/////////////////////////////////////////////////////////////////////////////
// Name: strconv.cpp
// Purpose: Unicode conversion classes
-// Author: Ove Kaaven, Robert Roebling, Vadim Zeitlin
+// Author: Ove Kaaven, Robert Roebling, Vadim Zeitlin, Vaclav Slavik
// Modified by:
// Created: 29/01/98
// RCS-ID: $Id$
-// Copyright: (c) 1999 Ove Kaaven, Robert Roebling, Vadim Zeitlin
-// Licence: wxWindows license
+// Copyright: (c) 1999 Ove Kaaven, Robert Roebling, Vaclav Slavik
+// (c) 2000-2003 Vadim Zeitlin
+// Licence: wxWindows licence
/////////////////////////////////////////////////////////////////////////////
// ============================================================================
// headers
// ----------------------------------------------------------------------------
-#ifdef __GNUG__
+#if defined(__GNUG__) && !defined(NO_GCC_PRAGMA)
#pragma implementation "strconv.h"
#endif
#pragma hdrstop
#endif
+#ifndef WX_PRECOMP
+ #include "wx/intl.h"
+ #include "wx/log.h"
+#endif // WX_PRECOMP
+
+#include "wx/strconv.h"
+
+#if wxUSE_WCHAR_T
+
#ifdef __WXMSW__
- #include "wx/msw/private.h"
+ #include "wx/msw/private.h"
+ #include "wx/msw/missing.h"
#endif
+#ifndef __WXWINCE__
#include <errno.h>
+#endif
+
#include <ctype.h>
#include <string.h>
#include <stdlib.h>
+#if defined(__WIN32__) && !defined(__WXMICROWIN__)
+ #define wxHAVE_WIN32_MB2WC
+#endif // __WIN32__ but !__WXMICROWIN__
+
+// ----------------------------------------------------------------------------
+// headers
+// ----------------------------------------------------------------------------
+
#ifdef __SALFORDC__
- #include <clib.h>
+ #include <clib.h>
#endif
-#ifdef HAVE_ICONV_H
- #include <iconv.h>
+#ifdef HAVE_ICONV
+ #include <iconv.h>
#endif
-#ifdef __WXMSW__
- #include <windows.h>
-#endif
+#include "wx/encconv.h"
+#include "wx/fontmap.h"
-#include "wx/debug.h"
-#include "wx/strconv.h"
-#include "wx/intl.h"
-#include "wx/log.h"
+#ifdef __WXMAC__
+#include "ATSUnicode.h"
+#include "TextCommon.h"
+#include "TextEncodingConverter.h"
-#if defined(WORDS_BIGENDIAN) || defined(__STDC_ISO_10646__)
-#define BSWAP_UCS4(str, len)
-#define BSWAP_UCS2(str, len)
-#else
-#define BSWAP_UCS4(str, len) { unsigned _c; for (_c=0; _c<len; _c++) str[_c]=wxUINT32_SWAP_ALWAYS(str[_c]); }
-#define BSWAP_UCS2(str, len) { unsigned _c; for (_c=0; _c<len; _c++) str[_c]=wxUINT16_SWAP_ALWAYS(str[_c]); }
-#define WC_NEED_BSWAP
+#include "wx/mac/private.h" // includes mac headers
#endif
-#define BSWAP_UTF32(str, len) BSWAP_UCS4(str, len)
-#define BSWAP_UTF16(str, len) BSWAP_UCS2(str, len)
+// ----------------------------------------------------------------------------
+// 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]); }
#if SIZEOF_WCHAR_T == 4
-#define WC_NAME "UCS4"
-#define WC_BSWAP BSWAP_UCS4
+ #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
-#define WC_UTF16
+ #define WC_NAME "UTF16"
+ #define WC_BSWAP BSWAP_UTF16
+ #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
-// ----------------------------------------------------------------------------
-// globals
-// ----------------------------------------------------------------------------
-
-WXDLLEXPORT_DATA(wxMBConv *) wxConvCurrent = &wxConvLibc;
-
// ============================================================================
// implementation
// ============================================================================
-#if wxUSE_WCHAR_T
+// ----------------------------------------------------------------------------
+// UTF-16 en/decoding to/from UCS-4
+// ----------------------------------------------------------------------------
-#ifdef WC_UTF16
-static size_t encode_utf16(wxUint32 input,wxUint16*output)
+static size_t encode_utf16(wxUint32 input, wxUint16 *output)
{
if (input<=0xffff)
{
- if (output) *output++ = input;
+ if (output)
+ *output = (wxUint16) input;
return 1;
}
else if (input>=0x110000)
{
if (output)
{
- *output++ = (input >> 10)+0xd7c0;
- *output++ = (input&0x3ff)+0xdc00;
+ *output++ = (wxUint16) ((input >> 10)+0xd7c0);
+ *output = (wxUint16) ((input&0x3ff)+0xdc00);
}
return 2;
}
}
-static size_t decode_utf16(wxUint16*input,wxUint32&output)
+static size_t decode_utf16(const wxUint16* input, wxUint32& output)
{
if ((*input<0xd800) || (*input>0xdfff))
{
}
}
-#endif // WC_UTF16
// ----------------------------------------------------------------------------
// wxMBConv
// ----------------------------------------------------------------------------
-WXDLLEXPORT_DATA(wxMBConv) wxConvLibc;
-
-size_t wxMBConv::MB2WC(wchar_t *buf, const char *psz, size_t n) const
-{
- return wxMB2WC(buf, psz, n);
-}
-
-size_t wxMBConv::WC2MB(char *buf, const wchar_t *psz, size_t n) const
+wxMBConv::~wxMBConv()
{
- return wxWC2MB(buf, psz, n);
+ // nothing to do here
}
const wxWCharBuffer wxMBConv::cMB2WC(const char *psz) const
{
- if (psz)
+ if ( psz )
{
- size_t nLen = MB2WC((wchar_t *) NULL, psz, 0);
- if (nLen == (size_t)-1)
- return wxWCharBuffer((wchar_t *) NULL);
- wxWCharBuffer buf(nLen);
- MB2WC((wchar_t *)(const wchar_t *) buf, psz, nLen);
- return buf;
+ // calculate the length of the buffer needed first
+ size_t nLen = MB2WC(NULL, psz, 0);
+ if ( nLen != (size_t)-1 )
+ {
+ // now do the actual conversion
+ wxWCharBuffer buf(nLen);
+ MB2WC(buf.data(), psz, nLen + 1); // with the trailing NUL
+
+ return buf;
+ }
}
- else
- return wxWCharBuffer((wchar_t *) NULL);
+
+ wxWCharBuffer buf((wchar_t *)NULL);
+
+ return buf;
}
-const wxCharBuffer wxMBConv::cWC2MB(const wchar_t *psz) const
+const wxCharBuffer wxMBConv::cWC2MB(const wchar_t *pwz) const
{
- if (psz)
+ if ( pwz )
{
- size_t nLen = WC2MB((char *) NULL, psz, 0);
- if (nLen == (size_t)-1)
- return wxCharBuffer((char *) NULL);
- wxCharBuffer buf(nLen);
- WC2MB((char *)(const char *) buf, psz, nLen);
- return buf;
- }
- else
- return wxCharBuffer((char *) NULL);
-}
-
-// ----------------------------------------------------------------------------
-// standard file conversion
-// ----------------------------------------------------------------------------
+ size_t nLen = WC2MB(NULL, pwz, 0);
+ if ( nLen != (size_t)-1 )
+ {
+ wxCharBuffer buf(nLen+3); // space for a wxUint32 trailing zero
+ WC2MB(buf.data(), pwz, nLen + 4);
-WXDLLEXPORT_DATA(wxMBConvFile) wxConvFile;
+ return buf;
+ }
+ }
-// just use the libc conversion for now
-size_t wxMBConvFile::MB2WC(wchar_t *buf, const char *psz, size_t n) const
-{
- return wxMB2WC(buf, psz, n);
-}
+ wxCharBuffer buf((char *)NULL);
