Doubt about Spec Section 5.2.4. Assignability of raw and parameterized types - On injection point without diamond operator

[nuno.godinhomatos]

Hi,

I have doubt about the following spec section:

https://docs.jboss.org/cdi/spec/1.2/cdi-spec.html#assignable_parameters

I am trying to understand what paragraph in the spec explains the reason why a parameterized bean would not be injectable into a raw type delcared injection point.

From what I understand in CDI 1.0 weld se core final, there was an implementation bug that has since been fixed, whereby the following type of injection would work:

@ Inject

Instance<Dao> allDaos;

Now Dao, is a paremeterized interface.

And in CDI 1.1 weld 2.3.2 final, the code to be made to work would need to fix the injection point to read as follows:

Instance<Dao<?>> allDaos;

So, one would jave to declare that he wants to inject a Dao where the type parameter is unbounded / wild carded.

Of course, the Dao class could be for example a

public interface Dao<T>{

}

The problem of course, is that as I try to upgrade, I am encountering some runtime blowups where essentially instances that used to be made accessible from injection points that were not using the diamon wild card fail.

This is fine, I will fix those ... I do need to spot them all... but life is like that. Mistakes have a price.

But in any case, i would like to see in the spec document the phrase that states that a class.

public class UserDao implements Dao<User>{

}

Is not assignable to the injection point Instance<Dao>.

I suspect the phrase is htis one:

"A parameterized bean type is considered assignable to a raw required type if the raw types are identical and all type parameters of the bean type are either unbounded type variables or java.lang.Object."

In this case:

A A parameterized bean type

UserDao

is considered assignbale to a raw requred type

@ Inject

Instance<Dao> allDaos;

if the raw types

"Dao"

and all type parameters of the bean type

"User"

either unbounded type

(it is not the case right it is bounded to "User")

or "java.lang.Object".

But I am not sure 100% that my interepretaiton is corect.

If it is, this means that to the old injection point only a class of the form:

UnAssigninedDao <T> extends Dao<T>{

}

Or a class

ObjectDao  extends Dao<Object>{

}

would be legal candaidates to the injection point above?

Thanks for the clarification, this is discovery is an unfortunate setback... that is blowing up in my face in runtime testing.

[mkouba]

Hi Nuno,

you're right, this is covered in the first paragraph of 5.2.4. Assignability of raw and parameterized types. This sentence is there since CDI 1.0 and it follows the rules for Java generics. @Inject Instance<Dao> allDaos is de facto @Inject Instance<Dao<Object>> allDaos and you can't assign Dao<Integer> to Dao<Object>. Imagine that there is a Dao.save(T instance) method. If it was permitted to inject Dao<Integer> bean into Dao<Object>, the client code would be able to Dao.save("stringInsteadOfInteger") which is obviously wrong.

[nuno.godinhomatos]

Hi Martin, thank you very much for the clarification.

It helped.