源码框架(九)Spring源码分析之AOP从解析到调用

在上一篇,我们对IOC核心部分流程已经分析完毕,相信小伙伴们有所收获,从这一篇开始,我们将会踏上新的旅程,即Spring的另一核心:AOP!

首先,为了让大家能更有效的理解AOP,先带大家过一下AOP中的术语:

  • 切面(Aspect):指关注点模块化,这个关注点可能会横切多个对象。事务管理是企业级Java应用中有关横切关注点的例子。在Spring AOP中,切面可以使用在普通类中以@Aspect注解来实现。
  • 连接点(Join point):在Spring AOP中,一个连接点总是代表一个方法的执行,其实就代表增强的方法。
  • 通知(Advice):在切面的某个特定的连接点上执行的动作。通知有多种类型,包括aroundbeforeafter等等。许多AOP框架,包括Spring在内,都是以拦截器做通知模型的,并维护着一个以连接点为中心的拦截器链。
  • 目标对象(Target):目标对象指将要被增强的对象。即包含主业务逻辑的类的对象。
  • 切点(Pointcut):匹配连接点的断言。通知和切点表达式相关联,并在满足这个切点的连接点上运行(例如,当执行某个特定名称的方法时)。切点表达式如何和连接点匹配是AOP的核心:Spring默认使用AspectJ切点语义。
  • 顾问(Advisor): 顾问是Advice的一种包装体现,Advisor是Pointcut以及Advice的一个结合,用来管理Advice和Pointcut。
  • 织入(Weaving):将通知切入连接点的过程叫织入
  • 引入(Introductions):可以将其他接口和实现动态引入到targetClass中

一个栗子

术语看完了,我们先上个Demo回顾一下吧~

  1. 首先,使用EnableAspectJAutoProxy注解开启我们的AOP

    @ComponentScan(basePackages = {"com.my.spring.test.aop"})
    @Configuration
    @EnableAspectJAutoProxy
    public class Main {
    
    	public static void main(String[] args) {
    		AnnotationConfigApplicationContext context = new AnnotationConfigApplicationContext(Main.class);
    		IService service = context.getBean("service", IService.class);
    		service.doService();
    	}
    }
    
  2. 写一个接口

    public interface IService {
    
    	void doService();
    }
    
  3. 写一个实现类

    @Service("service")
    public class ServiceImpl implements IService{
    
    	@Override
    	public void doService() {
    		System.out.println("do service ...");
    	}
    }
    
  4. 写一个切面

    @Aspect
    @Component
    public class ServiceAspect {
    
    	@Pointcut(value = "execution(* com.my.spring.test.aop.*.*(..))")
    	public void pointCut() {
    	}
    
    	@Before(value = "pointCut()")
    	public void methodBefore(JoinPoint joinPoint) {
    		String methodName = joinPoint.getSignature().getName();
    		System.out.println("执行目标方法 【" + methodName + "】 的【前置通知】,入参:" + Arrays.toString(joinPoint.getArgs()));
    	}
    
    	@After(value = "pointCut()")
    	public void methodAfter(JoinPoint joinPoint) {
    		String methodName = joinPoint.getSignature().getName();
    		System.out.println("执行目标方法 【" + methodName + "】 的【后置通知】,入参:" + Arrays.toString(joinPoint.getArgs()));
    	}
    
    	@AfterReturning(value = "pointCut()")
    	public void methodReturn(JoinPoint joinPoint) {
    		String methodName = joinPoint.getSignature().getName();
    		System.out.println("执行目标方法 【" + methodName + "】 的【返回通知】,入参:" + Arrays.toString(joinPoint.getArgs()));
    	}
    
    	@AfterThrowing(value = "pointCut()")
    	public void methodThrow(JoinPoint joinPoint) {
    		String methodName = joinPoint.getSignature().getName();
    		System.out.println("执行目标方法 【" + methodName + "】 的【异常通知】,入参:" + Arrays.toString(joinPoint.getArgs()));
    	}
    }
    
