Hi Mandy
Dan may provide more details later on, few comments below in meantime:
On 06/06/16 23:08, Mandy Warren wrote:
Hi,
We currently register a number of interceptors against the endpoint when our
Rest service starts up. I have a specific scenario whereby when I call a
backend system I need to create a new “session” and ensure this session is
deleted when the response is returned to the client. The session is only
created on some GETs (the first GET goes to the backend and subsequent ones go
to a cache) so I want to dynamically register an interceptor for a specific
request/response combination.
I came across https://issues.apache.org/jira/browse/CXF-2015
<https://issues.apache.org/jira/browse/CXF-2015> which helped me build the
following code to achieve what I wanted:-
Interceptor interceptor = new GenericOutboundInterceptor();
Collection<Interceptor> interceptorAsList = new ArrayList<>(1);
interceptorAsList.add(interceptor);
// register the specified CallbackAction only on the out & fault
interceptors for this request/response
// to ensure no impact on other messages
Message message = PhaseInterceptorChain.getCurrentMessage();
Collection<Interceptor> outInterceptors =
(Collection<Interceptor>)message.get(Message.OUT_INTERCEPTORS);
if (null!=outInterceptors) {
outInterceptors.add(interceptor);
} else {
message.put(Message.OUT_INTERCEPTORS, interceptorAsList);
}
Collection<Interceptor> faultInterceptors =
(Collection<Interceptor>)message.get(Message.FAULT_OUT_INTERCEPTORS);
if (null!=faultInterceptors) {
faultInterceptors.add(interceptor);
} else {
message.put(Message.FAULT_OUT_INTERCEPTORS, interceptorAsList);
}
}
It all works but I just had a few questions:
1. Am I using the correct way to access the message i.e. via
PhaseInterceptorChain.getCurrentMessage()?
Where do you access it from, from inside a service method which creates
a client proxy in scope of the current invocation ?
PhaseInterceptorChain.getCurrentMessage() is the right way to access the
current message outside of the CXF interceptor code, but it will return
a non-null message only (AFAIK) when the code is running in scope of the
current request. Furthermore, this message would be the current inbound
message which will not be used by the client proxy doing the follow-up
outbound call.
Perhaps I'm missing something, may be you can prototype some pseudo code
which will make it clearer how the end to end flow should work ?
2. I added the null checks in as the message seemed to have no existing
interceptors which surprised me as I had assumed it would have referenced the
ones registered on the endpoint but I guess they are different?
I believe this property represents only the dynamically registered
interceptors
3. Do I need to register the new interceptor on the fault chain too?
Not sure, probably not
4. Is there any need to manually remove the interceptor?
I don't think so, it will be GC-ed once the exchange completes
Thanks, Sergey
Many thanks
Mandy
