move virtual GetPath() implementation out of line to work around an apparent Fedora...

[wxWidgets.git] / include / wx / strconv.h

diff --git a/include/wx/strconv.h b/include/wx/strconv.h
index 72423f8c771bc6602da17812d8e6d37b2b585819..0061aa770c524c6a0a70689ba21d3d0686bd0eed 100644 (file)
--- a/include/wx/strconv.h
+++ b/include/wx/strconv.h
@@ -1,23 +1,20 @@
 ///////////////////////////////////////////////////////////////////////////////
 // Name:        strconv.h
 // Purpose:     conversion routines for char sets any Unicode
 ///////////////////////////////////////////////////////////////////////////////
 // Name:        strconv.h
 // Purpose:     conversion routines for char sets any Unicode

-// Author:      Robert Roebling, Ove Kaaven
+// Author:      Ove Kaaven, Robert Roebling, Vadim Zeitlin

 // Modified by:
 // Created:     29/01/98
 // RCS-ID:      $Id$
 // Modified by:
 // Created:     29/01/98
 // RCS-ID:      $Id$

-// Copyright:   (c) 1998 Ove Kaaven, Robert Roebling, Vadim Zeitlin
+// Copyright:   (c) 1998 Ove Kaaven, Robert Roebling
+//              (c) 1998-2006 Vadim Zeitlin

 // Licence:     wxWindows licence
 ///////////////////////////////////////////////////////////////////////////////
 
 // Licence:     wxWindows licence
 ///////////////////////////////////////////////////////////////////////////////
 

-#ifndef _WX_WXSTRCONVH__
-#define _WX_WXSTRCONVH__
-
-#if defined(__GNUG__) && !defined(NO_GCC_PRAGMA)
-  #pragma interface "strconv.h"
-#endif
+#ifndef _WX_STRCONV_H_
+#define _WX_STRCONV_H_

 
 #include "wx/defs.h"
 
 #include "wx/defs.h"

-#include "wx/wxchar.h"
+#include "wx/chartype.h"

 #include "wx/buffer.h"
 
 #ifdef __DIGITALMARS__
 #include "wx/buffer.h"
 
 #ifdef __DIGITALMARS__


@@ -28,34 +25,88 @@
 #  undef __BSEXCPT__
 #endif
 
 #  undef __BSEXCPT__
 #endif
 

+#ifndef __WXPALMOS5__

 #include <stdlib.h>
 #include <stdlib.h>

+#endif // ! __WXPALMOS5__

 
 #if wxUSE_WCHAR_T
 
 
 #if wxUSE_WCHAR_T
 

+class WXDLLIMPEXP_FWD_BASE wxString;
+
+// the error value returned by wxMBConv methods
+#define wxCONV_FAILED ((size_t)-1)
+

 // ----------------------------------------------------------------------------
 // wxMBConv (abstract base class for conversions)
 // ----------------------------------------------------------------------------
 
 // ----------------------------------------------------------------------------
 // wxMBConv (abstract base class for conversions)
 // ----------------------------------------------------------------------------
 

+// When deriving a new class from wxMBConv you must reimplement ToWChar() and
+// FromWChar() methods which are not pure virtual only for historical reasons,
+// don't let the fact that the existing classes implement MB2WC/WC2MB() instead
+// confuse you.
+//
+// You also have to implement Clone() to allow copying the conversions
+// polymorphically.
+//
+// And you might need to override GetMBNulLen() as well.

 class WXDLLIMPEXP_BASE wxMBConv
 {
 public:
 class WXDLLIMPEXP_BASE wxMBConv
 {
 public:

-    // the actual conversion takes place here
+    // The functions doing actual conversion from/to narrow to/from wide
+    // character strings.
+    //
+    // On success, the return value is the length (i.e. the number of
+    // characters, not bytes) of the converted string including any trailing
+    // L'\0' or (possibly multiple) '\0'(s). If the conversion fails or if
+    // there is not enough space for everything, including the trailing NUL
+    // character(s), in the output buffer, wxCONV_FAILED is returned.
+    //
+    // In the special case when dstLen is 0 (outputBuf may be NULL then) the
+    // return value is the length of the needed buffer but nothing happens
+    // otherwise. If srcLen is wxNO_LEN, the entire string, up to and
+    // including the trailing NUL(s), is converted, otherwise exactly srcLen
+    // bytes are.
+    //
+    // Typical usage:

