We Keep Coding

Spring Custom Scope Lifecycle Bean Termination

Question: How can I tell Spring that a set of beans with a custom scope should all be considered garbage, so that the next request on the same thread would not re-use their state?
What I've done: I've implemented a custom scope in Spring, to mimic the lifecycle of a request scope (HttpRequest) but for TcpRequests. It is very similar what is found here.
Many examples of custom scopes which I am finding are variants on prototype or singleton with no explicit termination of beans occurring, or, alternatively, they based around a thread local or ThreadScope but they do not describe telling Spring that the lifecycle has ended and that all beans should be destroyed.
Things I have tried (perhaps incorrectly):
Event + Listener to indicate the beginning and end of the scope (these occur when message is received and just before response is sent); in listener, the scope is explicitly cleared which clears the entire map used by the thread local implementation (scope.clear()). Clearing scope does result in the next call to context.getBean() returning a new instance when handled manually in tests, but my bean which is autowired in a singleton class does not get a new bean--it uses the same bean over and over.
Listener which implements: BeanFactoryPostProcessor, BeanPostProcessor, BeanFactoryAware, DisposableBean and attempt to call destroy() on all Disposable bean instances; something like this but for my custom scope only. This seems to fail in that nothing is explicitly ending the lifecycle of the beans, despite the fact that I'm calling customScope.clear() when I receive the scope ending event; ending the scope doesn't seem to translate to "end all beans associated with this scope".
I've read Spring documentation extensively and it seems to be clear that Spring doesn't manage the lifecycle of these custom beans in that it doesn't know when or how they should be destroyed, which means that it must be told when and how to destroy them; I've tried to read and understand the Session and Request scopes as provided by Spring so that I can mimic this but am missing something (again, these are not available to me since this is not a web-aware application and I'm not using HttpRequests and it is a non-trivial change in our application's structure)
Is anyone out there able to point me in the right direction?
I have the following code examples:
Xml Context Configuration:
<int-ip:tcp-connection-factory id="serverConnectionFactory" type="server" port="19000"
serializer="javaSerializer" deserializer="javaDeserializer"/>
<int-ip:tcp-inbound-gateway id="inGateway" connection-factory="serverConnectionFactory"
request-channel="incomingServerChannel" error-channel="errorChannel"/>
<int:channel id="incomingServerChannel" />
<int:chain input-channel="incomingServerChannel">
<int:service-activator ref="transactionController"/>
</int:chain>
TransactionController (handles request):
#Component("transactionController")
public class TransactionController {
#Autowired
private RequestWrapper requestWrapper;
#ServiceActivator
public String handle(final Message<?> requestMessage) {
// object is passed around through various phases of application
// object is changed, things are added, and finally, a response is generated based upon this data
tcpRequestCompletePublisher.publishEvent(requestWrapper, "Request lifecycle complete.");
return response;
}
}
TcpRequestScope (scope definition):
#Component
public class TcpRequestScope implements Scope {
private final ThreadLocal<ConcurrentHashMap<String, Object>> scopedObjects =
new InheritableThreadLocal<ConcurrentHashMap<String, Object>>({
#Override
protected ConcurrentHashMap<String, Object> initialValue(){
return new ConcurrentHashMap<>();
}
};
private final Map<String, Runnable> destructionCallbacks =
Collections.synchronizedMap(new HashMap<String, Runnable>());
#Override
public Object get(final String name, final ObjectFactory<?> objectFactory) {
final Map<String, Object> scope = this.scopedObjects.get();
Object object = scope.get(name);
if (object == null) {
object = objectFactory.getObject();
scope.put(name, object);
}
return object;
}
#Override
public Object remove(final String name) {
final Map<String, Object> scope = this.scopedObjects.get();
return scope.remove(name);
}
#Override
public void registerDestructionCallback(final String name, final Runnable callback) {
destructionCallbacks.put(name, callback);
}
#Override
public Object resolveContextualObject(final String key) {
return null;
}
#Override
public String getConversationId() {
return String.valueOf(Thread.currentThread().getId());
}
public void clear() {
final Map<String, Object> scope = this.scopedObjects.get();
scope.clear();
}
}
TcpRequestCompleteListener:
#Component
public class TcpRequestCompleteListener implements ApplicationListener<TcpRequestCompleteEvent> {
#Autowired
private TcpRequestScope tcpRequestScope;
#Override
public void onApplicationEvent(final TcpRequestCompleteEvent event) {
// do some processing
// clear all scope related data (so next thread gets clean slate)
tcpRequestScope.clear();
}
}
RequestWrapper (object we use throughout request lifecycle):
#Component
#Scope(scopeName = "tcpRequestScope", proxyMode =
ScopedProxyMode.TARGET_CLASS)
public class RequestWrapper implements Serializable, DisposableBean {
// we have many fields here which we add to and build up during processing of request
// actual request message contents will be placed into this class and used throughout processing
#Override
public void destroy() throws Exception {
System.out.print("Destroying RequestWrapper bean");
}
}

