On Nov 28, 1:31 pm, "Giovanni Dicanio" <giovanni.dica***@invalid.it>
wrote:
> "Arnshea" <arns***@gmail.com> ha scritto nel messaggionews:d105694e-b83f-463c-8a6d-4192d7965***@p69g2000hsa.googlegroups.com...
>
> > I'm working with an MFC based COM object.  From C# I'd like to be able
> > to call a method on the COM object that takes a string array and
> > modifies the contents.
>
> > Is there any way to do this with a variable length array?
>
> > I've only been able to get it to work with a fixed size array.
>
> I don't know about C# side.
>
> But I believe that - at least for COM and MFC side - if you use a SAFEARRAY
> containing BSTRs, you should have no problem at all.
>
> Have you tried to use your MFC COM object e.g. with Visual Basic 6? VB6
> understands COM very well.
> You could also try to test it from a MFC client, as well.
>
> My suggestion is to first make it work correctly with native COM (so you may
> test your COM object from a VB6 or MFC client), only after then, you may try
> to fix the C# Interop side.
>
> BTW: COM programming is *full* of lots of *details* (see other posts in this
> MFC newsgroup, for some example... e.g. there was a bug with a VT_UI8 flag
> used instead of correct VT_UI1, immersed in other well-working code), so
> posting the C++/MFC code snippet you gave us is useless.
>
> Just my 2 cents.
>
> Giovanni

On Nov 28, 2:07 pm, "Giovanni Dicanio" <giovanni.dica***@invalid.it>
wrote:
> "Arnshea" <arns***@gmail.com> ha scritto nel messaggionews:4b6908e9-df68-4c50-825b-48e296245***@s8g2000prg.googlegroups.com...
>
> > The full C++/MFC method is below.  The method works (the input array
> > is modified and the modifications are visible from C#).
>
> OK. So my understanding was wrong (I thought you had problems and bugs, e.g.
> about array resizing...).
> I apologize.
>
> > I'm mainly
> > looking for ideas for improving this, e.g., alternative
> > implementation, style, potential pitfalls and the like.
>
> OK. I put some comments in code below:
>
> > HRESULT CMyClass::StrArrayFunc(VARIANT &vArray)
> > {
> >  AFX_MANAGE_STATE(AfxGetStaticModuleState());
>
> I would put an HRESULT variable, local to the method, e.g.
>
>   HRESULT hr;
>
> So, you can assign it as the result of every function call you do in the
> method, and if you get an error (FAILED(hr)), then you can return this 'hr'
> value as function return value. So, the caller (client) can have better
> information about the cause of the error condition.
> e.g.
>
>   hr = SomeCOMCall(...);
>   if ( FAILED(hr) )
>   {
>       return hr;
>   }
>
> >  // make sure it's a string array
> >  if (V_VT(&vArray) != (VT_BYREF | VT_ARRAY | VT_BSTR))
> >    AfxThrowOleDispatchException(1001, "Type Mismatch in Parameter.
> > Pass a string array by reference");
>
> When you use AfxThrowOleDispatchException, I would prefer not using a string
> literal, but instead an integer reference into a string-table.
>
> AfxThrowOleDispatchException (MFC):http://msdn2.microsoft.com/en-us/library/hf2kbh8z(VS.80).aspx
>
> You could use the following overload, which takes an integer instead of a
> string:
>
> void AFXAPI AfxThrowOleDispatchException(
>    WORD wCode,
>    UINT nDescriptionID,
>    UINT nHelpID =  -1
> );
>
> Using an integer into a string table has several advantages, like making
> localization easier (you can just change the string error description in the
> string table, without modifying string literals in source code).
> In general, I prefer storing strings in string table resources, and
> reference the string via an integer resource ID, and not embedd string
> literals in code.
>
> >  // get the safearray from the variant
> >   SAFEARRAY **ppsa = V_ARRAYREF(&vArray);
>
> You may want to check the pointer value here.
>
>   if ( ppsa == NULL )
>      ... error: pointer not valid
>
> (you may return an E_POINTER error as HRESULT).
