Default method survey results

[Yuval Shavit]

There have been discussions within the last few days, on this mailing list,
about how people wish that Java had a keyword for package-private
visibility. Insofar as we should learn from the past, this seems to me a
nudge towards the direction of having one for defenders. That said, my
argument for where to place the "default" keyword (assuming there is one)
can also back the no-keyword camp.

I agree with the survey's description of what "default" would mean in each
place. But I voted in the minority, exactly because to me, the concept of a
"default method" is not very useful.

As a user of an interface with such a method, I don't care if it has a
default implementation -- I just care that it has some implementation. As
an implementer of such an interface, I don't care what "kind" of method it
is as much as whether I need to implement it; many times I'll be
implementing it via an abstract class like AbstractList, anyway, so default
methods are not the only ones for which I don't need to provide an
implementation. Or do I need to switch my mindset, such that if
AbstractList provides an implementation of iterator(), that is now a
"pseudo-default" method from my perspective? And as the interface writer, I
don't even really care about the fact that it has a default implementation
(unless final defaults come in later). I only care that the method exists
at all.

In other words, what does the concept of a "default method" add to how
anybody thinks about the interface?

With abstract methods, I always viewed the "abstract" keyword as a
confirmation to the compiler: "my semicolon isn't a typo -- this is really
a pure virtual function." I would view the "default" keyword the same way,
and in that context it seems to fit well between the signature and the body.

On Mon, Aug 13, 2012 at 10:25 AM, Collin Fagan <collin.fagan at gmail.com>wrote:

> As much as developers would like them to be, and as much as modern
> frameworks treat them as such, annotations are not keywords.  Annotations
> were intended only to add information and not to actually change the
> meaning of anything in the language. Frameworks can blur the lines if they
> wish but I think the Java language itself should not break it's own rules
> for this one case.
>
> Collin
>
> On Mon, Aug 13, 2012 at 9:09 AM, Jose A. Illescas <
> jantonio.illescas at rbc.com
> > wrote:
>
> > Gregg,
> >
> >    Compiler can add "default" keyword for every method (on interface)
> > with implementation (delimiteded with "{...}" instead of "default")
> >
> >    Demarcations on developer hands are ugly, (I think) must be resolved
> > by compiler...
> >
> >
> > On 13/08/2012 15:52, Gregg Wonderly wrote:
> > > In this day and age of the Java language, reusing existing keywords,
> > > trying to shoehorn them into use for a new purpose, can feel a little
> > > clumsy.  Annotations can provide a much better mechanism in many
> > > cases, simply because the right nomenclature is helpful.
> > >
> > > However, annotations also have a very specific usage scenario with
> > > tooling and moment-of-use issues.  For me, this means that you can't
> > > use annotations in places that annotations wouldn't work with the
> > > normal annotation tooling.  You certainly don't want to have to have a
> > > special class of annotation that further confuses the use of
> > > annotations in other cases.
> > >
> > > The use of "default" isn't awesome, but it does provide a usable
> > > demarcation which the compiler and run-time class loading can use to
> > > make the right decision for how to build the v-tables to dispatch the
> > > default method should it be involved some how.
> > >
> > > Gregg Wonderly
> > >
> > > On 8/13/2012 8:35 AM, Jose A. Illescas wrote:
> > >> On 13/08/2012 15:23, Deepak S Patwardhan wrote:
> > >>>> -----Original Message-----
> > >>>> From: lambda-dev-bounces at openjdk.java.net [mailto:lambda-dev-
> > >>>> bounces at openjdk.java.net] On Behalf Of Jose A. Illescas
> > >>>> Sent: 13 August 2012 16:40
> > >>>> To: Brian Goetz
> > >>>> Cc: lambda-dev at openjdk.java.net
> > >>>> Subject: Re: Default method survey results
> > >>>>
> > >>>> Brian, Why not add one thirth option (without "default" keyword) to
> > >>>> poll?
> > >>>>
> > >>>> 1.  void foo() default {
> > >>>>             System.out.println("foo");
> > >>>>         }
> > >>>>
> > >>>> 2.  default void foo() {
> > >>>>             System.out.println("foo");
> > >>>>         }
> > >>>>
> > >>>> 3.     void foo() {
> > >>>>
> > >>>>             System.out.println("foo");
> > >>>>         }
> > >>>>
> > >>> If we explore going down this path, I think we should also explore
> > >>> using an
> > >>> annotation,
> > >>>
> > >>> @java.lang.Default
> > >>> void foo() {
> > >>>       System.out.println("foo");
> > >>> }
> > >>>
> > >> Deepak, why proposed annotation?
> > >>
> > >>       I would like see an implementation as any other java method
> > >> (without new keywords, annotations, ...)
> > >>
> > >>       Implements a interface method ("default") must be equals to
> > >> implement another class method (I want "complexity" out of developer
> > >> hands, into compiler...)
> > >>
> > >>
> > >>
> > >>> I would be interested in knowing if using such an annotation has any
> > >>> demerits.
> > >>>
> > >>> Regards,
> > >>> Deepak S Patwardhan.
> > >>>
> > >>>>
> > >>>> On 12/08/2012 00:30, Brian Goetz wrote:
> > >>>>> Poll is closed, and results can be seen here:
> > >>>>>
> > >>>>>
> > >>>>>
> > >>>> https://www.surveymonkey.com/sr.aspx?sm=xPw4AxfWK2TTQLQtbIpQCg0l
> > >>>> bkw_2fuYDiZnrBDB_2bPyd4_3d
> > >>>>>
> > >>>
> > >>
> > >>
> > >
> > >
> >
> >
> >
>
>