     //
     //

-    // note that outputSize is the size of the output buffer, not the length of input
-    // (the latter is always supposed to be NUL-terminated)
-    virtual size_t MB2WC(wchar_t *outputBuf, const char *psz, size_t outputSize) const = 0;
-    virtual size_t WC2MB(char *outputBuf, const wchar_t *psz, size_t outputSize) const = 0;
+    //          size_t dstLen = conv.ToWChar(NULL, 0, src);
+    //          if ( dstLen != wxCONV_FAILED )
+    //              ... handle error ...
+    //          wchar_t *wbuf = new wchar_t[dstLen];
+    //          conv.ToWChar(wbuf, dstLen, src);
+    //
+    virtual size_t ToWChar(wchar_t *dst, size_t dstLen,
+                           const char *src, size_t srcLen = wxNO_LEN) const;
+
+    virtual size_t FromWChar(char *dst, size_t dstLen,
+                             const wchar_t *src, size_t srcLen = wxNO_LEN) const;

 
 

-    // MB <-> WC
-    const wxWCharBuffer cMB2WC(const char *psz) const;
-    const wxCharBuffer cWC2MB(const wchar_t *psz) const;

 
 

-    // MB <-> WC for strings with embedded null characters
+    // Convenience functions for translating NUL-terminated strings: returns
+    // the buffer containing the converted string or NULL pointer if the
+    // conversion failed.
+    const wxWCharBuffer cMB2WC(const char *in) const;
+    const wxCharBuffer cWC2MB(const wchar_t *in) const;
+
+    // Convenience functions for converting strings which may contain embedded
+    // NULs and don't have to be NUL-terminated.
+    //
+    // inLen is the length of the buffer including trailing NUL if any or
+    // wxNO_LEN if the input is NUL-terminated.

     //
     //

-    // pszLen length of the input string
-    // pOutSize gets the final size of the converted string
-    const wxWCharBuffer cMB2WC(const char *psz, size_t pszLen, size_t* pOutSize) const;
-    const wxCharBuffer cWC2MB(const wchar_t *psz, size_t pszLen, size_t* pOutSize) const;
+    // outLen receives, if not NULL, the length of the converted string or 0 if
+    // the conversion failed (returning 0 and not -1 in this case makes it
+    // difficult to distinguish between failed conversion and empty input but
+    // this is done for backwards compatibility). Notice that the rules for
+    // whether outLen accounts or not for the last NUL are the same as for
+    // To/FromWChar() above: if inLen is specified, outLen is exactly the
+    // number of characters converted, whether the last one of them was NUL or
+    // not. But if inLen == wxNO_LEN then outLen doesn't account for the last
+    // NUL even though it is present.
+    const wxWCharBuffer
+        cMB2WC(const char *in, size_t inLen, size_t *outLen) const;
+    const wxCharBuffer
+        cWC2MB(const wchar_t *in, size_t inLen, size_t *outLen) const;

 
     // convenience functions for converting MB or WC to/from wxWin default
 #if wxUSE_UNICODE
 
     // convenience functions for converting MB or WC to/from wxWin default
 #if wxUSE_UNICODE


@@ -70,6 +121,53 @@ public:
     const wxWCharBuffer cWX2WC(const char *psz) const { return cMB2WC(psz); }
 #endif // Unicode/ANSI
 
     const wxWCharBuffer cWX2WC(const char *psz) const { return cMB2WC(psz); }
 #endif // Unicode/ANSI
 

