BT
x Your opinion matters! Please fill in the InfoQ Survey about your reading habits!

What's New in Spring 2.5: Part 1

Posted by Mark Fisher on Nov 19, 2007 |

Introduction

Since its inception the Spring Framework has consistently focused on the goal of simplifying enterprise application development while providing powerful, non-invasive solutions to complex problems. With the release of Spring 2.0 just over a year ago, these themes advanced to a new level. XML Schema support and custom namespaces reduce the amount of XML-based configuration. Developers using Java 5 or greater are able to take advantage of Spring libraries that exploit new language features such as generics and annotations. The close integration with AspectJ's expression language enables the non-invasive addition of behavior across well-defined groupings of Spring-managed objects.

The newly released Spring 2.5 continues this trend by offering further simplifications and powerful new features especially for those who are using Java 5 or greater. These features include annotation-driven dependency injection, auto-detection of Spring components on the classpath using annotations rather than XML for metadata, annotation support for lifecycle methods, a new web controller model for mapping requests to annotated methods, support for Junit 4 in the test framework, new additions to the Spring XML namespaces, and more.

This article is the first of a three-part series exploring these new features. The current article will focus on simplified configuration and new annotation-based functionality in the core of the Spring application context. The second article will cover new features available in the web-tier, and the final article will highlight additional features available for integration and testing. Most of the examples depicted within this article series are based upon the Spring PetClinic sample application. That sample has recently been refactored to serve as a showcase of the latest Spring functionality and is included within the Spring 2.5 distribution available at the Spring Framework Download page. View the 'readme.txt' file within the 'samples/petclinic' directory for instructions on building and deploying the PetClinic application. Experimenting with the showcased features in the PetClinic application is probably the best way to master the new techniques discussed here.

Spring support for JSR-250 annotations

The 'Common Annotations for the Java Platform' were introduced with Java EE version 5 and are also included out-of-the-box beginning with Java SE version 6. In May 2006, BEA Systems announced their collaboration with Interface21 on a project called Pitchfork that provided a Spring-based implementation of the Java EE 5 programming model including support for JSR-250 annotations and EJB 3 annotations (JSR-220) for injection, interception, and transactions. As of version 2.5, the Spring Framework core now provides support for the following JSR-250 annotations:

  • @Resource
  • @PostConstruct
  • @PreDestroy

With Spring these annotations are supported in any environment - with or without an Application Server - and even for integration testing. Enabling this support is just a matter of registering a single Spring post-processor:

<bean class="org.springframework.context.annotation.CommonAnnotationBeanPostProcessor"/> 

The @Resource Annotation

The @Resource annotation enables dependency injection of a "named resource". Within a Java EE application, that typically translates to an object bound to the JNDI context. Spring does support this usage of @Resource for resolving objects through JNDI lookups, but by default the Spring-managed object whose "bean name" matches the name provided to the @Resource annotation will be injected. In the following example, Spring would pass a reference to the Spring-managed object with a bean name of "dataSource" to the annotated setter method.

@Resource(name="dataSource")
public void setDataSource(DataSource dataSource) {
this.dataSource = dataSource;
}

It is also possible to annotate a field directly with @Resource. By not exposing a setter method, the code is more concise and also provides the additional benefit of enforcing immutability. As demonstrated below, the @Resource annotation does not even require an explicit String value. When none is provided, the name of the field will be used as a default.

@Resource
private DataSource dataSource; // inject the bean named 'dataSource'

When applied to a setter method, the default name would be derived from the corresponding property. In other words, a method named 'setDataSource' would also resolve to the property named 'dataSource'.

private DataSource dataSource;
@Resource
public void setDataSource(DataSource dataSource) {
this.dataSource = dataSource;
}

When using @Resource without an explicitly provided name, if no Spring-managed object is found for the default name, the injection mechanism will fallback to a type-match. If there is exactly one Spring-managed object matching the dependency's type, then it will be injected. This feature can be disabled by setting the 'fallbackToDefaultTypeMatch' property of the CommonAnnotationBeanPostProcessor to 'false' (it is 'true' by default).