  5. 测试运行

    执行目标方法 【doService】 的【前置通知】,入参:[]
    do service ...
    执行目标方法 【doService】 的【返回通知】,入参:[]
    执行目标方法 【doService】 的【后置通知】,入参:[]
    

以上

Demo看完了,运行效果也出来了,AOP已生效,但如何生效的呢?相比于我们普通使用Bean的Demo,在这里,我们只不过加上了一个@EnableAspectJAutoProxy注解以及一个标识了@Aspectj的类,那么我们先看看@EnableAspectJAutoProxy这个注解做了什么吧~

开启AOP

以下是笔者所画的大致流程图

EnabelAspectJ.png

​ 其中AspectJAutoProxyRegistrar实现了ImportBeanDefinitionRegistrar,所以在处理BeanFactoryPostProcessor逻辑时将会调用registerBeanDefinitions方法,此时就会把AnnotationAwareAspectJAutoProxyCreator注册到容器中,其中BeanFactoryPostProcessor的逻辑就不再说了,往期文章有过详细分析。而AnnotationAwareAspectJAutoProxyCreator的类图如下:

AnnotationAwareAspectJ.png

我们发现AnnotationAwareAspectJAutoProxyCreator是实现了BeanPostProcessor接口的类,所以它其实是一个后置处理器,那么,还记得在创建Bean过程中的BeanPostProcessor九次调用时机吗?不记得也没关系,AnnotationAwareAspectJAutoProxyCreator起作用的地方是在bean的实例化前以及初始化后,分别对应着解析切面和创建动态代理的过程,现在,就让我们先来看看解析切面的过程吧~

解析切面

解析切面的流程如下图所示:

切面解析过程.png

我们已经了解到切面解析的过程是由AnnotationAwareAspectJAutoProxyCreator完成的,而AnnotationAwareAspectJAutoProxyCreator又继承了AbstractAutoProxyCreator,所以首先,我们先会来到AbstractAutoProxyCreator#postProcessBeforeInstantiation

public Object postProcessBeforeInstantiation(Class<?> beanClass, String beanName) {
	// class类型是否为(Advice, Pointcut, Advisor, AopInfrastructureBean)
  // shouldSkip中将会解析切面
  if (isInfrastructureClass(beanClass) || shouldSkip(beanClass, beanName)) {
    this.advisedBeans.put(cacheKey, Boolean.FALSE);
    return null;
  }
}

调用到子类的AspectJAwareAdvisorAutoProxyCreator#shouldSkip

@Override
protected boolean shouldSkip(Class<?> beanClass, String beanName) {
  // 寻找advisor
  List<Advisor> candidateAdvisors = findCandidateAdvisors();
  for (Advisor advisor : candidateAdvisors) {
    if (advisor instanceof AspectJPointcutAdvisor &&
        ((AspectJPointcutAdvisor) advisor).getAspectName().equals(beanName)) {
      return true;
    }
  }
  return super.shouldSkip(beanClass, beanName);
}

findCandidateAdvisors

protected List<Advisor> findCandidateAdvisors() {
  // 寻找实现了Advisor接口的类, 由于我们一般不会以接口的方式实现切面,这里返回null
  List<Advisor> advisors = super.findCandidateAdvisors();
  if (this.aspectJAdvisorsBuilder != null) {
    // 这里将解析出所有的切面
    advisors.addAll(this.aspectJAdvisorsBuilder.buildAspectJAdvisors());
  }
  return advisors;
}