+    // this function is used in the implementation of cMB2WC() to distinguish
+    // between the following cases:
+    //
+    //      a) var width encoding with strings terminated by a single NUL
+    //         (usual multibyte encodings): return 1 in this case
+    //      b) fixed width encoding with 2 bytes/char and so terminated by
+    //         2 NULs (UTF-16/UCS-2 and variants): return 2 in this case
+    //      c) fixed width encoding with 4 bytes/char and so terminated by
+    //         4 NULs (UTF-32/UCS-4 and variants): return 4 in this case
+    //
+    // anything else is not supported currently and -1 should be returned
+    virtual size_t GetMBNulLen() const { return 1; }
+
+    // return the maximal value currently returned by GetMBNulLen() for any
+    // encoding
+    static size_t GetMaxMBNulLen() { return 4 /* for UTF-32 */; }
+
+#if wxUSE_UNICODE_UTF8
+    // return true if the converter's charset is UTF-8, i.e. char* strings
+    // decoded using this object can be directly copied to wxString's internal
+    // storage without converting to WC and than back to UTF-8 MB string
+    virtual bool IsUTF8() const { return false; }
+#endif
+
+    // The old conversion functions. The existing classes currently mostly
+    // implement these ones but we're in transition to using To/FromWChar()
+    // instead and any new classes should implement just the new functions.
+    // For now, however, we provide default implementation of To/FromWChar() in
+    // this base class in terms of MB2WC/WC2MB() to avoid having to rewrite all
+    // the conversions at once.
+    //
+    // On success, the return value is the length (i.e. the number of
+    // characters, not bytes) not counting the trailing NUL(s) of the converted
+    // string. On failure, (size_t)-1 is returned. In the special case when
+    // outputBuf is NULL the return value is the same one but nothing is
+    // written to the buffer.
+    //
+    // Note that outLen is the length of the output buffer, not the length of
+    // the input (which is always supposed to be terminated by one or more
+    // NULs, as appropriate for the encoding)!
+    virtual size_t MB2WC(wchar_t *out, const char *in, size_t outLen) const;
+    virtual size_t WC2MB(char *out, const wchar_t *in, size_t outLen) const;
+
+
+    // make a heap-allocated copy of this object
+    virtual wxMBConv *Clone() const = 0;
+

     // virtual dtor for any base class
     virtual ~wxMBConv();
 };
     // virtual dtor for any base class
     virtual ~wxMBConv();
 };


@@ -84,12 +182,64 @@ class WXDLLIMPEXP_BASE wxMBConvLibc : public wxMBConv
 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;
 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;

+
+    virtual wxMBConv *Clone() const { return new wxMBConvLibc; }
+
+#if wxUSE_UNICODE_UTF8
+    virtual bool IsUTF8() const { return wxLocaleIsUtf8; }
+#endif

 };
 
 };
 

-// not very accurately named because it is not necessarily of type wxMBConvLibc
-// (but the name can't eb changed because of backwards compatibility) default
-// conversion
-WXDLLIMPEXP_DATA_BASE(extern wxMBConv&) wxConvLibc;
+#ifdef __UNIX__
+
+// ----------------------------------------------------------------------------
+// wxConvBrokenFileNames is made for Unix 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 on disk.
+// ----------------------------------------------------------------------------
+
+class WXDLLIMPEXP_BASE wxConvBrokenFileNames : public wxMBConv
+{
+public:
+    wxConvBrokenFileNames(const wxString& charset);
+    wxConvBrokenFileNames(const wxConvBrokenFileNames& conv)
+        : wxMBConv(),
+          m_conv(conv.m_conv ? conv.m_conv->Clone() : NULL)
+    {
+    }
+    virtual ~wxConvBrokenFileNames() { delete m_conv; }
+
+    virtual size_t MB2WC(wchar_t *out, const char *in, size_t outLen) const
+    {
+        return m_conv->MB2WC(out, in, outLen);
+    }
+
+    virtual size_t WC2MB(char *out, const wchar_t *in, size_t outLen) const
+    {
+        return m_conv->WC2MB(out, in, outLen);
+    }
+
+    virtual size_t GetMBNulLen() const
+    {
+        // cast needed to call a private function
+        return m_conv->GetMBNulLen();
+    }
+
+#if wxUSE_UNICODE_UTF8
+    virtual bool IsUTF8() const { return m_conv->IsUTF8(); }
+#endif
+
+    virtual wxMBConv *Clone() const { return new wxConvBrokenFileNames(*this); }
+
+private:
+    // the conversion object we forward to
+    wxMBConv *m_conv;
+
+    wxDECLARE_NO_ASSIGN_CLASS(wxConvBrokenFileNames);
+};
+
+#endif // __UNIX__

 
 // ----------------------------------------------------------------------------
 // wxMBConvUTF7 (for conversion using UTF7 encoding)
 
 // ----------------------------------------------------------------------------
 // wxMBConvUTF7 (for conversion using UTF7 encoding)


@@ -98,67 +248,232 @@ WXDLLIMPEXP_DATA_BASE(extern wxMBConv&) wxConvLibc;
 class WXDLLIMPEXP_BASE wxMBConvUTF7 : public wxMBConv
 {
 public:
 class WXDLLIMPEXP_BASE wxMBConvUTF7 : public wxMBConv
 {
 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;
-};
+    wxMBConvUTF7() { }
+
+    // compiler-generated copy ctor, assignment operator and dtor are ok
+    // (assuming it's ok to copy the shift state -- not really sure about it)

 
 