After many months and a few more attempts, I finally stumbled across some articles which pointed me in the right direction. Specifically, references in David Winterfeldt's blog post helped me understand the SimpleThreadScope which I had previously read, and was well aware of the fact that Spring makes no attempt to clear the scope after its lifecycle is complete, however, his article demonstrated the missing link for all previous implementations I had seen.
Specifically, the missing links were static references to ThreadScopeContextHolder in ThreadScope class in his implementation (in my proposed implementation above I called mine TcpRequestScope; the rest of this answer uses David Winterfeldt's terms since his reference documentation will prove most useful, and he wrote it).
Upon closer inspection of the Custom Thread Scope Module I noticed I was missing the ThreadScopeContextHolder, which contained a static reference to a ThreadLocal, which contains a ThreadScopeAttributes object which is what holds in-scope objects.
Some minor differences between David's implementation and my final one were, after Spring Integration sends its response, I use a ChannelInterceptor to clear the thread scope, since I'm using Spring Integration. In his examples, he extended threads which included a call to the context holder as part of a finally block.
How I'm clearing the scope attributes / beans:
public class ThreadScopeInterceptor extends ChannelInterceptorAdapter {
#Override
public void afterSendCompletion(final Message<?> message, final MessageChannel channel, final boolean sent,
#Nullable final Exception exception) {
// explicitly clear scope variables
ThreadScopeContextHolder.clearThreadScopeState();
}
Additionally, I added a method in the ThreadScopeContextHolder which clears the ThreadLocal:
public class ThreadScopeContextHolder {
// see: reference document for complete ThreadScopeContextHolder class
/**
* Clears all tcpRequest scoped beans which are stored on the current thread's ThreadLocal instance by calling
* {#link ThreadLocal#remove()}.
*/
public static void clearThreadScopeState() {
threadScopeAttributesHolder.remove();
}
}
While I'm not absolutely certain that there will not be memory leaks due to the ThreadLocal usage, I believe this will work as expected since I am calling ThreadLocal.remove(), which will remove the only reference to the ThreadScopeAttributes object, and therefore open it up to garbage collection.
Any improvements are welcomed, especially in terms of usage of ThreadLocal and how this might cause problems down the road.
Sources:
David Winterfeldt's Custom Thread Scope Module
Spring By Example Custom Thread Scope Module github (See David Winterfeldt's example above)
jyore's spring scopes (specifically, thread scope)
David Noel's (Devbury) Spring Boot Starter Thread Scope

Why and How to mock Model-Helperclasses?

