Database transactions and concurrency - POJOs in Action: Developing Enterprise Applications with Lightweight Frameworks - page 484

call if the exception was recoverable. The trouble with this approach is that it clut-
ters the client code with exception-handling logic. Every call that could poten-
tially throw a recoverable exception would be wrapped with the same code. A
much better approach is to centralize the retry logic in a Spring AOP interceptor
that retries transactions automatically. Let's take a look.

12.4.1 Using an AOP interceptor to retry transactions

A Spring application can use a custom AOP interceptor to catch the Concurrency-
FailureException and automatically retry the transaction. Here is a custom
Spring AOP interceptor that does this:

public class TransactionRetryInterceptor
implements MethodInterceptor {

protected int maxRetryCount = 3;

public void setMaxRetryCount(int maxRetryCount) {
this.maxRetryCount = maxRetryCount;
}

public Object invoke(MethodInvocation invocation)
throws Throwable {
int retryCount = 0;
while (true)
try {
ReflectiveMethodInvocation inv =
(ReflectiveMethodInvocation) invocation;
MethodInvocation anotherInvocation = inv.invocableClone();
return anotherInvocation.proceed();
} catch (ConcurrencyFailureException e) {
if (retryCount++ > maxRetryCount)
throw e;
else {
continue;
}
}
}

}
This class has a maxRetryCount property whose default value is 3 . The invoke()
method catches the ConcurrencyFailureException and loops at most maxRetry-
Count times, retrying the transaction. Because proceed() can only be called once,
invoke() clones the MethodInvocation before calling it. The TransactionRetry-
Interceptor is a good example of the power and flexibility of the Spring framework.