[JSR308] Proposal: Multiple Instance of same Annotation, Annotation Inheritance

[Niko Matsakis]

> I like your example of XML as a motivation for the need for users to
> specify annotations that express data-structure-like information.  I
> suspect you're right that users will come up with unforeseen  
> purposes for
> such annotations.  However, I would be much more comfortable with the
> proposal if it provided some real, concrete use case and benefits for
> substantial numbers of programmers today.  Such compelling examples  
> aren't
> really optional:  they are a necessary counterweight to the risks and
> effort inherent in any change to Java.

I completely agree that this section of my proposal was lacking.  In  
fact, I delayed sending it for some time because I wanted to work in a  
better example: however, I eventually decided to send it "as is" in  
the hope that it might prompt a discussion from which I could draw  
better use cases.  It seems that there have in fact been many e-mails  
recently calling for annotation inheritance which can serve as a basis  
for those use cases.

With respect to the timing, I am not sure what is the expected time  
span for JSR-308.  I assume there is some sort of deadline for a final  
draft of the specification, reference implementation, etc?

>> [In] the current annotation scheme
>> ... it would be possible to have @Nodes that contain either @Nodes or
>> @Leafs, but not both.
>
> but in Java 6, having @Nodes that contain @Nodes is illegal.

Yes, good point.  To handle my example, it would be necessary to  
permit recursive/cyclic references that occur within arrays.  As  
discussed in the JSR-107 FAQ, this does indeed raise the possibility  
of infinite annotation instances through the use of defaults, unless a  
more involved restriction were put in place.


regards,
Niko Matsakis