-WXDLLIMPEXP_DATA_BASE(extern wxMBConvUTF7&) wxConvUTF7;
+    virtual size_t ToWChar(wchar_t *dst, size_t dstLen,
+                           const char *src, size_t srcLen = wxNO_LEN) const;
+    virtual size_t FromWChar(char *dst, size_t dstLen,
+                             const wchar_t *src, size_t srcLen = wxNO_LEN) const;
+
+    virtual wxMBConv *Clone() const { return new wxMBConvUTF7; }
+
+private:
+    // UTF-7 decoder/encoder may be in direct mode or in shifted mode after a
+    // '+' (and until the '-' or any other non-base64 character)
+    struct StateMode
+    {
+        enum Mode
+        {
+            Direct,     // pass through state
+            Shifted     // after a '+' (and before '-')
+        };
+    };
+
+    // the current decoder state: this is only used by ToWChar() if srcLen
+    // parameter is not wxNO_LEN, when working on the entire NUL-terminated
+    // strings we neither update nor use the state
+    class DecoderState : private StateMode
+    {
+    private:
+        // current state: this one is private as we want to enforce the use of
+        // ToDirect/ToShifted() methods below
+        Mode mode;
+
+    public:
+        // the initial state is direct
+        DecoderState() { mode = Direct; }
+
+        // switch to/from shifted mode
+        void ToDirect() { mode = Direct; }
+        void ToShifted() { mode = Shifted; accum = bit = 0; isLSB = false; }
+
+        bool IsDirect() const { return mode == Direct; }
+        bool IsShifted() const { return mode == Shifted; }
+
+
+        // these variables are only used in shifted mode
+
+        unsigned int accum; // accumulator of the bit we've already got
+        unsigned int bit;   // the number of bits consumed mod 8
+        unsigned char msb;  // the high byte of UTF-16 word
+        bool isLSB;         // whether we're decoding LSB or MSB of UTF-16 word
+    };
+
+    DecoderState m_stateDecoder;
+
+
+    // encoder state is simpler as we always receive entire Unicode characters
+    // on input
+    class EncoderState : private StateMode
+    {
+    private:
+        Mode mode;
+
+    public:
+        EncoderState() { mode = Direct; }
+
+        void ToDirect() { mode = Direct; }
+        void ToShifted() { mode = Shifted; accum = bit = 0; }
+
+        bool IsDirect() const { return mode == Direct; }
+        bool IsShifted() const { return mode == Shifted; }
+
+        unsigned int accum;
+        unsigned int bit;
+    };
+
+    EncoderState m_stateEncoder;
+};

 
 // ----------------------------------------------------------------------------
 // wxMBConvUTF8 (for conversion using UTF8 encoding)
 // ----------------------------------------------------------------------------
 
 
 // ----------------------------------------------------------------------------
 // wxMBConvUTF8 (for conversion using UTF8 encoding)
 // ----------------------------------------------------------------------------
 

-class WXDLLIMPEXP_BASE wxMBConvUTF8 : public wxMBConv
+// this is the real UTF-8 conversion class, it has to be called "strict UTF-8"
+// for compatibility reasons: the wxMBConvUTF8 class below also supports lossy
+// conversions if it is created with non default options
+class WXDLLIMPEXP_BASE wxMBConvStrictUTF8 : public wxMBConv

 {
 public:
 {
 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;
+    // compiler-generated default ctor and other methods are ok
+
+    virtual size_t ToWChar(wchar_t *dst, size_t dstLen,
+                           const char *src, size_t srcLen = wxNO_LEN) const;
+    virtual size_t FromWChar(char *dst, size_t dstLen,
+                             const wchar_t *src, size_t srcLen = wxNO_LEN) const;
+
+    virtual wxMBConv *Clone() const { return new wxMBConvStrictUTF8(); }
+
+#if wxUSE_UNICODE_UTF8
+    // NB: other mapping modes are not, strictly speaking, UTF-8, so we can't
+    //     take the shortcut in that case
+    virtual bool IsUTF8() const { return true; }
+#endif

 };
 
 };
 