I am reading a lot about unit testing, mocking, and all that stuff. I am currently also reading the book "Growing Object-Oriented Software Guided by Tests" by Steve Freeman and Nat Pryce.
I am starting to understand a lot of stuff but missing one crucial point, where I tried to find the answer anywhere online but am not satisfied yet.
In the following example I have an Online Shop, which receives messages from a third-party library, translates those, interpret them and eventually persist them into the database if needed. In a concrete case, I receive a message about a change of the address of a credit card of a user and want to store that information into a database.
The structure looks like this:
src/
domain/
MessageTranslator.java
ShopEventListener.java
ShopHandler.java
model/
CreditCard.java
CreditCardBase.java
CreditCardBuilder.java
User.java
UserBase.java
UserBuilder.java
test/
MessageTranslatorTest.java
ShopHandlerTest.java
MessageTranslatorTest
public class MessageTranslatorTest {
#Test
public void notifiesCCAddressChangedWhenChangeCCAddressMessageReceived() throws Exception {
ShopEventListener listenerMock = mock(ShopEventListener.class);
MessageTranslator messageTranslator = new MessageTranslator(listenerMock);
messageTranslator.processMessage("action=changeCCAddress; firstname=John; lastname=Doe; address=foobar3");
verify(listenerMock).ccAddressChanged("John", "Doe", "foobar3");
}
}
MessageTranslator (very simple for now)
public class MessageTranslator {
private final ShopEventListener listener;
public MessageTranslator(ShopEventListener userEventListener) {
listener = userEventListener;
}
public void processMessage(String message) throws Exception {
String[] attributes = message.split(";");
listener.ccAddressChanged(attributes[1].split("=")[1].trim(), attributes[2].split("=")[1].trim(), attributes[3].split("=")[1].trim());
}
}
ShopHandler
public class ShopHandler implements ShopEventListener {
#Override
public void ccAddressChanged(String firstname, String lastname, String newAddress) throws Exception {
// find a user (especially userid) in the Database for given firstname and lastname
UserBase userBase = new UserBase();
User user = userBase.find(aUser().withFirstname(firstname).withLastname(lastname).build());
if (user == null) {
throw new Exception();
}
// find the matching CreditCard for the userid in the database
Integer userid = user.getUserid();
CreditCardBase ccBase = new CreditCardBase();
CreditCard cc = ccBase.find(aCreditCard().withUserid(userid).build());
if (cc == null) {
throw new Exception();
}
// change address locally and then write it back to the database
cc.setAddress(newAddress);
cc.persist();
}
}
ShopHandlerTest
public class ShopHandlerTest {
#Test
public void changesCCAddressWhenChangeCCAddressEventReceived() throws Exception {
ShopHandler shop = new ShopHandler();
shop.ccAddressChanged("John", "Doe", "foobar3");
// TODO: How to test the changes in inner object?
}
}
This is where I always stumble.
Do I want to mock the helper classes UserBase and CreditCardBase to not perform any database queries but just return a prepared fake object?
Do I want to mock the persist-method to not write any real data to the database but maybe just test the parameters of the object to be persisted and have other (integration) tests test the database operations?
If 1. and 2. will be answered with yes, then what am I actually testing here? Is it worth unittesting this unit then?
Does the structure make sense this way?
If 1. and 2. will be answered with yes, then how do I mock the inner objects? I feel like dependency injection is the wront approach here, because first its no real dependency, but some helper classes, second (and more important imo) the ShopHandler class could be flooded with dependencies, as it might need alot of different helper classes and model classes to perform all the different actions. What if I just want to update the birthdate of a user based on an external message, do I still have to path all the dependencies like CreditCardBase and stuff?
Sorry for the long post, but it would be really awesome if you could push me in the right direction.
If you need more code for the above to understand, let me know.

