Dne 16.5.2016 v 13:34 John D. Ament napsal(a):
>
> Martin,
>
>
> On Mon, May 16, 2016 at 3:54 AM Martin Kouba <mkouba@redhat.com
> <mailto:mkouba@redhat.com>> wrote:
>
> Dne 15.5.2016 v 17:14 John D. Ament napsal(a):
> > Romain,
> >
> > On Sun, May 15, 2016 at 11:05 AM Romain Manni-Bucau
> > <rmannibucau@gmail.com <mailto:rmannibucau@gmail.com>
> <mailto:rmannibucau@gmail.com <mailto:rmannibucau@gmail.com>>> wrote:
> >
> > Hi Jon
> >
> > Le 15 mai 2016 16:15, "John D. Ament" <john.d.ament@gmail.com
> <mailto:john.d.ament@gmail.com>
> > <mailto:john.d.ament@gmail.com
> <mailto:john.d.ament@gmail.com>>> a écrit :
> > >
> > > Hey guys
> > >
> > > Seems like we have some issues in JIRA all focused on managing
> > the lifecycle of Dependent scoped beans. It also seems like
> we have
> > many differing opinions about how to manage them.
> > >
> > > - Martin raised a PR to add a release() method to Instance to
> > help destroy a dependent bean
> https://github.com/cdi-spec/cdi/pull/286
> > > - I raised a PR https://github.com/cdi-spec/cdi/pull/289 to
> > update the spec to clarify how to manage a dependent scoped bean.
> > >
> > > Right now, it seems that the big disagreement is whether
> > Instance.destroy() can destroy objects not created by it (the
> case
> > being around the CDI utility class, being an impl of
> Instance). I'm
> > currently heavily against Martin's proposed changes, but want
> to get
> > input from others on the group to understand their perspective.
> > >
> > > - Does the spec require destroy() to be called only on
> instances
> > that it created? When I read 5.6.1 the only requirement I see is
> > that it has to be a dependent scoped bean. Note when I ask
> this I'm
> > asking from the spec perspective, its a different problem if
> there's
> > some issues with implementations following suite (I would imagine
> > there needs to be some shared global registry of dependent scoped
> > beans for this to work).
> > >
> >
> > Sound the only clean impl. Any other is not symmetric and
> > potentially lead to "oops this time it didnt work". I also not
> > seeing any use case limitation with that so think it is the same
> > solution
> >
> >
> > I'm not sure I follow or if this isn't an answer to "Does the spec
> > require destroy() to be called only on instances that it created?" ?
> >
> > Anyways I did look a bit closer and it seems that Martin's
> statement is
> > consistent with how OWB works,
> >
> https://github.com/apache/openwebbeans/blob/trunk/webbeans-impl/src/main/java/org/apache/webbeans/inject/instance/InstanceImpl.java#L293
> so
> > I wonder if there's a part of the spec I'm missing, or if there
> was some
> > offline agreement on how to understand it.
>
> John, I believe Instance CANNOT be used to destroy a dependent bean
> instance it didn't created, because a dependent bean instance doesn't
> know the dependent objects it depends on - that's what CreationalContext
> is for.
>
>
> This is the area I'm looking for clarification around. Where in the
> spec is this mandated?
I think it's implied. When you look at "6.1.1. The CreationalContext
interface", there is:
"Contextual.create() should use the given CreationalContext when
obtaining contextual references to inject, as defined in Contextual
reference for a bean, in order to ensure that any dependent objects are
associated with the contextual instance that is being created."
and also:
"Contextual.destroy() should call release() to allow the container to
destroy dependent objects of the contextual instance."
and "6.2. The Context interface":
"The context object must pass the same instance of CreationalContext to
Contextual.destroy() that it passed to Contextual.create() when it
created the instance."
And for dependent beans there is no real context which could hold a
reference to a CreationalContext. Each Instance<T> has its own
CreationalContext which only tracks the dependent instances produced by
a given Instance. Instance<T> does not know anything about
CreationalContexts of other dependent instances...
I think I'm starting to see your point. However, if its mandated that Instance uses a creational context to create a bean, we should call that out. Right now the text says that it will retrieve a bean, but realistically for dependent it's creating a bean.
Maybe something along the lines of
"In the case of the target bean being a dependent scoped bean, the instance object used to retrieve that bean will retain a reference to the creational context used to create that bean. That creational context will be used to destroy the bean when calling destroy()"
I would still like us to explore ways to do this without requiring the original instance, for the case of CDI.current() usage.
John
>
> So if you pass any dependent instance to Instance.destroy() there is no
> CreationalContext apart from the one Instance<> has. In other words you
> wouldn't be able to destroy the @Dependent dependencies of a @Dependent
> bean instane. Does it make sense?
>
> See also https://issues.jboss.org/browse/CDI-519 (cdi-spec/cdi/pull/278
> is already merged).
>
>
> >
> > John
> >
> > > - Do we want two methods that effectively do the same
> thing? I
> > don't see a strong difference between the two.
> > >
> > > On the flipside, my change is more a spec clarification. I'm
> > thinking more now that it belongs as a reword of 5.6.1 to clarify
> > how to use destroy() on dependent beans, rather than where I put
> > it. I think realistically we have all of the tools needed to
> manage
> > the lifecycle of these classes, just need to clarify them for
> people
> > to use.
> > >
> > > John
> > >
> >
> > > _______________________________________________
> > > cdi-dev mailing list
> > > cdi-dev@lists.jboss.org <mailto:cdi-dev@lists.jboss.org>
> <mailto:cdi-dev@lists.jboss.org <mailto:cdi-dev@lists.jboss.org>>
> > > https://lists.jboss.org/mailman/listinfo/cdi-dev
> > >
> > > Note that for all code provided on this list, the provider
> > licenses the code under the Apache License, Version 2
> > (http://www.apache.org/licenses/LICENSE-2.0.html). For all other
> > ideas provided on this list, the provider waives all patent and
> > other intellectual property rights inherent in such information.
> >
> >
> >
> > _______________________________________________
> > cdi-dev mailing list
> > cdi-dev@lists.jboss.org <mailto:cdi-dev@lists.jboss.org>
> > https://lists.jboss.org/mailman/listinfo/cdi-dev
> >
> > Note that for all code provided on this list, the provider
> licenses the code under the Apache License, Version 2
> (http://www.apache.org/licenses/LICENSE-2.0.html). For all other
> ideas provided on this list, the provider waives all patent and
> other intellectual property rights inherent in such information.
> >
>
> --
> Martin Kouba
> Software Engineer
> Red Hat, Czech Republic
>
--
Martin Kouba
Software Engineer
Red Hat, Czech Republic