Why do I need a setter for autowired / injected field?

问题

I have a bean:

    <bean id="BasketLogic" class="efco.logic.EfcoBasketLogic" autowire="byType">
        <property name="documentLogic" ref="DocumentLogic" />
        <property name="stateAccess" ref="StateAccess" />
        <property name="contextAccess" ref="ContextAccess" />
    </bean>

  <bean id="EfcoErpService" autowire="byType" class="efco.erp.service.EfcoErpServiceImpl">
    <constructor-arg ref="ErpConnector"/>
  </bean>

documentLogic, stateAccess and contextAccess are fields on BasketLogicImpl

And I do not have <context:component-scan />

EfcoBasketLogic.java:

public class EfcoBasketLogic extends BasketLogicImpl {

        @Inject
        private EfcoErpService erpService;
    ...
    ...
    ...
}

erpService is null, unless I provide a setter. But why? I thought a setter isn't needed where autowiring is taking place? Could it be that BasketLogicImpl is responsible for that?

回答1:

You need to use a setter because annotations are not detected unless spring is told so through either <context:component-scan /> or <context:annotation-config />. Setter is detected because you specified autowire="byType".

You may find this question and answer helpful as well: When to use autowiring in Spring

回答2:

First of all, the use of <context:component-scan /> or <context:annotation-config /> enables Spring to scan your code for eligible beans to meet dependencies, which will greatly improve it's ability to wire them up correctly, so I suggest adding them to your context file.

Second, you should be aware that @Inject is a standard (meaning JSR-330 specification) annotation. It is okay to mix and match Spring annotations with standard ones, but behavior may vary when doing so. @Named is commonly paired with @Inject to match components with dependencies (both JSR-330). See this reference for details, and refer to Table 4.6 for usage comments.

But to directly answer your question, "why do I need a setter when not using component-scan", is because you are not using component-scan. You are asking Spring to inject a dependency "byType", but not allowing Spring to scan your code for components which are of that type. The reason the setter works is that the type of the setter argument being injected can be discovered by Spring in the compiled bytecode (i.e. meta-data), and so it successfully resolves your request.

回答3:

My understanding is that XML configuration overrides annotation config.The fact that autowire="byType" specified overrides the auto injection, which looks for a presence of setter method for injecting the dependency.

来源：https://stackoverflow.com/questions/13193655/why-do-i-need-a-setter-for-autowired-injected-field

标签

Spring

setter

autowired

inject