-size_t wxMBConvFile::WC2MB(char *buf, const wchar_t *psz, size_t n) const
-{
- return wxWC2MB(buf, psz, n);
+ return buf;
}
// ----------------------------------------------------------------------------
-// standard gdk conversion
+// wxMBConvLibc
// ----------------------------------------------------------------------------
-#ifdef __WXGTK12__
-
-WXDLLEXPORT_DATA(wxMBConvGdk) wxConvGdk;
-
-#include <gdk/gdk.h>
-
-size_t wxMBConvGdk::MB2WC(wchar_t *buf, const char *psz, size_t n) const
+size_t wxMBConvLibc::MB2WC(wchar_t *buf, const char *psz, size_t n) const
{
- if (buf)
- {
- return gdk_mbstowcs((GdkWChar *)buf, psz, n);
- }
- else
- {
- GdkWChar *nbuf = new GdkWChar[n=strlen(psz)];
- size_t len = gdk_mbstowcs(nbuf, psz, n);
- delete[] nbuf;
- return len;
- }
+ return wxMB2WC(buf, psz, n);
}
-size_t wxMBConvGdk::WC2MB(char *buf, const wchar_t *psz, size_t n) const
+size_t wxMBConvLibc::WC2MB(char *buf, const wchar_t *psz, size_t n) const
{
- char *mbstr = gdk_wcstombs((GdkWChar *)psz);
- size_t len = mbstr ? strlen(mbstr) : 0;
- if (buf)
- {
- if (len > n)
- len = n;
- memcpy(buf, psz, len);
- if (len < n)
- buf[len] = 0;
- }
- return len;
+ return wxWC2MB(buf, psz, n);
}
-#endif // GTK > 1.0
-
// ----------------------------------------------------------------------------
// UTF-7
// ----------------------------------------------------------------------------
-WXDLLEXPORT_DATA(wxMBConvUTF7) wxConvUTF7;
-
#if 0
static char utf7_setD[]="ABCDEFGHIJKLMNOPQRSTUVWXYZ"
"abcdefghijklmnopqrstuvwxyz"
// UTF-8
// ----------------------------------------------------------------------------
-WXDLLEXPORT_DATA(wxMBConvUTF8) wxConvUTF8;
-
static wxUint32 utf8_max[]=
{ 0x7f, 0x7ff, 0xffff, 0x1fffff, 0x3ffffff, 0x7fffffff, 0xffffffff };
return (size_t)-1;
}
#ifdef WC_UTF16
- size_t pa = encode_utf16(res, buf);
+ // 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;
-#else
+#else // !WC_UTF16
if (buf)
*buf++ = res;
len++;
-#endif
+#endif // WC_UTF16/!WC_UTF16
}
}
}
{
wxUint32 cc;
#ifdef WC_UTF16
- size_t pa = decode_utf16(psz,cc);
+ // cast is ok for WC_UTF16
+ size_t pa = decode_utf16((const wxUint16 *)psz, cc);
psz += (pa == (size_t)-1) ? 1 : pa;
#else
cc=(*psz++) & 0x7fffffff;
{
// plain ASCII char
if (buf)
- *buf++ = cc;
+ *buf++ = (char) cc;
len++;
}
len += cnt + 1;
if (buf)
{
- *buf++ = (-128 >> cnt) | ((cc >> (cnt * 6)) & (0x3f >> cnt));
+ *buf++ = (char) ((-128 >> cnt) | ((cc >> (cnt * 6)) & (0x3f >> cnt)));
while (cnt--)
- *buf++ = 0x80 | ((cc >> (cnt * 6)) & 0x3f);
+ *buf++ = (char) (0x80 | ((cc >> (cnt * 6)) & 0x3f));
}
}
}
if (buf && (len<n)) *buf = 0;
+
return len;
}
-// ----------------------------------------------------------------------------
-// specified character set
-// ----------------------------------------------------------------------------
-WXDLLEXPORT_DATA(wxCSConv) wxConvLocal((const wxChar *)NULL);
-#include "wx/encconv.h"
-#include "wx/fontmap.h"
-// TODO: add some tables here
-// - perhaps common encodings to common codepages (for Win32)
-// - perhaps common encodings to objects ("UTF8" -> wxConvUTF8)
-// - move wxEncodingConverter meat in here
+// ----------------------------------------------------------------------------
+// UTF-16
+// ----------------------------------------------------------------------------
-#ifdef __WIN32__
-#include "wx/msw/registry.h"
-// this should work if M$ Internet Exploiter is installed
-static long CharsetToCodepage(const wxChar *name)
-{
- if (!name)
- return GetACP();
+#ifdef WORDS_BIGENDIAN
+ #define wxMBConvUTF16straight wxMBConvUTF16BE
+ #define wxMBConvUTF16swap wxMBConvUTF16LE
+#else
+ #define wxMBConvUTF16swap wxMBConvUTF16BE
+ #define wxMBConvUTF16straight wxMBConvUTF16LE
+#endif
- long CP=-1;
- wxString cn(name);
- do {
- wxString path(wxT("MIME\\Database\\Charset\\"));
- path += cn;
- wxRegKey key(wxRegKey::HKCR, path);
+#ifdef WC_UTF16
- if (!key.Exists()) continue;
+// copy 16bit MB to 16bit String
+size_t wxMBConvUTF16straight::MB2WC(wchar_t *buf, const char *psz, size_t n) const
+{
+ size_t len=0;
- // two cases: either there's an AliasForCharset string,
- // or there are Codepage and InternetEncoding dwords.
- // The InternetEncoding gives us the actual encoding,
- // the Codepage just says which Windows character set to
- // use when displaying the data.
- if (key.HasValue(wxT("InternetEncoding")) &&
- key.QueryValue(wxT("InternetEncoding"), &CP)) break;
+ while (*(wxUint16*)psz && (!buf || len < n))
+ {
+ if (buf)
+ *buf++ = *(wxUint16*)psz;
+ len++;
- // no encoding, see if it's an alias
- if (!key.HasValue(wxT("AliasForCharset")) ||
- !key.QueryValue(wxT("AliasForCharset"), cn)) break;
- } while (1);
+ psz += sizeof(wxUint16);
+ }
+ if (buf && len<n) *buf=0;
- return CP;
+ return len;
}
-#endif
-class wxCharacterSet
-{
-public:
- wxCharacterSet(const wxChar*name)
- : cname(name) {}
- virtual ~wxCharacterSet()
- {}
- virtual size_t MB2WC(wchar_t *buf, const char *psz, size_t n)
- { return (size_t)-1; }
- virtual size_t WC2MB(char *buf, const wchar_t *psz, size_t n)
- { return (size_t)-1; }
- virtual bool usable()
- { return FALSE; }
-public:
- const wxChar*cname;
-};
-class ID_CharSet : public wxCharacterSet
+// copy 16bit String to 16bit MB
+size_t wxMBConvUTF16straight::WC2MB(char *buf, const wchar_t *psz, size_t n) const
{
-public:
- ID_CharSet(const wxChar *name,wxMBConv *cnv)
- : wxCharacterSet(name), work(cnv) {}
-
- size_t MB2WC(wchar_t *buf, const char *psz, size_t n)
- { return work ? work->MB2WC(buf,psz,n) : (size_t)-1; }
-
- size_t WC2MB(char *buf, const wchar_t *psz, size_t n)
- { return work ? work->WC2MB(buf,psz,n) : (size_t)-1; }
-
- bool usable()
- { return work!=NULL; }
-public:
- wxMBConv*work;
-};
+ size_t len=0;
+ while (*psz && (!buf || len < n))
+ {
+ if (buf)
+ {
+ *(wxUint16*)buf = *psz;
+ buf += sizeof(wxUint16);
+ }
+ len += sizeof(wxUint16);
+ psz++;
+ }
+ if (buf && len<=n-sizeof(wxUint16)) *(wxUint16*)buf=0;
-#ifdef HAVE_ICONV_H
+ return len;
+}
-// VS: glibc 2.1.3 is broken in that iconv() conversion to/from UCS4 fails with E2BIG
-// if output buffer is _exactly_ as big as needed. Such case is (unless there's
-// yet another bug in glibc) the only case when iconv() returns with (size_t)-1
-// (which means error) and says there are 0 bytes left in the input buffer --
-// when _real_ error occurs, bytes-left-in-input buffer is non-zero. Hence,
-// this alternative test for iconv() failure.
-// [This bug does not appear in glibc 2.2.]
