I have an approach that uses factories that does not require CGLib if your interested.
On 9/13/07, Adam Taft <[EMAIL PROTECTED]> wrote: > Tom McGee wrote: > > Should this in the appicationContext.xml: > > <bean id="fooResource" scope="request" class="FooResource" /> > > be this: > > <bean id="fooResource" scope="request" class="FooResource" > > scope="prototype"/> > > It should be either one of two ways: > > a) <bean id="fooResource" scope="request" class="FooResource" /> > > b) <bean id="fooResource" scope="prototype" class="FooResource" /> > > -a- should "theoretically" work in a servlet environment (like he's > using). I say theoretically, because this is getting into some edge > case uses of spring. > > -b- would be the only reasonable choice outside of the servlet > environment or if things didn't work with -a-. > > Of course, as I've mentioned before, using lookup-method implies using > CGLib proxies in Spring. The problem would ideally not have to be > solved in this way. I have found, for example, that using a > lookup-method declaration (which creates a CGLib proxy) will clash with > other proxies such as Spring's database transaction management features. > Your transactions will never roll back automagically even on failure. > > Anyway, the point is, one should tread carefully when using > lookup-method, as it's definitely a work around and not a main stream > way to use Spring. > > > On 9/13/07, Makunas, Michael <[EMAIL PROTECTED]> wrote: > > haven't tested it in anything close to production. > > Break out Apache bench and start testing. Make sure you're testing and > confirming your complete stack is working right, including database > transactions that both succeed and fail. Again, as mentioned, > generating proxies on top of proxies will likely lead to edge case > goofiness and hard to track down bugs, so pay particular attention to this. > > Adam >
