我是AspectJ的新手,我试图弄清楚如何保持/跟踪多个异步方法调用的上下文。想象以下代码:
@TimerStart
public void doSomething() throws InterruptedException {
Thread.sleep(1000);
MyCallable callable = new MyCallable();
Future future = executorService.submit(callable );
}
private class MyCallable implements Callable {
@Override
public Object call() throws Exception {
someOtherMethod();
return null;
}
@TimerEnd
private void someOtherMethod() throws InterruptedException {
Thread.sleep(1000);
}
}
我想测量@TimerStart和@TimerEnd之间经过的时间。我现在正面临两个问题:
我如何在各个方面之间保持对象。一个方面中的字段似乎都是静态的,那么并发问题呢?
如何获得两条建议,一条在@TimerStart之前执行,另一条在@TimerEnd之后执行。
目前,我有一些类似的思路:
public aspect TimerAspect {
pointcut timerStart(Object object, TimerStart timed):
execution(@TimerStart * *(..)) && this(object) && @annotation(timed);
pointcut timerStop(Object object, TimerEnd timed):
cflow(execution(@TimerEnd * *(..)) && this(object) && @annotation(timed) && !within(FlowTimerAspect));
before(Object object, TimerStart timed): timerStart(object, timed) {
System.out.println("##### Flow timer START");
}
after(Object object, TimerEnd timed): timerStop(object, timed) {
System.out.println("##### Flow timer STOP");
}
但是,我现在唯一得到的是StackOverflowException(是的,我知道-这就是为什么我在这里询问)。
编辑:
我偶然发现
percflow
似乎只有在@TimerStart和@TimerEnd出现在同一线程中时才有效。建议高度赞赏!public aspect TimerAspect percflow(timerStart(Object, TimerStart)) {
private long context;
pointcut timerStart(Object object, TimerStart timed):
execution(@TimerStart * *(..)) && this(object) && @annotation(timed);
pointcut timerStop(Object object, TimerEnd timed):
execution(@TimerEnd * *(..)) && this(object) && @annotation(timed);
before(Object object, TimerStart timed): timerStart(object, timed) {
context = System.currentTimeMillis();
}
after(Object object, TimerEnd timed): timerStop(object, timed) {
long passed = System.currentTimeMillis() - context;
System.out.println("passed time: " + passed);
}
}
最佳答案
由于您打算在测量时切换线程,因此percflow
实例化方法将无济于事。您必须坚持默认的单例纵横比,并将感兴趣对象的计时值保留在WeakHashMap
中。这样,只要与该时间关联的对象/线程处于活动状态,就可以保留时间。
我们将需要另一个注释来标记将新对象(在此示例中为Callable
)与您的时间相关联的事件。我们称之为@TimerJoin
。 @TimerJoin
注释将类似于您现有的@TimerStart
和@TimerEnd
注释。您的测量方面将如下所示。
import java.util.Map;
import java.util.WeakHashMap;
public aspect TimerAspect {
private final Map<Object, Timer> objectTiming = new WeakHashMap<>();
private final ThreadLocal<Timer> currentThreadTimer = new ThreadLocal<>();
pointcut timerStart(Object object):
execution(@TimerStart * *(..)) && this(object);
pointcut timerStop(Object object):
execution(@TimerEnd * *(..)) && this(object);
pointcut timerJoin(Object object):
(execution(@TimerJoin * *(..)) || execution(@TimerJoin *.new(..)) )
&& this(object);
before(Object object): timerStart(object) {
Timer timer = new Timer();
timer.start();
objectTiming.put(object, timer);
currentThreadTimer.set(timer);
System.out.println("##### Flow timer START");
}
before(Object object): timerJoin(object) {
Timer timing = currentThreadTimer.get();
objectTiming.put(object, timing);
System.out.println("##### Flow timer JOIN");
}
after(Object object): timerStop(object) {
Timer timing = objectTiming.get(object);
timing.stop();
System.out.println("##### Flow timer STOP");
System.out.println("Elapsed: " + timing.getElapsed());
}
}
和简单的
Timer.java
类:public class Timer {
private long start;
private long stop;
public long getStart() {
return start;
}
public long getStop() {
return stop;
}
public void start() {
start = System.currentTimeMillis();
}
public void stop() {
stop = System.currentTimeMillis();
}
public long getElapsed() {
return stop-start;
}
}
修改可调用对象以将其标记为加入当前线程上的计时器:
private class MyCallable implements Callable {
@TimerJoin
public MyCallable() {
}
@Override
public Object call() throws Exception {
someOtherMethod();
return null;
}
@TimerEnd
private void someOtherMethod() throws InterruptedException {
Thread.sleep(1000);
}
}
您的其余代码将相同。
您可能会注意到,方面正在使用
ThreadLocal
作为当前计时器的存储方式,以便能够将其与新对象相关联。您可以为此选择另一种存储,但是出于示例的原因,我尝试使其保持简单。同样,再次为简单起见,我在方面没有对空值进行任何安全检查。您需要自己处理极端情况。