-#if defined(__GLIBC__) && __GLIBC__ == 2 && __GLIBC_MINOR__ <= 1
-#define ICONV_FAILED(cres, bufLeft) ((cres == (size_t)-1) && \
- (errno != E2BIG || bufLeft != 0))
-#else
-#define ICONV_FAILED(cres, bufLeft) (cres == (size_t)-1)
-#endif
-class IC_CharSet : public wxCharacterSet
+// swap 16bit MB to 16bit String
+size_t wxMBConvUTF16swap::MB2WC(wchar_t *buf, const char *psz, size_t n) const
{
-public:
- IC_CharSet(const wxChar *name)
- : wxCharacterSet(name)
- {
- m2w = iconv_open(WC_NAME, wxConvLibc.cWX2MB(cname));
- w2m = iconv_open(wxConvLibc.cWX2MB(cname), WC_NAME);
- }
-
- ~IC_CharSet()
- {
- if ( m2w != (iconv_t)-1 )
- iconv_close(m2w);
- if ( w2m != (iconv_t)-1 )
- iconv_close(w2m);
- }
+ size_t len=0;
- size_t MB2WC(wchar_t *buf, const char *psz, size_t n)
+ while (*(wxUint16*)psz && (!buf || len < n))
{
- size_t inbuf = strlen(psz);
- size_t outbuf = n * SIZEOF_WCHAR_T;
- size_t res, cres;
- // VS: Use these instead of psz, buf because iconv() modifies its arguments:
- wchar_t *bufPtr = buf;
- const char *pszPtr = psz;
-
if (buf)
{
- // have destination buffer, convert there
-#ifdef WX_ICONV_TAKES_CHAR
- cres = iconv(m2w, (char**)&pszPtr, &inbuf, (char**)&bufPtr, &outbuf);
-#else
- cres = iconv(m2w, &pszPtr, &inbuf, (char**)&bufPtr, &outbuf);
-#endif
- res = n - (outbuf / SIZEOF_WCHAR_T);
- // convert to native endianness
-#ifdef WC_NEED_BSWAP
- WC_BSWAP(buf /* _not_ bufPtr */, res)
-#endif
- }
- else
- {
- // no destination buffer... convert using temp buffer
- // to calculate destination buffer requirement
- wchar_t tbuf[8];
- res = 0;
- do {
- bufPtr = tbuf; outbuf = 8*SIZEOF_WCHAR_T;
-#ifdef WX_ICONV_TAKES_CHAR
- cres = iconv( m2w, (char**)&pszPtr, &inbuf, (char**)&bufPtr, &outbuf );
-#else
- cres = iconv( m2w, &pszPtr, &inbuf, (char**)&bufPtr, &outbuf );
-#endif
- res += 8-(outbuf/SIZEOF_WCHAR_T);
- } while ((cres==(size_t)-1) && (errno==E2BIG));
+ ((char *)buf)[0] = psz[1];
+ ((char *)buf)[1] = psz[0];
+ buf++;
}
+ len++;
+ psz += sizeof(wxUint16);
+ }
+ if (buf && len<n) *buf=0;
- if (ICONV_FAILED(cres, inbuf))
- return (size_t)-1;
+ return len;
+}
- return res;
- }
- size_t WC2MB(char *buf, const wchar_t *psz, size_t n)
- {
-#if defined(__BORLANDC__) && (__BORLANDC__ > 0x530)
- size_t inbuf = std::wcslen(psz) * SIZEOF_WCHAR_T;
-#else
- size_t inbuf = ::wcslen(psz) * SIZEOF_WCHAR_T;
-#endif
- size_t outbuf = n;
- size_t res, cres;
+// swap 16bit MB to 16bit String
+size_t wxMBConvUTF16swap::WC2MB(char *buf, const wchar_t *psz, size_t n) const
+{
+ size_t len=0;
-#ifdef WC_NEED_BSWAP
- // 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
- // could be in read-only memory, or be accessed in some other thread)
- wchar_t *tmpbuf=(wchar_t*)malloc((inbuf+1)*SIZEOF_WCHAR_T);
- memcpy(tmpbuf,psz,(inbuf+1)*SIZEOF_WCHAR_T);
- WC_BSWAP(tmpbuf, inbuf)
- psz=tmpbuf;
-#endif
+ while (*psz && (!buf || len < n))
+ {
if (buf)
{
- // have destination buffer, convert there
-#ifdef WX_ICONV_TAKES_CHAR
- cres = iconv( w2m, (char**)&psz, &inbuf, &buf, &outbuf );
-#else
- cres = iconv( w2m, (const char**)&psz, &inbuf, &buf, &outbuf );
-#endif
- res = n-outbuf;
- }
- else
- {
- // no destination buffer... convert using temp buffer
- // to calculate destination buffer requirement
- char tbuf[16];
- res = 0;
- do {
- buf = tbuf; outbuf = 16;
-#ifdef WX_ICONV_TAKES_CHAR
- cres = iconv( w2m, (char**)&psz, &inbuf, &buf, &outbuf );
-#else
- cres = iconv( w2m, (const char**)&psz, &inbuf, &buf, &outbuf );
-#endif
- res += 16 - outbuf;
- } while ((cres==(size_t)-1) && (errno==E2BIG));
+ *buf++ = ((char*)psz)[1];
+ *buf++ = ((char*)psz)[0];
}
-#ifdef WC_NEED_BSWAP
- free(tmpbuf);
-#endif
- if (ICONV_FAILED(cres, inbuf))
- return (size_t)-1;
-
- return res;
+ len += sizeof(wxUint16);
+ psz++;
}
+ if (buf && len<=n-sizeof(wxUint16)) *(wxUint16*)buf=0;
- bool usable()
- { return (m2w != (iconv_t)-1) && (w2m != (iconv_t)-1); }
+ return len;
+}
-public:
- iconv_t m2w, w2m;
-};
-#endif
-#ifdef __WIN32__
-class CP_CharSet : public wxCharacterSet
+#else // WC_UTF16
+
+
+// copy 16bit MB to 32bit String
+size_t wxMBConvUTF16straight::MB2WC(wchar_t *buf, const char *psz, size_t n) const
{
-public:
- CP_CharSet(const wxChar*name)
- : wxCharacterSet(name), CodePage(CharsetToCodepage(name)) {}
+ size_t len=0;
- size_t MB2WC(wchar_t *buf, const char *psz, size_t n)
+ while (*(wxUint16*)psz && (!buf || len < n))
{
- size_t len =
- MultiByteToWideChar(CodePage, 0, psz, -1, buf, buf ? n : 0);
- return len ? len : (size_t)-1;
- }
+ wxUint32 cc;
+ size_t pa=decode_utf16((wxUint16*)psz, cc);
+ if (pa == (size_t)-1)
+ return pa;
- size_t WC2MB(char *buf, const wchar_t *psz, size_t n)
- {
- size_t len = WideCharToMultiByte(CodePage, 0, psz, -1, buf,
- buf ? n : 0, NULL, NULL);
- return len ? len : (size_t)-1;
+ if (buf)
+ *buf++ = cc;
+ len++;
+ psz += pa * sizeof(wxUint16);
}
+ if (buf && len<n) *buf=0;
- bool usable()
- { return CodePage != -1; }
+ return len;
+}
-public:
- long CodePage;
-};
-#endif
-class EC_CharSet : public wxCharacterSet
+// copy 32bit String to 16bit MB
+size_t wxMBConvUTF16straight::WC2MB(char *buf, const wchar_t *psz, size_t n) const
{
-public:
- // temporarily just use wxEncodingConverter stuff,
- // so that it works while a better implementation is built
- EC_CharSet(const wxChar*name) : wxCharacterSet(name),
- enc(wxFONTENCODING_SYSTEM)
- {
- if (name)
- enc = wxTheFontMapper->CharsetToEncoding(name, FALSE);
- m2w.Init(enc, wxFONTENCODING_UNICODE);
- w2m.Init(wxFONTENCODING_UNICODE, enc);
- }
+ size_t len=0;
- size_t MB2WC(wchar_t *buf, const char *psz, size_t n)
+ while (*psz && (!buf || len < n))
{
- size_t inbuf = strlen(psz);
- if (buf)
- m2w.Convert(psz,buf);
- return inbuf;
- }
+ wxUint16 cc[2];
+ size_t pa=encode_utf16(*psz, cc);
+
+ if (pa == (size_t)-1)
+ return pa;
- size_t WC2MB(char *buf, const wchar_t *psz, size_t n)
- {
