+1 Andrey, can you please file a ticket and put the proposed design to it?
--Yakov 2015-08-13 21:37 GMT+03:00 Vladimir Ozerov <voze...@gridgain.com>: > +1 > > It is hard to understand why do we have so much different annotations for > injections while requierd resource type could be clearly derived from field > or method types. Looks like a single annotation will be enough. And it > should not be hard to abstract out injection logic into SPI. > > On Thu, Aug 13, 2015 at 6:10 PM, Andrey Kornev <andrewkor...@hotmail.com> > wrote: > > > Hello, > > > > Long story short, I'd love to be able to intercept the resource injection > > calls. > > > > For example, in our application we do not use Spring, and also, we do not > > expect the developers to use Ignite-specific annotations in the > application > > code. Instead, we use the standard javax.inject API's @Inject which makes > > integration with Ignite a bit painful since every application-level > > closure/class needs to be un-/wrapped before and after serialization. On > > the receiving node, the wrapper code looks up a local instance of > injector > > (we use Guice) and finally injects the dependencies. > > > > It would be nice to be able to define a custom injection callback (an > SPI) > > to which Ignite would delegate at specific points during the resource > > injection flow. The current implementation of GridResourceProcessor makes > > no provision for such thing. > > > > I also think this feature would allow to make the injector support > > pluggable: Spring, Guice, Dagger, whatever... Going forward, it would > also > > be possible to gradually phase out all injection related Ignite > annotations > > in favor of the standard javax.inject API. > > > > Please share your thoughts! > > > > Andrey > > >
