Using method injection makes the singleton-bean class difficult to unit-test (you need to create a subclass to implement the method which gives out the dependency). Plus it's less reusable because you can't directly instantiate it, so if you're not using Spring and want to use this class, you'll need to subclass and provide the bean-returning method.
A better approach IMHO is to use a proxy, a prototype target source and a prototype target bean, as follows. Such a singleton-bean class is easily unit-testable and better reusable.
<bean id="targetPooledObject" class="pool.PooledObject" scope="prototype">
<constructor-arg value="42" />
</bean>
<bean id="prototypeTargetSource" class="org.springframework.aop.target.PrototypeTargetSource">
<property name="targetBeanName" value="targetPooledObject" />
</bean>
<bean id="pooledObject" class="org.springframework.aop.framework.ProxyFactoryBean">
<property name="targetSource" ref="prototypeTargetSource" />
</bean>
<bean id="poolConsumer" class="pool.PoolConsumer">
<property name="pooledObject" ref="pooledObject" />
</bean>
Now we can inject pooledObject into a singleton bean (poolConsumer as shown above), and for every method call that we make on that singleton bean, (e.g. every time we call poolConsumer.callPooledObjectMethod() which in turn calls pooledObject.foo()) we get a new PooledObject bean.
Following is the corresponding code:
public class PooledObject
{
private int x;
public PooledObject(int x)
{
this.x = x;
}
public void foo()
{
System.out.println("foo called");
}
}
public class PoolConsumer
{
private PooledObject pooledObject;
public PooledObject getPooledObject()
{
return pooledObject;
}
public void setPooledObject(PooledObject pooledObject)
{
this.pooledObject = pooledObject;
}
public void callPooledObjectMethod()
{
pooledObject.foo();
}
}