Do I want to mock the helper classes UserBase and CreditCardBase to not perform any database queries but just return a prepared fake object?
Looks like your "helper classes" are actually repositories/DAOs. You normally want to test your business logic separately from DAOs, without the real database access. So yes, you should probably mock these DAOs and prepare the calls to them as they would work in reality. Prepared fake object is OK in most cases. You may also want to verify that your mocked DAO was actually called.
Do I want to mock the persist-method to not write any real data to the database but maybe just test the parameters of the object to be persisted and have other (integration) tests test the database operations?
I find it a bit strange that you seem to have the persist method in your business entity. Normally DAOs implement this type of methods.
Yes, if you test business logic you should mock the persist call to DAOs as well. If you don't do this, you'll be making tests of the business logic much heavier that they should be.
Yes, you should test your DAOs as well but separately from the business logic.
If 1. and 2. will be answered with yes, then what am I actually testing here? Is it worth unittesting this unit then?
You're testing you business logic. Just what is implemented in your ccAddressChanged method. Roughly:
if the user could not be found, an exception is thrown.
if user is found but users credit card could not be found, an exception is thrown.
if both could be found then credit card is persisted with an updated address.
Does the structure make sense this way?
It is not quite what I'm used to. You seem to have data access logic in entities, then you also have this "base" helper classess...
If 1. and 2. will be answered with yes, then how do I mock the inner objects?
With "inner objects" you probaby mean these helper classes. They are actually more that "helper classes", they are DAOs providing access to the database. You can pass or inject them from the outside. Basically this is dependency injection, your business logic depends on these DAO components. If you are able to pass them from the outside then in your test you can mock DAOs and pass mocks to your business service. With DI frameworks like Spring you'll have framework support for this.
Here'a a rough sketch of how a test for your ShopHandler class could look like with Spring and Mockito:
#RunWith(SpringJUnit4ClassRunner.class)
#ContextConfiguration(classes = {ShopHandler.class})
public class ShopHandlerTest {
#Autowired
private ShopHandler sut;
#MockBean
private UserRepository userRepository;
#MockBean
private CreditCardRepository creditCardRepository;
#Test(expected = UserNotFoundException.class)
public void throwsUserNotFoundExceptionIfUserIsUnknown() {
when(userRepository.findUserByFirstNameAndLastName("Scott", "Tiger").thenReturn(null);
sut.ccAddressChanged("Scott", "Tiger", "Some Address");
}
#Test
public void successFullyUpdatesCreditCardAddress() {
when(userRepository.findUserByFirstNameAndLastName("Scott", "Tiger").thenReturn(new User("userId", ...));
when(creditCardRepository.findByUserId("userId")).thenReturn(new CreditCard(...));
ArgumentCaptor<CreditCard> creditCardCaptor = ArgumentCaptor.forClass(CreditCard.class);
verify(creditCardRepository).save(creditCardCaptor.capture());
sut.ccAddressChanged("Scott", "Tiger", "Some Address");
asserthThat(creditCardCaptor.getValue().getAddress()).isEqualTo("Some Address");
}
}
I feel like dependency injection is the wront approach here,
Dependency injection is a very sensible approach here.
because first its no real dependency,
Well, of course these are real dependencies.
but some helper classes,
Where do you think it end being a "helper class" and starts being a "real dependency"? What you call "helper classes" pretty much resemble DAOs which absolutely are "real dependencies".
second (and more important imo) the ShopHandler class could be flooded with dependencies, as it might need alot of different helper classes and model classes to perform all the different actions.
If you need to perform all these actions and need all these dependencies to do this, then this is the reality. The question is, however - do you really have to implement all of these actions in just one business service? Can't you divide this into many business services? You'll get smaller more focused classes then, and they will only need a few dependencies.

Since you're creating UserBase and CreditCard instances using new keyword in the method ccAddressChanged() - you cannot mock them!
In order to be able to mock them use DI - Dependency Injection (also called IoC - Inversion Of Control) by injecting instances of these class to ccAddressChanged():
change the signature of the class from:
public void ccAddressChanged(String firstname, String lastname, String newAddress)
to:
public void ccAddressChanged(String firstname, String lastname, String newAddress, UserBase userBase, CreditCard creditCard)
This way, you'll be able to mock them (using Mockito or any other mocking framework) and sending the mocks to the method.
Example of how the test will look, using Mockito:
#Test
public void changesCCAddressWhenChangeCCAddressEventReceived() throws Exception {
ShopHandler shop = new ShopHandler();
// mock UserBase and its behavior
UserBase mockedUserBase = mock(UserBase.class)
when(mockedUserBase.find(any()).thenReturns(mock(User.class));
// mock CreditCard
CreditCard mockedCreditCard = mock(CreditCard.class);
shop.ccAddressChanged("John", "Doe", "foobar3");
}
I feel like dependency injection is the wrong approach here, because
first its no real dependency, but some helper classes, second (and
more important imo) the ShopHandler class could be flooded with
dependencies
DI is not wrong:
It seems that ShopHandler class does have a real dependency on UserBase and CreditCardBase
To avoid "flooded" scenario you can inject them into the constructor of ShopHandler and save them into private fields. This way it's done only once during initialization and does not burden the user as well as doesn't expose implementation details.
Further, assuming that you refactored your code and now you're assigning UserBase and CreditCardBase in the constructor. I would refactor the code from:
#Override
public void ccAddressChanged(String firstname, String lastname, String newAddress) throws Exception {
// find a user (especially userid) in the Database for given firstname and lastname
UserBase userBase = new UserBase();
User user = userBase.find(aUser().withFirstname(firstname).withLastname(lastname).build());
if (user == null) {
throw new Exception();
}
// find the matching CreditCard for the userid in the database
Integer userid = user.getUserid();
CreditCardBase ccBase = new CreditCardBase();
CreditCard cc = ccBase.find(aCreditCard().withUserid(userid).build());
if (cc == null) {
throw new Exception();
}
// change address locally and then write it back to the database
cc.setAddress(newAddress);
cc.persist();
}
to:
#Override
public void ccAddressChanged(String firstname, String lastname, String newAddress) throws Exception {
User user = getUserByName(firstname, lastname);
CreditCard creditCard = getCCByUser(user);
setAddress(creditCard, newAddress);
}
and now you don't have to unit-test this ccAddressChanged() anymore. What you should do is test, and each one of the three methods: getUserByName, getCCByUser and setAddress. And each one of them is easy to mock and test!

