As far as I understood, Spring manages autowiring mechanism with AutowiredAnnotationBeanPostProcessor on postProcessBeforeInitialization stage. But how does it inject proxies that ought to be created on postProcessAfterInitialization stage?
EDIT 1
Suppose I have this Spring configuration
#Service
class RegularBean {
// injected on postProcessBeforeInitialization stage
#Autowired
private TransactionBean tBean;
// invoked in between of postProcessBeforeInitialization and postProcessAfterInitialization
#PostConstruct
void init() {
tBean.transactionMethod();
}
}
#Service
class TransactionBean {
// transactional proxy is created on postProcessAfterInitialization stage
#Transactional
public void transactionMethod() { ... }
}
Transactional proxy is created on postProcessAfterInitialization stage. But #PostConstruct is called right before it. Is injected tBean wrapped with transactional proxy? If it so, then why? Because it should not be. If it is not wrapped, then how transactions are going to be handled in the future?
Suppose that I replace field-injection with constructor-injection. Will it change the behavior somehow?
When you use autowiring on method or field ,Spring Container not always create and inject the required field/attribute instance. Spring internally create smart proxies and inject the proxies to your bean. This smart proxy will resolve the bean at later point and delegate call to actual bean during method invocation. This is a common strategy we use to resolve request and session scoped beans to singleton instance (Say service layer beans) using Scope annotation.
Adding small snippet to show Spring internally create proxies for Object. Consider a classic circular dependency case when we use Constructor injection.
#Component
public class CircularDependencyA {
private CircularDependencyB circB;
public CircularDependencyA(#Lazy CircularDependencyB circB) {
System.out.println("CircularDependencyA Ctr ->"+circB.getClass().getName());
this.circB = circB;
}
}
#Component
public class CircularDependencyB {
private CircularDependencyA circA;
public CircularDependencyB(CircularDependencyA circA) {
System.out.println("CircularDependencyB Ctr ->"+circA.getClass().getName());
this.circA = circA;
}
}
#Configuration
#ComponentScan(basePackages = { "com.example.springdemo.cd" })
public class TestConfig {
}
public class TestCircularDependency {
public static void main(String[] args) {
try(AnnotationConfigApplicationContext context
= new AnnotationConfigApplicationContext(TestConfig.class);){
}
}
}
Based on our hints Spring is creating Proxy(using CGLIB) for CircularDependencyB Object and see the op from the CircularDependencyA Constructor
CircularDependencyA Ctr ->com.example.springdemo.cd.CircularDependencyB$$EnhancerBySpringCGLIB$$e6be3b79
Thanks
Related
This question already has answers here:
Injecting beans into a class outside the Spring managed context
(8 answers)
Closed 3 years ago.
Introduction
I have some business logic properties in the application.yml file.
They are loaded into the application via a #ConfigurationProperties annotated class.
How could I use these properties in a class which is not a Spring Bean? It cannot be a singleton, because many objects of it must be created during run-time.
Example
application.yml
business.foo: 2
BusinessProperties.java
#ConfigurationProperties("business")
#Getter // lombok
#Setter // lombok
public class BusinessProperties {
private int foo;
}
TypicalBean.java
#Component
public class TypicalBean {
private final BusinessProperties properties;
#Autowired
public TypicalBean(BusinessProperties properties) {
this.properties = properties;
}
#PostConstruct
public void printFoo() {
System.out.println("Foo: " + properties.getFoo()); // "Foo: 2"
}
}
NonBean.java
public class NonBean {
public void printFoo() {
System.out.println("Foo: ???"); // How to access the property?
}
}
Is there some way to create a non-singleton class, which can have access to configuration (or even other Spring beans) but otherwise works the same as a regular java class? Meaning that I can control its creation, it is collected by the garbage collector if not used anymore, etc.
You can still define the NonBean.class as a Component with Scope.Prototype
#Component
#Scope(value = BeanDefinition.SCOPE_PROTOTYPE)
public class NonBean {
#Autowired
public TypicalBean(BusinessProperties properties) {
this.properties = properties;
}
public void printFoo() {
System.out.println("Foo: " + properties.getFoo());
}
}
The trick is how you create an instance of NonBean.class. In the code where you'll be creating an instance of NonBean.class, use Spring's ObjectFactory<T>
private final ObjectFactory<NonBean> nonBeanFactory;
...
