Java EE 6 @javax.annotation.ManagedBean vs. @javax.inject.Named vs. @javax.faces.ManagedBean

First of all let me do some clarifications:

Managed bean definition : generally a managed bean is an object that its life cycle (construction, destruction, etc) is managed by a container.

In Java ee we have many containers that manage life cycle of their objects, like JSF container, EJB container, CDI container, Servlet container, etc.

All of these containers work kind of independent, they boot in application server initialization and scan classes of all artifacts including jar, ejb-jar, war and ear files in deployment time and gather and store some metadata about them, then when you need an object of a class at runtime they will give you instances of those classes and after finishing the job, they will destroy them.

So we can say that we have:

• JSF managed beans
• CDI managed beans
• EJB managed beans
• And even Servlets are managed beans because they are instantiated and destroyed by a container, which is a servlet container.

So when you see the Managed Bean word, you should ask about the context or type of it.(JSF, CDI, EJB, etc.)

Then you might ask why we have many of these containers: AFAIK, Java EE guys wanted to have a dependency injection framework, but they could not gather all requirements in one specification because they could not predict the future requirements and they made EJB 1.0 and then 2.0 and then 3.0 and now 3.1 but EJB's target was for just some requirements (transaction, distributed component model, etc).

At the same time (in parallel) they realized that they need to support JSF too, then they made JSF managed beans and another container for JSF beans and they considered it a mature DI container, but still it was not complete and mature container.

After that Gavin King and some other nice guys ;) made CDI which is the most mature DI container I've seen. CDI (inspired by Seam2, Guice and Spring) was made to fill the gap between JSF and EJB and lots of other useful stuff like pojo injection, producer methods, interceptors, decorators, integration SPI, very flexible, etc. and it can even do what EJB and JSF managed beans are doing then we can have just one mature and powerful DI container. But for some backward compatibility and political reasons Java EE guys want to keep them!!!

Here you can find the difference and use cases for each of these types:

JSF Managed Beans, CDI Beans and EJBs

JSF was initially developed with its own managed bean and dependency injection mechanism which was enhanced for JSF 2.0 to include annotation based beans. When CDI was released with Java EE 6, it was regarded as the managed bean framework for that platform and of course, EJBs outdated them all having been around for well over a decade.

The problem of course is knowing which one to use and when use them.

Let’s start with the simplest, JSF Managed beans.

JSF Managed Beans

In short, don’t use them if you are developing for Java EE 6 and using CDI. They provide a simple mechanism for dependency injection and defining backing beans for web pages, but they are far less powerful than CDI beans.

They can be defined using the @javax.faces.bean.ManagedBean annotation which takes an optional name parameter. This name can be used to reference the bean from JSF pages.

Scope can be applied to the bean using one of the different scopes defined in the javax.faces.bean package which include the request, session, application, view and custom scopes.

@ManagedBean(name="someBean")
@RequestScoped
public class SomeBean {
    ....
    ....
}

JSF beans cannot be mixed with other kinds of beans without some kind of manual coding.

CDI Beans

CDI is the bean management and dependency injection framework that was released as part of Java EE 6 and it includes a complete, comprehensive managed bean facility. CDI beans are far more advanced and flexible than simple JSF managed beans. They can make use of interceptors, conversation scope, Events, type safe injection, decorators, stereotypes and producer methods.

To deploy CDI beans, you must place a file called beans.xml in a META-INF folder on the classpath. Once you do this, then every bean in the package becomes a CDI bean. There are a lot of features in CDI, too many to cover here, but as a quick reference for JSF-like features, you can define the scope of the CDI bean using one of the scopes defined in the javax.enterprise.context package (namely, request, conversation, session and application scopes). If you want to use the CDI bean from a JSF page, you can give it a name using the javax.inject.Named annotation. To inject a bean into another bean, you annotate the field with javax.inject.Inject annotation.

@Named("someBean")
@RequestScoped
public class SomeBean {

    @Inject
    private SomeService someService;
}

Automatic injection like that defined above can be controlled through the use of Qualifiers that can help match the specific class that you want injected. If you have multiple payment types, you might add a qualifier for whether it is asynchronous or not. While you can use the @Named annotation as a qualifier, you shouldn’t as it is provided for exposing the beans in EL.