Here is how I would write integration tests for ShopHandler (as shown in the question, with no changes):
public class ShopHandlerTest {
#Tested(fullyUnitialized = true) AppDB appDB;
#Tested ShopHandler sut;
#Test(expected = UserNotFoundException.class)
public void throwsUserNotFoundExceptionIfUserIsUnknown() {
sut.ccAddressChanged("Unknown", "user", "...");
}
#Test
public void successFullyUpdatesCreditCardAddress() {
User user = new User("Scott", "Tiger");
appDB.persist(user);
CreditCard cc = new CreditCard(user, ...);
appDB.persist(cc);
String newAddress = "New address";
sut.ccAddressChanged(user.getFirstName(), user.getLastName(), newAddress);
appDB.refresh(cc);
assertEquals(newAddress, cc.getAddress());
}
}
Above, #Tested is a JMockit annotation with full DI support, and also JPA/EJB/etc. support. It can be used as a meta-annotation, so you could create a #SUT or #TestUtil annotation to simplify its use in tests.
The fully-reusable test utility class AppDB would be something like this:
public final class AppDB {
#PersistenceContext private EntityManager em;
#PostConstruct
private void startTransaction() { ... using em... }
#PreDestroy
private void endTransaction() { ... rollback using em... }
public void persist(Object entity) { em.persist(entity); }
public void refresh(Object entity) { em.refresh(entity); }
}
Notice how nice and simple those integration tests look. They basically only contain high-level code, essentially the same kind of code you would see in the production (SUT) code. No complicated mocking APIs to get you in trouble. They are also fast and stable.

Using the provider from two different scopes

I have the following problem with Guice: a singleton service, is injected with provider of context-sensitive information. Until now, context was related only to servlet requests, so I used a #RequestScoped provider, and I was injecting this provider in service like so:
#RequestScoped
public class ContextProvider<IContext> implements Provider<IContext> {
#Override
public IContext get() { ... } // returns context
}
#Singleton
public class ServiceImpl implements IService {
#Inject
private Provider<IContext> contextProvider;
}
That works fine. Now, I'm working on adding background task processing to the application. Background tasks are not initiated from web-requests, so I can not use ServletScopes.scopeRequest(..). I have written a custom scope (almost exact copy of BatchScoped from Giuce doc) to make each Task run in it's own scope. Now the question is - how to make BatchScoped ContextProvider and configure Guice to use it?
I've made this attempt with binding EDSL:
line 1 : bind(IContext.class).toProvider(ContextProvider.class).in(RequestScoped.class);
line 2 : bind(IContext.class).toProvider(BatchContextProvider.class).in(BatchScoped.class);
but Guice tells me at line 2 that 'A binding to IContext was already configured at line 1'.
The question is: what's the right way of doing such injection with Guice?

A similar question: Getting multiple guice singletons of the same type
In general the problem here is that you want to bind the same class to two different providers (and scopes, but that's actually beside the point). That is only possible if you use unique binding annotations for each one, like so:
bind(IContext.class)
.annotatedWith(MyAnnotation1.class)
.toProvider(ContextProvider.class)
.in(RequestScoped.class);
bind(IContext.class)
.annotatedWith(MyAnnotation2.class)
.toProvider(BatchContextProvider.class)
.in(BatchScoped.class);
And change injection sites to include relevant annotation:
#Inject
#MyAnnotationX
private Provider<IContext> contextProvider;

