}
// as MB<->WC conversion would fail "randomly".
wxMutexLocker lock(wxConstCast(this, wxMBConv_iconv)->m_iconvMutex);
#endif
-
+
size_t inbuf = strlen(psz);
size_t outbuf = n * SIZEOF_WCHAR_T;
size_t res, cres;
if (ms_wcNeedsSwap)
{
// convert to native endianness
- WC_BSWAP(buf /* _not_ bufPtr */, res)
+ for ( unsigned i = 0; i < res; i++ )
+ buf[n] = WC_BSWAP(buf[i]);
}
// NB: iconv was given only strlen(psz) characters on input, and so
if (ICONV_FAILED(cres, inbuf))
{
//VS: it is ok if iconv fails, hence trace only
- wxLogTrace(wxT("strconv"), wxT("iconv failed: %s"), wxSysErrorMsg(wxSysErrorCode()));
+ wxLogTrace(TRACE_STRCONV, wxT("iconv failed: %s"), wxSysErrorMsg(wxSysErrorCode()));
return (size_t)-1;
}
// NB: explained in MB2WC
wxMutexLocker lock(wxConstCast(this, wxMBConv_iconv)->m_iconvMutex);
#endif
-
- size_t inbuf = wxWcslen(psz) * SIZEOF_WCHAR_T;
+
+ size_t inlen = wxWcslen(psz);
+ size_t inbuf = inlen * SIZEOF_WCHAR_T;
size_t outbuf = n;
size_t res, cres;
if (ms_wcNeedsSwap)
{
// need to copy to temp buffer to switch endianness
- // this absolutely doesn't rock!
- // (no, doing WC_BSWAP twice on the original buffer won't help, as it
+ // (doing WC_BSWAP twice on the original buffer won't help, as it
// could be in read-only memory, or be accessed in some other thread)
- tmpbuf=(wchar_t*)malloc((inbuf+1)*SIZEOF_WCHAR_T);
- memcpy(tmpbuf,psz,(inbuf+1)*SIZEOF_WCHAR_T);
- WC_BSWAP(tmpbuf, inbuf)
- psz=tmpbuf;
+ tmpbuf = (wchar_t *)malloc(inbuf + SIZEOF_WCHAR_T);
+ for ( size_t i = 0; i < inlen; i++ )
+ tmpbuf[n] = WC_BSWAP(psz[i]);
+ tmpbuf[inlen] = L'\0';
+ psz = tmpbuf;
}
if (buf)
if (ICONV_FAILED(cres, inbuf))
{
- //VS: it is ok if iconv fails, hence trace only
- wxLogTrace(wxT("strconv"), wxT("iconv failed: %s"), wxSysErrorMsg(wxSysErrorCode()));
+ wxLogTrace(TRACE_STRCONV, wxT("iconv failed: %s"), wxSysErrorMsg(wxSysErrorCode()));
return (size_t)-1;
}
// 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 )
+ {
+ wxCharBuffer mbBuf(n);
+ if ( ::WideCharToMultiByte
+ (
+ m_CodePage,
+ 0,
+ buf,
+ -1,
+ mbBuf.data(),
+ n,
+ 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
+ }
+
long m_CodePage;
};
#if SIZEOF_WCHAR_T == 4
UniChar* szUniCharBuffer = new UniChar[nOutSize];
#endif
-
+
CFStringGetCharacters(theString, theRange, szUniCharBuffer);
-
+
CFRelease(theString);
szUniCharBuffer[nOutLength] = '\0' ;
converter.MB2WC(szOut, (const char*)szUniCharBuffer , nOutSize ) ;
delete[] szUniCharBuffer;
#endif
-
+
return nOutLength;
}
size_t WC2MB(char *szOut, const wchar_t *szUnConv, size_t nOutSize) const
{
wxASSERT(szUnConv);
-
+
size_t nRealOutSize;
size_t nBufSize = wxWcslen(szUnConv);
UniChar* szUniBuffer = (UniChar*) szUnConv;
#if SIZEOF_WCHAR_T == 4
- wxMBConvUTF16BE converter ;
+ wxMBConvUTF16 converter ;
nBufSize = converter.WC2MB( NULL , szUnConv , 0 );
szUniBuffer = new UniChar[ (nBufSize / sizeof(UniChar)) + 1] ;
converter.WC2MB( (char*) szUniBuffer , szUnConv, nBufSize + sizeof(UniChar)) ;
{
if (szOut != NULL)
CFStringGetCharacters(theString, CFRangeMake(0, nOutSize - 1), (UniChar*) szOut);
-
+
nRealOutSize = CFStringGetLength(theString) + 1;
}
else
//0 tells CFString to return NULL if it meets such a character
false, //not an external representation
(UInt8*) szOut,
- nOutSize,
+ nOutSize,
(CFIndex*) &nRealOutSize
);
}
bool IsOk() const
{
- return m_encoding != kCFStringEncodingInvalidId &&
+ return m_encoding != kCFStringEncodingInvalidId &&
CFStringIsEncodingAvailable(m_encoding);
}
// we have to terminate here, because n might be larger for the trailing zero, and if UniChar