<bean class="org.springframework.context.annotation.CommonAnnotationBeanPostProcessor">
<property name="fallbackToDefaultTypeMatch" value="false"/>
</bean>

As mentioned above, Spring does provide support for JNDI-lookups when resolving dependencies that are annotated with @Resource. To force direct JNDI lookups for all dependencies annotated with @Resource, set the 'alwaysUseJndiLookup' flag of the CommonAnnotationBeanPostProcessor to 'true' (it is 'false' by default).

<bean class="org.springframework.context.annotation.CommonAnnotationBeanPostProcessor">
<property name="alwaysUseJndiLookup" value="true"/>
</bean>

Alternatively, to enable lookup based upon global JNDI names specified as 'resource-ref-mappings', provide the 'mappedName' attribute within the @Resource annotation. Even when the target object is actually a JNDI resource, it is the recommended practice to still reference a Spring-managed object thereby providing a level of indirection and hence a lesser degree of coupling. With the namespace additions available since Spring 2.0, a bean definition that delegates to Spring for handling the JNDI lookup is trivial and concise:

<jee:jndi-lookup id="dataSource" jndi-name="java:comp/env/jdbc/petclinic"/> 

The advantage of this approach is that the level of indirection provides for greater deployment flexibility. For example, a standalone system test environment should not require a JNDI registry. In such a case, the following alternate bean definition could be provided within the system test configuration:

<bean id="dataSource" class="org.springframework.jdbc.datasource.DriverManagerDataSource"
p:driverClassName="${jdbc.driverClassName}"
p:url="${jdbc.url}"
p:username="${jdbc.username}"
p:password="${jdbc.password}"/>

As an aside, in the example above the actual JDBC connection properties are resolved from a properties file where the keys match the provided ${placeholder} tokens. This is accomplished by registering a Spring BeanFactoryPostProcessor implementation called PropertyPlaceholderConfigurer. This is a commonly used technique for externalizing those properties - often environment-specific ones - that may need to change more frequently than the rest of the configuration.

<bean class="org.springframework.beans.factory.config.PropertyPlaceholderConfigurer">
<property name="location" value="classpath:jdbc.properties"/>
</bean>

With the addition of the 'context' namespace in Spring 2.5, a more concise alternative for the property placeholder configuration is available:

<context:property-placeholder location="classpath:jdbc.properties"/> 

Lifecycle Annotations: @PostConstruct and @PreDestroy

The @PostConstruct and @PreDestroy annotations can be used to trigger Spring initialization and destruction callbacks respectively. This feature extends but does not replace the two options for providing such callbacks in Spring versions prior to 2.5. The first option is to implement one or both of Spring's InitializingBean and DisposableBean interfaces. Each of those interfaces requires a single callback method implementation (afterPropertiesSet() and destroy() respectively). The interface-based approach takes advantage of Spring's ability to automatically recognize any Spring managed object implementing those interfaces and therefore requires no additional configuration. On the other hand, a key goal of Spring is to be as non-invasive as possible. Therefore instead of implementing Spring-specific interfaces, many Spring users have taken advantage of the second option which is to provide their own initialization and destruction methods. While less invasive, the drawback of that approach is that it requires explicit declaration of 'init-method' and/or 'destroy-method' attributes on the 'bean' element. That explicit configuration is sometimes necessary, such as when the callbacks need to be invoked on code that is outside of the developer's control. The PetClinic application demonstrates this scenario. When it is run with its JDBC configuration, a third party DataSource is used, and a 'destroy-method' is declared explicitly. Also notice that the standalone connection-pooling DataSource is yet another deployment option for the 'dataSource' and does not require any code changes.