-WXDLLIMPEXP_DATA_BASE(extern wxMBConvUTF8&) wxConvUTF8;
+class WXDLLIMPEXP_BASE wxMBConvUTF8 : public wxMBConvStrictUTF8
+{
+public:
+    enum
+    {
+        MAP_INVALID_UTF8_NOT = 0,
+        MAP_INVALID_UTF8_TO_PUA = 1,
+        MAP_INVALID_UTF8_TO_OCTAL = 2
+    };
+
+    wxMBConvUTF8(int options = MAP_INVALID_UTF8_NOT) : m_options(options) { }
+
+    virtual size_t ToWChar(wchar_t *dst, size_t dstLen,
+                           const char *src, size_t srcLen = wxNO_LEN) const;
+    virtual size_t FromWChar(char *dst, size_t dstLen,
+                             const wchar_t *src, size_t srcLen = wxNO_LEN) const;
+
+    virtual wxMBConv *Clone() const { return new wxMBConvUTF8(m_options); }
+
+#if wxUSE_UNICODE_UTF8
+    // NB: other mapping modes are not, strictly speaking, UTF-8, so we can't
+    //     take the shortcut in that case
+    virtual bool IsUTF8() const { return m_options == MAP_INVALID_UTF8_NOT; }
+#endif
+
+private:
+    int m_options;
+};
+
+// ----------------------------------------------------------------------------
+// wxMBConvUTF16Base: for both LE and BE variants
+// ----------------------------------------------------------------------------
+
+class WXDLLIMPEXP_BASE wxMBConvUTF16Base : public wxMBConv
+{
+public:
+    enum { BYTES_PER_CHAR = 2 };
+
+    virtual size_t GetMBNulLen() const { return BYTES_PER_CHAR; }
+
+protected:
+    // return the length of the buffer using srcLen if it's not wxNO_LEN and
+    // computing the length ourselves if it is; also checks that the length is
+    // even if specified as we need an entire number of UTF-16 characters and
+    // returns wxNO_LEN which indicates error if it is odd
+    static size_t GetLength(const char *src, size_t srcLen);
+};

 
 // ----------------------------------------------------------------------------
 // wxMBConvUTF16LE (for conversion using UTF16 Little Endian encoding)
 // ----------------------------------------------------------------------------
 
 
 // ----------------------------------------------------------------------------
 // wxMBConvUTF16LE (for conversion using UTF16 Little Endian encoding)
 // ----------------------------------------------------------------------------
 

-class WXDLLIMPEXP_BASE wxMBConvUTF16LE : public wxMBConv
+class WXDLLIMPEXP_BASE wxMBConvUTF16LE : public wxMBConvUTF16Base

 {
 public:
 {
 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;
+    virtual size_t ToWChar(wchar_t *dst, size_t dstLen,
+                           const char *src, size_t srcLen = wxNO_LEN) const;
+    virtual size_t FromWChar(char *dst, size_t dstLen,
+                             const wchar_t *src, size_t srcLen = wxNO_LEN) const;
+    virtual wxMBConv *Clone() const { return new wxMBConvUTF16LE; }

 };
 
 // ----------------------------------------------------------------------------
 // wxMBConvUTF16BE (for conversion using UTF16 Big Endian encoding)
 // ----------------------------------------------------------------------------
 
 };
 
 // ----------------------------------------------------------------------------
 // wxMBConvUTF16BE (for conversion using UTF16 Big Endian encoding)
 // ----------------------------------------------------------------------------
 

-class WXDLLIMPEXP_BASE wxMBConvUTF16BE : public wxMBConv
+class WXDLLIMPEXP_BASE wxMBConvUTF16BE : public wxMBConvUTF16Base

 {
 public:
 {
 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;
+    virtual size_t ToWChar(wchar_t *dst, size_t dstLen,
+                           const char *src, size_t srcLen = wxNO_LEN) const;
+    virtual size_t FromWChar(char *dst, size_t dstLen,
+                             const wchar_t *src, size_t srcLen = wxNO_LEN) const;
+    virtual wxMBConv *Clone() const { return new wxMBConvUTF16BE; }

 };
 
 // ----------------------------------------------------------------------------
 };
 
 // ----------------------------------------------------------------------------

-// wxMBConvUCS4LE (for conversion using UTF32 Little Endian encoding)
+// wxMBConvUTF32Base: base class for both LE and BE variants

 // ----------------------------------------------------------------------------
 
 // ----------------------------------------------------------------------------
 

-class WXDLLIMPEXP_BASE wxMBConvUTF32LE : public wxMBConv
+class WXDLLIMPEXP_BASE wxMBConvUTF32Base : public wxMBConv

 {
 public:
 {
 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;
+    enum { BYTES_PER_CHAR = 4 };
+
+    virtual size_t GetMBNulLen() const { return BYTES_PER_CHAR; }
+
+protected:
+    // this is similar to wxMBConvUTF16Base method with the same name except
+    // that, of course, it verifies that length is divisible by 4 if given and
+    // not by 2
+    static size_t GetLength(const char *src, size_t srcLen);