You need a fake request that starts with your background task and remains for all of it. That is what ServletScopes.scopeRequest does.
public class MyBackgroundTask extends Thread {
#Override
public void run() {
RequestScoper scope = ServletScopes.scopeRequest(Collections.emptyMap());
try ( RequestScoper.CloseableScope ignored = scope.open() ) {
doTask();
}
}
private void doTask() {
}
}
Oh, don't forget to use providers so you delay the retrieval of your dependencies. For example, expading the previous example so the background task uses your IContext.
public class MyBackgroundTask extends Thread {
private Provider<IContext> contextProvider;
#Inject
public MyBackgroundTask(Provider<IContext> contextProvider) {
this.contextProvider = contextProvider;
}
#Override
public void run() {
RequestScoper scope = ServletScopes.scopeRequest(Collections.emptyMap());
try ( RequestScoper.CloseableScope ignored = scope.open() ) {
doTask();
}
}
private void doTask() {
}
}
If you don't use providers the injection, in this example, will be done from the thread that creates the background task which could be inside another scope.
BONUS: You may have noticed the empty map sent as a parameter to the scopeRequest method. Check the Guice javadocs. Those are the instances that you want already present in your fake request scope. Depending on your IContext you may need it.

Implementing GWT RequestFactory service for non-entity requests

I have the following Java servlet that performs what I call the "Addition Service":
public class AdditionService extends HttpServlet {
#Override
public void doGet(HttpServletRequest request, HttpServletResponse response) {
// The request will have 2 Integers inside its body that need to be
// added together and returned in the response.
Integer addend = extractAddendFromRequest(request);
Integer augend = extractAugendFromRequest(request);
Integer sum = addend + augend;
PrintWriter writer = response.getWriter();
writer.write(sum);
}
}
I am trying to get GWT's RequestFactory to do the same thing (adding two numbers on the app server and returning the sum as a response) using a ValueProxy and AdditionService, and am running into a few issues.
Here's the AdditionRequest (client tier) which is a value object holding two Integers to be added:
// Please note the "tier" (client, shared, server) I have placed all of my Java classes in
// as you read through the code.
public class com.myapp.client.AdditionRequest {
private Integer addend;
private Integer augend;
public AdditionRequest() {
super();
this.addend = 0;
this.augend = 0;
}
// Getters & setters for addend/augend.
}
Next my proxy (client tier):
#ProxyFor(value=AdditionRequest.class)
public interface com.myapp.client.AdditionRequestProxy extends ValueProxy {
public Integer getAddend();
public Integer getAugend();
public void setAddend(Integer a);
public void setAugend(Integer a);
}
Next my service API (in the shared tier):
#Service(value=DefaultAdditionService.class)
public interface com.myapp.shared.AdditionService extends RequestContext {
Request<Integer> sum(AdditionRequest request);
}
Next my request factory (shared tier):
public class com.myapp.shared.ServiceProvider implements RequestFactory {
public AdditionService getAdditionService() {
return new DefaultAdditionService();
}
// ... but since I'm implementing RequestFactory, there's about a dozen
// other methods GWT is forcing me to implement: find, getEventBus, fire, etc.
// Do I really need to implement all these?
}
Finally where the magic happens (server tier):
public class com.myapp.server.DefaultAdditionService implements AdditionService {
#Override
public Request<Integer> sum(AdditionRequest request) {
Integer sum = request.getAddend() + request.getAugend();
return sum;
}
// And because AdditionService extends RequestContext there's another bunch of
// methods GWT is forcing me to implement here: append, create, isChanged, etc.
// Do I really need to implement all these?
}
Here are my questions:
Is my "tier" strategy correct? Have I packaged all the types in the correct client/shared/server packages?
I don't think my setup is correct because AdditionService (in shared) references DefaultAdditionService, which is on the server, which it shouldn't be doing. Shared types should be able to live both on the client and the server, but not have dependencies on either...
Should ServiceProvider be a class that implements RequestFactory, or should it be an interface that extends it? If the latter, where do I define the ServiceProvider impl, and how do I link it back to all these other classes?
What about all these methods in ServiceProvider and DefaultAdditionService? Do I need to implement all 20+ of these core GWT methods? Or am I using the API incorrectly or not as simply as I could be using it?
Where does service locator factor in here? How?

