On Feb 26, 2007, at 2:25 PM, Jeremy Boynes wrote:
Resending a previous reply as my ISP's MTA seems to be acting up...
On 2/26/07, Jim Marino <[EMAIL PROTECTED]> wrote:
FYI,
Now that we have the AutowireResolver in place, I'm going to look
into adding support for full SCA 1.0 autowire semantics in trunk.
This will mean we can get rid of the @Autowire tag (using @Reference
instead).
It might be an idea to deprecate that now so that users know that it
will be going away.
Sure, I just deprecated it and checked it in (r512090). Once we get
the new functionality, I'd like to just get rid of the old way of
doing things if others agree since the impact will be minimal.
Will they need to use @Reference or is default introspection typically
going to ensure that there is no need for any annotation (which would
even better)?
Good question...We have a couple of options:
1. We use normal heuristics and introspection rules that are
implemented for References. This will mean that in many cases setters
will need to be annotated with @Reference unless it can be determined
from the type that it is a service (e.g. it is decorated with
@Remotable, @Service, or @Conversational).
2. We do something fancy similar to PicoContainer where the the
AutoWireResolver will attempt to resolve any ambiguous complex type
if autowire is true for the reference, component, or composite.
I haven't thought much about the implications of #2 but I've found
PicoContainer incredibly useful, powerful, and concise (not to
mention small) and think we should borrow ideas from it. What do
people think?
Jim
--
Jeremy
---------------------------------------------------------------------
To unsubscribe, e-mail: [EMAIL PROTECTED]
For additional commands, e-mail: [EMAIL PROTECTED]
---------------------------------------------------------------------
To unsubscribe, e-mail: [EMAIL PROTECTED]
For additional commands, e-mail: [EMAIL PROTECTED]