NonBean nonBean = nonBeanFactory.getObject();
The instantiated nonBean object will have been autowired.
All spring-beans creates by SpringApplicationContext. Bean - it's simple POJO-object, but created by Spring and saved in his container. If you want to get access to bean from outside of container - see this:
Getting Spring Application Context
Spring beans are really meant to be used within the application context but you might be able to achieve what you want by autowiring the properties to a static field in a Spring bean.
#Component
public class BusinessPropertiesUtils {
public static BusinessProperties INSTANCE;
#Autowired
public setBusinessProperties(BusinessProperties properties) {
this.INSTANCE = properties;
}
}
And then:
public class NonBean {
public void printFoo() {
System.out.println("Foo: " + BusinessPropertiesUtils.INSTANCE.getFoo());
}
}
PS: this is very hacky and definitely not the "Spring way".
You can configure beans with the prototype scope, which will give you a new instance of the bean every time it's requested.
From the Spring documentation:
In contrast to the other scopes, Spring does not manage the complete lifecycle of a prototype bean. The container instantiates, configures, and otherwise assembles a prototype object and hands it to the client, with no further record of that prototype instance.
...
In some respects, the Spring container’s role in regard to a prototype-scoped bean is a replacement for the Java new operator. All lifecycle management past that point must be handled by the client.
Example of how you can convert the TypicalBean class to a prototype scoped bean:
#Component
#Scope("prototype")
public class TypicalBean {
...
}
Another alternative is to manually instantiate the bean class (or any POJO) and injecting the dependencies (configuration, spring beans, etc.) through the constructor or setter methods, if you have them available or can get them from the Spring Context.
new TypicalBean(properties);
I would like to use spring transaction management capabilities within a prototype bean. I did the following:
I've used javax.inject.Provider to create my prototype bean.
I've annotated the method of the prototyped bean with the #Transactional annotation.
Is this the right way of doing it?
#Service
public class SomeService {
#Autowired
private Provider<SomePrototype> myPrototypeProvider;
public void execute() {
SomePrototype somePrototype = myPrototypeProvider.get();
somePrototype.someMethod();
}
}
#Component
#Scope(BeanDefinition.SCOPE_PROTOTYPE)
public class SomePrototype {
#Autowired
private SomeSpringBean someSpringBean;
#Autowired
private SomeRepository someRepository;
#Transactional(propagation = Propagation.REQUIRED)
public void someMethod() {
Result result = someSpringBean.doSomething();
someRepository.save(result);
}
}
The initialisation of transaction-scoped bean requires a proxy. Therefore, if we define a transactional bean as prototype, every that bean is requested, a new proxy is created, and that is not efficient.
What is reason behind this requirement (to have transactional prototype bean)
I have a instance of a class that is created outside of Spring that I'd like to have access to Spring beans so that it may fire an event and be observed by Spring beans. I'm not using Spring web, my application is running from the command-line via spring boot.
The only option you have is to expose the Spring context of your application using a static method so that the object that is not managed by Spring can use it to get references to managed beans it needs.
Start with a wrapper for the context. Create a regular managed bean which required reference to the context in its constructor. The reference is assigned to a static class field, which also has a static getter:
#Service
class ContextWrapper {
private static ApplicationContext context;
#Autowired
public ContextWrapper(ApplicationContext ac) {
context = ac;
}
public static ApplicationContext getContext() {
return context;
}
}
Use the static getter to get access to context in the object which is not managed by Spring and get reference to beans using methods available in the context:
SomeBean bean = ContextWrapper.getContext().getBean("someBean", SomeBean.class);
// do something with the bean
Last thing you need is communication channel from Spring beans to non-managed object. For instance, the SomeBean can expose a setter which will accept the non-managed object as a parameter and store the reference in a field for future use. The object mast get references to managed beans using the static context accessor mentioned above and use the setter to make the bean aware of its existence.
#Service
class SomeBean {
// ... your bean stuff
private SomeClass someclass;
public void setSomeClass(Someclass someclass) {
this.someclass = someclass;
}
private void sendEventToSomeClass() {
// communicate with the object not managed by Spring
if (someClass == null) return;
someClass.sendEvent();
}
}
You can inject by constructor that spring beans, something like:
#Service
class Bean {
...
