RE: Re: [picocontainer-dev] Pico2: ASM ImplementationHiding ComponentAdapter

Jörg Schaible wrote on Monday, May 14, 2007 11:00 PM:

> Paul Hammant wrote:
> 
>> HSCA is caching implicitly. So I've no need to wrap with CachingCAF.
>> 
>> Maybe CA should optionally extend a new Caching interface.  That may
>> be marker-esque (bad) unless we add a flush() method.
>> 
>> Though if it is as you say I can do:
>> 
>>    CA ca = pico.getComponentAdapter(MyClass.class).getDelegate();
>>    Object ldInstance = ca.swapRealInstance(newInstance)
>> 
>> Well, it would involve some casts to get it to compile :-)
> 
> Well, CachingCA was only an example. Maybe you should have a look at
> the EJBClientCA for more complex real-life scenarios.

You're right the EJBCA was not the right example. Scenario was:

pico.registerComponent(new CachingCA(new HSCA(new CICA(A.class))));

vs.

pico.registerComponent(new HSCA(new CachingCA(new CICA(A.class))));


With the first expression, you can exchange the A instance globally, while in the second scenario you can exchange a global default instance of A for a specific component. Therefore I don't like a combination of caching and HS. ANd consider that 

pico.registerComponent(new ThreadLocalCA(new HSCA(new CICA(A.class))));

vs.

pico.registerComponent(new HSCA(new ThreadLocalCA(new CICA(A.class))));

makes perfectly sence. Still need the HS functionality, but caching is completely different (or replace ThreadLocalCA with PooledCA).

- Jörg



> 
> - Jörg
> 
> 
> ---------------------------------------------------------------------
> To unsubscribe from this list please visit:
> 
>     http://xircles.codehaus.org/manage_email

---------------------------------------------------------------------
To unsubscribe from this list please visit:

    http://xircles.codehaus.org/manage_email