Spring源码分析:AOP源码解析(下篇)

AspectJAwareAdvisorAutoProxyCreator及为Bean生成代理时机分析

上篇文章说了,org.springframework.aop.aspectj.autoproxy.AspectJAwareAdvisorAutoProxyCreator这个类是Spring提供给开发者的AOP的核心类,就是AspectJAwareAdvisorAutoProxyCreator完成了【类/接口–>代理】的转换过程,首先我们看一下AspectJAwareAdvisorAutoProxyCreator的层次结构:

Spring源码分析:AOP源码解析(下篇)

这里最值得注意的一点是最左下角的那个方框,我用几句话总结一下:

  1. AspectJAwareAdvisorAutoProxyCreator是BeanPostProcessor接口的实现类
  2. postProcessBeforeInitialization方法与postProcessAfterInitialization方法实现在父类AbstractAutoProxyCreator中
  3. postProcessBeforeInitialization方法是一个空实现
  4. 逻辑代码在postProcessAfterInitialization方法中

基于以上的分析,将Bean生成代理的时机已经一目了然了:在每个Bean初始化之后,如果需要,调用AspectJAwareAdvisorAutoProxyCreator中的postProcessBeforeInitialization为Bean生成代理

代理对象实例化—-判断是否为<bean>生成代理

上文分析了Bean生成代理的时机是在每个Bean初始化之后,下面把代码定位到Bean初始化之后,先是AbstractAutowireCapableBeanFactory的initializeBean方法进行初始化:

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
protected Object initializeBean(final String beanName, final Object bean, RootBeanDefinition mbd) {
    if (System.getSecurityManager() != null) {
        AccessController.doPrivileged(new PrivilegedAction<Object>() {
            public Object run() {
                invokeAwareMethods(beanName, bean);
                return null;
            }
        }, getAccessControlContext());
    }
    else {
        invokeAwareMethods(beanName, bean);
    }
 
    Object wrappedBean = bean;
    if (mbd == null || !mbd.isSynthetic()) {
        wrappedBean = applyBeanPostProcessorsBeforeInitialization(wrappedBean, beanName);
    }
 
    try {
    invokeInitMethods(beanName, wrappedBean, mbd);
    }
    catch (Throwable ex) {
        throw new BeanCreationException(
                (mbd != null ? mbd.getResourceDescription() : null),
                beanName, "Invocation of init method failed", ex);
    }
 
    if (mbd == null || !mbd.isSynthetic()) {
        wrappedBean = applyBeanPostProcessorsAfterInitialization(wrappedBean, beanName);
    }
    return wrappedBean;
}

初始化之前是第16行的applyBeanPostProcessorsBeforeInitialization方法,初始化之后即29行的applyBeanPostProcessorsAfterInitialization方法:

1
2
3
4
5
6
7
8
9
10
11
12
public Object applyBeanPostProcessorsAfterInitialization(Object existingBean, String beanName)
        throws BeansException {
 
    Object result = existingBean;
    for (BeanPostProcessor beanProcessor : getBeanPostProcessors()) {
        result = beanProcessor.postProcessAfterInitialization(result, beanName);
        if (result == null) {
            return result;
        }
    }
    return result;
}

这里调用每个BeanPostProcessor的postProcessBeforeInitialization方法。按照之前的分析,看一下AbstractAutoProxyCreator的postProcessAfterInitialization方法实现:

1
2
3
4
5
6
7
8
9
public Object postProcessAfterInitialization(Object bean, String beanName) throws BeansException {
    if (bean != null) {
        Object cacheKey = getCacheKey(bean.getClass(), beanName);
        if (!this.earlyProxyReferences.contains(cacheKey)) {
            return wrapIfNecessary(bean, beanName, cacheKey);
        }
    }
    return bean;
}

跟一下第5行的方法wrapIfNecessary:

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
protected Object wrapIfNecessary(Object bean, String beanName, Object cacheKey) {
    if (this.targetSourcedBeans.contains(beanName)) {
        return bean;
    }
    if (this.nonAdvisedBeans.contains(cacheKey)) {
        return bean;
    }
    if (isInfrastructureClass(bean.getClass()) || shouldSkip(bean.getClass(), beanName)) {
        this.nonAdvisedBeans.add(cacheKey);
        return bean;
    }
 
