Stephane Bailliez <[EMAIL PROTECTED]> writes:

> Hello,
>
> I would appreciate some enlightments about the rational of having static
> methods on peers in Torque.
>
> With time, I think that a better approach would be to have singleton on
> these instances to benefit from an interface and method overriding rather
> than method hiding. We should be able to get the singleton peer from the
> object via a static getPeer method or such.
>
> The main problem right now, is that we cannot design a generic set of
> operation on Peers. (or at least I'm unable to do so) Say I have A and B and
> each of them has a customized peer APeer, BPeer which method doSelect hide
> the base one because I want to add some precondition/postcondition.
>
> Now say I want to perform a batch selection of A and B via a generic select
> operation, I would use instances of APeer and BPeer in an operation that
> takes a BasePeer[] argument.
> Now due to method hiding I'm in trouble to perform the appropriate
> doSelect() on APeer and BPeer. How to call the right static method ? yay !
> Using reflection is the only option AFAIK.
>
> Of course at some points it is problematic for retrieveByPK signatures since
> Java as of now does not allow covariant return type (JDK 1.5 I'm afraid). So
> these methods must return an Object.
>
> Any stupid things in what I'm saying ? :o)

Stephane, this has been discussed to death (see the archives).
Everyone is in favor of moving Torque away from statics.

Dan

--
To unsubscribe, e-mail:   <mailto:[EMAIL PROTECTED]>
For additional commands, e-mail: <mailto:[EMAIL PROTECTED]>

Reply via email to