Re: [aspectj-users] declare parents syntax

[Adrian Colyer <adrian.colyer@xxxxxxxxx>]

Named type patterns have been on the radar for some time now. We  
decided not to make any language changes in the 1.5.x stream (the  
jump to 1.5.0 was big enough). For AspectJ 1.6 support for named type  
patterns (which in turn allows us to introduce things like matching  
on type modifiers without the syntax getting too unwieldy) is high on  
the list of considerations. See e.g. https://bugs.eclipse.org/bugs/ 
show_bug.cgi?id=86818.

Regards, Adrian.


On 2 Oct 2006, at 01:27, Ron Bodkin wrote:

> That's a good clarification, thanks. Often this pattern is adequate  
> when
> working in application code where you expect to have subtypes in  
> the same
> packages (and the warning is a good enforcement/reminder tool).
>
> However, these caveats just further remind me of the improved value  
> that
> having a named reusable type pattern would provide. I think a  
> reusable type
> pattern is what Eugene wants also. And named type patterns (and method
> signatures too) would also allow more readable versions of complex
> expressions through composition of named parts.
>
> -----Original Message-----
> From: aspectj-users-bounces@xxxxxxxxxxx
> [mailto:aspectj-users-bounces@xxxxxxxxxxx] On Behalf Of Wes Isberg
> Sent: Saturday, September 30, 2006 12:19 PM
> To: aspectj-users@xxxxxxxxxxx
> Subject: Re: [aspectj-users] declare parents syntax
>
> Just remember that
>
>   within(XyzType+)
>
> can pick out subtypes outside com.xyz..* that extend
> a type in com.xyz..*
>
> You can name the pointcut to reuse it:
>
>   pointcut inScope() : within(com.xyz..*);
>
> then
>   before(): something() && inScope() {..}
>
> To address the problem of typo's (i.e., redundancy across
> type patterns and pointcuts), try something like
>
>   declare parents : com.xyz..* : implements X;
>
>   declare warning : staticinitialization(X+)
>      && !inScope() : "X not in scope";
>
> The latter would also pick out those types not in
> com.xyz..*
> i.e., the subtypes outside the package.
>
> Wes
>
> On Fri, 29 Sep 2006 11:14:00 -0700
>  "Ron Bodkin" <rbodkin@xxxxxxxxxxxxxx> wrote:
> > Hi Eugene,
> >
> > It would be great to have a way to reuse type patterns in
> > AspectJ... that's
> > something I've advocated for a while. That's how I'd
> > improve on redundancy.
> >
> > But if you are using declare parents you could use the
> > new type in your
> > pointcut, i.e., use this pointcut execution(* *(..)) &&
> > within(XyzType+)
> > instead of execution(* *(..)) && within(com.xyz..*)
> >
> >
> >
> > -----Original Message-----
> > From: aspectj-users-bounces@xxxxxxxxxxx
> > [mailto:aspectj-users-bounces@xxxxxxxxxxx] On Behalf Of
> > Eugene Kuleshov
> > Sent: Friday, September 29, 2006 11:04 AM
> > To: aspectj-users@xxxxxxxxxxx
> > Subject: Re: [aspectj-users] declare parents syntax
> >
> >
> >   Thanks Ron, I know that. But the thing is that the same
> > pattern is
> > used in pointcuts to apply additional aspects to the very
> > same classes.
> > Hence a duplication that may lead to some unexpected
> > typos.
> >
> >   regards,
> >   Eugene
> >
> >
> > Ron Bodkin wrote:
> > > Hi Eugene,
> > >
> > > declare parents accepts a type pattern, not a pointcut.
> > A type pattern
> > picks
> > > out types, whereas a pointcut picks out join points.
> > The expression inside
> > a
> > > within pointcut are is a type pattern.
> > >
> > > Here's an example of using declare parents to match a
> > type pattern, which
> > is
> > > surely what you wanted to do with within:
> > >
> > > declare parents: com.xyz..* implements XyzType;
> > >
> > >
> > > -----Original Message-----
> > > From: aspectj-users-bounces@xxxxxxxxxxx
> > > [mailto:aspectj-users-bounces@xxxxxxxxxxx] On Behalf Of
> > Eugene Kuleshov
> > > Sent: Thursday, September 28, 2006 12:06 PM
> > > To: aspectj-users@xxxxxxxxxxx
> > > Subject: [aspectj-users] declare parents syntax
> > >
> > > Hi,
> > >
> > >   I wonder why "declare parents" introduction syntax
> > does not allow to
> > > use within(*) pointcut expression?
> > >
> > >   This limitation lead to significant amount of
> > duplication in the
> > > aspect code and/or makes xml-based aspect definitions
> > more verbose. The
> > > only workaround I know is to "bridge" trough annotation
> > introduction,
> > > but that approach does not work on 1.4 and looks like a
> > hack.
> > >   I thought I saw that it was possible to use within
> > pointcuts for this
> > > in some old version of AspectJ but it was removed. I
> > might be mistaken
> > > though...
> > >
> > >   Thanks
> > >
> > >   Eugene
> >
> > _______________________________________________
> > aspectj-users mailing list
> > aspectj-users@xxxxxxxxxxx
> > https://dev.eclipse.org/mailman/listinfo/aspectj-users
> >
> > _______________________________________________
> > aspectj-users mailing list
> > aspectj-users@xxxxxxxxxxx
> > https://dev.eclipse.org/mailman/listinfo/aspectj-users
>
> _______________________________________________
> aspectj-users mailing list
> aspectj-users@xxxxxxxxxxx
> https://dev.eclipse.org/mailman/listinfo/aspectj-users
>
> _______________________________________________
> aspectj-users mailing list
> aspectj-users@xxxxxxxxxxx
> https://dev.eclipse.org/mailman/listinfo/aspectj-users