On Wed, Jul 22, 2009 at 7:13 AM, Pete Muir <pmuir@redhat.com> wrote:

On 22 Jul 2009, at 03:18, Dan Allen wrote:

After the recent revisions of the JSR-299 spec, I'm finding that I'm a tad confused about the bean selection algorithm when a producer method exists.

Consider the following:

class PotatoChip {
}

class SnackFactory() {
public @Produces PotatoChip createPotatoChip() {
return new PotatoChip();
}
}

1) Does the produced PotatoChip take precedence over the raw dependent type when being selected for injection?

No.

2) How do I get the Bean<PotatoChip> that represents the producer method?

BeanManager#getBeans(PotatoChip.class) returns two beans that are indistinguishable from the public API.

either:

@Alternative

class SnackFactory() {
public @Produces PotatoChip createPotatoChip() {
return new PotatoChip();
}
}

<beans>
<alternatives>
<alternative>SnackFactory</alternative>
</alternatives>
</beans>

or

class SnackFactory() {
public @Produces @Snack PotatoChip createPotatoChip() {
return new PotatoChip();
}
}

BeanManager.getBeans(PotatoChip.class, new AnnotationLiteral<Snack>() {});

where @Snack is a binding type

3) What if the producer method is @RequestScoped? How do I get the request-scoped produce instead of the dependent bean?

Bean<PotatoChip> potatoChipBean = null;
for (Bean<PotatoChip> candidate : getBeans(PotatoChip.class))
{
if (candidate.getScopeType().equals(RequestScoped.class))
{
potatoChipBean = candidate;
break;
}
}

Sorry if this sounds like a dumb question. The spec has changed a lot and my memory is getting mixed in with the current set of facts.

As above.

It's possible the RI doesn't actually reflect this behavior atm - if you have a problem, make sure there is a test, then change the RI or file a WBRI.