On Tuesday 23 September 2008 11:10:57 am Quilleash, Michael (IT) wrote: > Looking at the code it looks like the following would happen with an > exception on the out chain... > > - OutgoingChainInterceptor invoked > - Outgoing chain interceptor chain runs > - Exception > - Outgoing chain is unwound > - OutgoingChainInterceptor completes > - Any other "in" interceptors then run (usually none) > > So it looks like an in interceptor after the OutgoingChainInterceptor will > always run, unless there's an exception on the inbound chain, in which case > handleFault() would be called on my main in interceptor :)
The one case it doesn't "completely" cover is an exception on the inbound chain. In that case, your handleFault will be called, but then the OutFault chain will be constructed and your interceptor wouldn't have any actions on that chain. Not sure if that's a big deal. > One thing I noticed, should the PhaseInterceptorChain.unwind() method be > more defensive when calling handleFault(), by catching/logging/swallowing > any exceptions thrown from handleFault()? Probably not a bad idea. :-) Feel free to log a jira (with a patch). :-) Dan > > Cheers. > > Michael Quilleash > Morgan Stanley | Technology > 20 Cabot Square | Canary Wharf | Floor 01 > London, E14 4QW > Phone: +44 20 7677-4543 > [EMAIL PROTECTED] > > -----Original Message----- > From: Quilleash, Michael (IT) > Sent: 23 September 2008 15:27 > To: Daniel Kulp; [email protected] > Subject: RE: "Around" interceptor > > Ah that sounds good. Be ideal if I only had to register one interceptor > and have it drive everything. > > So would I be right that if I did the following: > > - Add an in interceptor to the start of the in chain > - Implement the handleFault() to do the "finally" code > - In handleMessage() add a new interceptor right at the end of the in > chain, which also does the "finally" code. > > Does this cover every case? What happens if an exception occurs in the > outbound chain, does this propogate back and always unwind the in chain > too? > > Cheers. > > Michael Quilleash > Morgan Stanley | Technology > 20 Cabot Square | Canary Wharf | Floor 01 London, E14 4QW > Phone: +44 20 7677-4543 > [EMAIL PROTECTED] > > -----Original Message----- > From: Daniel Kulp [mailto:[EMAIL PROTECTED] > Sent: 23 September 2008 14:15 > To: [email protected] > Cc: Quilleash, Michael (IT) > Subject: Re: "Around" interceptor > > On Tuesday 23 September 2008 9:04:23 am Quilleash, Michael (IT) wrote: > > Thanks Dan, > > > > I see what you mean. > > > > But I guess that wouldn't work as a "finally" check tho as an > > exeception in the chain causes the rest of the interceptors not to be > > run. I would still have to add something to the fault chains to do the > > finally stuff. > > Yep. > > Note: your interceptor at the start of the "in" chain can also implement > the handleFault message which is called when the chain is unwinding but > before the fault chain is called. You can then get timing information or > something for the "in" chain part separate from the fault stuff. > > Dan > > > Cheers. > > > > Michael Quilleash > > Morgan Stanley | Technology > > 20 Cabot Square | Canary Wharf | Floor 01 London, E14 4QW > > Phone: +44 20 7677-4543 > > [EMAIL PROTECTED] the contents of an OutputStream for the out interceptor. I looked at your logging interceptor and found that you used a CacheAndWriteOutputStream. I tried using this in my out interceptor. I registered a CachedOutputStreamCallback that handles the transformation in the onClose method. The problem I have is that when I call message.setContent() and pass it an output stream with my modified message, it is not used. I'm assuming that I'm using the CacheAndWriteOutputStream incorrectly. Do you have any suggestions on how I should be transforming the contents of the OutputStream? > > > > -----Original Message----- > > From: Daniel Kulp [mailto:[EMAIL PROTECTED] > > Sent: 22 September 2008 19:39 > > To: [email protected] > > Cc: Quilleash, Michael (IT) > > Subject: Re: "Around" interceptor > > > > On Sunday 21 September 2008 8:13:54 am Quilleash, Michael (IT) wrote: > > > Thanks for the reply Ian. > > > > > > I did see this approach being used by the default interceptors, > > > however I found out that the out/outFault messages are always null > > > when the "in" interceptor is invoked so the interceptor chains > > > aren't available to add new interceptors to. > > > > > > I haven't found an example of any interceptors that add interceptors > > > to different interceptor chains. > > > > Yea, that's very hard to do. The chains aren't created until they are > > actually needed. In a "one way" case, there would never even be an > > "out" chain created or called. > > > > On the server side, the "simple" way to do it is to add your "stop" > > interceptor to the VERY VERY end of the chain. If there is an out > > message, the last interceptor on the "in" chain is the one that sets up > > the out message/chain and calls it. Thus, your interceptor would be > > invoked after that. > > > > Dan > > > > > Is there a way of doing what you are saying, perhaps by creating the > > > out messages earlier? > > > > > > Cheers. > > > > > > -----Original Message----- > > > From: Ian Roberts [mailto:[EMAIL PROTECTED] > > > Sent: 21 September 2008 12:50 > > > To: [email protected] > > > Subject: Re: "Around" interceptor > > > > > > Quilleash, Michael (IT) wrote: > > > > I would need to have: > > > > > > > > In Interceptor - start timing > > > > Out/InFault/OutFault interceptors - stop timing > > > > > > > > To catch all the possible code paths. Is there any easier way of > > > > doing this? > > > > > > A trick used in a number of CXF built-in interceptors is to have a > > > main interceptor which is added to the chain explicitly, but then > > > have that interceptor add others to the relevant chains for that > > > exchange during its handleMessage. For example (pseudo-code): > > > > > > class TimingInterceptor { > > > private Interceptor stopTiming = new StopTimingInterceptor(); > > > > > > handleMessage(Message m) { > > > Exchange ex = m.getExchange(); > > > ex.put("startTime", System.currentTimeMillis()); > > > ex.getOutMessage().getInterceptorChain().add(stopTiming); > > > ex.getOutFaultMessage().getInterceptorChain().add(stopTiming); > > > } > > > > > > static class StopTimingInterceptor { > > > handleMessage(Message m) { > > > Exchange ex = m.getExchange(); > > > long startTime = ex.get("startTime"); > > > // ... > > > } > > > } > > > } > > > > > > You only need to add the TimingInterceptor to your input chain and > > > it will manage the rest for you. > > > > > > Ian > > > > > > -- > > > Ian Roberts | Department of Computer Science > > > [EMAIL PROTECTED] | University of Sheffield, UK > > > -------------------------------------------------------- > > > > > > NOTICE: If received in error, please destroy and notify sender. > > > Sender does not intend to waive confidentiality or privilege. Use of > > > this email is prohibited when received in error. > > > > -- > > Daniel Kulp > > [EMAIL PROTECTED] > > http://www.dankulp.com/blog > > -------------------------------------------------------- > > > > NOTICE: If received in error, please destroy and notify sender. Sender > > does not intend to waive confidentiality or privilege. Use of this > > email is prohibited when received in error. > > -- > Daniel Kulp > [EMAIL PROTECTED] > http://www.dankulp.com/blog > -------------------------------------------------------- > > NOTICE: If received in error, please destroy and notify sender. Sender does > not intend to waive confidentiality or privilege. Use of this email is > prohibited when received in error. > -------------------------------------------------------- > > NOTICE: If received in error, please destroy and notify sender. Sender does > not intend to waive confidentiality or privilege. Use of this email is > prohibited when received in error. -- Daniel Kulp [EMAIL PROTECTED] http://www.dankulp.com/blog
