How to Convert from ANSI to Unicode & Unicode to ANSI for OLE

2.00    | 2.00

The information in this article applies to:

- Microsoft OLE libraries included with:

    - Microsoft Windows NT 3.51
    - Microsoft Windows 95

SUMMARY

All strings that are passed to and received from 32-bit OLE APIs and interface methods use Unicode. This requires applications that use ANSI strings to convert them to Unicode before passing them to OLE and to convert the Unicode strings that are received from OLE to ANSI. This article demonstrates how these conversions can be done.

MORE INFORMATION

Windows NT implements Unicode (or wide character) and ANSI versions of Win32 functions that take string parameters. However Windows 95 does not implement the Unicode version of most Win32 functions that take string parameters. Instead it implements only the ANSI versions of these functions.

A major exception to this rule is 32-bit OLE. 32-bit OLE APIs and interface methods on Windows NT and Windows 95 use Unicode exclusively. ANSI versions of these functions are not implemented either on Windows NT or Windows 95.

This means that a 32-bit application which needs to run on both Windows 95 and Windows NT must use the ANSI versions of the non-OLE Win32 functions and must convert ANSI strings to Unicode before they are passed to OLE.

A 32-bit Unicode application that runs only on Windows NT need not use any ANSI/Unicode conversion functions.

Win32 provides MultiByteToWideChar and WideCharToMultiByte to convert ANSI strings to Unicode and Unicode strings to ANSI. This article provides AnsiToUnicode and UnicodeToAnsi which uses these functions for ANSI/Unicode conversion.

/*
 * AnsiToUnicode converts the ANSI string pszA to a Unicode string
 * and returns the Unicode string through ppszW. Space for the
 * the converted string is allocated by AnsiToUnicode.
 */

    ULONG cCharacters;
    DWORD dwError;

    // If input is null then just return the same.
    if (NULL == pszA)
    {
        *ppszW = NULL;
        return NOERROR;
    }

    // Determine number of wide characters to be allocated for the
    // Unicode string.
    cCharacters =  strlen(pszA)+1;

    // Use of the OLE allocator is required if the resultant Unicode
    // string will be passed to another COM component and if that
    // component will free it. Otherwise you can use your own allocator.
    *ppszW = (LPOLESTR) CoTaskMemAlloc(cCharacters*2);
    if (NULL == *ppszW)
        return E_OUTOFMEMORY;

    // Covert to Unicode.
    if (0 == MultiByteToWideChar(CP_ACP, 0, pszA, cCharacters,
                  *ppszW, cCharacters))
    {
        dwError = GetLastError();
        CoTaskMemFree(*ppszW);
        *ppszW = NULL;
        return HRESULT_FROM_WIN32(dwError);
    }

    return NOERROR;

/*
 * UnicodeToAnsi converts the Unicode string pszW to an ANSI string
 * and returns the ANSI string through ppszA. Space for the
 * the converted string is allocated by UnicodeToAnsi.
 */

    ULONG cbAnsi, cCharacters;
    DWORD dwError;

    // If input is null then just return the same.
    if (pszW == NULL)
    {
        *ppszA = NULL;
        return NOERROR;
    }

    cCharacters = wcslen(pszW)+1;
    // Determine number of bytes to be allocated for ANSI string. An
    // ANSI string can have at most 2 bytes per character (for Double
    // Byte Character Strings.)
    cbAnsi = cCharacters*2;

    // Use of the OLE allocator is not required because the resultant
    // ANSI  string will never be passed to another COM component. You
    // can use your own allocator.
    *ppszA = (LPSTR) CoTaskMemAlloc(cbAnsi);
    if (NULL == *ppszA)
        return E_OUTOFMEMORY;

    // Convert to ANSI.
    if (0 == WideCharToMultiByte(CP_ACP, 0, pszW, cCharacters, *ppszA,
                  cbAnsi, NULL, NULL))
    {
        dwError = GetLastError();
        CoTaskMemFree(*ppszA);
        *ppszA = NULL;
        return HRESULT_FROM_WIN32(dwError);
    }

    return NOERROR;

Sample use of these functions is as follows. CoTaskMemFree is used to free the converted string if CoTaskMemAlloc was used to allocate the string. The converted string must not be freed if it is returned through an out-parameter to another OLE component, because that component is responsible for freeing the string. LPOLESTR is a pointer to a Unicode string.

// The following code gets an ANSI filename that is specified by the
// user in the OpenFile common dialog. This file name is converted into
// a Unicode string and is passed to the OLE API CreateFileMoniker. The
// Unicode string is then freed.

// Get file name from OpenFile Common Dialog. The ANSI file name will
// be placed in ofn.lpstrFile

// The following code implements IOleInPlaceFrame::SetStatusText.
// The lpszStatusText string, that is received from another OLE
// component, uses Unicode. The string is converted to ANSI before it is
// passed to the ANSI version of SetWindowText. Windows 95 supports only
// the ANSI version of SetWindowText.