buildAspectJAdvisors

public List<Advisor> buildAspectJAdvisors() {
  // aspectBeanNames有值则说明切面已解析完毕
  List<String> aspectNames = this.aspectBeanNames;
  // Double Check
  if (aspectNames == null) {
    synchronized (this) {
      aspectNames = this.aspectBeanNames;
      if (aspectNames == null) {
        List<Advisor> advisors = new ArrayList<>();
        aspectNames = new ArrayList<>();
        // 取出是Object子类的bean,其实就是所有的bean
        String[] beanNames = BeanFactoryUtils.beanNamesForTypeIncludingAncestors(
          this.beanFactory, Object.class, true, false);
        for (String beanName : beanNames) {
          // 获得该bean的class
          Class<?> beanType = this.beanFactory.getType(beanName);
          // 判断是否有标识@AspectJ注解
          if (this.advisorFactory.isAspect(beanType)) {
            // 将beanName放入集合中
            aspectNames.add(beanName);
            // 将beanType和beanName封装到AspectMetadata中
            AspectMetadata amd = new AspectMetadata(beanType, beanName);
            // Kind默认为SINGLETON
            if (amd.getAjType().getPerClause().getKind() == PerClauseKind.SINGLETON) {
              MetadataAwareAspectInstanceFactory factory =
                new BeanFactoryAspectInstanceFactory(this.beanFactory, beanName);
              // 这里会通过@Before @After等标识的方法获取到所有的advisor
              List<Advisor> classAdvisors = this.advisorFactory.getAdvisors(factory);
              if (this.beanFactory.isSingleton(beanName)) {
                // 将获取到的所有advisor放入缓存
                this.advisorsCache.put(beanName, classAdvisors);
              }
              advisors.addAll(classAdvisors);
            }
          }
        }
        // 将所有解析过的beanName赋值
        this.aspectBeanNames = aspectNames;
        return advisors;
      }
    }
  }
  // aspectNames不为空,意味有advisor,取出之前解析好的所有advisor
  List<Advisor> advisors = new ArrayList<>();
  // 获取到所有解析好的advisor
  for (String aspectName : aspectNames) {
    List<Advisor> cachedAdvisors = this.advisorsCache.get(aspectName);
    if (cachedAdvisors != null) {
      advisors.addAll(cachedAdvisors);
    }
		return advisors;
	}

advisorFactory.getAdvisors

public List<Advisor> getAdvisors(MetadataAwareAspectInstanceFactory aspectInstanceFactory) {
	// 获取到标识了@AspectJ的class,其实就是刚刚封装的class
  Class<?> aspectClass = aspectInstanceFactory.getAspectMetadata().getAspectClass();
  // 获取className
  String aspectName = aspectInstanceFactory.getAspectMetadata().getAspectName();
  
  List<Advisor> advisors = new ArrayList<>();
  
  // 拿出该类除了标识@PointCut的所有方法进行遍历 getAdvisorMethods时会对method进行一次排序
  // 排序顺序 Around, Before, After, AfterReturning, AfterThrowing
  for (Method method : getAdvisorMethods(aspectClass)) {
    // 获取到advisor
    Advisor advisor = getAdvisor(method, lazySingletonAspectInstanceFactory, 0, aspectName);
    if (advisor != null) {
      // 加入到集合中
      advisors.add(advisor);
    }
  }
}

我们先看下getAdvisorMethods方法

private List<Method> getAdvisorMethods(Class<?> aspectClass) {
  final List<Method> methods = new ArrayList<>();
  // 循环遍历该类和父类的所有方法
  ReflectionUtils.doWithMethods(aspectClass, method -> {
    // 排除@PointCut标识的方法
    if (AnnotationUtils.getAnnotation(method, Pointcut.class) == null) {
      methods.add(method);
    }
  }, ReflectionUtils.USER_DECLARED_METHODS);
  if (methods.size() > 1) {
    // 以Around, Before, After, AfterReturning, AfterThrowing的顺序自定义排序
    methods.sort(METHOD_COMPARATOR);
  }
  return methods;
}

不知道小伙伴们对ReflectionUtils.doWithMethods这个工具类熟不熟悉呢,这个工具类在之前分析Bean创建过程时可是出现了好多次呢,并且我们也是可以使用的

