Groups | Search | Server Info | Keyboard shortcuts | Login | Register [http] [https] [nntp] [nntps]

Groups > comp.lang.java.programmer > #33797

Re: To deprecate or not

Show all headers | View raw

mike <mikaelpetterson@hotmail.com> wrote:
> I have an interface call it InterfaceTop that looks like this:
> public InterfaceTop {  
>   public String getValue(String arg1);
>
> }
> Then we have two sub interfaces call them InterfaceSub1 and InterfaceSub2.
> They both extend IntefaceTop.
> Now we discovered that the method in Inteface top produces erroneous output in
> in implementation of InterfaceSub1. We need an additional parameter to get value:
> public String getValue(String arg1, String arg2);

If ISub1 implementations need the 2-args method, and ISub2 implementations
don't need it, then the question is, whether ITop should even remain a
parent interface of ISub1 - or if the method should even remain in ITop 
at all.

Interfaces are used to deal with a variety of implementations by a common 
API.  But if the variety of implementations doesn't agree on that common
API, then you should consider redesigning the interface hierarchy.
(and bumping the major version number, if it is in a non-private library.)

> Maybe let the "faulty" method throw UnsupportedOperationException.

Imho, that is a very bad idea. It's a way to defer errors to runtime that 
could otherwise (with refined interfaces) be detectable at compile time.

PS: I know that the JSL uses a similar pattern for unmodifiable* classes,
  but that's not quite the same, imho.

Back to comp.lang.java.programmer | Previous | Next — Previous in thread | Find similar | Unroll thread

Thread

To deprecate or  not mike <mikaelpetterson@hotmail.com> - 2016-02-12 01:16 -0800
  Re: To deprecate or  not Eric Douglas <e.d.programmer@gmail.com> - 2016-02-12 05:59 -0800
    Re: To deprecate or not Arne Vajhøj <arne@vajhoej.dk> - 2016-02-13 11:02 -0500
      Re: To deprecate or not Eric Douglas <e.d.programmer@gmail.com> - 2016-02-15 05:12 -0800
        Re: To deprecate or not Arne Vajhøj <arne@vajhoej.dk> - 2016-02-15 11:42 -0500
  Re: To deprecate or not Arne Vajhøj <arne@vajhoej.dk> - 2016-02-13 11:34 -0500
  Re: To deprecate or  not Andreas Leitgeb <avl@logic.at> - 2016-02-15 00:53 +0000

csiph-web