>> And, oGadget isn't defined... shouldn't it be a string literal?
>I was assuming oGadget was a property of the class containing the Doit() method - should have written 'this.oGadget' for clarity but the 'this' isn't required.
Hm... interesting, and makes the code less readable. You have to know all PEMs of 'this'. Pretty much like referencing a field in the current alias, qualification is implied.
>>The AS operator is unintuitive but neat... implicit cast returning a null if it fails? Though, this sounds more like a try-catch "see if this works" code, instead of having an IsA(toObj, tcClass) function. What happens if you try to cast a twice subclassed timer as a timer, do you still get a reference to the original object?
>
>Yes - you can cast to any ancestor class....
That's cool. I wish we had casting objects in VFP.
>>>I'd do a bit more type checking in the first one if it was for production - but even as it stands it is safer than the VFP version...
>>
>>And what's unsafe in it? In the C# version ?
>Firstly, whilst 'pi' not being null guarantees that the property exists and is public it does not guarantee that it is writeable (i.e. not readonly). Also there's no check that the type of the 'oengine' property allows 'oGadget' to be assigned to it. Something like this would be a bit safer:
if (pi != null
> {
> if (pi.PropertyType == oGadget.GetType())
> pi.SetValue(toObj, oGadget, null);
> }
Another thing that would be nice to have in VFP, the .canWrite - I remember writing builders and other framework code, where I'd hit some "read-only at runtime" property and then have to look for another way.