    LPSTR pszStatusTextA;
    UnicodeToAnsi(pszStatusTextW, &pszStatusTextA);
    SetWindowText(m_hwndStatus, pszStatusTextA);
    CoTaskMemFree(pszStatusTextA);

Note the comments in AnsiToUnicode and UnicodeToAnsi regarding the allocator that is used to allocate the converted string. CoTaskMemAlloc (the OLE allocator) is required to be used only if the resultant string will be passed to another OLE component and if that component can free the string. This means that strings that are passed as in-parameters to OLE interface methods need not use the OLE allocator. Strings that are passed as in-out-parameters or returned through out-parameters or in- out-parameters must be allocated using the OLE allocator.

String constants can be converted to Unicode at compile time by using the OLESTR macro. For example: CreateFileMoniker(OLESTR(“c:\\foo\\bar.doc”), &pmk);

Another example of ANSI/Unicode conversion routines can be found in the Microsoft Foundation Classes (MFC) source code which ships with the Visual C++ 4.0 compiler. These routines are described in MFC Technote 59: ‘Using MFC MBCS/Unicode Conversion Macros’. The definition these macros OLE2T, T2OLE, OLE2CT, T2COLE, A2W, W2A, A2CW, W2CA and USES_CONVERSION are in \msdev\mfc\include\afxpriv.h. Also see AfxA2WHelper and AfxW2AHelper in the MFC source code in \msdev\mfc\src and the use of OLE2T, T2OLE, OLE2CT and T2COLE in the MFC source code in \msdev\mfc\src. These functions allow code to be compiled either for Unicode or ANSI depending on whether the _UNICODE preprocessor definition has been made. For example, the CreateFileMoniker call in the above example can be made as follows with the MFC macros:

USES_CONVERSION; GetOpenFileName(&ofn); CreateFileMoniker(T2OLE(ofn.lpstrFile), &pmk);

If _UNICODE is defined, T2OLE is defined as follows: inline LPOLESTR T2OLE(LPTSTR lp) { return lp; } If _UNICODE is not defined, T2OLE is defined as follows:

#define T2OLE(lpa) A2W(lpa)

#define A2W(lpa) (\
        ((LPCSTR)lpa == NULL) ? NULL : (\
            _convert = (strlen(lpa)+1),\
            AfxA2WHelper((LPWSTR) alloca(_convert*2), lpa, _convert)\
        )\
)

The MFC conversion macros use _alloca to allocate space from the program stack for the converted string. The space is automatically deallocated when the procedure call has completed. OLE requires the OLE allocator to be used for all strings (data) that will be allocated by one component and freed by another. This means that strings passed through out- parameters and in-out-parameters of OLE interfaces must be allocated with the OLE allocator. In-parameters need not be allocated with the OLE allocator because the caller is responsible for freeing them. Most Linking/Embedding OLE interfaces and API pass strings as in-parameters. Consequently the MFC conversion macros can be used in most cases. The MFC conversion macros cannot be used for in-out parameters or for returning values through out-parameters because they do not allocate space using the OLE allocator. AnsiToUnicode and UnicodeToAnsi can be used in these cases.

Yet another set of Unicode/ANSI conversion routines can be found in Don Box’s column on OLE in Microsoft Systems Journal, August 1995, Vol. 10 No. 8, Page 86. Don Box defines a C++ class with a cast operator which will return a Unicode/ANSI converted string. The allocated space is automatically freed when the object goes out of scope. This class can be modified to allocate using the OLE allocator and to not free the allocated space for strings that are passed through in-out or out- parameters.

One of the classes, String16, from Don Box’s column which converts an ANSI string to Unicode, follows. Another class, String8, that is similar to this one is used for ANSI to Unicode conversion. The CreateFileMoniker call from the previous example can be made as follows with this class:

GetOpenFileName(&ofn); CreateFileMoniker(String16(ofn.lpstrFile), &pmk);

In the above code, an instance of String16 is created. The constructor of the class will convert the ANSI string to Unicode. The language implementation will call the cast operator, operator const wchar_t *, to cast this parameter to the type of of CreateFileMoniker’s first parameter. The cast operator will return the Unicode string which is passed to CreateFileMoniker. The object will destruct when in goes out of scope.

// String16 ////////////////////////////////////////////////////////

// Shim class that converts both 8-bit (foreign) and
// 16-bit (native) strings to 16-bit wideness

// native and foreign constructors
    String16(const char *p8);
    String16(const wchar_t *p16);

// non-virtual destructor (this class is concrete)
  ~String16(void);

// native conversion operator
  operator const wchar_t * (void) const;

// native wideness string
    wchar_t *m_sz;
// is foreign??
    BOOL m_bIsForeign;

// protect against assignment!
  String16(const String16&);
    String16& operator=(const String16&);

// native constructor is a pass-through

// simply give out the native wideness string

  return m_sz;

// foreign constructor requires allocation of a native
// string and conversion

// calculate string length
  size_t len = strlen(p8);

// calculate required buffer size (some characters may
// already occupy 16-bits under DBCS)
  size_t size = mbstowcs(0, p8, len) + 1;

// alloc native string and convert
  if (m_sz = new wchar_t[size])
    mbstowcs(m_sz, p8, size);

// delete native string only if synthesized in foreign constructor

  if (m_bIsForeign)
    delete[] m_sz;