using interfaces as reference variable

[Shay Levy]

Hi,


(1) interface A {}
(2) class B implements A{}

(3) A a = new B();
(4) B b = new B();

is there any diffrence between statement (3) and (4)? Is the declaration of (3) better (due to polomorphism) ? do I gain anything by using (3) instead of (4)? or did I just miss something really basic?

Thanks

[Campbell Ritchie]

Welcome to JavaRanch

You find (3) has an advantage when you have

public class C implements A
{
 . . .
}

and

public class D implements A
{
 . . .
}

Then you can easily change "new B()" to "new C()" or "new D()" and retain all your functionality.

[Shay Levy]

Thanks for the reply.

[Embla Tingeling]

Shay Levy wrote:Is the declaration of (3) better (due to polomorphism) ?

There's a design criterion saying: Use the highest abstraction level (upwards in the inheritance tree) that meets your needs.

Writing code using A variables will definately make your code more flexible and general because it will work with any subtype of A like B (and C and D and ....).

On the other hand the subtypes B and C and D may not be equally efficient when objects of these types are plugged into an algorithm written using A variables. Say B is excellent but C and D are disasters. In that case the A abstraction is too high and you should write your code using B variables instead.