现在,已经获取到切面中的所有方法了,那么接下来就该对这些方法解析并进行封装成advisor了~

getAdvisor

public Advisor getAdvisor(Method candidateAdviceMethod, MetadataAwareAspectInstanceFactory aspectInstanceFactory,
			int declarationOrderInAspect, String aspectName) {
	// 获取方法上的切点表达式
  AspectJExpressionPointcut expressionPointcut = getPointcut(
    candidateAdviceMethod, aspectInstanceFactory.getAspectMetadata().getAspectClass());
  // 封装成对象返回,创建对象时将会解析方法创建advice
  return new InstantiationModelAwarePointcutAdvisorImpl(expressionPointcut, candidateAdviceMethod,
                                                        this, aspectInstanceFactory, declarationOrderInAspect, aspectName);
}

获取切点表达式的过程其实非常简单,即是解析方法上的注解,取出注解上的value即可

getPointcut

private AspectJExpressionPointcut getPointcut(Method candidateAdviceMethod, Class<?> candidateAspectClass) {
  // 查找方法上和AspectJ相关注解
  AspectJAnnotation<?> aspectJAnnotation =
    AbstractAspectJAdvisorFactory.findAspectJAnnotationOnMethod(candidateAdviceMethod);
  // 设置切点表达式
  AspectJExpressionPointcut ajexp =
    new AspectJExpressionPointcut(candidateAspectClass, new String[0], new Class<?>[0]);
  // PointcutExpression 为注解上value属性的值
  ajexp.setExpression(aspectJAnnotation.getPointcutExpression());
  if (this.beanFactory != null) {
    ajexp.setBeanFactory(this.beanFactory);
  }
  return ajexp;
}

new InstantiationModelAwarePointcutAdvisorImpl,在这里,才会真正创建出advice

public InstantiationModelAwarePointcutAdvisorImpl(){
  //...省略赋值过程...
  // 实例化出advice
  this.instantiatedAdvice = instantiateAdvice(this.declaredPointcut);
}
private Advice instantiateAdvice(AspectJExpressionPointcut pointcut) {
  // 获取advice,aspectJAdviceMethod为方法,aspectName为切面类
  Advice advice = this.aspectJAdvisorFactory.getAdvice(this.aspectJAdviceMethod, pointcut,
                                                       this.aspectInstanceFactory, this.declarationOrder, this.aspectName);
  return (advice != null ? advice : EMPTY_ADVICE);
}
public Advice getAdvice(){
  // 根据方法获取到注解信息
  AspectJAnnotation<?> aspectJAnnotation =
				AbstractAspectJAdvisorFactory.findAspectJAnnotationOnMethod(candidateAdviceMethod);
  AbstractAspectJAdvice springAdvice;
  // 根据注解类型返回对象,创建对象的过程都是一样的,都是调用父类的构造方法
  // candidateAdviceMethod为切面的方法,expressionPointcut是切点
  switch (aspectJAnnotation.getAnnotationType()) {
    case AtPointcut
      return null;
    case AtAround:
      springAdvice = new AspectJAroundAdvice(
        candidateAdviceMethod, expressionPointcut, aspectInstanceFactory);
      break;
    case AtBefore:
      springAdvice = new AspectJMethodBeforeAdvice(
        candidateAdviceMethod, expressionPointcut, aspectInstanceFactory);
      break;
    case AtAfter:
      springAdvice = new AspectJAfterAdvice(
        candidateAdviceMethod, expressionPointcut, aspectInstanceFactory);
      break;
      //...省略其他的advice
    default:
      throw new UnsupportedOperationException(
        "Unsupported advice type on method: " + candidateAdviceMethod);
  }
  return springAdvice;
}

springAdvice已创建完毕,意味着切面中的某个方法已经解析完毕了,其他的方法解析过程大致也是相似的

小结

其实解析切面本身并不复杂,只是Spring中将切面类封装来封装去容易使人混乱,如buildAspectJAdvisors方法中,封装了一个AspectMetadata amd = new AspectMetadata(beanType, beanName);,又立即发起判定amd.getAjType().getPerClause().getKind() == PerClauseKind.SINGLETON,其实这里完全可以变为AjTypeSystem.getAjType(currClass).getPerClause().getKind() == PerClauseKind.SINGLETONAjTypeSystem.getAjType(currClass)new AspectMetadata的一部分逻辑,笔者这里给大家总结一下吧。

