Auto return type inference issue?

[Matthew Dudley via Digitalmars-d-learn]

On Wednesday, 23 April 2014 at 00:02:29 UTC, Jesse Phillips wrote:
> On Tuesday, 22 April 2014 at 07:54:34 UTC, Matthew Dudley wrote:
>> Here's the gist of what I'm trying to do.
>>
>> https://gist.github.com/pontifechs/11169069
>>
>> I'm getting an error I don't understand:
>>
>> tinker.d(42): Error: mismatched function return type inference 
>> of tinker.B and tinker.A
>> tinker.d(55): Error: template instance tinker.DynamicTuple!(A, 
>> B).DynamicTuple.notopDispatch!"one" error instantiating
>> Failed: 'dmd' '-v' '-o-' 'tinker.d' '-I.'
>>
>> Also, as an aside, why can't tuples be indexed dynamically? 
>> Most of my problems with this have been because you apparently 
>> can't.
>
> Might have this a little bit wrong, just a quick hand 
> translation of:
>
>     foreach (i, elem; tuple)
>         if (names[i] == s)
>         {
>             return elem;
>         }
>
> This becomes something like:
>
>     if (names[0] == "one")
>         return A;
>     if (names[1] == "one")
>         return B;
>
> As said I could have this mixed up, but /elem/ is a type. But 
> even if they were an object, you are saying that this method 
> may return an A or a B, it is not legal for a function to 
> return different types.
>
> I'd suggest a static if, but since /names/ is dynamic, that 
> won't work.

tuple in this case would be the member variable of type T (from 
T...) So wouldn't elem be the actual object, and not the type?

The effective lowering I was hoping for would be something like 
this

if (names[0] == "one")
     return tuple[0];
if (names[1] == "one")
     return tuple[1];

etc...