Why are CodeUnitsShould have..Type() and notHave..Type() methods not contravariant?

[daniel-shuy]

For example, FieldsShould#haveRawType(DescribedPredicate) is contravariant (takes in a DescribedPredicate<? super JavaClass>), whereas CodeUnitsShould#haveRawReturnType(DescribedPredicate) is invariant (takes in a DescribedPredicate<JavaClass>).

This makes it possible to do:

ArchRuleDefinition.fields()
    .should().haveRawType(CanBeAnnotated.Predicates.annotatedWith(Annotation.class))

(because CanBeAnnotated.Predicates#annotatedWith(Class) returns a DescribedPredicate<CanBeAnnotated>, and JavaClass implements CanBeAnnotated)

but not possible to do:

ArchRuleDefinition.methods()
    .should().haveRawReturnType(CanBeAnnotated.Predicates.annotatedWith(Annotation.class))

Is this intended?

[codecholeric]

No, sorry, this is not intended ☹️ I think the second one should have a contravariant predicate type as well. Must have slipped through. Maybe I should write some sort of ArchRule for that 😉
There is some "convenience" method annotatedWith(Foo.class).<JavaClass>forSubtype() to mitigate such pain points (even though in this place it should just work out of the box).

[codecholeric]

I've adjusted the API, there were a couple of places where this slipped through.
Also added an ArchTest to make sure future methods will always be written correctly with respect to DescribedPredicate 😉
Thanks for reporting this issue and helping to make ArchUnit better 😃