-#if defined(__BORLANDC__) && (__BORLANDC__ > 0x530)
- size_t inbuf = std::wcslen(psz);
-#else
- size_t inbuf = ::wcslen(psz);
-#endif
if (buf)
- w2m.Convert(psz,buf);
+ {
+ *(wxUint16*)buf = cc[0];
+ buf += sizeof(wxUint16);
+ if (pa > 1)
+ {
+ *(wxUint16*)buf = cc[1];
+ buf += sizeof(wxUint16);
+ }
+ }
- return inbuf;
+ len += pa*sizeof(wxUint16);
+ psz++;
}
+ if (buf && len<=n-sizeof(wxUint16)) *(wxUint16*)buf=0;
- bool usable()
- { return (enc!=wxFONTENCODING_SYSTEM) && (enc!=wxFONTENCODING_DEFAULT); }
+ return len;
+}
-public:
- wxFontEncoding enc;
- wxEncodingConverter m2w, w2m;
-};
-static wxCharacterSet *wxGetCharacterSet(const wxChar *name)
+// swap 16bit MB to 32bit String
+size_t wxMBConvUTF16swap::MB2WC(wchar_t *buf, const char *psz, size_t n) const
{
- wxCharacterSet *cset = NULL;
- if (name)
- {
- if (wxStricmp(name, wxT("UTF8")) == 0 || wxStricmp(name, wxT("UTF-8")) == 0)
- {
- cset = new ID_CharSet(name, &wxConvUTF8);
- }
- else
- {
-#ifdef HAVE_ICONV_H
- cset = new IC_CharSet(name); // may not take NULL
-#endif
- }
- }
+ size_t len=0;
- if (cset && cset->usable()) return cset;
- if (cset)
+ while (*(wxUint16*)psz && (!buf || len < n))
{
- delete cset;
- cset = NULL;
- }
+ wxUint32 cc;
+ char tmp[4];
+ tmp[0]=psz[1]; tmp[1]=psz[0];
+ tmp[2]=psz[3]; tmp[3]=psz[2];
-#ifdef __WIN32__
- cset = new CP_CharSet(name); // may take NULL
- if (cset->usable())
- return cset;
+ size_t pa=decode_utf16((wxUint16*)tmp, cc);
+ if (pa == (size_t)-1)
+ return pa;
- delete cset;
-#endif // __WIN32__
+ if (buf)
+ *buf++ = cc;
- cset = new EC_CharSet(name);
- if (cset->usable())
- return cset;
+ len++;
+ psz += pa * sizeof(wxUint16);
+ }
+ if (buf && len<n) *buf=0;
- delete cset;
- wxLogError(_("Unknown encoding '%s'!"), name);
- return NULL;
+ return len;
}
-wxCSConv::wxCSConv(const wxChar *charset)
-{
- m_name = (wxChar *)NULL;
- m_cset = (wxCharacterSet *) NULL;
- m_deferred = TRUE;
-
- SetName(charset);
-}
-wxCSConv::~wxCSConv()
+// swap 32bit String to 16bit MB
+size_t wxMBConvUTF16swap::WC2MB(char *buf, const wchar_t *psz, size_t n) const
{
- free(m_name);
- delete m_cset;
-}
+ size_t len=0;
-void wxCSConv::SetName(const wxChar *charset)
-{
- if (charset)
+ while (*psz && (!buf || len < n))
{
- m_name = wxStrdup(charset);
- m_deferred = TRUE;
- }
-}
+ wxUint16 cc[2];
+ size_t pa=encode_utf16(*psz, cc);
-void wxCSConv::LoadNow()
-{
- if (m_deferred)
- {
- if ( !m_name )
+ if (pa == (size_t)-1)
+ return pa;
+
+ if (buf)
{
- wxString name = wxLocale::GetSystemEncodingName();
- if ( !name.empty() )
- SetName(name);
+ *buf++ = ((char*)cc)[1];
+ *buf++ = ((char*)cc)[0];
+ if (pa > 1)
+ {
+ *buf++ = ((char*)cc)[3];
+ *buf++ = ((char*)cc)[2];
+ }
}
- m_cset = wxGetCharacterSet(m_name);
- m_deferred = FALSE;
+ len += pa*sizeof(wxUint16);
+ psz++;
}
+ if (buf && len<=n-sizeof(wxUint16)) *(wxUint16*)buf=0;
+
+ return len;
}
-size_t wxCSConv::MB2WC(wchar_t *buf, const char *psz, size_t n) const
-{
- ((wxCSConv *)this)->LoadNow(); // discard constness
+#endif // WC_UTF16
- if (m_cset)
- return m_cset->MB2WC(buf, psz, n);
- // latin-1 (direct)
- size_t len = strlen(psz);
+// ----------------------------------------------------------------------------
+// UTF-32
+// ----------------------------------------------------------------------------
- if (buf)
+#ifdef WORDS_BIGENDIAN
+#define wxMBConvUTF32straight wxMBConvUTF32BE
+#define wxMBConvUTF32swap wxMBConvUTF32LE
+#else
+#define wxMBConvUTF32swap wxMBConvUTF32BE
+#define wxMBConvUTF32straight wxMBConvUTF32LE
+#endif
+
+
+WXDLLIMPEXP_DATA_BASE(wxMBConvUTF32LE) wxConvUTF32LE;
+WXDLLIMPEXP_DATA_BASE(wxMBConvUTF32BE) wxConvUTF32BE;
+
+
+#ifdef WC_UTF16
+
+// copy 32bit MB to 16bit String
+size_t wxMBConvUTF32straight::MB2WC(wchar_t *buf, const char *psz, size_t n) const
+{
+ size_t len=0;
+
+ while (*(wxUint32*)psz && (!buf || len < n))
{
- for (size_t c = 0; c <= len; c++)
- buf[c] = (unsigned char)(psz[c]);
+ wxUint16 cc[2];
+
+ size_t pa=encode_utf16(*(wxUint32*)psz, cc);
+ if (pa == (size_t)-1)
+ return pa;
+
+ if (buf)
+ {
+ *buf++ = cc[0];
+ if (pa > 1)
+ *buf++ = cc[1];
+ }
+ len += pa;
+ psz += sizeof(wxUint32);
}
+ if (buf && len<n) *buf=0;
return len;
}
-size_t wxCSConv::WC2MB(char *buf, const wchar_t *psz, size_t n) const
+
+// copy 16bit String to 32bit MB
+size_t wxMBConvUTF32straight::WC2MB(char *buf, const wchar_t *psz, size_t n) const
{
- ((wxCSConv *)this)->LoadNow(); // discard constness
+ size_t len=0;
- if (m_cset)
- return m_cset->WC2MB(buf, psz, n);
+ while (*psz && (!buf || len < n))
+ {
+ wxUint32 cc;
- // latin-1 (direct)
-#if defined(__BORLANDC__) && (__BORLANDC__ > 0x530)
- size_t len=std::wcslen(psz);
+ // cast is ok for WC_UTF16
+ size_t pa = decode_utf16((const wxUint16 *)psz, cc);
+ if (pa == (size_t)-1)
+ return pa;
+
+ if (buf)
+ {
+ *(wxUint32*)buf = cc;
+ buf += sizeof(wxUint32);
+ }
+ len += sizeof(wxUint32);
+ psz += pa;
+ }
+
+ if (buf && len<=n-sizeof(wxUint32))
+ *(wxUint32*)buf=0;
+
+ return len;
+}
+
+
+
+// swap 32bit MB to 16bit String
+size_t wxMBConvUTF32swap::MB2WC(wchar_t *buf, const char *psz, size_t n) const
+{
+ size_t len=0;
+
+ while (*(wxUint32*)psz && (!buf || len < n))
+ {
+ char tmp[4];
+ tmp[0] = psz[3]; tmp[1] = psz[2];
+ tmp[2] = psz[1]; tmp[3] = psz[0];
+
+
+ wxUint16 cc[2];
+
+ size_t pa=encode_utf16(*(wxUint32*)tmp, cc);
+ if (pa == (size_t)-1)
+ return pa;
+
+ if (buf)
+ {
+ *buf++ = cc[0];
+ if (pa > 1)
+ *buf++ = cc[1];
+ }
+ len += pa;
+ psz += sizeof(wxUint32);
+ }
+
+ if (buf && len<n)
+ *buf=0;
+
+ return len;
+}
+
+
+// swap 16bit String to 32bit MB
+size_t wxMBConvUTF32swap::WC2MB(char *buf, const wchar_t *psz, size_t n) const
+{
+ size_t len=0;
+
+ while (*psz && (!buf || len < n))
+ {
+ char cc[4];
+
+ // cast is ok for WC_UTF16
+ size_t pa=decode_utf16((const wxUint16 *)psz, *(wxUint32*)cc);
+ if (pa == (size_t)-1)
+ return pa;
+
+ if (buf)
+ {
+ *buf++ = cc[3];
+ *buf++ = cc[2];