CDI handles the injection of beans with mismatched scopes through the use of proxies. Because of this you can inject a request scoped bean into a session scoped bean and the reference will still be valid on each request because for each request, the proxy re-connects to a live instance of the request scoped bean.

CDI also has support for interceptors, events, the new conversation scope and many other features which makes it a much better choice over JSF managed beans.

EJB

EJBs predate CDI beans and are in someways similar to CDI beans and in other ways very different. Primarily, the differences between CDI beans and EJBs is that EJBs are :

• Transactional
• Remote or local
• Able to passivate stateful beans freeing up resources
• Able to make use of timers
• Can be asynchronous

The two types of EJBs are called stateless and stateful. Stateless EJBs can be thought of as thread safe single-use beans that don’t maintain any state between two web requests. Stateful EJBs do hold state and can be created and sit around for as long as they are needed until they are disposed of.

Defining an EJB is simple, you just add either a javax.ejb.Stateless or javax.ejb.Stateful annotation to the class.

@Stateless
public class BookingService {

  public String makeReservation(Item Item, Customer customer) {
    ...
    ...
  }
}

Stateless beans must have a dependent scope while a stateful session bean can have any scope. By default they are transactional, but you can use the transaction attribute annotation.

While EJBs and CDI beans are very different in terms of features, writing the code to integrate them is very similar since CDI beans can be injected into EJBs and EJBs can be injected into CDI beans. There is no need to make any distinction when injecting one into the other. Again, the different scopes are handled by CDI through the use of proxying. One exception to this is that CDI does not support the injection of remote EJBs but that can be implemented by writing a simple producer method for it.

The javax.inject.Named annotation as well as any Qualifiers can be used on an EJB to match it to an injection point.

When to use which bean

How do you know when to use which bean? Simple.

Never use JSF managed beans unless you are working in a servlet container and don’t want to try and get CDI working in Tomcat (although there are some Maven archetypes for that so there’s no excuse).

In general, you should use CDI beans unless you need the advanced functionality available in the EJBs such as transactional functions. You can write your own interceptor to make CDI beans transactional, but for now, it's simpler to use an EJB until CDI gets transactional CDI beans which is just around the corner. If you are stuck in a servlet container and are using CDI, then either hand written transactions or your own transaction interceptor is the only option without EJBs.

If you need to use @ViewScoped in CDI you should

• use seam-faces or MyFaces CODI module. just add one of them to your classpath and @ViewScoped will work in CDI. MyFaces CODI has an even more solid support of @ViewScoped
• use MyFaces CODI's @ViewAccessScoped, it is an extension written on top of CDI by Apache, just download it and use @ViewAccessScoped annotation instead of @ViewScoped.
• Use CDI @ConversationScoped and make it long running. See here for more info.
• Use Omnifaces @ViewScoped annotation

Some parts pilfered from here.

Yep, this can be confusing.

For some ehm historical reasons JSF and CDI are using the same annotations for scopes, but from different packages.

As you are probably guessing those from javax.faces.bean are from the JSF spec, and are not related to CDI. Do not use them unless you have a very good reason to do so. And do never ever mix them with CDI annotations from javax.ejb. This will produce a sheer endless lists of bugs and subtle anomalies.

I generally recommend that you skim the first few (or even more) pages of the excellent Weld documentation. This should put you on track for Java EE 6.

And feel free to post further questions here.

Since there are no replies specifically about @javax.annotation.ManagedBean, here's a link to the answer of a similar question: Backing beans (@ManagedBean) or CDI Beans (@Named)?. The spec can be found at http://download.oracle.com/otndocs/jcp/managed_beans-1.0-fr-eval-oth-JSpec/. So it looks to me that @javax.annotation.ManagedBean was meant to be a generalization of @javax.faces.bean.ManagedBean.

From what I gathered, JSF Managed Beans are being phased out in favour of CDI Beans (maybe getting deprecated from JSF 2.3?), so I guess @javax.annotation.ManagedBean is all the more becoming obsolete now.