If you want to use RF as a simple RPC mechanism [*] you can (and you are right: only ValueProxys), but you need something more: ServiceLocators (i.e., GWT 2.1.1).
With ServiceLocator you can simply put your service implementation (like your servlet) into a real service instance, instead into an entity object (as you will use only ValueProxys, with no static getXyz() methods) as required by the RF protocol. Note the existence also of Locators, used to externalize all those methods from your server-side entities: not needed if you use ValueProxy everywhere.
A ServiceLocator looks something like (taken from official docs):
public class DefaultAdditionServiceLocator implements ServiceLocator {
#Override
public Object getInstance(Class<?> clazz) {
try {
return clazz.newInstance();
} catch (InstantiationException e) {
throw new RuntimeException(e);
} catch (IllegalAccessException e) {
throw new RuntimeException(e);
}
}
}
You need to annotate your DefaultAdditionService also with a locator param, so RF knows on what to rely when it comes to dispatch your request to your service. Something like:
#Service(value = DefaultAdditionService.class, locator = DefaultAdditionServiceLocator.class)
public interface com.myapp.shared.AdditionService extends RequestContext {
// Note here, you need to use the proxy type of your AdditionRequest.
Request<Integer> sum(AdditionRequestProxy request);
}
Your service will then be the simplest possible thing on Earth (no need to extend/implement anything RF-related):
public class com.myapp.server.DefaultAdditionService {
// The server-side AdditionRequest type.
public Integer sum(AdditionRequest request) {
Integer sum = request.getAddend() + request.getAugend();
return sum;
}
}
If you mispell sum() or you do not implement a method declared in your RequestContext you will get an error.
To instantiate RequestContexts you need to extend the RequestFactory interface, with a public factory method for com.myapp.shared.AdditionService. Something like:
public interface AdditionServiceRequestFactory extends RequestFactory {
public com.myapp.shared.AdditionService createAdditionServiceRequestContext();
}
All your client calls will start from this. See the docs, if not already.
Now, RF works by totally separating the objects your want to pass from client (using EntityProxy and ValueProxy) and server (the real objects, either Entity values or simple DTO classes). You will use proxy types (i.e., interfaces whom implementations are automatically generated) everywhere in client/shared tier, and you use the relative domain object (the one referenced with #ProxyFor) only on server side. RF will take care of the rest. So your AdditionRequest will be on your server side, while AdditionRequestProxy will be on your client side (see the note in the RequestContext). Also note that, if you simply use primitive/boxed types as your RequestContext params or return types, you will not even need to create ValueProxys at all, as they are default transportable.
The last bit you need, is to wire the RequestFactoryServlet on your web.xml. See the docs here. Note that you can extend it if you want to, say, play around with custom ExceptionHandlers or ServiceLayerDecorators, but you don't need to.
Speaking about where to put everything:
Locators, ServiceLocators, service instances, domain objects, and RequestFactoryServlet extensions, will be on your server-side;
The RequestContext, RequestFactory extensions and all your proxy types will be on the shared-side;
client side will initialize the RequestFactory extension and use it to obtain the factory instance for your service requests.
All in all... to create a simple RPC mechanism with RF, just:
create your service along with ServiceLocator;
create a RequestContext for your requests (annotated with service and locator values);
create a RequestFactory extension to return your RequestContext;
if you want to use more than primitive types in your RequestContext (like simple DTOs), just create client proxy interfaces for them, annotated with #ProxyFor, and remember where to use each type;
wire everything.
Much like that. Ok, I wrote too much and probably forgot something :)
For reference, see:
Official RF documentation;
Thomas Broyer's articles [1], [2];
RF vs GWT-RPC from the RF author point of view.
[*]: In this approach you shift your logic from data-oriented to service-oriented app. You give up using Entitys, IDs, versions and, of course, all the complex diff logic between client and server, when it comes to CRUD operations.

spring two-way rmi callback from server executing on client side

On the server-side I have a ListenerManager which fires callbacks to its Listeners. The manager is exported using a Spring RmiServiceExporter
On the client-side I have a proxy to the manager created by an RmiProxyFactoryBean, and a Listener implementation registered through this proxy with the manager on the server side.
So far so good: the ListenerManager is given a Listener and it invokes its callbacks, however since the listener is just a deserialized copy of the client-side object, the callback runs on the server side, not the client side.
How can I get Spring to generate a proxy on the server-side to the client-side listener so that the callback invoked by the server is executed remotely on the client-side? Surely I don't need another (exporter, proxy factory) pair in the opposite direction?