+ *buf++ = cc[1];
+ *buf++ = cc[0];
+ }
+ len += sizeof(wxUint32);
+ psz += pa;
+ }
+
+ if (buf && len<=n-sizeof(wxUint32))
+ *(wxUint32*)buf=0;
+
+ return len;
+}
+
+#else // WC_UTF16
+
+
+// copy 32bit MB to 32bit String
+size_t wxMBConvUTF32straight::MB2WC(wchar_t *buf, const char *psz, size_t n) const
+{
+ size_t len=0;
+
+ while (*(wxUint32*)psz && (!buf || len < n))
+ {
+ if (buf)
+ *buf++ = *(wxUint32*)psz;
+ len++;
+ psz += sizeof(wxUint32);
+ }
+
+ if (buf && len<n)
+ *buf=0;
+
+ return len;
+}
+
+
+// copy 32bit String to 32bit MB
+size_t wxMBConvUTF32straight::WC2MB(char *buf, const wchar_t *psz, size_t n) const
+{
+ size_t len=0;
+
+ while (*psz && (!buf || len < n))
+ {
+ if (buf)
+ {
+ *(wxUint32*)buf = *psz;
+ buf += sizeof(wxUint32);
+ }
+
+ len += sizeof(wxUint32);
+ psz++;
+ }
+
+ if (buf && len<=n-sizeof(wxUint32))
+ *(wxUint32*)buf=0;
+
+ return len;
+}
+
+
+// swap 32bit MB to 32bit String
+size_t wxMBConvUTF32swap::MB2WC(wchar_t *buf, const char *psz, size_t n) const
+{
+ size_t len=0;
+
+ while (*(wxUint32*)psz && (!buf || len < n))
+ {
+ if (buf)
+ {
+ ((char *)buf)[0] = psz[3];
+ ((char *)buf)[1] = psz[2];
+ ((char *)buf)[2] = psz[1];
+ ((char *)buf)[3] = psz[0];
+ buf++;
+ }
+ len++;
+ psz += sizeof(wxUint32);
+ }
+
+ if (buf && len<n)
+ *buf=0;
+
+ return len;
+}
+
+
+// swap 32bit String to 32bit MB
+size_t wxMBConvUTF32swap::WC2MB(char *buf, const wchar_t *psz, size_t n) const
+{
+ size_t len=0;
+
+ while (*psz && (!buf || len < n))
+ {
+ if (buf)
+ {
+ *buf++ = ((char *)psz)[3];
+ *buf++ = ((char *)psz)[2];
+ *buf++ = ((char *)psz)[1];
+ *buf++ = ((char *)psz)[0];
+ }
+ len += sizeof(wxUint32);
+ psz++;
+ }
+
+ if (buf && len<=n-sizeof(wxUint32))
+ *(wxUint32*)buf=0;
+
+ return len;
+}
+
+
+#endif // WC_UTF16
+
+
+// ============================================================================
+// The classes doing conversion using the iconv_xxx() functions
+// ============================================================================
+
+#ifdef HAVE_ICONV
+
+// VS: glibc 2.1.3 is broken in that iconv() conversion to/from UCS4 fails with E2BIG
+// if output buffer is _exactly_ as big as needed. Such case is (unless there's
+// yet another bug in glibc) the only case when iconv() returns with (size_t)-1
+// (which means error) and says there are 0 bytes left in the input buffer --
+// when _real_ error occurs, bytes-left-in-input buffer is non-zero. Hence,
+// this alternative test for iconv() failure.
+// [This bug does not appear in glibc 2.2.]
+#if defined(__GLIBC__) && __GLIBC__ == 2 && __GLIBC_MINOR__ <= 1
+#define ICONV_FAILED(cres, bufLeft) ((cres == (size_t)-1) && \
+ (errno != E2BIG || bufLeft != 0))
#else
- size_t len=::wcslen(psz);
+#define ICONV_FAILED(cres, bufLeft) (cres == (size_t)-1)
#endif
+
+#define ICONV_CHAR_CAST(x) ((ICONV_CONST char **)(x))
+
+// ----------------------------------------------------------------------------
+// wxMBConv_iconv: encapsulates an iconv character set
+// ----------------------------------------------------------------------------
+
+class wxMBConv_iconv : public wxMBConv
+{
+public:
+ wxMBConv_iconv(const wxChar *name);
+ virtual ~wxMBConv_iconv();
+
+ virtual size_t MB2WC(wchar_t *buf, const char *psz, size_t n) const;
+ 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); }
+
+protected:
+ // the iconv handlers used to translate from multibyte to wide char and in
+ // the other direction
+ iconv_t m2w,
+ w2m;
+
+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;
+
+ // 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;
+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];
+
+ // check for charset that represents wchar_t:
+ if (ms_wcCharsetName == NULL)
+ {
+ ms_wcNeedsSwap = false;
+
+ // try charset with explicit bytesex info (e.g. "UCS-4LE"):
+ ms_wcCharsetName = WC_NAME_BEST;
+ m2w = iconv_open(ms_wcCharsetName, cname);
+
+ if (m2w == (iconv_t)-1)
+ {
+ // try charset w/o bytesex info (e.g. "UCS4")
+ // and check for bytesex ourselves:
+ ms_wcCharsetName = WC_NAME;
+ m2w = iconv_open(ms_wcCharsetName, cname);
+
+ // 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);
+ }
+
+ if (m2w != (iconv_t)-1)
+ {
+ 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
+ {
+ ms_wcNeedsSwap = wbuf[0] != (wchar_t)buf[0];
+ }
+ }
+ else
+ {
+ ms_wcCharsetName = NULL;
+
+ // VS: we must not output an error here, since wxWindows 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(
+ }
+ }
+ wxLogTrace(wxT("strconv"), wxT("wchar_t charset is '%s', needs swap: %i"), ms_wcCharsetName, ms_wcNeedsSwap);
+ }
+ else // we already have ms_wcCharsetName
+ {
+ m2w = iconv_open(ms_wcCharsetName, cname);
+ }
+
+ // NB: don't ever pass NULL to iconv_open(), it may crash!
+ if ( ms_wcCharsetName )
+ {
+ w2m = iconv_open( cname, ms_wcCharsetName);
+ }
+ else
+ {
+ w2m = (iconv_t)-1;
+ }
+}
+
+wxMBConv_iconv::~wxMBConv_iconv()
+{
+ if ( m2w != (iconv_t)-1 )
+ iconv_close(m2w);
+ if ( w2m != (iconv_t)-1 )
+ iconv_close(w2m);
+}
+
+size_t wxMBConv_iconv::MB2WC(wchar_t *buf, const char *psz, size_t n) const
+{
+ size_t inbuf = strlen(psz);
+ size_t outbuf = n * SIZEOF_WCHAR_T;
+ size_t res, cres;
+ // VS: Use these instead of psz, buf because iconv() modifies its arguments:
+ wchar_t *bufPtr = buf;
+ const char *pszPtr = psz;
+
if (buf)
{
- for (size_t c = 0; c <= len; c++)
- buf[c] = (psz[c] > 0xff) ? '?' : psz[c];
+ // have destination buffer, convert there
+ cres = iconv(m2w,
+ ICONV_CHAR_CAST(&pszPtr), &inbuf,
+ (char**)&bufPtr, &outbuf);
+ res = n - (outbuf / SIZEOF_WCHAR_T);
+
+ if (ms_wcNeedsSwap)
+ {
+ // convert to native endianness
+ WC_BSWAP(buf /* _not_ bufPtr */, res)
+ }
+
+ // NB: iconv was given only strlen(psz) characters on input, and so
+ // it couldn't convert the trailing zero. Let's do it ourselves
+ // if there's some room left for it in the output buffer.