 };
 
 // ----------------------------------------------------------------------------
 };
 
 // ----------------------------------------------------------------------------

-// wxMBConvUCS4BE (for conversion using UTF32 Big Endian encoding)
+// wxMBConvUTF32LE (for conversion using UTF32 Little Endian encoding)

 // ----------------------------------------------------------------------------
 
 // ----------------------------------------------------------------------------
 

-class WXDLLIMPEXP_BASE wxMBConvUTF32BE : public wxMBConv
+class WXDLLIMPEXP_BASE wxMBConvUTF32LE : public wxMBConvUTF32Base

 {
 public:
 {
 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;
+    virtual size_t ToWChar(wchar_t *dst, size_t dstLen,
+                           const char *src, size_t srcLen = wxNO_LEN) const;
+    virtual size_t FromWChar(char *dst, size_t dstLen,
+                             const wchar_t *src, size_t srcLen = wxNO_LEN) const;
+    virtual wxMBConv *Clone() const { return new wxMBConvUTF32LE; }
+};
+
+// ----------------------------------------------------------------------------
+// wxMBConvUTF32BE (for conversion using UTF32 Big Endian encoding)
+// ----------------------------------------------------------------------------
+
+class WXDLLIMPEXP_BASE wxMBConvUTF32BE : public wxMBConvUTF32Base
+{
+public:
+    virtual size_t ToWChar(wchar_t *dst, size_t dstLen,
+                           const char *src, size_t srcLen = wxNO_LEN) const;
+    virtual size_t FromWChar(char *dst, size_t dstLen,
+                             const wchar_t *src, size_t srcLen = wxNO_LEN) const;
+    virtual wxMBConv *Clone() const { return new wxMBConvUTF32BE; }

 };
 
 // ----------------------------------------------------------------------------
 };
 
 // ----------------------------------------------------------------------------


@@ -172,7 +487,7 @@ class WXDLLIMPEXP_BASE wxCSConv : public wxMBConv
 public:
     // we can be created either from charset name or from an encoding constant
     // but we can't have both at once
 public:
     // we can be created either from charset name or from an encoding constant
     // but we can't have both at once

-    wxCSConv(const wxChar *charset);
+    wxCSConv(const wxString& charset);

     wxCSConv(wxFontEncoding encoding);
 
     wxCSConv(const wxCSConv& conv);
     wxCSConv(wxFontEncoding encoding);
 
     wxCSConv(const wxCSConv& conv);


@@ -180,10 +495,22 @@ public:
 
     wxCSConv& operator=(const wxCSConv& conv);
 
 
     wxCSConv& operator=(const wxCSConv& conv);
 

-    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;
+    virtual size_t ToWChar(wchar_t *dst, size_t dstLen,
+                           const char *src, size_t srcLen = wxNO_LEN) const;
+    virtual size_t FromWChar(char *dst, size_t dstLen,
+                             const wchar_t *src, size_t srcLen = wxNO_LEN) const;
+    virtual size_t GetMBNulLen() const;
+
+#if wxUSE_UNICODE_UTF8
+    virtual bool IsUTF8() const;
+#endif

 
 

-    void Clear() ;
+    virtual wxMBConv *Clone() const { return new wxCSConv(*this); }
+
+    void Clear();
+
+    // return true if the conversion could be initialized successfully
+    bool IsOk() const;

 
 private:
     // common part of all ctors
 
 private:
     // common part of all ctors


@@ -196,13 +523,19 @@ private:
     wxMBConv *DoCreate() const;
 
     // set the name (may be only called when m_name == NULL), makes copy of
     wxMBConv *DoCreate() const;
 
     // set the name (may be only called when m_name == NULL), makes copy of

-    // charset string
-    void SetName(const wxChar *charset);
+    // the charset string
+    void SetName(const char *charset);

 
 
 
 

+    // m_name may be NULL in which case m_encoding should be used
+    //

     // note that we can't use wxString here because of compilation
     // dependencies: we're included from wx/string.h
     // note that we can't use wxString here because of compilation
     // dependencies: we're included from wx/string.h

-    wxChar *m_name;
+    char *m_name;
+
+    // may be wxFONTENCODING_SYSTEM in which case m_name is used
+    //
+    // if m_name is NULL, then we should use the default system encoding

     wxFontEncoding m_encoding;
 
     // use CreateConvIfNeeded() before accessing m_convReal!
     wxFontEncoding m_encoding;
 
     // use CreateConvIfNeeded() before accessing m_convReal!