<bean id="dataSource"
class="org.apache.commons.dbcp.BasicDataSource"
destroy-method="close"
p:driverClassName="${jdbc.driverClassName}"
p:url="${jdbc.url}"
p:username="${jdbc.username}"
p:password="${jdbc.password}"/>

With Spring 2.5, if an object requires invocation of a callback method upon initialization, that method can be annotated with the @PostConstruct annotation. For example, imagine a background task that needs to start polling a file directory upon startup.

public class FilePoller {

@PostConstruct
public void startPolling() {
...
}
...
}

Similarly, a method annotated with @PreDestroy on a Spring-managed object will be invoked when the application context hosting that object is closed.

public class FilePoller {

@PreDestroy
public void stopPolling() {
...
}
...
}

With the addition of support for the JSR-250 annotations, Spring 2.5 now combines the advantages of its two previous lifecycle method alternatives. Adding @PostConstruct and @PreDestroy as method-level annotations is sufficient for triggering the callbacks within a Spring managed context. In other words, no additional XML-based configuration is necessary. At the same time, the annotations are part of the Java language itself (and even included within Java SE as of version 6) and thus require no Spring-specific imports. The annotations have the added benefit of indicating semantics that should be understood within other environments, and over time Java developers can likely expect to see these annotations used more frequently within third-party libraries. Finally, one interesting consequence of the annotation-based lifecycle callbacks is that more than one method may carry either annotation, and all annotated methods will be invoked

To enable all of the behavior as described above for the @Resource, @PostConstruct, and @PreDestroy annotations, provide a single bean definition for Spring's CommonAnnotationBeanPostProcessor as shown previously. An even more concise option is possible with the new 'context' namespace in 2.5:

<context:annotation-config/> 

Including that single element will not only register a CommonAnnotationBeanPostProcessor, but it will also enable the autowiring behavior as described in the section that follows. The CommonAnnotationBeanPostProcessor even provides support for @WebServiceRef and @EJB annotations. These will be covered in the third article of this series along with other new Spring 2.5 features for enterprise integration.

Fine-Grained Autowiring with Annotations

Documentation covering Spring's support for autowiring has often been accompanied with caveats due to the coarse granularity of the autowiring mechanism. Prior to Spring 2.5, autowiring could be configured for a number of different approaches: constructor, setters by type, setters by name, or autodetect (where Spring chooses to autowire either a constructor or setters by type). These various options do offer a large degree of flexibility, but none of them offer very fine-grained control. In other words, prior to Spring 2.5, it has not been possible to autowire a specific subset of an object's setter methods or to autowire some of its properties by type and others by name. As a result, many Spring users have recognized the benefits of autowiring for prototyping and testing, but when it comes to maintaining and supporting systems in production, most agree that the added verbosity of explicit configuration is well worth the clarification it affords.

However, Spring 2.5 dramatically changes the landscape. As described above, the autowiring choices have now been extended with support for the JSR-250 @Resource annotation to enable autowiring of named resources on a per-method or per-field basis. However, the @Resource annotation alone does have some limitations. Spring 2.5 therefore introduces an @Autowired annotation to further increase the level of control. To enable the behavior described in this section, register a single bean definition:

<bean class="org.springframework.beans.factory.annotation.AutowiredAnnotationBeanPostProcessor"/> 

Alternatively, the 'context' namespace provides a more concise alternative as shown previously. This will enable both post-processors discussed in this article (AutowiredAnnotationBeanPostProcessor and CommonAnnotationBeanPostProcessor) as well as the annotation-based post-processors that were introduced in Spring 2.0: RequiredAnnotationBeanPostProcessor and PersistenceAnnotationBeanPostProcessor.

<context:annotation-config/> 

With the @Autowired annotation, it is possible to inject dependencies that match by type. This behavior is enabled for fields, constructors, and methods. In fact, autowired methods do not have to be 'setter' methods and can even accept multiple parameters. The following is perfectly acceptable:

@Autowired
public void setup(DataSource dataSource, AnotherObject o) { ... }

By default, dependencies marked with the @Autowired annotation are treated as required. However, it is also possible to declare any of them as optional by setting the 'required' attribute to 'false'. In the following example, DefaultStrategy will be used if no Spring-managed object of type SomeStrategy is found within the context.

@Autowired(required=false)
private SomeStrategy strategy = new DefaultStrategy();

Autowiring by type can obviously result in ambiguities when the Spring context contains more than one object of the expected type. By default, the autowiring mechanism will fail if there is not exactly one bean for a required dependency. Likewise for any optional properties, it will fail if more than one candidate is available (if optional and zero candidates are available, then it will simply skip the property). There are a number of configuration options for avoiding these conflicts.