+ if (res < n)
+ buf[res] = 0;
+ }
+ else
+ {
+ // no destination buffer... convert using temp buffer
+ // to calculate destination buffer requirement
+ wchar_t tbuf[8];
+ res = 0;
+ do {
+ bufPtr = tbuf;
+ outbuf = 8*SIZEOF_WCHAR_T;
+
+ cres = iconv(m2w,
+ ICONV_CHAR_CAST(&pszPtr), &inbuf,
+ (char**)&bufPtr, &outbuf );
+
+ res += 8-(outbuf/SIZEOF_WCHAR_T);
+ } while ((cres==(size_t)-1) && (errno==E2BIG));
}
- return len;
+ if (ICONV_FAILED(cres, inbuf))
+ {
+ //VS: it is ok if iconv fails, hence trace only
+ wxLogTrace(wxT("strconv"), wxT("iconv failed: %s"), wxSysErrorMsg(wxSysErrorCode()));
+ return (size_t)-1;
+ }
+
+ return res;
}
-#ifdef HAVE_ICONV_H
+size_t wxMBConv_iconv::WC2MB(char *buf, const wchar_t *psz, size_t n) const
+{
+ size_t inbuf = wxWcslen(psz) * SIZEOF_WCHAR_T;
+ size_t outbuf = n;
+ size_t res, cres;
+
+ wchar_t *tmpbuf = 0;
-class IC_CharSetConverter
+ 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
+ // 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;
+ }
+
+ if (buf)
+ {
+ // have destination buffer, convert there
+ cres = iconv( w2m, ICONV_CHAR_CAST(&psz), &inbuf, &buf, &outbuf );
+
+ res = n-outbuf;
+
+ // NB: iconv was given only wcslen(psz) characters on input, and so
+ // it couldn't convert the trailing zero. Let's do it ourselves
+ // if there's some room left for it in the output buffer.
+ if (res < n)
+ buf[0] = 0;
+ }
+ else
+ {
+ // no destination buffer... convert using temp buffer
+ // to calculate destination buffer requirement
+ char tbuf[16];
+ res = 0;
+ do {
+ buf = tbuf; outbuf = 16;
+
+ cres = iconv( w2m, ICONV_CHAR_CAST(&psz), &inbuf, &buf, &outbuf );
+
+ res += 16 - outbuf;
+ } while ((cres==(size_t)-1) && (errno==E2BIG));
+ }
+
+ if (ms_wcNeedsSwap)
+ {
+ free(tmpbuf);
+ }
+
+ if (ICONV_FAILED(cres, inbuf))
+ {
+ //VS: it is ok if iconv fails, hence trace only
+ wxLogTrace(wxT("strconv"), wxT("iconv failed: %s"), wxSysErrorMsg(wxSysErrorCode()));
+ return (size_t)-1;
+ }
+
+ return res;
+}
+
+#endif // HAVE_ICONV
+
+
+// ============================================================================
+// Win32 conversion classes
+// ============================================================================
+
+#ifdef wxHAVE_WIN32_MB2WC
+
+// from utils.cpp
+extern WXDLLIMPEXP_BASE long wxCharsetToCodepage(const wxChar *charset);
+extern WXDLLIMPEXP_BASE long wxEncodingToCodepage(wxFontEncoding encoding);
+
+class wxMBConv_win32 : public wxMBConv
{
public:
- IC_CharSetConverter(IC_CharSet *from, IC_CharSet *to)
+ wxMBConv_win32()
{
- cnv = iconv_open(wxConvLibc.cWX2MB(to->cname),
- wxConvLibc.cWX2MB(from->cname));
+ m_CodePage = CP_ACP;
}
- ~IC_CharSetConverter()
+ wxMBConv_win32(const wxChar* name)
{
- if (cnv != (iconv_t)-1)
- iconv_close(cnv);
+ m_CodePage = wxCharsetToCodepage(name);
}
- size_t Convert(char *buf, const char *psz, size_t n)
+ wxMBConv_win32(wxFontEncoding encoding)
{
- size_t inbuf = strlen(psz);
- size_t outbuf = n;
-#ifdef WX_ICONV_TAKES_CHAR
- size_t res = iconv( cnv, (char**)&psz, &inbuf, &buf, &outbuf );
-#else
- size_t res = iconv( cnv, &psz, &inbuf, &buf, &outbuf );
-#endif
- if (res == (size_t)-1)
+ m_CodePage = wxEncodingToCodepage(encoding);
+ }
+
+ size_t MB2WC(wchar_t *buf, const char *psz, size_t n) const
+ {
+ const size_t len = ::MultiByteToWideChar
+ (
+ m_CodePage, // code page
+ 0, // flags (none)
+ psz, // input string
+ -1, // its length (NUL-terminated)
+ buf, // output string
+ buf ? n : 0 // size of output buffer
+ );
+
+ // 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;
+ }
+
+ size_t WC2MB(char *buf, const wchar_t *pwz, size_t n) const
+ {
+ /*
+ we have a problem here: by default, WideCharToMultiByte() may
+ replace characters unrepresentable in the target code page with bad
+ quality approximations such as turning "1/2" symbol (U+00BD) into
+ "1" for the code pages which don't have it and we, obviously, want
+ to avoid this at any price
+
+ the trouble is that this function does it _silently_, i.e. it won't
+ even tell us whether it did or not... Win98/2000 and higher provide
+ WC_NO_BEST_FIT_CHARS but it doesn't work for the older systems and
+ we have to resort to a round trip, i.e. check that converting back
+ results in the same string -- this is, of course, expensive but
+ otherwise we simply can't be sure to not garble the data.
+ */
+
+ // determine if we can rely on WC_NO_BEST_FIT_CHARS: according to MSDN
+ // it doesn't work with CJK encodings (which we test for rather roughly
+ // here...) nor with UTF-7/8 nor, of course, with Windows versions not
+ // supporting it
+ BOOL usedDef wxDUMMY_INITIALIZE(false),
+ *pUsedDef;
+ int flags;
+ if ( CanUseNoBestFit() && m_CodePage < 50000 )
+ {
+ // it's our lucky day
+ flags = WC_NO_BEST_FIT_CHARS;
+ pUsedDef = &usedDef;
+ }
+ else // old system or unsupported encoding
+ {
+ flags = 0;
+ pUsedDef = NULL;
+ }
+
+ const size_t len = ::WideCharToMultiByte
+ (
+ m_CodePage, // code page
+ flags, // either none or no best fit
+ pwz, // input string
+ -1, // it is (wide) NUL-terminated
+ buf, // output buffer
+ buf ? n : 0, // and its size
+ NULL, // default "replacement" char
+ pUsedDef // [out] was it used?
+ );
+
+ if ( !len )
+ {
+ // function totally failed
return (size_t)-1;
- return (n - outbuf);
+ }
+
+ // if we were really converting, check if we succeeded
+ if ( buf )
+ {
+ if ( flags )
+ {
+ // check if the conversion failed, i.e. if any replacements
+ // were done
+ if ( usedDef )
+ return (size_t)-1;
+ }
+ else // we must resort to double tripping...
