ClassOrInterfaceType currently has a single image for the entire type. E.g. the type Map.Entry<K,V> is parsed as

ClassOrInterfaceType "Map.Entry"
   - TypeArguments
     - TypeArgument "K"
     - TypeArgument "V"

This looks ok. Now, the invented type Foo<K>.Bar.Brew<V> is parsed as

ClassOrInterfaceType "Foo.Bar.Brew"
   - TypeArguments
     - TypeArgument "K"
   - TypeArguments
     - TypeArgument "V"

and now we don't know to which type the type arguments apply. Moreover, Java 1.8+ makes it possible to add annotations to the individual segments, such that e.g. Map.@Foo Entry<K,V> is perfectly legal. But our parser chokes on it for now (related to #997). If we are to add support for that, we'd better do it in a sensible way, such that the annotations don't end up all mixed whenever there's several of them.

I think for those reasons, we'd have interest in making ClassOrInterfaceType conform to the JLS (our ClassOrInterfaceType would be the JLS's ClassType), which would make the above examples be parsed recursively, (see examples at the end). I think this may help type resolution too.

Map.Entry<K,V>

ClassOrInterfaceType "Entry"
   - ClassOrInterfaceType "Map"
   - TypeArguments
     - TypeArgument "K"
     - TypeArgument "V"

Foo<K>.Bar.Brew<V>

ClassOrInterfaceType "Brew"
   - ClassOrInterfaceType "Bar"
      - ClassOrInterfaceType "Foo"
          - TypeArguments
             - TypeArgument "K"
   - TypeArguments
     - TypeArgument "V"

Map.@Foo Entry<K,V>

ClassOrInterfaceType "Entry"
   - ClassOrInterfaceType "Map"
   - Annotation
      - MarkerAnnotation "Foo"
   - TypeArguments
     - TypeArgument "K"
     - TypeArgument "V"