When there is one primary instance of a given type within the context, the bean definition for that type should contain the 'primary' attribute. This approach works well when other instances may be available in the context, yet those non-primary instances are always explicitly configured.

<bean id="dataSource" primary="true" ... /> 

When more control is needed, any autowired field, constructor argument, or method parameter may be further annotated with a @Qualifier annotation. The qualifier may contain a String value in which case Spring will attempt to match by name.

@Autowired
@Qualifier("primaryDataSource")
private DataSource dataSource;

The main reason that @Qualifier exists as a separate annotation is so that it can be applied at the level of a constructor argument or method parameter while the @Autowired annotation is available on the constructor or method itself.

@Autowired
public void setup(@Qualifier("primaryDataSource") DataSource dataSource, AnotherObject o) { ... }

The fact that @Qualifier is a separate annotation provides even more benefits with regard to customization. User-defined annotations may also play the role of qualifier in the autowiring process. The simplest way to accomplish this is to annotate the custom annotation with @Qualifier itself as a meta-annotation.

@Target({ElementType.FIELD, ElementType.PARAMETER, ElementType.TYPE, ElementType.ANNOTATION_TYPE})
@Retention(RetentionPolicy.RUNTIME)
@Qualifier
public @interface VetSpecialty { ... }

Custom annotations may optionally include a value for matching by name but more commonly would be used as "marker" annotations or define a value that provides some further meaning to the qualifier process. For example, the excerpt below depicts a field that should be autowired with a qualified candidate among those that match by type.

@Autowired
@VetSpecialty("dentistry")
private Clinic dentistryClinic;

When using XML configuration for the target of this dependency resolution, 'qualifier' sub-elements may be added to the bean definition. In the next section on component-scanning, a non-XML alternative will be presented.

<bean id="dentistryClinic" class="samples.DentistryClinic">
<qualifier type="example.VetSpecialty" value="dentistry"/>
</bean>

To avoid any dependency on the @Qualifier annotation whatsoever, provide a CustomAutowireConfigurer bean definition within the Spring context and register any custom annotation types directly:

<bean class="org.springframework.beans.factory.annotation.CustomAutowireConfigurer">
<property name="customQualifierTypes">
<set>
<value>example.VetSpecialty</value>
</set>
</property>
</bean>

Now that the custom qualifier has been explicitly declared, the @Qualifier meta-annotation is no longer required.

@Target({ElementType.FIELD, ElementType.PARAMETER, ElementType.TYPE, ElementType.ANNOTATION_TYPE})
@Retention(RetentionPolicy.RUNTIME)
public @interface VetSpecialty { ... }

On a related note, it is even possible to replace the @Autowired annotation itself when configuring the AutowiredAnnotationBeanPostProcessor.

<bean class="org.springframework.beans.factory.annotation.AutowiredAnnotationBeanPostProcessor">
<property name="autowiredAnnotationType" value="example.Injected"/>
</bean>

In a majority of cases, the ability to define custom "marker" annotations combined with the options of matching by name or other semantic value should be sufficient for achieving fine-grained control of the autowiring process. However, Spring also provides support for any number of arbitrary attributes on qualifier annotations. For example, the following is a hypothetical example of a very fine-grained qualifier.

@SpecializedClinic(species="dog", breed="poodle")
private Clinic poodleClinic;

The custom qualifier implementation would define those attributes.

@Target({ElementType.FIELD, ElementType.PARAMETER, ElementType.TYPE, ElementType.ANNOTATION_TYPE})
@Retention(RetentionPolicy.RUNTIME)
@Qualifier
public @interface SpecializedClinic {

String species();

String breed();

}

The custom qualifier's attributes can then match against 'attribute' sub-elements of the 'qualifier' annotation within the XML of a bean definition. These elements are used to provide key/value pairs.

<bean id="poodleClinic" class="example.PoodleClinic">
<qualifier type="example.SpecializedClinic">
<attribute key="species" value="dog"/>
<attribute key="breed" value="poodle"/>
</qualifier>
</bean>

All of the autowiring demonstrated so far has been for single instances, but collections are supported as well. Any time it's necessary to get all Spring-managed objects of a certain type within the context, simply add the @Autowired annotation to a strongly-typed Collection.