+ {
+ wxWCharBuffer wcBuf(n);
+ if ( MB2WC(wcBuf.data(), buf, n) == (size_t)-1 ||
+ wcscmp(wcBuf, pwz) != 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;
+ }
+ }
+ }
+
+ // see the comment above for the reason of "len - 1"
+ return len - 1;
+ }
+
+ bool IsOk() const { return m_CodePage != -1; }
+
+private:
+ static bool CanUseNoBestFit()
+ {
+ static int s_isWin98Or2k = -1;
+
+ if ( s_isWin98Or2k == -1 )
+ {
+ int verMaj, verMin;
+ switch ( wxGetOsVersion(&verMaj, &verMin) )
+ {
+ case wxWIN95:
+ s_isWin98Or2k = verMaj >= 4 && verMin >= 10;
+ break;
+
+ case wxWINDOWS_NT:
+ s_isWin98Or2k = verMaj >= 5;
+ break;
+
+ default:
+ // unknown, be conseravtive by default
+ s_isWin98Or2k = 0;
+ }
+
+ wxASSERT_MSG( s_isWin98Or2k != -1, _T("should be set above") );
+ }
+
+ return s_isWin98Or2k == 1;
}
+ long m_CodePage;
+};
+
+#endif // wxHAVE_WIN32_MB2WC
+
+// ============================================================================
+// Mac conversion classes
+// ============================================================================
+
+#if defined(__WXMAC__) && defined(TARGET_CARBON)
+
+class wxMBConv_mac : public wxMBConv
+{
public:
- iconv_t cnv;
+ wxMBConv_mac()
+ {
+ Init(CFStringGetSystemEncoding()) ;
+ }
+
+ wxMBConv_mac(const wxChar* name)
+ {
+ Init( wxMacGetSystemEncFromFontEnc(wxFontMapper::Get()->CharsetToEncoding(name, FALSE) ) ) ;
+ }
+
+ wxMBConv_mac(wxFontEncoding encoding)
+ {
+ Init( wxMacGetSystemEncFromFontEnc(encoding) );
+ }
+
+ ~wxMBConv_mac()
+ {
+ OSStatus status = noErr ;
+ status = TECDisposeConverter(m_MB2WC_converter);
+ status = TECDisposeConverter(m_WC2MB_converter);
+ }
+
+
+ void Init( TextEncodingBase encoding)
+ {
+ OSStatus status = noErr ;
+ m_char_encoding = encoding ;
+#if SIZEOF_WCHAR_T == 4
+ m_unicode_encoding = CreateTextEncoding(kTextEncodingUnicodeDefault,0,kUnicode32BitFormat) ;
+#else
+ m_unicode_encoding = CreateTextEncoding(kTextEncodingUnicodeDefault,0,kUnicode16BitFormat) ;
+#endif
+ status = TECCreateConverter(&m_MB2WC_converter,
+ m_char_encoding,
+ m_unicode_encoding);
+ status = TECCreateConverter(&m_WC2MB_converter,
+ m_unicode_encoding,
+ m_char_encoding);
+ }
+
+ size_t MB2WC(wchar_t *buf, const char *psz, size_t n) const
+ {
+ OSStatus status = noErr ;
+ ByteCount byteOutLen ;
+ ByteCount byteInLen = strlen(psz) ;
+ wchar_t *tbuf = NULL ;
+
+ if (buf == NULL)
+ {
+ n = byteInLen ;
+ tbuf = (wchar_t*) malloc( n * SIZEOF_WCHAR_T) ;
+ }
+
+ ByteCount byteBufferLen = n * SIZEOF_WCHAR_T ;
+ status = TECConvertText(m_MB2WC_converter, (ConstTextPtr) psz , byteInLen, &byteInLen,
+ (TextPtr) (buf ? buf : tbuf) , byteBufferLen, &byteOutLen);
+
+ if ( buf == NULL )
+ free(tbuf) ;
+
+ size_t res = byteOutLen / SIZEOF_WCHAR_T ;
+ if ( buf && res < n)
+ buf[res] = 0;
+
+ return res ;
+ }
+
+ size_t WC2MB(char *buf, const wchar_t *psz, size_t n) const
+ {
+ OSStatus status = noErr ;
+ ByteCount byteOutLen ;
+ ByteCount byteInLen = wxWcslen(psz) * SIZEOF_WCHAR_T ;
+
+ char *tbuf = NULL ;
+
+ if (buf == NULL)
+ {
+ // worst case
+ n = byteInLen * 2 ;
+ tbuf = (char*) malloc( n ) ;
+ }
+
+ ByteCount byteBufferLen = n ;
+ status = TECConvertText(m_WC2MB_converter, (ConstTextPtr) psz , byteInLen, &byteInLen,
+ (TextPtr) ( buf ? buf : tbuf ) , byteBufferLen, &byteOutLen);
+
+ if ( buf == NULL )
+ free(tbuf) ;
+
+ size_t res = byteOutLen ;
+ if ( buf && res < n)
+ buf[res] = 0;
+
+ return res ;
+ }
+
+ bool IsOk() const
+ { return m_MB2WC_converter != NULL && m_WC2MB_converter != NULL ; }
+
+private:
+ TECObjectRef m_MB2WC_converter ;
+ TECObjectRef m_WC2MB_converter ;
+
+ TextEncodingBase m_char_encoding ;
+ TextEncodingBase m_unicode_encoding ;
};
-#endif // HAVE_ICONV_H
+#endif // defined(__WXMAC__) && defined(TARGET_CARBON)
+
+// ============================================================================
+// wxEncodingConverter based conversion classes
+// ============================================================================
-class EC_CharSetConverter
+#if wxUSE_FONTMAP
+
+class wxMBConv_wxwin : public wxMBConv
{
+private:
+ void Init()
+ {
+ m_ok = m2w.Init(m_enc, wxFONTENCODING_UNICODE) &&
+ w2m.Init(wxFONTENCODING_UNICODE, m_enc);
+ }
+
public:
- EC_CharSetConverter(EC_CharSet*from,EC_CharSet*to)
- { cnv.Init(from->enc,to->enc); }
+ // temporarily just use wxEncodingConverter stuff,
+ // so that it works while a better implementation is built
+ wxMBConv_wxwin(const wxChar* name)
+ {
+ if (name)
+ m_enc = wxFontMapper::Get()->CharsetToEncoding(name, false);
+ else
+ m_enc = wxFONTENCODING_SYSTEM;
+
+ Init();
+ }
- size_t Convert(char*buf, const char*psz, size_t n)
+ wxMBConv_wxwin(wxFontEncoding enc)
+ {
+ m_enc = enc;
+
+ Init();
+ }
+
+ size_t MB2WC(wchar_t *buf, const char *psz, size_t WXUNUSED(n)) const
{
size_t inbuf = strlen(psz);
- if (buf) cnv.Convert(psz,buf);
+ if (buf)
+ m2w.Convert(psz,buf);
return inbuf;
}
+ size_t WC2MB(char *buf, const wchar_t *psz, size_t WXUNUSED(n)) const
+ {
+ const size_t inbuf = wxWcslen(psz);
+ if (buf)
+ w2m.Convert(psz,buf);
+
+ return inbuf;
+ }
+
+ bool IsOk() const { return m_ok; }
+
public:
- wxEncodingConverter cnv;
+ wxFontEncoding m_enc;
+ wxEncodingConverter m2w, w2m;
+
+ // were we initialized successfully?
