Propagate named interface assignments of types to return and parameter types of methods declared

[ivangsa]

closes #1264

[odrotbohm]

This looks rather involved. I might be missing some detail, but essentially, I was expecting NamedInterface.ofAnnotatedTypes(…) to be extended to not only consider the types themselves but related ones, in case an annotation is found. That should do the trick, shouldn't it? NamedInterfaces shouldn't need to be tweaked at all, I think. We're also gonna need tests.

I'd still love to see a flag on the @NamedInterface annotation that allows restoring the legacy behavior, i.e., something like propagate defaulting to true, which could then be evaluated in the tweaked ….ofAnnotatedTypes(…).

Please let me know if you would want to pursue this further. Totally fine for me to take over, but I guess I'd start from scratch then.

[ivangsa]

ok, I refactored to just extend NamedInterfaces.ofAnnotatedTypes... this was the first implementation and I got it working earlier and it works for me..

[odrotbohm]

Thanks again for contributing. I think I'm gonna take this as inspiration to give it a shot myself, as it looks like I'd need to invest the same amount of time course correcting the efforts.

[ivangsa]

this implementation is very simple right now and it doesn't touch any public interface.. I think it's exaclty what you say which was my first approach (bc it covered my use case..)

[ivangsa]

have a look at this test project with the current 1.4.1 implementation... which can not be validated because of related types
https://github.com/ivangsa/modulith-test

I thought (and think) that project should validate... and that would require a much involved implementation bc of the many places NamedInterfaces are build (open, root, named type, named package...)

[odrotbohm]

Superseded by the fixes made for GH-1264.