@Autowired
private List<Clinic> allClinics;

One final feature that is worth pointing out in this section is the use of autowiring in place of Spring's "Aware" interfaces. Prior to Spring 2.5, if an object requires a reference to the Spring context's ResourceLoader, it can implement ResourceLoaderAware thereby allowing Spring to provide this dependency via the setResourceLoader(ResourceLoader resourceLoader) method. This same technique applies for obtaining a reference to the Spring-managed MessageSource and even the ApplicationContext itself. For Spring 2.5 users, this behavior is now fully supported through autowiring (note that the inclusion of these Spring-specific dependencies should always be carefully considered and typically only used within "infrastructure" code that is clearly separated from business logic).

@Autowired
private MessageSource messageSource;

@Autowired
private ResourceLoader resourceLoader;

@Autowired
private ApplicationContext applicationContext;

Auto-Detection of Spring Components

Beginning with version 2.0, Spring introduced the concept of "stereotype" annotations with the @Repository annotation serving as a marker for data access code. Spring 2.5 adds two new annotations - @Service and @Controller - to complete the role designations for a common three-tier architecture (data access objects, services, and web controllers). Spring 2.5 also introduces the generic @Component annotation which the other stereotypes logically extend. By clearly indicating application roles, these stereotypes facilitate the use of Spring AOP and post-processors for providing additional behavior to the annotated objects based on those roles. For example, Spring 2.0 introduced the PersistenceExceptionTranslationPostProcessor to automatically enable data access exception translation for any object carrying the @Repository annotation.

These same annotations can also be used in conjunction with another new feature of Spring 2.5: auto-detection of components on the classpath. Although XML has traditionally been the most popular format for Spring metadata, it is not the only option. In fact, the Spring container's internal metadata representation is pure Java, and when XML is used to define Spring-managed objects, those definitions are parsed and converted to Java objects prior to the instantiation process. One significant new capability of Spring 2.5 is the support for reading that metadata from source-level annotations. The autowiring mechanisms described thus far make use of annotation metadata for injecting dependencies but still require registration of at least a minimal "bean definition" in order to provide the implementation class of each Spring-managed object. The component scanning functionality can remove the need for even that minimal bean definition in XML.

As seen above, Spring's annotation-driven autowiring can significantly reduce the amount of XML without sacrificing fine-grained control. The component detection mechanism takes this even further. It is not necessary to completely supplant configuration in XML, rather the component scanning can operate alongside XML metadata to simplify the overall configuration. This possibility of combining XML and annotation-driven techniques can lead to a well-balanced approach as demonstrated by the 2.5 version of the PetClinic sample application. There, the infrastructural components (data source, transaction manager, etc) are defined in XML along with externalized properties as described above. The data access tier objects are also defined partially in XML, but their configuration also takes advantage of the @Autowired annotations to simplify the injection of dependencies. Finally, the web tier "controllers" are not explicitly defined in XML at all. Instead the following configuration is used to trigger the auto-detection of all web controllers:

<context:component-scan base-package="org.springframework.samples.petclinic.web"/> 

Notice that the 'base-package' attribute is provided. The default matching rules for component-scanning will recursively detect any of Spring's stereotype annotations on classes within that package (multiple packages can be provided in a comma-separated list). Therefore, the various controller implementations for the PetClinic sample application are all annotated with @Controller (one of Spring's built-in stereotypes). Here is an example:

@Controller
public class ClinicController {

private final Clinic clinic;

@Autowired
public ClinicController(Clinic clinic) {
this.clinic = clinic;
}
...

Auto-detected components are registered with the Spring container just as if they had been defined in XML. As depicted above, those objects can in turn make use of annotation-driven autowiring.

The component scanner's matching rules can also be customized with filters for including or excluding components based on type, annotation, AspectJ expression, or regular expressions for name patterns. The default stereotypes can also be disabled. For example, a test configuration may ignore the default stereotypes and instead auto-detect any class whose name starts with Stub or which includes the @Mock annotation:

<context:component-scan base-package="example" use-default-filters="false">
<context:include-filter type="aspectj" expression="example..Stub*"/>
<context:include-filter type="annotation" expression="example.Mock"/>
</context:component-scan>

Type matching restrictions can be controlled with exclusion filters as well. For example, to rely on the default filters except for the @Repository annotation, then add an exclude-filter.

