Mostly for Scott: bean load order definition

[Cantor, Scott]

On 10/9/15, 10:35 AM, "dev on behalf of Rod Widdowson" <dev-bounces at shibboleth.net on behalf of rdw at steadingsoftware.com> wrote:



>> Or is this the problem again of not being able to access even the parent
>> ApplicationContext while creating the child?
>
>I'm pretty sure that that’s the problem.

That's incredibly annoying. I don't really understand what they're thinking there. A child context's behavior is self-evidently something that could need the parent's beans.

>OTOH according to your touchstone I'm not sure whether saying
>
><AttributeDefinition  xsi:type="ad:Script" id="scriptedInline">
>
>Is really substantially better than saying
>
><AttributeDefinition  xsi:type="ad:Script" id="scriptedInline"
>customObjectRef="shibboleth.CustomScriptObject" />
>
>In all the other scripting cases we leave the injection up to the person
>configuring.

It's actually arguably better in some ways because it means you don't have to inject a specific named object, but an arbitrary one. With views, you have to inject multiple objects using a map so the flow files can be written ahead of time, but here you can just directly define different beans and inject them into specific scripts as needed, making the syntax in the script simpler (no indirection through a map, and no changes to the scripts needed if you start with no map but then make it a map later to add more objects).

-- Scott