    // Create proxy if we have advice.
    Object[] specificInterceptors = getAdvicesAndAdvisorsForBean(bean.getClass(), beanName, null);
    if (specificInterceptors != DO_NOT_PROXY) {
        this.advisedBeans.add(cacheKey);
        Object proxy = createProxy(bean.getClass(), beanName, specificInterceptors, new SingletonTargetSource(bean));
        this.proxyTypes.put(cacheKey, proxy.getClass());
        return proxy;
    }
 
    this.nonAdvisedBeans.add(cacheKey);
    return bean;
}

第2行~第11行是一些不需要生成代理的场景判断,这里略过。首先我们要思考的第一个问题是:哪些目标对象需要生成代理?因为配置文件里面有很多Bean,肯定不能对每个Bean都生成代理,因此需要一套规则判断Bean是不是需要生成代理,这套规则就是第14行的代码getAdvicesAndAdvisorsForBean:

1
2
3
4
5
6
7
8
9
protected List<Advisor> findEligibleAdvisors(Class beanClass, String beanName) {
    List<Advisor> candidateAdvisors = findCandidateAdvisors();
    List<Advisor> eligibleAdvisors = findAdvisorsThatCanApply(candidateAdvisors, beanClass, beanName);
    extendAdvisors(eligibleAdvisors);
    if (!eligibleAdvisors.isEmpty()) {
        eligibleAdvisors = sortAdvisors(eligibleAdvisors);
    }
    return eligibleAdvisors;
}

顾名思义,方法的意思是为指定class寻找合适的Advisor。

第2行代码,寻找候选Advisors,根据上文的配置文件,有两个候选Advisor,分别是<aop:aspect>节点下的<aop:before>和<aop:after>这两个,这两个在XML解析的时候已经被转换生成了RootBeanDefinition。

跳过第3行的代码,先看下第4行的代码extendAdvisors方法,之后再重点看一下第3行的代码。第4行的代码extendAdvisors方法作用是向候选Advisor链的开头(也就是List.get(0)的位置)添加一个org.springframework.aop.support.DefaultPointcutAdvisor

第3行代码,根据候选Advisors,寻找可以使用的Advisor,跟一下方法实现:

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
public static List<Advisor> findAdvisorsThatCanApply(List<Advisor> candidateAdvisors, Class<?> clazz) {
    if (candidateAdvisors.isEmpty()) {
        return candidateAdvisors;
    }
    List<Advisor> eligibleAdvisors = new LinkedList<Advisor>();
    for (Advisor candidate : candidateAdvisors) {
        if (candidate instanceof IntroductionAdvisor && canApply(candidate, clazz)) {
            eligibleAdvisors.add(candidate);
        }
    }
    boolean hasIntroductions = !eligibleAdvisors.isEmpty();
    for (Advisor candidate : candidateAdvisors) {
        if (candidate instanceof IntroductionAdvisor) {
            // already processed
            continue;
        }
        if (canApply(candidate, clazz, hasIntroductions)) {
            eligibleAdvisors.add(candidate);
        }
    }
    return eligibleAdvisors;
}

整个方法的主要判断都围绕canApply展开方法:

1
2
3
4
5
6
7
8
9
10
11
12
13
public static boolean canApply(Advisor advisor, Class<?> targetClass, boolean hasIntroductions) {
    if (advisor instanceof IntroductionAdvisor) {
        return ((IntroductionAdvisor) advisor).getClassFilter().matches(targetClass);
    }
    else if (advisor instanceof PointcutAdvisor) {
        PointcutAdvisor pca = (PointcutAdvisor) advisor;
        return canApply(pca.getPointcut(), targetClass, hasIntroductions);
    }
    else {
        // It doesn't have a pointcut so we assume it applies.
        return true;
    }
}

第一个参数advisor的实际类型是AspectJPointcutAdvisor,它是PointcutAdvisor的子类,因此执行第7行的方法:

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
public static boolean canApply(Pointcut pc, Class<?> targetClass, boolean hasIntroductions) {
    if (!pc.getClassFilter().matches(targetClass)) {
        return false;
    }
 
    MethodMatcher methodMatcher = pc.getMethodMatcher();
    IntroductionAwareMethodMatcher introductionAwareMethodMatcher = null;
    if (methodMatcher instanceof IntroductionAwareMethodMatcher) {
        introductionAwareMethodMatcher = (IntroductionAwareMethodMatcher) methodMatcher;
    }
 