<context:component-scan base-package="example">
<context:exclude-filter type="annotation" expression="org.springframework.stereotype.Repository"/>
</context:component-scan>

Clearly it is possible to extend the component scanning in a number of ways to register your own custom types. The stereotype annotations are the simplest option, so the notion of stereotype is therefore extensible itself. As mentioned earlier, @Component is the generic stereotype indicator that the @Repository, @Service, and @Controller annotations "logically" extend. It just so happens that @Component can be provided as a meta-annotation (i.e. an annotation declared on another annotation), and any custom annotation that has the @Component meta-annotation will be automatically detected by the default matching rules of the scanner. An example will hopefully reveal that this is much simpler than it sounds.

Recall the hypothetical background task that was described in the section above covering the @PostConstruct and @PreDestroy lifecycle annotations. Perhaps an application has a number of such background tasks, and those task instances would typically require XML bean definitions in order to be registered with the Spring context and have their lifecycle methods invoked at the right time. With component scanning, there is no longer a need for those explicit XML bean definitions. If the background tasks all implement the same interface or follow a naming convention, then include-filters could be used. However, an even simpler approach is to create an annotation for these task objects and provide the @Component meta-annotation.

@Target({ElementType.TYPE})
@Retention(RetentionPolicy.RUNTIME)
@Documented
@Component
public @interface BackgroundTask {
String value() default "";
}

Then provide the custom stereotype annotation in any background task's class definitions.

@BackgroundTask
public class FilePoller {

@PostConstruct
public void startPolling() {
...
}

@PreDestroy
public void stopPolling() {
...
}
...
}

The generic @Component annotation could just as easily have been provided instead, but the custom annotation technique provides an opportunity for using meaningful, domain-specific names. These domain-specific annotations then provide further opportunities such as using an AspectJ pointcut expression to identify all background tasks for the purpose of adding advice that monitors the activity of those tasks.

By default, when a component is detected, Spring will automatically generate a "bean name" using the non-qualified class name. In the previous example, the generated bean name would be "filePoller". However, for any class that is annotated with one of Spring's stereotype annotations (@Component, @Repository, @Service, or @Controller) or any other annotation that is annotated with @Component as a meta-annotation (such as @BackgroundTask in the above example), the 'value' attribute can be explicitly specified for the stereotype annotation, and the instance will then be registered within the context with that value as its "bean name". In the following example the name would be "petClinic" instead of the default generated name of "simpleJdbcClinic".

@Service("petClinic")
public class SimpleJdbcClinic {
...
}

Likewise, the bean name generated for the following revised version of the FilePoller would be "poller" instead of "filePoller".

@BackgroundTask("poller")
public class FilePoller {
...
}

While all Spring-managed objects are treated as singleton instances by default, it is sometimes necessary to specify an alternate "scope" for an object. For example, in the web-tier a Spring-managed object may be bound to 'request' or 'session' scope. As of version 2.0, Spring's scope mechanism is even extensible so that custom scopes can be registered with the application context. Within an XML configuration, it's simply a matter of including the 'scope' attribute and the name of the scope.

<bean id="shoppingCart" class="example.ShoppingCart" scope="session">
...
</bean>

With Spring 2.5, the same can be accomplished for a scanned component by providing the @Scope annotation.

@Component
@Scope("session")
public class ShoppingCart {
...
}

One final topic to address here is the simplification of qualifier annotations when using component-scanning. In the previous section, the following object was used as an example of autowiring with a custom qualifier annotation:

@VetSpecialty("dentistry")
private Clinic dentistryClinic;

That same example then featured use of a 'qualifier' element within the XML on the intended target bean definition for that dependency. When using component-scanning, the XML metadata is not necessary. Instead, the custom qualifier may be included as a type-level annotation in the target class definition. An alternative example with a scanned @Repository instance as the dependency would thus appear as follows:

@Repository
@VetSpecialty("dentistry")
public class DentistryClinic implements Clinic {
...
}