// is not properly terminated we get random characters at the end
ubuf[byteOutLen / sizeof( UniChar ) ] = 0 ;
- wxMBConvUTF16BE converter ;
+ wxMBConvUTF16 converter ;
res = converter.MB2WC( (buf ? buf : tbuf) , (const char*)ubuf , n ) ;
free( ubuf ) ;
#else
ByteCount byteBufferLen = n ;
UniChar* ubuf = NULL ;
#if SIZEOF_WCHAR_T == 4
- wxMBConvUTF16BE converter ;
+ wxMBConvUTF16 converter ;
size_t unicharlen = converter.WC2MB( NULL , psz , 0 ) ;
byteInLen = unicharlen ;
ubuf = (UniChar*) malloc( byteInLen + 2 ) ;
if ( buf && res < n)
{
buf[res] = 0;
-
+
//we need to double-trip to verify it didn't insert any ? in place
//of bogus characters
wxWCharBuffer wcBuf(n);
DECLARE_NO_COPY_CLASS(wxMBConv_wxwin)
};
+// make the constructors available for unit testing
+WXDLLIMPEXP_BASE wxMBConv* new_wxMBConv_wxwin( const wxChar* name )
+{
+ wxMBConv_wxwin* result = new wxMBConv_wxwin( name );
+ if ( !result->IsOk() )
+ {
+ delete result;
+ return 0;
+ }
+ return result;
+}
+
#endif // wxUSE_FONTMAP
// ============================================================================
SetName(charset);
}
+#if wxUSE_FONTMAP
+ m_encoding = wxFontMapperBase::GetEncodingFromName(charset);
+#else
m_encoding = wxFONTENCODING_SYSTEM;
+#endif
}
wxCSConv::wxCSConv(wxFontEncoding encoding)
}
}
+#if wxUSE_FONTMAP
+#include "wx/hashmap.h"
+
+WX_DECLARE_HASH_MAP( wxFontEncoding, wxString, wxIntegerHash, wxIntegerEqual,
+ wxEncodingNameCache );
+
+static wxEncodingNameCache gs_nameCache;
+#endif
+
wxMBConv *wxCSConv::DoCreate() const
{
+#if wxUSE_FONTMAP
+ wxLogTrace(TRACE_STRCONV,
+ wxT("creating conversion for %s"),
+ (m_name ? m_name
+ : wxFontMapperBase::GetEncodingName(m_encoding).c_str()));
+#endif // wxUSE_FONTMAP
+
// 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;
#endif // !wxUSE_FONTMAP
{
wxString name(m_name);
+ wxFontEncoding encoding(m_encoding);
+
+ if ( !name.empty() )
+ {
+ wxMBConv_iconv *conv = new wxMBConv_iconv(name);
+ if ( conv->IsOk() )
+ return conv;
+
+ delete conv;
#if wxUSE_FONTMAP
- if ( name.empty() )
- name = wxFontMapperBase::Get()->GetEncodingName(m_encoding);
+ encoding =
+ wxFontMapperBase::Get()->CharsetToEncoding(name, false);
#endif // wxUSE_FONTMAP
+ }
+#if wxUSE_FONTMAP
+ {
+ const wxEncodingNameCache::iterator it = gs_nameCache.find(encoding);
+ if ( it != gs_nameCache.end() )
+ {
+ if ( it->second.empty() )
+ return NULL;
- wxMBConv_iconv *conv = new wxMBConv_iconv(name);
- if ( conv->IsOk() )
- return conv;
+ wxMBConv_iconv *conv = new wxMBConv_iconv(it->second);
+ if ( conv->IsOk() )
+ return conv;
- delete conv;
+ delete conv;
+ }
+
+ const wxChar** names = wxFontMapperBase::GetAllEncodingNames(encoding);
+
+ for ( ; *names; ++names )
+ {
+ wxMBConv_iconv *conv = new wxMBConv_iconv(*names);
+ if ( conv->IsOk() )
+ {
+ gs_nameCache[encoding] = *names;
+ return conv;
+ }
+
+ delete conv;
+ }
+
+ gs_nameCache[encoding] = _T(""); // cache the failure
+ }
+#endif // wxUSE_FONTMAP
}
#endif // HAVE_ICONV
#if defined(__WXMAC__)
{
// leave UTF16 and UTF32 to the built-ins of wx
- if ( m_name || ( m_encoding < wxFONTENCODING_UTF16BE ||
+ if ( m_name || ( m_encoding < wxFONTENCODING_UTF16BE ||
( m_encoding >= wxFONTENCODING_MACMIN && m_encoding <= wxFONTENCODING_MACMAX ) ) )
{
WXDLLIMPEXP_DATA_BASE(wxMBConv *) wxConvCurrent = &wxConvLibcObj;
WXDLLIMPEXP_DATA_BASE(wxMBConv *) wxConvFileName = &
#ifdef __WXOSX__
- wxConvUTF8Obj;
+ wxConvUTF8Obj;
#else
- wxConvLibcObj;
+ wxConvLibcObj;
#endif
wxConvUTF8;
#endif // wxUSE_WCHAR_T/!wxUSE_WCHAR_T
-
-
projects / wxWidgets.git / blobdiff