[nem-en] macro to define a new kind of member

[Kamil Dworakowski]

Kamil Skalski wrote:
> I don't really like how you hardcoded the logic of passing modifiers
> into a macro call.
Sorry :>
> I think it would be much better to even create a dummy member (field,
> method) without name instead:
>
> public pointcut bla () blu : bli ();
>
> --->
>
> [pointcut (bla () blu : bli ())]
> public __dummy : int;
A junk method will be created, but shouldn't cause any problems. It will 
definitely make a clearer implementation.

After parsing modifiers we can parse top level extensions targeted at a 
WithinClass, and if sth was actually eaten and parser is at the end of 
the loose group, create a method with some unique name. This method will 
be given this macro as a yet one more attribute. Or maybe even resign 
from WithinClass target and gen this meth whenever all the loosegroup is 
eaten after parsing top extensions. That would be very short indeed.

One problem is that ParseClassMember initially collects all the custom 
attributes that are targeted at anything but parameter. Should they be 
filtered out before the dummy method is created?
>
> This is much more elegant and could be probably more useful in other 
> cases.
Some example of another case?

Kamil Dworakowski