Finally, for the previous example that featured custom qualifier annotations with attributes, the non-XML equivalent for that dependency's target would be:

@Repository
@SpecializedClinic(species="dog", breed="poodle")
public class PoodleClinic implements Clinic {
...
}

Conclusion

Spring 2.5 offers significant new functionality in a number of areas. The primary focus of this article has been on simplifying configuration by harnessing the power of Java annotations. Spring supports 'Common Annotations' as defined in JSR-250 while providing additional annotations for even more fine-grained control of the autowiring process. Spring 2.5 also extends the 'stereotype' annotations that began with Spring 2.0's @Repository, and all of these stereotypes can be used in conjunction with the new component-scanning functionality. XML-based configuration is still fully supported, and Spring 2.5 introduces a new 'context' namespace that offers more concise syntax for common configuration scenarios. In fact, the support for seamlessly combining XML and annotation-based configuration enables a well-balanced overall approach. Complex configuration of infrastructure can be defined in modular XML files while the progressively higher layers of an application stack can benefit from more annotation-based techniques - all within the same Spring 2.5 application context.

Stay tuned for the next article in this series, which will cover powerful new annotation-based functionality in the Spring web tier.

Hello stranger!

You need to Register an InfoQ account or or login to post comments. But there's so much more behind being registered.

Get the most out of the InfoQ experience.

Tell us what you think

Allowed html: a,b,br,blockquote,i,li,pre,u,ul,p

Email me replies to any of my messages in this thread

Missing/hidden XML by Matt Raible

There seems to be a fair amount of missing or hidden XML in this article. If you look at the HTML source, you'll see where some XML has not been properly escaped.

Re: Missing/hidden XML by Mark Fisher

Matt,

You are right. Someone must have edited this recently, because the XML was showing up properly before. I just sent a mail to the editor.

Thanks,
Mark

Re: Missing/hidden XML by Diana Baciu

Hi Matt, Mark

The XML has been fixed now.

Best
Diana

Good Stuff by Ray Krueger

I am really looking forward to using these new features, great job guys.

By the way Mark, your sudden and unexplained use of the p: namespace might freak people out. In your explanation of the lifecycle annotations you declare the datasource using the p: namespace trick from blog.interface21.com/main/2006/11/25/xml-syntax... Unfortunately, many people probably aren't aware of this feature and it isn't actually mentioned in the article.

Grate Guice alternative by Thom Nichols

This is great -- I always liked the annotation-driven style of Guice but didn't want to abandon Spring just for that. Hooray for less XML!

Re: Good Stuff by Mark Fisher

Ray,

Thanks for pointing out that blog for the 'p' namespace. Hopefully that will serve as a 'footnote' now for anyone who may be confused by its usage. Those examples are taken directly from the PetClinic application by the way.

Thanks,
Mark

Annotations vs. XML by Geoffrey Wiseman

I have to admit, I'm content to use XML for wiring, myself -- although perhaps when I try the annotations, I'll discover more advantages than I expect. I seem to be in the minority here.

That said, I was more impressed by the features here than I expected to be, so next time I start some Spring config., I might give this a try.

Sorry about the XML; that might have been my fault, I changed some metadata after the initial publishing, and there are some quirks in the publishing process that Diana's better at handling. ;)

I'm looking forward to the next two parts of the article.

Re: Annotations vs. XML by Rod Johnson

Geoffrey
I have to admit, I'm content to use XML for wiring, myself -- although perhaps when I try the annotations, I'll discover more advantages than I expect. I seem to be in the minority here.

The goal of Spring is to be the ultimate component model. That component model can be configured in different ways. There is no perfect one size fits all approach to configuration. Different contributions are merged together by the container. Certainly, annotations have an important place. However, my experience in practice (long predating my creation of Spring) has been that you need to externalize significant parts of your configuration from Java code.

The annotation support in Spring 2.5 is very slick and definitely makes Spring a better product. You can mix and match annotation-driven and XML (and other) configuration so that you can use the appropriate solution for each problem. I'm proud that each version of Spring has made applications easier to build.

