Hi Paul,
On Fri, 2008-06-06 at 10:56 +0200, Paul Sandoz wrote:
> Hi Martin,
>
> Perhaps we could pass a ComponentContext to a new getInstance method:
>
> <T> T getInstance(ComponentContext cc, Scope scope, Class<T> c)
>
> This ComponentContext could present the AccessibleObject, Annotation[]
> and Annotation (if any are present).
Sounds good!
>
> I think it necessary that @Inject would need to be modified anyway if
> you want to support it on constructor/method params, since the parameter
> name is not available (plus there is no direct accessible object, the
> closest is the constructor or method).
What do you think how @Inject would have to be modified / should look
like?
Cheers,
Martin
>
> This ComponentContext is similar to
> com.sun.jersey.spi.inject.InjectableContext so we could unify that.
>
> Paul.
>
> Martin Grotzke wrote:
> > Hi,
> >
> > the @Inject has currently one limitation: it just can fetch a bean based
> > on the field's type. E.g. when we have this resource:
> >
> > @Path("/foo")
> > class Resource {
> > private @Inject Properties _props1;
> > ....
> > }
> >
> > the ComponentProvider will be asked for a bean of type Properties. If
> > there are two beans of type Properties configured in the spring
> > applicationContext (or the IoC-config), the ComponentProvder cannot
> > decide which bean to return and therefore will fail.
> >
> > Regarding the spring-integration it would be good to get some
> > information about the @Inject annotation or even the field.
> >
> > Then one could extend @Inject with some "value" or "hint" property that
> > can be evaluated by the ComponentProvider, of, if the field would be
> > handed to the CP, one could e.g. use the field.name as a hint regarding
> > the desired bean. I would prefer the first option, as it's more explicit
> > and more self-explaining.
> >
> > I think we have two choices:
> > 1) Extend the ComponentProvider
> > 2) Create some @InjectBean annotation just for the spring-integration
> > that will be handled by the SpringComponentProvider
> >
> > What do you think?
> >
> > Cheers,
> > Martin
> >
> >