mapping from complex response objects to custom value objects - is it worth it? what is the clearest/easiest/most maintainable way to do it?

[Gary Weaver <ga...@duke.edu>]

Hello,

Though this isn't really a CXF question, I thought it might be good to 
ask here to see people's thoughts...

In short, my question is "What is the easiest and most maintainable way 
(and able to handle a small but not unsignificant load) of mapping 
cxf-codegen-plugin v2.1 + JAXB generated client request and response 
value objects (or even the SOAP response itself) to custom-written value 
objects? (or is it even worth it?)"

Background:

Have been using Apache CXF cxf-codegen-plugin v2.1 in a project to 
generate client code (using the default JAXB data binding) and currently 
I have an overly complex and ugly response coming back from one service 
that I've integrated with.

Several years ago, I had thought to myself that I would never write 
another unnecessary layer of value objects on top of client value 
objects autogenerated by a tool like wsdl2java. The reason being that 
you have to typically would have to change things in more places when 
the service interface changed, despite initial efforts to the contrary 
to "buffer" from service interface changes.

However, I'm currently integrating with a few services that 
nasty-enough-looking and overly-complicated-enough that I really 
couldn't help but to write my own value objects to map to the 
cxf-codegen-plugin v2.1 + JAXB generated value objects.

For the first few services, they were fairly simple (comparatively) so I 
just mapped these manually in the code (several gets here, several sets 
there, and presto). However, the current service I'm working on has a 
real bear of a response. If the service interface changes, it is going 
to be a good deal of work to redo the additional layer of value objects 
in addition to the autogenerated ones.

I looked briefly into using Dozer, but the response classes (because of 
the way the service is defined) are so ugly and hard to read that unless 
IDea supports auto-completion in the dozer mapping file (like it does 
for Spring configs), I don't think I want to go there.

I've also thought about spending more time looking into how I could 
configure CXF (or whichever data binding- JAXB, Aegis, MTOM, even though 
I'm using JAXB at the moment) to possibly unmarshal the response to my 
custom classes, in an attempt to clarify the request/response value 
objects by ignoring the fluff and simplifying structure. However, I 
don't know if this is the best use of time, and the resulting code would 
likely be even less clear, would it not?

Thanks in advance for any advice you can provide,

Gary