I recently presented on Configuring the Spring Container at QCon San Francisco, discussing alternative configuration options and best practices. The slides are available as PDF.

Re: Annotations vs. XML by Jörg Gottschling

The annotation @Resource is a little confusing in the context of spring. As I saw it I first thought it will be used for ressources and not for beans. Like that:

@Ressource("file:out/example.txt")
public void setOutput(Ressource ressource) {...}

Re: Annotations vs. XML by Rod Johnson

Jorg

I agree that @Resource is a poor name for the annotation, but we didn't choose it...

Rgds
Rod

lament the passing of design and OO programming by James Richardson

as now i can turn everything into a FactoryBeanImpl.

@Bonkers(Retention.FORALLTIME)
afterPropertiesSet()

Great Job by Khaled Habiburahman

Great Job, keep it up

Thanks

Re: Missing/hidden XML by mohan raj

Good Artical...

Re: Great Job by mohan raj

Of course...Great

the 2nd part and 3rd part by Nantian Lotus

Good material. But where can I find the 2nd part and 3rd part of this series.
Thanks.

Re: Good Stuff by Darryl Pentz

Not to mention how it's never mentioned how to actually create an instance of the object that is injected by Spring. We see the annotated class, and the <bean.../> configuration XML, but no idea how to construct an instance of the annotated class. Does it appear as if by magic? Is it just me, but why are there no examples of this. Everything is just assumed to originate as some implicit class of the Spring framework, like a controller or some such.

I'd like to use this uber-magic of Spring, but how do I create instances of my own 'controller-type' classes automagically injected by Spring. Is this a state secret perhaps? Sorry, just frustrated by the lack of examples, clearly.</bean.../>

Re: the 2nd part and 3rd part by Lou Sacco

Agreed...this would be very useful.

Re: Good Stuff by Archie Cobbs

Darryl Pentz writes:

Not to mention how it's never mentioned how to actually create an instance of the object that is injected by Spring. We see the annotated class, and the configuration XML, but no idea how to construct an instance of the annotated class. Does it appear as if by magic?


I had the same question the first time I read this article (too quickly). The answer lies in the "Auto-Detection of Spring Components". The key "magic" is this tag:


<context:component-scan ... />


which causes Spring to automatically go hunting through your JARs looking for specially-annotated classes. When it finds them, it auto-constructs instances and adds them to your application context. The net effect is a purely annotation-driven (zero XML) way to add and configure individual beans.

2nd and 3rd parts by Gregor Morrison

Mark, great work, are parts 2 and 3 available yet?

Re: 2nd and 3rd parts by Mohammad Naqvi

Hey Greg, Did you hear anything about the 2nd and 3rd parts yet?

the last part by john wang

"While all Spring-managed objects are treated as singleton instances by default, it is sometimes necessary to specify an alternate "scope" for an object. "

I believe with the @PostConstruct @PreDestroy, the default should change to "request" or "prototype", not singleton any more... otherwise it is very possibly not thread-safe....

And is there any plan to support more lifecycle/AOP like annoations? like @PreInvoke @PostInvoke....

Re: the last part by Lukasz Budnik

I wrote a series of posts called "Spring for JEE developers"

I describe some of Spring 2.5 new features like partial implementations of JSR 220 and JSR 250.

If someone is interested here they are:

jee-bpel-soa.blogspot.com/2008/11/spring-for-je...
jee-bpel-soa.blogspot.com/2008/11/spring-for-je...
jee-bpel-soa.blogspot.com/2008/11/spring-for-je...

best regards

Re: 2nd and 3rd parts by Johan Pelgrim

Here's the second part (haven't found the third part yet)

www.infoq.com/articles/spring-2.5-ii-spring-mvc

Allowed html: a,b,br,blockquote,i,li,pre,u,ul,p

Email me replies to any of my messages in this thread

Allowed html: a,b,br,blockquote,i,li,pre,u,ul,p

Email me replies to any of my messages in this thread

23 Discuss

Educational Content

General Feedback
Bugs
Advertising
Editorial
InfoQ.com and all content copyright © 2006-2014 C4Media Inc. InfoQ.com hosted at Contegix, the best ISP we've ever worked with.
Privacy policy
BT