Raymond,

Thanks for giving this some thought.

Yeah, it might be worth adding the input/outputs as well to a factored-out
util as well.   I guess you could put the
Message body handling (e.g. setBody or setFaultBody) into a common util as
well, but I wasn't pushing that.

As I'm looking this over I'm noticing how we rely on the intf introspectors
to set the IDL_INPUT but set the IDL_OUTPUT
ourselves here.   Maybe it doesn't matter (unless we want to refactor), but
wondering if there was a reason for the asymmetry there.

Not sure if any of the transformers actually use the 'wire' in the
'metadata'..  if they don't might be better to cut out of a util.

Scott

On Wed, Jul 23, 2008 at 1:22 PM, Raymond Feng <[EMAIL PROTECTED]> wrote:

>  Hi,
>
> It's ok to refactor the code out. Would it be better to extract the
> transformation logic for the payload (input/output/fault) out
> of DataTransformationInterceptor to become OperationDataTransformer (or a
> better name) and expose it as a utility service? This way, binding or
> implementation providers can use it to transform input, output and fault
> data.
>
> Thanks,
> Raymond
>
>  *From:* Scott Kurz <[EMAIL PROTECTED]>
> *Sent:* Tuesday, July 22, 2008 11:41 AM
> *To:* [email protected]
> *Subject:* refactor fault matching in DataTransformationInterceptor for
> non-interceptor usage
>
> In experimenting with using the Mediator from a binding impl, rather than
> only via the DataTransformationInterceptor on the wire, I found myself
> needing to use the exact same logic (that the
> DataTransformationInterceptor  uses) to map an Message's fault body to both
> the source/target fault DataType.
>
> I coded up a refactoring of the fault matching function and added it to a
> sandbox:
>
> https://svn.apache.org/repos/asf/tuscany/sandbox/scottkurz/core-databinding
>
> I'd be interested to hear any comments.   (I got the 1.3 itests to pass
> with my change though I have some JDK issues and can't get HEAD to build)
>
> Thanks,
> Scott
>
>

Reply via email to