@@ -210,15 +543,68 @@ private:
     bool m_deferred;
 };
 
     bool m_deferred;
 };
 

-#ifdef __WXOSX__
-#define wxConvFile wxConvUTF8
-#else
-#define wxConvFile wxConvLocal
-#endif

 
 

-WXDLLIMPEXP_DATA_BASE(extern wxCSConv&) wxConvLocal;
-WXDLLIMPEXP_DATA_BASE(extern wxCSConv&) wxConvISO8859_1;
-WXDLLIMPEXP_DATA_BASE(extern wxMBConv *) wxConvCurrent;
+// ----------------------------------------------------------------------------
+// declare predefined conversion objects
+// ----------------------------------------------------------------------------
+
+// Note: this macro is an implementation detail (see the comment in
+// strconv.cpp). The wxGet_XXX() and wxGet_XXXPtr() functions shouldn't be
+// used by user code and neither should XXXPtr, use the wxConvXXX macro
+// instead.
+#define WX_DECLARE_GLOBAL_CONV(klass, name)                             \
+    extern WXDLLIMPEXP_DATA_BASE(klass*) name##Ptr;                     \
+    extern WXDLLIMPEXP_BASE klass* wxGet_##name##Ptr();                 \
+    inline klass& wxGet_##name()                                        \
+    {                                                                   \
+        if ( !name##Ptr )                                               \
+            name##Ptr = wxGet_##name##Ptr();                            \
+        return *name##Ptr;                                              \
+    }
+
+
+// conversion to be used with all standard functions affected by locale, e.g.
+// strtol(), strftime(), ...
+WX_DECLARE_GLOBAL_CONV(wxMBConv, wxConvLibc)
+#define wxConvLibc wxGet_wxConvLibc()
+
+// conversion ISO-8859-1/UTF-7/UTF-8 <-> wchar_t
+WX_DECLARE_GLOBAL_CONV(wxCSConv, wxConvISO8859_1)
+#define wxConvISO8859_1 wxGet_wxConvISO8859_1()
+
+WX_DECLARE_GLOBAL_CONV(wxMBConvStrictUTF8, wxConvUTF8)
+#define wxConvUTF8 wxGet_wxConvUTF8()
+
+WX_DECLARE_GLOBAL_CONV(wxMBConvUTF7, wxConvUTF7)
+#define wxConvUTF7 wxGet_wxConvUTF7()
+
+// conversion used for the file names on the systems where they're not Unicode
+// (basically anything except Windows)
+//
+// this is used by all file functions, can be changed by the application
+//
+// by default UTF-8 under Mac OS X and wxConvLibc elsewhere (but it's not used
+// under Windows normally)
+extern WXDLLIMPEXP_DATA_BASE(wxMBConv *) wxConvFileName;
+
+// backwards compatible define
+#define wxConvFile (*wxConvFileName)
+
+// the current conversion object, may be set to any conversion, is used by
+// default in a couple of places inside wx (initially same as wxConvLibc)
+extern WXDLLIMPEXP_DATA_BASE(wxMBConv *) wxConvCurrent;
+
+// the conversion corresponding to the current locale
+WX_DECLARE_GLOBAL_CONV(wxCSConv, wxConvLocal)
+#define wxConvLocal wxGet_wxConvLocal()
+
+// the conversion corresponding to the encoding of the standard UI elements
+//
+// by default this is the same as wxConvLocal but may be changed if the program
+// needs to use a fixed encoding
+extern WXDLLIMPEXP_DATA_BASE(wxMBConv *) wxConvUI;
+
+#undef WX_DECLARE_GLOBAL_CONV

 
 // ----------------------------------------------------------------------------
 // endianness-dependent conversions
 
 // ----------------------------------------------------------------------------
 // endianness-dependent conversions


@@ -236,27 +622,20 @@ WXDLLIMPEXP_DATA_BASE(extern wxMBConv *) wxConvCurrent;
 // filename conversion macros
 // ----------------------------------------------------------------------------
 
 // filename conversion macros
 // ----------------------------------------------------------------------------
 

-// filenames are multibyte on Unix and probably widechar on Windows?
-#if defined(__UNIX__) || defined(__BORLANDC__) || defined(__WXMAC__ )
-    #define wxMBFILES 1
-#else
-    #define wxMBFILES 0
-#endif
-
+// filenames are multibyte on Unix and widechar on Windows

