Proxies done right with Guava's AbstractInvocationHandler
 |
| Snarøya |
java.lang.reflect.Proxy. There is really no magic in this mechanism and it's worth knowing even you will never really use it - because Java proxies are ubiquitous in various frameworks and libraries.The idea is quite simple: dynamically create an object that implements one or more interfaces but every time any method of these interfaces is called our custom callback handler is invoked. This handler receives a handle to a method that was called (
java.lang.reflect.Method instance) and is free to behave in any way. Proxies are often used to implement seamless mocking, caching, transactions, security - i.e. they are a foundation for AOP.Before I explain what the purpose of
com.google.common.reflect.AbstractInvocationHandler from the title, let's start from a simple example. Say we want to transparently run methods of given interface asynchronously in a thread pool. Popular stacks like Spring (see: 27.4.3 The @Async Annotation) and Java EE (see: Asynchronous Method Invocation) already support this using the same technique.Imagine we have the following service:
public interface MailServer {
void send(String msg);
int unreadCount();
}
Our goal is to run send() asynchronously so that several subsequent invocations are not blocking but queue up and are executed in external thread pool concurrently rather than in calling thread. First we need factory code that will create a proxy instance:public class AsyncProxy {
public static <T> T wrap(T underlying, ExecutorService pool) {
final ClassLoader classLoader = underlying.getClass().getClassLoader();
final Class<T> intf = (Class<T>) underlying.getClass().getInterfaces()[0];
return (T)Proxy.newProxyInstance(
classLoader,
new Class<?>[] {intf},
new AsyncHandler<T>(underlying, pool));
}
}
Code above makes few bold assumptions, for example that an underlying object (real instance that we are proxying) implements exactly one interface. In real life a class can of course implement multiple interfaces, so can proxies - but we simplify this a bit for educational purposes. Now for starters we shall create no-op proxy that delegates to underlying object without any added value:class AsyncHandler<T> implements InvocationHandler {
private static final Logger log = LoggerFactory.getLogger(AsyncHandler.class);
private final T underlying;
private final ExecutorService pool;
AsyncHandler1(T underlying, ExecutorService pool) {
this.underlying = underlying;
this.pool = pool;
}
@Override
public Object invoke(Object proxy, final Method method, final Object[] args) throws Throwable {
return method.invoke(underlying, args);
}
}
ExecutorService pool will be used later. The last line is crucial - we invoke method on underlying instance with the same args. At this point we can:- invoke
underlyingor not (e.g. if given call is cached/memoized) - change arguments (i.e. for security purposes)
- run code before/after/around/on exception
- alter result by returning different value (it must match the type of
method.getReturnType()) - ...and much more
method.invoke() with Callable and run it asynchronously:class AsyncHandler<T> implements InvocationHandler {
private final T underlying;
private final ExecutorService pool;
AsyncHandler(T underlying, ExecutorService pool) {
this.underlying = underlying;
this.pool = pool;
}
@Override
public Object invoke(Object proxy, final Method method, final Object[] args) throws Throwable {
final Future<Object> future = pool.submit(new Callable<Object>() {
@Override
public Object call() throws Exception {
return method.invoke(underlying, args);
}
});
return handleResult(method, future);
}
private Object handleResult(Method method, Future<Object> future) throws Throwable {
if (method.getReturnType() == void.class)
return null;
try {
return future.get();
} catch (ExecutionException e) {
throw e.getCause();
}
}
}
Extra handleResult() method was extracted in order to properly handle non-void methods. Using such a proxy is straightforward:final MailServer mailServer = new RealMailServer();Now even if
final ExecutorService pool = Executors.newFixedThreadPool(10);
final MailServer asyncMailServer = AsyncProxy.wrap(mailServer, pool);
RealMailServer.send() takes a second to complete, invoking it twice via asyncMailServer.send() takes no time because both invocations run asynchronously in background.Broken equals(), hashCode() and toString()
Some developers are not aware of potential issues with default InvocationHandler implementation. Quoting the official documentation:An invocation of theIn our case case this means that for examplehashCode,equals, ortoStringmethods declared injava.lang.Objecton a proxy instance will be encoded and dispatched to the invocation handler'sinvokemethod in the same manner as interface method invocations are encoded and dispatched, as described above.
toString() is executed in the same thread pool as other methods of MailServer, quite surprising. Now imagine you have a local proxy where every method invocation triggers remote call. Dispatching equals(), hashCode() and toString() via network is definitely not what we want.Fixing with AbstractInvocationHandler
AbstractInvocationHandler from Guava is a simple abstract class that correctly deals with issues above. By default it dispatches equals(), hashCode() and toString() to Object class rather than passing it to invocation handler. Refactoring from straight InvocationHandler to AbstractInvocationHandler is dead simple:import com.google.common.reflect.AbstractInvocationHandler;That's it! I decided to override
class AsyncHandler<T> extends AbstractInvocationHandler {
//...
@Override
protected Object handleInvocation(Object proxy, final Method method, final Object[] args) throws Throwable {
//...
}
@Override
public String toString() {
return "Proxy of " + underlying;
}
}
toString() to help debugging. equals() and hashCode() are inherited from Object which is fine for the beginning. Now please look around your code base and search for custom proxies. If you were not using AbstractInvocationHandler or similar so far, chances are you introduces few subtle bugs.
Tags: design patterns, guava, multithreading