+ bool m_ok;
+
+ DECLARE_NO_COPY_CLASS(wxMBConv_wxwin)
};
-#else // !wxUSE_WCHAR_T
+#endif // wxUSE_FONTMAP
+
+// ============================================================================
+// wxCSConv implementation
+// ============================================================================
+
+void wxCSConv::Init()
+{
+ m_name = NULL;
+ m_convReal = NULL;
+ m_deferred = true;
+}
+
+wxCSConv::wxCSConv(const wxChar *charset)
+{
+ Init();
+
+ if ( charset )
+ {
+ SetName(charset);
+ }
+
+ m_encoding = wxFONTENCODING_SYSTEM;
+}
+
+wxCSConv::wxCSConv(wxFontEncoding encoding)
+{
+ if ( encoding == wxFONTENCODING_MAX || encoding == wxFONTENCODING_DEFAULT )
+ {
+ wxFAIL_MSG( _T("invalid encoding value in wxCSConv ctor") );
+
+ encoding = wxFONTENCODING_SYSTEM;
+ }
+
+ Init();
+
+ m_encoding = encoding;
+}
+
+wxCSConv::~wxCSConv()
+{
+ Clear();
+}
+
+wxCSConv::wxCSConv(const wxCSConv& conv)
+ : wxMBConv()
+{
+ Init();
+
+ SetName(conv.m_name);
+ m_encoding = conv.m_encoding;
+}
+
+wxCSConv& wxCSConv::operator=(const wxCSConv& conv)
+{
+ Clear();
+
+ SetName(conv.m_name);
+ m_encoding = conv.m_encoding;
+
+ return *this;
+}
+
+void wxCSConv::Clear()
+{
+ free(m_name);
+ delete m_convReal;
+
+ m_name = NULL;
+ m_convReal = NULL;
+}
+
+void wxCSConv::SetName(const wxChar *charset)
+{
+ if (charset)
+ {
+ m_name = wxStrdup(charset);
+ m_deferred = true;
+ }
+}
+
+wxMBConv *wxCSConv::DoCreate() const
+{
+ // 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 )
+ {
+ // don't convert at all
+ return NULL;
+ }
+
+ // we trust OS to do conversion better than we can so try external
+ // conversion methods first
+ //
+ // the full order is:
+ // 1. OS conversion (iconv() under Unix or Win32 API)
+ // 2. hard coded conversions for UTF
+ // 3. wxEncodingConverter as fall back
+
+ // step (1)
+#ifdef HAVE_ICONV
+#if !wxUSE_FONTMAP
+ if ( m_name )
+#endif // !wxUSE_FONTMAP
+ {
+ wxString name(m_name);
+
+#if wxUSE_FONTMAP
+ if ( name.empty() )
+ name = wxFontMapper::Get()->GetEncodingName(m_encoding);
+#endif // wxUSE_FONTMAP
+
+ wxMBConv_iconv *conv = new wxMBConv_iconv(name);
+ if ( conv->IsOk() )
+ return conv;
+
+ delete conv;
+ }
+#endif // HAVE_ICONV
+
+#ifdef wxHAVE_WIN32_MB2WC
+ {
+ wxMBConv_win32 *conv = m_name ? new wxMBConv_win32(m_name)
+ : new wxMBConv_win32(m_encoding);
+ if ( conv->IsOk() )
+ return conv;
+
+ delete conv;
+ }
+#endif // wxHAVE_WIN32_MB2WC
+#if defined(__WXMAC__)
+ {
+ if ( m_name || ( m_encoding < wxFONTENCODING_UTF16BE ) )
+ {
+
+ wxMBConv_mac *conv = m_name ? new wxMBConv_mac(m_name)
+ : new wxMBConv_mac(m_encoding);
+ if ( conv->IsOk() )
+ return conv;
+
+ delete conv;
+ }
+ }
+#endif
+ // step (2)
+ wxFontEncoding enc = m_encoding;
+#if wxUSE_FONTMAP
+ if ( enc == wxFONTENCODING_SYSTEM && m_name )
+ {
+ // use "false" to suppress interactive dialogs -- we can be called from
+ // anywhere and popping up a dialog from here is the last thing we want to
+ // do
+ enc = wxFontMapper::Get()->CharsetToEncoding(m_name, false);
+ }
+#endif // wxUSE_FONTMAP
+
+ switch ( enc )
+ {
+ case wxFONTENCODING_UTF7:
+ return new wxMBConvUTF7;
+
+ case wxFONTENCODING_UTF8:
+ return new wxMBConvUTF8;
+
+ case wxFONTENCODING_UTF16BE:
+ return new wxMBConvUTF16BE;
+
+ case wxFONTENCODING_UTF16LE:
+ return new wxMBConvUTF16LE;
+
+ case wxFONTENCODING_UTF32BE:
+ return new wxMBConvUTF32BE;
+
+ case wxFONTENCODING_UTF32LE:
+ return new wxMBConvUTF32LE;
+
+ default:
+ // nothing to do but put here to suppress gcc warnings
+ ;
+ }
+
+ // step (3)
+#if wxUSE_FONTMAP
+ {
+ wxMBConv_wxwin *conv = m_name ? new wxMBConv_wxwin(m_name)
+ : new wxMBConv_wxwin(m_encoding);
+ if ( conv->IsOk() )
+ return conv;
+
+ delete conv;
+ }
+#endif // wxUSE_FONTMAP
+
+ // NB: This is a hack to prevent deadlock. What could otherwise happen
+ // in Unicode build: wxConvLocal creation ends up being here
+ // because of some failure and logs the error. But wxLog will try to
+ // attach timestamp, for which it will need wxConvLocal (to convert
+ // time to char* and then wchar_t*), but that fails, tries to log
+ // error, but wxLog has a (already locked) critical section that
+ // guards static buffer.
+ static bool alreadyLoggingError = false;
+ if (!alreadyLoggingError)
+ {
+ alreadyLoggingError = true;
+ wxLogError(_("Cannot convert from the charset '%s'!"),
+ m_name ? m_name
+ :
+#if wxUSE_FONTMAP
+ wxFontMapper::GetEncodingDescription(m_encoding).c_str()
+#else // !wxUSE_FONTMAP
+ wxString::Format(_("encoding %s"), m_encoding).c_str()
+#endif // wxUSE_FONTMAP/!wxUSE_FONTMAP
+ );
+ alreadyLoggingError = false;
+ }
+
+ return NULL;
+}
+
+void wxCSConv::CreateConvIfNeeded() const
+{
+ if ( m_deferred )
+ {
+ wxCSConv *self = (wxCSConv *)this; // const_cast
+
+#if wxUSE_INTL
+ // if we don't have neither the name nor the encoding, use the default
+ // encoding for this system
+ if ( !m_name && m_encoding == wxFONTENCODING_SYSTEM )
+ {
+ self->m_name = wxStrdup(wxLocale::GetSystemEncodingName());
+ }
+#endif // wxUSE_INTL
+
+ self->m_convReal = DoCreate();
+ self->m_deferred = false;
+ }
+}
+
+size_t wxCSConv::MB2WC(wchar_t *buf, const char *psz, size_t n) const
+{
+ CreateConvIfNeeded();
+
+ if (m_convReal)
+ return m_convReal->MB2WC(buf, psz, n);
+
+ // latin-1 (direct)
+ size_t len = strlen(psz);
+
+ if (buf)
+ {
+ for (size_t c = 0; c <= len; c++)
+ buf[c] = (unsigned char)(psz[c]);
+ }
+
+ return len;
+}
+
+size_t wxCSConv::WC2MB(char *buf, const wchar_t *psz, size_t n) const
+{
+ CreateConvIfNeeded();
+
+ if (m_convReal)
+ return m_convReal->WC2MB(buf, psz, n);
+
+ // latin-1 (direct)
+ const size_t len = wxWcslen(psz);
+ if (buf)
+ {
+ for (size_t c = 0; c <= len; c++)
+ {
+ if (psz[c] > 0xFF)
+ return (size_t)-1;
+ buf[c] = psz[c];
+ }
+ }
+ else
+ {
+ for (size_t c = 0; c <= len; c++)
+ {
+ if (psz[c] > 0xFF)
+ return (size_t)-1;
+ }
+ }
+
+ return len;
+}
// ----------------------------------------------------------------------------
-// stand-ins in absence of wchar_t
+// globals
// ----------------------------------------------------------------------------
-WXDLLEXPORT_DATA(wxMBConv) wxConvLibc, wxConvFile;
+#ifdef __WINDOWS__
+ static wxMBConv_win32 wxConvLibcObj;
+#else
+ static wxMBConvLibc wxConvLibcObj;
+#endif
+
+static wxCSConv wxConvLocalObj(wxFONTENCODING_SYSTEM);
+static wxCSConv wxConvISO8859_1Obj(wxFONTENCODING_ISO8859_1);
+static wxMBConvUTF7 wxConvUTF7Obj;
+static wxMBConvUTF8 wxConvUTF8Obj;
+
+
+WXDLLIMPEXP_DATA_BASE(wxMBConv&) wxConvLibc = wxConvLibcObj;
+WXDLLIMPEXP_DATA_BASE(wxCSConv&) wxConvLocal = wxConvLocalObj;
+WXDLLIMPEXP_DATA_BASE(wxCSConv&) wxConvISO8859_1 = wxConvISO8859_1Obj;
+WXDLLIMPEXP_DATA_BASE(wxMBConvUTF7&) wxConvUTF7 = wxConvUTF7Obj;
+WXDLLIMPEXP_DATA_BASE(wxMBConvUTF8&) wxConvUTF8 = wxConvUTF8Obj;
+WXDLLIMPEXP_DATA_BASE(wxMBConv *) wxConvCurrent = &wxConvLibcObj;
+
+#else // !wxUSE_WCHAR_T
+
+// stand-ins in absence of wchar_t
+WXDLLIMPEXP_DATA_BASE(wxMBConv) wxConvLibc,
+ wxConvISO8859_1,
+ wxConvLocal,
+ wxConvUTF8;
-#endif // wxUSE_WCHAR_T
+#endif // wxUSE_WCHAR_T/!wxUSE_WCHAR_T
projects / wxWidgets.git / blobdiff