>
> Moreover, considering that you use "SAFEARRAY *" parameters, I would get rid
> of double level of indirection (SAFEARRAY **), and just use one level of
> indirection, i.e.:
>
>   SAFEARRAY * psa = *ppsa;
>   ASSERT( psa != NULL ); // ASSERT in debug builds only
>   if ( psa == NULL ) // Check for release builds
>   {
>      // signal errors in release builds
>      return E_POINTER; // or something similar
>   }
>
> So, you can use 'psa' instead of '*ppsa' in following lines, e.g.:
>
> >  // get lower and upper bounds
> >  long lLBound, lUBound;
> >  SafeArrayGetLBound(*ppsa, 1, &lLBound);
> >  SafeArrayGetUBound(*ppsa, 1, &lUBound);
>
> becomes:
>
>   SafeArrayGetLBound( psa, 1, ... );
>   SafeArrayGetUBound( psa, ... );
>
> >  // access each element
> >  BSTR bstrCurrent;
> >  BSTR bstrNew;
> >  CString curStr;
> >  for (long i=lLBound; i <= lUBound; i++)
> >  {
> >    SafeArrayGetElement(*ppsa, &i, &bstrCurrent);
>
> >    SysFreeString(bstrCurrent);
> >    bstrNew = SysAllocString(L"replaced");
>
> In general, I would use a BSTR class wrapper (like CComBSTR or _bstr_t)
> instead of "raw" BSTR.
>
> >    HRESULT hr = SafeArrayPutElement(*ppsa, &i, bstrNew);
>
> 'hr' should be defined at the top of the method, so you can assign it from
> whatever function call you make.
>
> For example, also SafeArrayGetLBound and SafeArrayGetUBound return an
> HRESULT. For robust code, I think you should check hr also in these cases.
> So 'HRESULT hr' should be local to the whole method.
>
>  HRESULT hr; // <-- in the top of method
>
>  ...
>
>  hr = SafeArrayGetLBound( psa, ... );
>  if ( FAILED(hr) )
>  {
>     ...
>     return hr;
>  }
>
>  ...
>
>  hr = ...
>  if ( FAILED(hr) )
>  ...
>
>
>
>
>
> > error:
> ...
> >   return 0;
>
> If you have an error, you should not return 0. You should return at least an
> error code (generic) like E_FAIL. But you should return a more specific
> error code, if possible.
>
> Please consider all above as IMHO. I like freely expressing opinions (and
> elaborate them), but I don't like when personal opinions become "religious"
> wars or flames :)
>
> Giovanni

"Arnshea" <arns***@gmail.com> wrote in message
news:d105694e-b83f-463c-8a6d-4192d79652fb@p69g2000hsa.googlegroups.com...
> (apologies for the crosspost)
>
> I'm working with an MFC based COM object.  From C# I'd like to be able
> to call a method on the COM object that takes a string array and
> modifies the contents.
>
> Is there any way to do this with a variable length array?
>
> I've only been able to get it to work with a fixed size array.  The
> relevant code snippets are below.  Suggestions are greatly appreciated
> (e.g., there's a better way to do it, the code will cause memory
> leaks, etc...).
>
> MFC DISPATCH MAP ENTRY:
>
> DISP_FUNCTION_ID(CMyClass, "StrArrayFunc", dispidStrArrayFunc,
> StrArrayFunc, VT_I4, VTS_VARIANT)
>
>
> .H PROTOTYPE:
>
> HRESULT StrArrayFunc(VARIANT &vArray);
>
>
> .CPP:
>
> HRESULT CMyClass::StrArrayFunc(VARIANT &vArray)
> {
>   AFX_MANAGE_STATE(AfxGetStaticModuleState());
>
>   SAFEARRAY **ppsa = V_ARRAYREF(&vArray);
>
>   // ...
> }
>
>
> MANAGED/C# CLIENT:
>
> static int Main(string[] args)
> {
>   String[] ar = new string[30];
>
>   for (int i=0; i < ar.Length; i++)
>      ar[i] = "str";
>
>   Array a = (Array) ar;
>
>   MyCOMServer.MyClassClass mc = new MyClassClass();
>
>   mc.StrArrayFunc(ref a); // INTEROP call
>
>   string[] nAr = (string[]) a;
>
>   // ... remaining code accesses modified array through nAr
> }