A pure RMI solution: the client-side listener object needs to implement java.rmi.server.UnicastRemoteObject. If it does, and each of its methods throw RemoteException then when it is passed to the server through the manager proxy everything is wired up automatically, and method invocations on the server-side proxy to this listener are remote invocations of methods on the real client-side object.
This will do, but it's even better to be able to wrap the object for export without requiring a particular superclass. We can use a CGLIB Enhancer to "proxy" the listener as a subclass of UnicastRemoteObject that also implements the service interfaces. This still requires that the target object implement java.rmi.Remote and declare throws RemoteException.
Next step is a solution that can export arbitrary objects for remote invocation of their methods, without requiring that they implement Remote or declare throws RemoteException. We must integrate this proxying with the existing Spring infrastructure, which we can do with a new implementation of RmiBasedExporter modelled on the non-registry bits of RmiServiceExporter#prepare() to export the RMI stub of our proxy and on the invocation part of RmiClientInterceptor.doInvoke(MethodInvocation, RmiInvocationHandler). We need to be able to get hold of an exported proxy instance of our service interfaces. We can model this on the means used by Spring to apparently "export" non-RMI interfaces. Spring proxies the interface to generate a RmiInvocationWrapper for invocation of a non-RMI method, serialises the method details and arguments, then invokes this on the far side of the RMI connection.
Use a ProxyFactory and an RmiInvocationHandler implementation to proxy the target object.
Use a new implementation of RmiBasedExporter to getObjectToExport(), and export it using UnicastRemoteObject#export(obj, 0).
For the invocation handler, rmiInvocationHandler.invoke(invocationFactory.createRemoteInvocation(invocation)), with a DefaultRemoteInvocationFactory.
Handle exceptions and wrap appropriately to avoid seeing UndeclaredThrowableExceptions.
So, we can use RMI to export arbitrary objects. This means we can use one of these objects on the client-side as a parameter to an RMI method call on an RMI server-side object, and when the deserialised stub on the server-side has methods invoked, those methods will execute on the client-side. Magic.

Following Joe Kearney's explaination, I have created my RMIUtil.java. Hope there is nothing left.
BTW, please ref this
for "java.rmi.NoSuchObjectException: no such object in table"

Just add some code to Joe's answer.
Extends RmiServiceExporter and get access to exported object:
public class RmiServiceExporter extends org.springframework.remoting.rmi.RmiServiceExporter {
private Object remoteService;
private String remoteServiceName;
#Override
public Remote getObjectToExport() {
Remote exportedObject = super.getObjectToExport();
if (getService() instanceof Remote && (
getServiceInterface() == null || exportedObject.getClass().isAssignableFrom(getServiceInterface()))) {
this.remoteService = exportedObject;
}
else {
// RMI Invokers.
ProxyFactory factory = new ProxyFactory(getServiceInterface(),
new RmiServiceInterceptor((RmiInvocationHandler) exportedObject, remoteServiceName));
this.remoteService = factory.getProxy();
}
return exportedObject;
}
public Object getRemoteService() {
return remoteService;
}
/**
* Override to get access to the serviceName
*/
#Override
public void setServiceName(String serviceName) {
this.remoteServiceName = serviceName;
super.setServiceName(serviceName);
}
}
The interceptor used in the proxy (the remote service callback):
public class RmiServiceInterceptor extends RemoteInvocationBasedAccessor
implements MethodInterceptor, Serializable {
private RmiInvocationHandler invocationHandler;
private String serviceName;
public RmiServiceInterceptor(RmiInvocationHandler invocationHandler) {
this(invocationHandler, null);
}
public RmiServiceInterceptor(RmiInvocationHandler invocationHandler, String serviceName) {
this.invocationHandler = invocationHandler;
this.serviceName = serviceName;
}
/**
* {#inheritDoc}
*/
public Object invoke(MethodInvocation invocation) throws Throwable {
try {
return invocationHandler.invoke(createRemoteInvocation(invocation));
}
catch (RemoteException ex) {
throw RmiClientInterceptorUtils.convertRmiAccessException(
invocation.getMethod(), ex, RmiClientInterceptorUtils.isConnectFailure(ex),
extractServiceUrl());
}
}
/**
* Try to extract service Url from invationHandler.toString() for exception info
* #return Service Url
*/
private String extractServiceUrl() {
String toParse = invocationHandler.toString();
String url = "rmi://" + StringUtils.substringBefore(
StringUtils.substringAfter(toParse, "endpoint:["), "]");
if (serviceName != null)
url = StringUtils.substringBefore(url, ":") + "/" + serviceName;
return url;
}
}
When exporting the service with this RmiServiceExporter, we cand send a rmi callback with:
someRemoteService.someRemoteMethod(rmiServiceExporter.getRemoteService());