}
class NotBean {
private Bean bean;
public NotBean(Bean bean) {
this.bean = bean;
}
// your stuff (handle events, etc...)
}
I have the following components, in two different files:
#Component
public class Chauffeur {
Car car;
public Chauffeur(Car car){
this.car = car;
}
public void go(){
System.out.println("Chauffeur");
car.drive();
}
}
#Component
public class Car{
public void drive() {
System.out.println("Drive car");
}
}
the following configuration file:
#Configuration
#ComponentScan
public class DriveableConfiguration {
}
and the following test:
#RunWith(SpringJUnit4ClassRunner.class)
#ContextConfiguration(classes=DriveableConfiguration.class)
public class DriveableTest {
#Autowired
Chauffeur chauffeur;
#Test
public void chauffeurTest(){
chauffeur.go();
}
}
All the classes above are in the same package and the test is passing.
In the test I annotated chauffer with #Autowired, which should mean that the Spring container looks after the creation of the instance of Chauffeur without the developer needing to explicitly instantiate it.
Now, the constructor for Chauffer needs an instance of Car, so there is no default constructor for that class. Nonetheless the container creates it, injecting the required instance in the constructor.
Is the #Autowired saying to the container to instantiate the element with whatever (Components, Beans) it can provide, included parameters in the constructor? If so, in what case is it needed to use #Autowired to annotate a constructor?
Only if you use Spring 4.3+. In such a case #Autowired on constructor is optional if you have one non default constructor.
You can check the example here.
So as of 4.3, you no longer need to specify an explicit injection annotation in such a single-constructor scenario. This is particularly elegant for classes which otherwise do not carry any container annotations at all, for example when programmatically registered
For versions lower than 4.3 you will an exception will be thrown:
the container will throw an exception looking for a default
constructor, unless you explicitly indicate autowire mode
‘constructor’ in your bean definition setup (e.g. in an XML )
I'm using Jersey 2 and Spring, and I'm trying to initialize my Jersey application (i.e. the class derived from ResourceConfig) with parameters from the Spring context.
Background: I have a single Jersey application that I build (i.e. a single WAR) and I deploy it across a server cluster with different Spring configurations on different servers to enable or disable different parts of the server, e.g. some of the servers have /search resources turned on, etc. This was really easy in Jersey 1.0: I just put,
<context:component-scan base-package="com.mycompany.resources.search"/>
in a Spring config to have Jersey scan that particular package and enable the JAX-RS resource providers in it.
Now in Jersey 2.0 the Spring <context:component-scan ... /> doesn't work, so resources have to be programmatically registered in a startup class derived from ResourceConfig:
public class MyApplication extends ResourceConfig {
public MyApplication() {
packages("com.mycompany.resources.search");
}
}
So far so good, but I need to conditionally scan that package, and I can't figure out how to get any Spring configuration into the MyApplication class. I thought that constructor injection might work:
public class MyApplication extends ResourceConfig {
#Autowired
public MyApplication(#Qualifier("my-config") MyConfiguration myConfiguration) {
if (myConfiguration.isEnabled()) {
packages("com.mycompany.resources.search");
}
}
}
However HK2 complains that it can't find a default constructor to use... so this indicates to me that DI is in play in the construction of this class, but that the DI isn't using Spring.
Similarly, using the the Spring bean lifecycle doesn't work:
public class MyApplication extends ResourceConfig implements InitializingBean {
#Autowired
private MyConfiguration myConfiguration;
public MyApplication() {
}
#Override
public void afterPropertiesSet() throws Exception {
if (myConfiguration.isEnabled()) {
packages("com.mycompany.resources.search");
}
}
}
(The afterPropertiesSet method isn't called.)
So now I'm stuck: is there any way to configure a Jersey ResourceConfig application object using Spring?
UPDATE:
I accepted #JohnR's answer below but I'll also include my eventual solution which I think is a bit cleaner. #JohnR's answer was to have the object initialized twice: first by Spring and then by Jersey/HK2. When Spring initializes the object you cache the dependencies in a static member, and then when Jersey/HK2 initializes it later you can retrieve the dependencies.
