>>Ed, I'd be interested in whether or not you tried my code sample?
>
>Nancy, I was really surprised to read that AMEMBERS would grab the information in a COM object before the methods and or properties had been used in VFP, because I didn't think that VFP would read the type library until a request was made to access members. So I tried you code, and my results were as I expected: AMEMBERS didn't work at all.
>
>I wonder what could be different about our environments?
>
>I am using NT4, SP 4, IE5, VFP 5 (and VFP 6 SP2, I tried both).
>
>
>Ed, any insight into this? What do you get?
>

As you can see, there's been a lot of discussion. My results on Win98 and various flavors of NT with both VFP 5.0a and 6 match yours without running under the debugger. There's definitely something happening within the debugger when I single step through Nancy's code; AMEMBERS() populates with the methods automagically when I step through under the debugger, but not the properties. Outside of single stepping through code, though, the only PEMs that appear through AMEMBERS() are PEMs that I've explicitly referenced.

George has noted similar behavior, except that with his Win95 box, even single-stepping through the code doesn't populate AMEMBERS() with the methods from Shell.Application without explicit reference.

I could easily accept that running under the debugger, something probed the automation provider to extract the ODL information embedded in it. If that were the case, then why are only the methods picked up? It's easy enough to write a piece of code that probes an automation object using IDispatchEx (in fact, IDispatchEx specifically includes the capability to get at the list of an object's properties and enumerate the member elements of an object.) There's no reason that I can see for VFP not to extract all of the description if it extracts some of it prior to the first reference to a PEM.

I don't have a good enough understanding of what VFP is doing internally when it instantiates an ActiveX object, and even less understanding of what happens internally in the debugger. There's nothing obvious about configuration - the difference in behavior between my system and George's in the debugger might point at the shell object's implementation (I'm running a later version of the shell components on all my systems here than George is.) And the difference in the behavior on Nancy's system and everyone else's is even more of a mystery.

Any other ideas?

>
>>>Nancy, AFAIK, VFP's AMEMBERS() won't pull out the information on the Shell.Application object
>>