Injection and Persistence Context inheritance

[donatasc]

Hi CDI gurus,

Persistence Context propagation makes living with EJBs easier, but CDI specification does not explicitly state whether (and how) propagation rules work when @Inject is being used instead of @EJB.

In particular, my question is whether persistence context inheritance will take place, if one @SessionScoped stateful SB (with Extended Persistence Context) will @Inject another @SessionScoped stateful SB with Extended Persistence Context.

For example: let A1, A2 and B all be stateful @SessionScoped session beans with extended persistence contexts each. Then:

@Stateful
@SessionScoped
A1 {
  @Inject B b;
  ...
}

@Stateful
@SessionScoped
A2 {
  @Inject B b;
}

@Stateful
@SessionScoped
B {
...
}

Is persistence context inheritance happening?

Cheers,
Donatas

[pmuir]

Donatas Ciuksys wrote on Apr 14, 2010 20:20:


Hi CDI gurus,

Persistence Context propagation makes living with EJBs easier, but CDI specification does not explicitly state whether (and how) propagation rules work when @Inject is being used instead of @EJB.

In particular, my question is whether persistence context inheritance will take place, if one @SessionScoped stateful SB (with Extended Persistence Context) will @Inject another @SessionScoped stateful SB with Extended Persistence Context.

For example: let A1, A2 and B all be stateful @SessionScoped session beans with extended persistence contexts each. Then:

@Stateful
@SessionScoped
A1 {
  @Inject B b;
  ...
}

@Stateful
@SessionScoped
A2 {
  @Inject B b;
}

@Stateful
@SessionScoped
B {
...
}

Is persistence context inheritance happening?

Cheers,
Donatas

Persistence context propagation should happen as though you had injected with @EJB IIRC. If this isn't the case with Weld, please raise an issue in JIRA so we can review.

[donatasc]

Pete Muir wrote on Apr 15, 2010 13:03:


Persistence context propagation should happen as though you had injected with @EJB IIRC. If this isn't the case with Weld, please raise an issue in JIRA so we can review.

Well I believe my example above should create some problems for Weld implementers, if they would implement persistence context inheritance as it is specified by JPA specification. Citing (page 295):

If the component is a stateful session bean to which an extended persistence context has been
bound and there is a different persistence context bound to the JTA transaction, an EJBException
is thrown by the container.

Please look at my example case above: all EJBs are SessionScoped, so there is only one instance created for each of them (for a single user). Thus both A1 and A2 will inject one and the same instance of B.

My question is: what persistence context will be bound to B? A1's? A2's? Or B's own?

If all @Inject annotations would be replaced by @EJB, then we would have two instances of B, each having persistence context bound to corresponding A1/A2 component (no problems what so ever).

But here we have only one instance of B, so this should be suspicious at least, if not erroneous.