Experiments with command-line and web flow adaptor

[Cantor, Scott]

On 4/5/13 10:33 AM, "Marvin Addison" <marvin.addison at gmail.com> wrote:

>Not as pretty, but it cleans up nicely with a shorter key name for
>ProfileRequestContext:
>
><action-state id="StaticResolver">
>  <evaluate 
>expression="StaticResolverBean.doExecute(conversationScope.profileRequestC
>ontext)"
>/>
>  <transition to="OutputAttributes" />
></action-state>
>
>I would argue that's much more natural and even axiomatic for a flow,
>and it accomplishes the goal removing SWF dependencies just the same.

Tom expressed a strong distaste for anything that cluttered up the web
flow syntax. The other advantage is that the adaptor hides the state
management of the context object, vs. assuming it's in SWF conversation
scope. If that turns out to be a problem because of serialization, etc.,
the adaptor can accomodate that without changing any flow config.

I don't have as strong a preference myself, but I slightly prefer pushing
the complexity into the bean defs. I also like the fact that the adaptor
approach means you can mix beans that are POJOs with beans that are
SWF-aware, and they look identical in the flow config.

-- Scott