Spring AOP总览:灵魂画手图解Spring AOP实现原理
Spring AOP源码01:Jdk动态代理底层源码
Spring AOP源码02:ProxyFactory
Spring AOP源码03:JdkDynamicAopProxy
Spring AOP源码04:MethodInvocation 拦截器调用
Spring AOP源码05:DefaultAdvisorAutoProxyCreator
Spring期末考压轴题:当Spring AOP遇上循环依赖
git注释源码地址:https://github.com/chaitou/spring-framework-master.git
通过对JdkDynamicAopProxy
的invoke
方法的探究,发现invoke
的核心是:
advisor
,例如与当前方法所匹配的所有before、afterReturning等增强,再将这些增强组成拦截器链// 生产拦截器链
MethodInvocation invocation =
new ReflectiveMethodInvocation(proxy, target, method, args, targetClass, chain);
// 执行拦截器链
retVal = invocation.proceed();
before -> method -> after -> afterReturning
的顺序链式调用。但是在源代码中,看到的并不是for循环的调用,而是只有invoke(this)
这一行关键代码。而这一行代码是如何保证各个增强的有序调用的呢?我们将带着这个问题来探究一下拦截器的源码。典型的职责链模式public Object proceed() throws Throwable {
// We start with an index of -1 and increment early.
if (this.currentInterceptorIndex == this.interceptorsAndDynamicMethodMatchers.size() - 1) {
return invokeJoinpoint();
}
Object interceptorOrInterceptionAdvice =
this.interceptorsAndDynamicMethodMatchers.get(++this.currentInterceptorIndex);
if (interceptorOrInterceptionAdvice instanceof InterceptorAndDynamicMethodMatcher) {
// 省略动态匹配,不是本节重点 ...
}
else {
// It's an interceptor, so we just invoke it: The pointcut will have
// been evaluated statically before this object was constructed.
// 普通拦截器,直接触发
return ((MethodInterceptor) interceptorOrInterceptionAdvice).invoke(this);
}
}
Service、ServiceImpl
public interface UserService {
public void findUser();
}
@Component
public class UserServiceImpl implements UserService {
public void findUser() {
System.out.println("findUser");
}
}
2个带order的切面类
@Component
@Aspect
public class LogAspect1 implements Ordered {
@Pointcut("execution(* aop.UserServiceImpl.findUser())")
public void point() {
}
@Before("point()")
public void before(){
System.out.println("log before1...");
}
@AfterReturning("point()")
public void afterReturning() { System.out.println("log after returning1..."); }
@Override
public int getOrder() {
return 1;
}
}
@Component
@Aspect
public class LogAspect2 implements Ordered {
@Pointcut("execution(* aop.UserServiceImpl.findUser())")
public void point() {
}
@Before("point()")
public void before(){
System.out.println("log before2...");
}
@AfterReturning("point()")
public void afterReturning() { System.out.println("log after returning2..."); }
@Override
public int getOrder() {
return 2;
}
}
xml配置,自动扫描、开启AOP而已
<context:component-scan base-package="aop" />
<aop:aspectj-autoproxy/>
测试
@Test
public void test1() {
ApplicationContext applicationContext = new ClassPathXmlApplicationContext("aop.xml");
UserService userService = (UserService) applicationContext.getBean("userServiceImpl", UserService.class);
userService.findUser();
}
结果
log before1...
log before2...
findUser
log after returning2...
log after returning1...
看执行结果已经可以得出,拦截器执行过程跟栈差不多,先before的后after,大概就是下图所示
// 生产拦截器链
MethodInvocation invocation =
new ReflectiveMethodInvocation(proxy, target, method, args, targetClass, chain);
首先我们先看一下拦截器链到底长什么样,通过调试代码,查看invocation
变量
getOrder
获取优先级,然后排序afterReturing
居然排在了before
,这样怎么保证before
在after
之前调用呢?method
方法并不在列表中,又是何时调用method
呢?答案是,控制增强的调用顺序其实由每个拦截器负责,所以我们需要分析MethodBeforeAdviceInterceptor
和AfterReturningAdviceInterceptor
// 执行拦截器链
retVal = invocation.proceed();
public Object proceed() throws Throwable {
// 初始currentInterceptorIndex为-1,每调用一次proceed就把currentInterceptorIndex+1
if (this.currentInterceptorIndex == this.interceptorsAndDynamicMethodMatchers.size() - 1) {
// 当调用次数 = 拦截器个数时
// 1. 触发当前method方法
return invokeJoinpoint();
}
Object interceptorOrInterceptionAdvice =
this.interceptorsAndDynamicMethodMatchers.get(++this.currentInterceptorIndex);
if (interceptorOrInterceptionAdvice instanceof InterceptorAndDynamicMethodMatcher) {
// 省略动态匹配,不是本节重点 ...
}
else {
// It's an interceptor, so we just invoke it: The pointcut will have
// been evaluated statically before this object was constructed.
// 2. 普通拦截器,直接触发拦截器invoke方法
return ((MethodInterceptor) interceptorOrInterceptionAdvice).invoke(this);
}
}
当调用次数 = 拦截器个数
时,调用method
方法,可能读者会有疑惑?这样method
方法不就变成最后调用了吗?想解决这个疑问还是要看拦截器源码才能解决,因此继续往下看invoke
方法,所以重点还是在拦截器源码。这里涉及到3种拦截器MethodBeforeAdviceInterceptor
和AfterReturningAdviceInterceptor
我们将重点分析。ExposeInvocationInterceptor
是每个拦截器链都会加上的拦截器,用于曝光当前拦截器链到ThreadLocal的,好让同线程下可以共享拦截器链,不是本节重点就不赘述了MethodBeforeAdviceInterceptor
public class MethodBeforeAdviceInterceptor implements MethodInterceptor, BeforeAdvice, Serializable {
private final MethodBeforeAdvice advice;
@Override
public Object invoke(MethodInvocation mi) throws Throwable {
// 先调用before方法
this.advice.before(mi.getMethod(), mi.getArguments(), mi.getThis());
// 继续链式调用
return mi.proceed();
}
}
AfterReturningAdviceInterceptor
public class AfterReturningAdviceInterceptor implements MethodInterceptor, AfterAdvice, Serializable {
private final AfterReturningAdvice advice;
@Override
public Object invoke(MethodInvocation mi) throws Throwable {
// 先进行链式调用
Object retVal = mi.proceed();
// 再调用afterReturning方法
this.advice.afterReturning(retVal, mi.getMethod(), mi.getArguments(), mi.getThis());
return retVal;
}
}
看到这里不知道大家是不是有点眉目了,拦截器链执行的顺序正时在各个拦截器的invoke
方法中实现的。before
会先执行advice
增强方法再链式调用,这个比较好理解而after
则是先执行链式调用,再调用advice
增强方法,也就是一个递归的过程 ,这里灵魂画手出山,画一下调用顺序,相信有助于大家理解。(注:忽视掉ExposeInvocationInterceptor拦截器,只关心before和afterReturing)