Unsatisfied deps in migration to Weld2

[nickarls]

Jozef tried at one point explaining the reason to me on irc but I was unable to apply the knowledge so I'm back with an example

I have an interface

public interface Dummy
{
}

with two implementations

public class Foo implements Dummy
{
}

public class Bar implements Dummy
{
}

and an abstract manager

public abstract class DummyManager<F extends Dummy, B extends Dummy>
{
}

and then an implementation

@ApplicationScoped
public class MyDummyManager extends DummyManager<Foo, Bar>
{
}

But I'm unable to do a

   @Inject
   private DummyManager<Dummy, Dummy> dummyManager;

anywhere. How come? Why doesn't the MyDummyManager work? Foo and Bar both implement Dummy, right?

[luksa]

You also can't do this in Java, actually.

DummyManager<Dummy, Dummy> dummyManager = new MyDummyManager() will not compile. See http://docs.oracle.com/javase/tutorial/java/generics/inheritance.html

You should change the injection point to DummyManager<? extends Dummy, ? extends Dummy>.

[luksa]

A good explanation of why this shouldn't be possible is this:

Let's say DummyManager had a setter for F:

public abstract class DummyManager<F extends Dummy, B extends Dummy>
{
   private F f;



   public void setF(F f) {
      ...

   }
}

Now, if MyDummyManager was injected into injection point

   @Inject
   private DummyManager<Dummy, Dummy> dummyManager;

It would be possible for you to do:

   this.dummyManager.setF(new Bar());

So you'd end up with MyDummyManager having an instance of Bar stored in field f, which is supposed to be of type Foo.

[nickarls]

Indeed! Thanks for the clarification. My migration project now blows up at a fresh location