Re: [aspectj-dev] Question on static cross-cutting and queries

[Wes Isberg <wes@xxxxxxxxxxxxxx>]

This is a good question for the aspectj-users@xxxxxxxxxxx mail list.

> The idea would be to apply the interface Stringable to every class with
> a single-argument constructor that takes a String as an argument.

You're right: there's no way in AspectJ to specify a type pattern
based on the type's methods or constructors.

It's a curious rule, since a subclass of a Stringable class
might not have a String constructor, but is Stringable.

If it helps, you can find all such types:

  declare warning: execution(new(String)) : "Stringable?";

And also signal an error if they should be Stringable:

  declare error: execution((!Stringable).new(String)) :
      "Not Stringable!";

(I can't write an error when they are Stringable but
don't have a String constructor.)

> [...] I
> can imagine the difficulties with this sort of query as related to
> precendence of operation and other aspects that may have already been
> applied.

I see no problem with a declare parent for each target type.

Perhaps if you tell folks on the users list more about what
you're trying to do, an AspectJ solution can emerge.

Thanks -
Wes

Paul Brown wrote:

> I'd like to be able to write something like the following:
>
>   public aspect ApplyStringable {
> 	declare parents: *..* where new(String) implements Stringable;
>   }
>
> The idea would be to apply the interface Stringable to every class with
> a single-argument constructor that takes a String as an argument.
>
> > From what I know and have used of AspectJ, the only way to do this is to
> actually enumerate all of the classes that this would apply to.  (And I
> can imagine the difficulties with this sort of query as related to
> precendence of operation and other aspects that may have already been
> applied.)
>
> Has this sort of functionality been tried?  Is there another way to
> accomplish the same thing?  (I know how to do it with BCEL, but that's
> not the point...)
>
> Thanks in advance for any ideas.
>
> 	-- Paul
> _______________________________________________
> aspectj-dev mailing list
> aspectj-dev@xxxxxxxxxxx
> http://dev.eclipse.org/mailman/listinfo/aspectj-dev