 #if wxMBFILES && wxUSE_UNICODE
 #if wxMBFILES && wxUSE_UNICODE

-    #define wxFNCONV(name) wxConvFile.cWX2MB(name)
+    #define wxFNCONV(name) wxConvFileName->cWX2MB(name)

     #define wxFNSTRINGCAST wxMBSTRINGCAST
 #else
     #define wxFNSTRINGCAST wxMBSTRINGCAST
 #else

-#if defined( __WXOSX__ ) && wxMBFILES
-    #define wxFNCONV(name) wxConvFile.cWC2MB( wxConvLocal.cWX2WC(name) )
+#if defined( __WXOSX_OR_COCOA__ ) && wxMBFILES
+    #define wxFNCONV(name) wxConvFileName->cWC2MB( wxConvLocal.cWX2WC(name) )

 #else
     #define wxFNCONV(name) name
 #endif
     #define wxFNSTRINGCAST WXSTRINGCAST
 #endif
 
 #else
     #define wxFNCONV(name) name
 #endif
     #define wxFNSTRINGCAST WXSTRINGCAST
 #endif
 

-#else
-  // !wxUSE_WCHAR_T
+#else // !wxUSE_WCHAR_T

 
 // ----------------------------------------------------------------------------
 // stand-ins in absence of wchar_t
 
 // ----------------------------------------------------------------------------
 // stand-ins in absence of wchar_t


@@ -271,11 +650,11 @@ public:
 
 #define wxConvFile wxConvLocal
 
 
 #define wxConvFile wxConvLocal
 

-WXDLLIMPEXP_DATA_BASE(extern wxMBConv) wxConvLibc,
+extern WXDLLIMPEXP_DATA_BASE(wxMBConv) wxConvLibc,

                                        wxConvLocal,
                                        wxConvISO8859_1,
                                        wxConvUTF8;
                                        wxConvLocal,
                                        wxConvISO8859_1,
                                        wxConvUTF8;

-WXDLLIMPEXP_DATA_BASE(extern wxMBConv *) wxConvCurrent;
+extern WXDLLIMPEXP_DATA_BASE(wxMBConv *) wxConvCurrent;

 
 #define wxFNCONV(name) name
 #define wxFNSTRINGCAST WXSTRINGCAST
 
 #define wxFNCONV(name) name
 #define wxFNSTRINGCAST WXSTRINGCAST


@@ -290,12 +669,26 @@ WXDLLIMPEXP_DATA_BASE(extern wxMBConv *) wxConvCurrent;
 #if wxUSE_UNICODE
     #define wxConvertWX2MB(s)   wxConvCurrent->cWX2MB(s)
     #define wxConvertMB2WX(s)   wxConvCurrent->cMB2WX(s)
 #if wxUSE_UNICODE
     #define wxConvertWX2MB(s)   wxConvCurrent->cWX2MB(s)
     #define wxConvertMB2WX(s)   wxConvCurrent->cMB2WX(s)

+
+    // these functions should be used when the conversions really, really have
+    // to succeed (usually because we pass their results to a standard C
+    // function which would crash if we passed NULL to it), so these functions
+    // always return a valid pointer if their argument is non-NULL
+
+    // this function safety is achieved by trying wxConvLibc first, wxConvUTF8
+    // next if it fails and, finally, wxConvISO8859_1 which always succeeds
+    extern WXDLLIMPEXP_BASE wxWCharBuffer wxSafeConvertMB2WX(const char *s);
+
+    // this function uses wxConvLibc and wxConvUTF8(MAP_INVALID_UTF8_TO_OCTAL)
+    // if it fails
+    extern WXDLLIMPEXP_BASE wxCharBuffer wxSafeConvertWX2MB(const wchar_t *ws);

 #else // ANSI
     // no conversions to do
     #define wxConvertWX2MB(s)   (s)
     #define wxConvertMB2WX(s)   (s)
 #else // ANSI
     // no conversions to do
     #define wxConvertWX2MB(s)   (s)
     #define wxConvertMB2WX(s)   (s)

+    #define wxSafeConvertMB2WX(s) (s)
+    #define wxSafeConvertWX2MB(s) (s)

 #endif // Unicode/ANSI
 
 #endif // Unicode/ANSI
 

-#endif
-  // _WX_WXSTRCONVH__
+#endif // _WX_STRCONV_H_