    Set<Class> classes = new HashSet<Class>(ClassUtils.getAllInterfacesForClassAsSet(targetClass));
    classes.add(targetClass);
    for (Class<?> clazz : classes) {
        Method[] methods = clazz.getMethods();
        for (Method method : methods) {
            if ((introductionAwareMethodMatcher != null &&
                introductionAwareMethodMatcher.matches(method, targetClass, hasIntroductions)) ||
                    methodMatcher.matches(method, targetClass)) {
                return true;
            }
        }
    }
    return false;
}

这个方法其实就是拿当前Advisor对应的expression做了两层判断:

  1. 目标类必须满足expression的匹配规则
  2. 目标类中的方法必须满足expression的匹配规则,当然这里方法不是全部需要满足expression的匹配规则,有一个方法满足即可

如果以上两条都满足,那么容器则会判断该<bean>满足条件,需要被生成代理对象,具体方式为返回一个数组对象,该数组对象中存储的是<bean>对应的Advisor。

代理对象实例化—-为<bean>生成代理代码上下文梳理

上文分析了为<bean>生成代理的条件,现在就正式看一下Spring上下文是如何为<bean>生成代理的。回到AbstractAutoProxyCreator的wrapIfNecessary方法:

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
protected Object wrapIfNecessary(Object bean, String beanName, Object cacheKey) {
    if (this.targetSourcedBeans.contains(beanName)) {
        return bean;
    }
    if (this.nonAdvisedBeans.contains(cacheKey)) {
        return bean;
    }
    if (isInfrastructureClass(bean.getClass()) || shouldSkip(bean.getClass(), beanName)) {
        this.nonAdvisedBeans.add(cacheKey);
        return bean;
    }
 
    // Create proxy if we have advice.
    Object[] specificInterceptors = getAdvicesAndAdvisorsForBean(bean.getClass(), beanName, null);
    if (specificInterceptors != DO_NOT_PROXY) {
        this.advisedBeans.add(cacheKey);
        Object proxy = createProxy(bean.getClass(), beanName, specificInterceptors, new SingletonTargetSource(bean));
        this.proxyTypes.put(cacheKey, proxy.getClass());
        return proxy;
    }
 
    this.nonAdvisedBeans.add(cacheKey);
    return bean;
}

第14行拿到<bean>对应的Advisor数组,第15行判断只要Advisor数组不为空,那么就会通过第17行的代码为<bean>创建代理:

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
protected Object createProxy(
        Class<?> beanClass, String beanName, Object[] specificInterceptors, TargetSource targetSource) {
 
    ProxyFactory proxyFactory = new ProxyFactory();
    // Copy our properties (proxyTargetClass etc) inherited from ProxyConfig.
    proxyFactory.copyFrom(this);
 
    if (!shouldProxyTargetClass(beanClass, beanName)) {
        // Must allow for introductions; can't just set interfaces to
        // the target's interfaces only.
        Class<?>[] targetInterfaces = ClassUtils.getAllInterfacesForClass(beanClass, this.proxyClassLoader);
        for (Class<?> targetInterface : targetInterfaces) {
            proxyFactory.addInterface(targetInterface);
        }
    }
 
    Advisor[] advisors = buildAdvisors(beanName, specificInterceptors);
    for (Advisor advisor : advisors) {
        proxyFactory.addAdvisor(advisor);
    }
 
    proxyFactory.setTargetSource(targetSource);
    customizeProxyFactory(proxyFactory);
 
    proxyFactory.setFrozen(this.freezeProxy);
    if (advisorsPreFiltered()) {
        proxyFactory.setPreFiltered(true);
    }
 
    return proxyFactory.getProxy(this.proxyClassLoader);
}

第4行~第6行new出了一个ProxyFactory,Proxy,顾名思义,代理工厂的意思,提供了简单的方式使用代码获取和配置AOP代理。

第8行的代码做了一个判断,判断的内容是<aop:config>这个节点中proxy-target-class=”false”或者proxy-target-class不配置,即不使用CGLIB生成代理。如果满足条件,进判断,获取当前Bean实现的所有接口,讲这些接口Class对象都添加到ProxyFactory中。

第17行~第28行的代码没什么看的必要,向ProxyFactory中添加一些参数而已。重点看第30行proxyFactory.getProxy(this.proxyClassLoader)这句:

1
2
3
public Object getProxy(ClassLoader classLoader) {
return createAopProxy().getProxy(classLoader);
}

实现代码就一行,但是却明确告诉我们做了两件事情:

  1. 创建AopProxy接口实现类
  2. 通过AopProxy接口的实现类的getProxy方法获取<bean>对应的代理

就从这两个点出发,分两部分分析一下。

代理对象实例化—-创建AopProxy接口实现类

看一下createAopProxy()方法的实现,它位于DefaultAopProxyFactory类中:

1
2
3
4
5
6
protected final synchronized AopProxy createAopProxy() {
if (!this.active) {
activate();
}
return getAopProxyFactory().createAopProxy(this);
}

前面的部分没什么必要看,直接进入重点即createAopProxy方法:

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
public AopProxy createAopProxy(AdvisedSupport config) throws AopConfigException {
    if (config.isOptimize() || config.isProxyTargetClass() || hasNoUserSuppliedProxyInterfaces(config)) {
        Class targetClass = config.getTargetClass();
        if (targetClass == null) {
            throw new AopConfigException("TargetSource cannot determine target class: " +
                    "Either an interface or a target is required for proxy creation.");
        }
        if (targetClass.isInterface()) {
            return new JdkDynamicAopProxy(config);
        }
        if (!cglibAvailable) {
            throw new AopConfigException(
                    "Cannot proxy target class because CGLIB2 is not available. " +
                    "Add CGLIB to the class path or specify proxy interfaces.");
        }
        return CglibProxyFactory.createCglibProxy(config);
    }
    else {
        return new JdkDynamicAopProxy(config);
    }
}

平时我们说AOP原理三句话就能概括:

  1. 对类生成代理使用CGLIB
  2. 对接口生成代理使用JDK原生的Proxy
  3. 可以通过配置文件指定对接口使用CGLIB生成代理

这三句话的出处就是createAopProxy方法。看到默认是第19行的代码使用JDK自带的Proxy生成代理,碰到以下三种情况例外:

  1. ProxyConfig的isOptimize方法为true,这表示让Spring自己去优化而不是用户指定
  2. ProxyConfig的isProxyTargetClass方法为true,这表示配置了proxy-target-class=”true”
  3. ProxyConfig满足hasNoUserSuppliedProxyInterfaces方法执行结果为true,这表示<bean>对象没有实现任何接口或者实现的接口是SpringProxy接口

在进入第2行的if判断之后再根据目标<bean>的类型决定返回哪种AopProxy。简单总结起来就是:

  1. proxy-target-class没有配置或者proxy-target-class=”false”,返回JdkDynamicAopProxy
  2. proxy-target-class=”true”或者<bean>对象没有实现任何接口或者只实现了SpringProxy接口,返回Cglib2AopProxy

当然,不管是JdkDynamicAopProxy还是Cglib2AopProxy,AdvisedSupport都是作为构造函数参数传入的,里面存储了具体的Advisor。 

代理对象实例化—-通过getProxy方法获取<bean>对应的代理

其实代码已经分析到了JdkDynamicAopProxy和Cglib2AopProxy,剩下的就没什么好讲的了,无非就是看对这两种方式生成代理的熟悉程度而已。

Cglib2AopProxy生成代理的代码就不看了,对Cglib不熟悉的朋友可以看Cglib及其基本使用一文。

JdkDynamicAopProxy生成代理的方式稍微看一下:

1
2
3
4
5
6
7
8
public Object getProxy(ClassLoader classLoader) {
    if (logger.isDebugEnabled()) {
        logger.debug("Creating JDK dynamic proxy: target source is " + this.advised.getTargetSource());
    }
    Class[] proxiedInterfaces = AopProxyUtils.completeProxiedInterfaces(this.advised);
    findDefinedEqualsAndHashCodeMethods(proxiedInterfaces);
    return Proxy.newProxyInstance(classLoader, proxiedInterfaces, this);
}

这边解释一下第5行和第6行的代码,第5行代码的作用是拿到所有要代理的接口,第6行代码的作用是尝试寻找这些接口方法里面有没有equals方法和hashCode方法,同时都有的话打个标记,寻找结束,equals方法和hashCode方法有特殊处理。

最终通过第7行的Proxy.newProxyInstance方法获取接口/类对应的代理对象,Proxy是JDK原生支持的生成代理的方式。

代理方法调用原理

前面已经详细分析了为接口/类生成代理的原理,生成代理之后就要调用方法了,这里看一下使用JdkDynamicAopProxy调用方法的原理。

由于JdkDynamicAopProxy本身实现了InvocationHandler接口,因此具体代理前后处理的逻辑在invoke方法中:

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
public Object invoke(Object proxy, Method method, Object[] args) throws Throwable {
    MethodInvocation invocation;
    Object oldProxy = null;
    boolean setProxyContext = false;
 
    TargetSource targetSource = this.advised.targetSource;
    Class targetClass = null;
    Object target = null;
 
    try {
        if (!this.equalsDefined && AopUtils.isEqualsMethod(method)) {
            // The target does not implement the equals(Object) method itself.
            return equals(args[0]);
        }
        if (!this.hashCodeDefined && AopUtils.isHashCodeMethod(method)) {
            // The target does not implement the hashCode() method itself.
            return hashCode();
        }
        if (!this.advised.opaque && method.getDeclaringClass().isInterface() &&
                method.getDeclaringClass().isAssignableFrom(Advised.class)) {
            // Service invocations on ProxyConfig with the proxy config...
            return AopUtils.invokeJoinpointUsingReflection(this.advised, method, args);
        }
 
        Object retVal;
 
        if (this.advised.exposeProxy) {
            // Make invocation available if necessary.
            oldProxy = AopContext.setCurrentProxy(proxy);
            setProxyContext = true;
        }
 
        // May be null. Get as late as possible to minimize the time we "own" the target,
        // in case it comes from a pool.
        target = targetSource.getTarget();
        if (target != null) {
            targetClass = target.getClass();
        }
 
        // Get the interception chain for this method.
        List<Object> chain = this.advised.getInterceptorsAndDynamicInterceptionAdvice(method, targetClass);
 
        // Check whether we have any advice. If we don't, we can fallback on direct
        // reflective invocation of the target, and avoid creating a MethodInvocation.
        if (chain.isEmpty()) {
            // We can skip creating a MethodInvocation: just invoke the target directly
            // Note that the final invoker must be an InvokerInterceptor so we know it does
            // nothing but a reflective operation on the target, and no hot swapping or fancy proxying.
            retVal = AopUtils.invokeJoinpointUsingReflection(target, method, args);
        }
        else {
            // We need to create a method invocation...
            invocation = new ReflectiveMethodInvocation(proxy, target, method, args, targetClass, chain);
            // Proceed to the joinpoint through the interceptor chain.
            retVal = invocation.proceed();
        }
 
        // Massage return value if necessary.
        if (retVal != null && retVal == target && method.getReturnType().isInstance(proxy) &&
                !RawTargetAccess.class.isAssignableFrom(method.getDeclaringClass())) {
            // Special case: it returned "this" and the return type of the method
            // is type-compatible. Note that we can't help if the target sets
            // a reference to itself in another returned object.
            retVal = proxy;
        }
        return retVal;
    }
    finally {
        if (target != null && !targetSource.isStatic()) {
            // Must have come from TargetSource.
            targetSource.releaseTarget(target);
        }
        if (setProxyContext) {
            // Restore old proxy.
            AopContext.setCurrentProxy(oldProxy);
        }
    }
}

第11行~第18行的代码,表示equals方法与hashCode方法即使满足expression规则,也不会为之产生代理内容,调用的是JdkDynamicAopProxy的equals方法与hashCode方法。至于这两个方法是什么作用,可以自己查看一下源代码。

第19行~第23行的代码,表示方法所属的Class是一个接口并且方法所属的Class是AdvisedSupport的父类或者父接口,直接通过反射调用该方法。

第27行~第30行的代码,是用于判断是否将代理暴露出去的,由<aop:config>标签中的expose-proxy=”true/false”配置。

第41行的代码,获取AdvisedSupport中的所有拦截器和动态拦截器列表,用于拦截方法,具体到我们的实际代码,列表中有三个Object,分别是:

  • chain.get(0):ExposeInvocationInterceptor,这是一个默认的拦截器,对应的原Advisor为DefaultPointcutAdvisor
  • chain.get(1):MethodBeforeAdviceInterceptor,用于在实际方法调用之前的拦截,对应的原Advisor为AspectJMethodBeforeAdvice
  • chain.get(2):AspectJAfterAdvice,用于在实际方法调用之后的处理

第45行~第50行的代码,如果拦截器列表为空,很正常,因为某个类/接口下的某个方法可能不满足expression的匹配规则,因此此时通过反射直接调用该方法。

第51行~第56行的代码,如果拦截器列表不为空,按照注释的意思,需要一个ReflectiveMethodInvocation,并通过proceed方法对原方法进行拦截,proceed方法感兴趣的朋友可以去看一下,里面使用到了递归的思想对chain中的Object进行了层层的调用。