What's a smart way to organize classes in Spring 3 for component-scan?

Question

I've begun work on a new project using Spring 3 and I'm using annotations. I love that I can wire up my classes to get dependencies injected, but I know it's a bad practice to have context:component-scan start at the base package.

I'm using a DispatcherServlet which has its own xml configuration file. In that is also a context:component-scan. When I first started learning Spring I had overlap in my component scans and saw beans created multiple times. I'd like to avoid that.

What is a good way to organize either my packages or my component scans to cover all the beans without duplication?

Currently I have packages like this:

my.package.controller
my.package.dao
my.package.entity
my.package.service
my.package.util

If I have beans in all those packages then it seems like the easy way out would be to put <context:component-scan base-package="my.package"></context:component-scan> into applicationContext.xml and be done with it.

Would it be better to scan my.package.controller in the dispatcher's xml and the rest (excluding my.package.controller) in applicationContext.xml?

Or should I arrange all my annotated classes in one area and everything else in another? Something like:

my.package.spring.controller
my.package.spring.dao
my.package.spring.entity
my.package.spring.service
my.package.spring.util
my.package.notannotated
my.package.notannotated2

I'm using @Autowired to add logging to most if not all of my classes so I don't know that I'll have any classes that won't be annotated.

I hate getting stuck on configuration...I'd rather be stuck in code, so if someone can offer any tips I'd readily welcome them.

Thanks!

Answer 1

Yes - in your main context scan everything except controllers

 <context:component-scan base-package="my.package">
    <context:exclude-filter type="regex" expression="my\.package\.controller.*"/>
 </context:component-scan>

and in your DispatcherServlet context just scan the controller package.