I ended up doing this:
public class MyApplication extends ResourceConfig {
public MyApplication() {
ApplicationContext rootCtx = ContextLoader.getCurrentWebApplicationContext();
MyConfiguration myConfiguration = rootCtx.getBean(MyConfiguration.class);
if (myConfiguration.isEnabled()) {
packages("com.mycompany.resources.whatever");
}
}
}
Rather than having the object initialized twice, we let Jersey/HK2 initialize it but then we retrieve the dependencies from Spring.
Both solutions are vulnerable to timing: they both assume that Spring is initialized before Jersey/HK2.
Expanding on my previous comment:
Trying to extend ResourceConfig is dangerous if you don't know what you're doing. Jersey becomes unpredictable, and if you try to subclass it into an Abstract class, Jersey crashes.
Instead, the JAX-RS specification provides us with a very useful interface called Feature: It allows you to register any classes you want as if you were configuring your own application. Furthermore, you don't need to use the awkward AbstractBinder, you just specify what contracts you register your classes with.
import org.springframework.context.ApplicationContext;
import org.springframework.web.context.support.WebApplicationContextUtils;
import javax.ws.rs.container.DynamicFeature;
import javax.ws.rs.container.ContainerRequestFilter;
import javax.ws.rs.container.ContainerResponseFilter;
import javax.ws.rs.core.Context;
import javax.ws.rs.core.Feature;
import javax.ws.rs.core.FeatureContext;
// Don't use #Component here, we need to inject the Spring context manually.
public class MySpringFeature implements Feature {
#Context
private ServletContext servletContext;
private ApplicationContext applicationContext;
#Autowired
private MySecurityDAO mySecurityDAO;
#Autowired
private MySpringResponseFilter myResponseFilter;
#Override
public boolean configure(FeatureContext context) {
if(this.servletContext == null) {
return false; // ERROR!
}
this.applicationContext = WebApplicationContextUtils.getWebApplicationContext(servletContext);
if(this.applicationContext == null) {
return false; // ERROR!
}
// This is where the magic happens!
AutowireCapableBeanFactory bf = applicationContext.getAutowireCapableBeanFactory();
bf.autowireBean(this);
// From here you can get all the beans you need
// Now we take a Spring bean instance,
// and register it with its appropriate JAX-RS contract
context.register(myResponseFilter, ContainerResponseFilter.class);
// Or, we could do this instead:
SomeSecurityFilter mySecurityFilter = new SomeSecurityFilter();
mySecurityFilter.setSecurityDAO(mySecurityDAO);
context.register(mySegurityFilter, ContainerRequestFilter.class);
// Or even this:
SomeOtherSpringBean someOtherBean = applicationContext.getBean(SomeOtherSpringBean.class);
context.register(someOtherBean, SomeOtherJerseyContract.class);
// Success!
return true;
}
}
And in your ResourceConfig:
public class MyApplication extends ResourceConfig() {
public MyApplication() {
register(MySpringFeature.class);
}
}
Ta-da!
So now I'm stuck: is there any way to configure a Jersey
ResourceConfig application object using Spring?
I don't think you can configure Jersey to obtain your ResourceConfig from Spring as a Spring managed bean. It's a bit hackish, but you could do something like this. Note that you'll end up with two instance of your ResourceConfig: one managed by Spring and another by Jersey:
public class MyApplication extends ResourceConfig {
// static, available to all instances
private static MyConfiguration myConfiguration;
public MyApplication() {
// when Spring creates the first instance of MyApplication, myConfiguration
// will be null because the setter wasn't called yet
if (myConfiguration != null)
{
// second instance created by jersey... Spring will have autowired
// the first instance, and myConfiguration is static
if (myConfiguration.isEnabled())
packages("com.mycompany.resources.search");
}
}
#Autowired
public void setMyConfiguration(MyConfiguration config)
{
// instance level setter saves to a static variable to make it available for
// future instances (i.e. the one created by jersey)
MyApplication.myConfiguration = config;
}
}
Again, this is fairly hackish. You'll want to make sure Spring is initialized before Jersey and look closely at any threading issues that could occur during initialization.