首先,循环所有的beanName,找到带有@Aspectj注解的class, 获取到class中的所有方法进行遍历解析,取出方法注解上的值(切点:pointcut),然后把方法,切点表达式,封装了BeanFactory,BeanName的factory封装成相应的SpringAdvice, 由SpringAdvice和pointcut组合成一个advisor。

创建代理对象

切面已经解析完毕,接下来,我们就来看看如何把解析出的切面织入到目标方法中吧

但,在这之前,还有必要给小伙伴们补充一点前置知识。

我们知道,一个bean是否能够被aop代理,取决于它是否满足代理条件,即为是否能够被切点表达式所命中,而在Spring AOP中,bean与切点表达式进行匹配的是AspectJ实现的,并非Spring所完成的,所以我们先来看看AspectJ如何匹配出合适的bean的吧

栗子

首先需要引入org.aspectj:aspectjweaver依赖

一个Service,包名为com.my.spring.test.aop

package com.my.spring.test.aop;

/**
 * 切点表达式可以匹配的类
 *
 */
public class ServiceImpl{
	/**
	 * 切点表达式可以匹配的方法
	 */
  public void doService() {
    System.out.println("do service ...");
  }
	public void matchMethod() {
		System.out.println("ServiceImpl.notMatchMethod");
	}
}

然后,我们自己封装一个用于匹配的工具类,具体功能大家看注释哈哈

package com.my.spring.test.aspectj;

import org.aspectj.weaver.tools.PointcutExpression;
import org.aspectj.weaver.tools.PointcutParser;
import org.aspectj.weaver.tools.ShadowMatch;

import java.lang.reflect.Method;

/**
 * aop工具
 */
public class AOPUtils {
	// AspectJ的固定写法,获取一个切点解析器
	static PointcutParser parser = PointcutParser
			.getPointcutParserSupportingSpecifiedPrimitivesAndUsingSpecifiedClassLoaderForResolution(
					PointcutParser.getAllSupportedPointcutPrimitives(), ClassLoader.getSystemClassLoader());
	// 切点表达式
	private static PointcutExpression pointcutExpression;

	/**
	 * 初始化工具类,我们需要先获取一个切点表达式
	 *
	 * @param expression 表达式
	 */
	public static void init(String expression){
		// 解析出一个切点表达式
		pointcutExpression =  parser.parsePointcutExpression(expression);
	}

	/**
	 * 第一次筛选,根据类筛选,也叫做粗筛
	 *
	 * @param targetClass 目标类
	 * @return 是否匹配
	 */
	public static boolean firstMatch(Class<?> targetClass){
    // 根据类筛选
		return pointcutExpression.couldMatchJoinPointsInType(targetClass);
	}

	/**
	 * 第二次筛选,根据方法筛选,也叫做精筛,精筛通过则说明完全匹配
	 * ps: 也可以使用该方法进行精筛,粗筛的目的是提高性能,第一次直接过滤掉不合适的类再慢慢精筛
	 * 
	 * @param method 方法
	 * @return 是否匹配
	 */
	public static boolean lastMatch(Method method){
    // 根据方法筛选
		ShadowMatch shadowMatch = pointcutExpression.matchesMethodExecution(method);
		return shadowMatch.alwaysMatches();
	}

}

测试

public class AOPUtilsTest {

	public static void main(String[] args) throws NoSuchMethodException {
		// 定义表达式
		String expression = "execution(* com.my.spring.test.aop.*.*(..))";
		// 初始化工具类
		AOPUtils.init(expression);
		// 粗筛
		boolean firstMatch = AOPUtils.firstMatch(ServiceImpl.class);
		if(firstMatch){
			System.out.println("第一次筛选通过");
			// 正常情况应该是获取所有方法进行遍历,我这里偷懒了~
			Method doService = ServiceImpl.class.getDeclaredMethod("doService");
			// 精筛
			boolean lastMatch = AOPUtils.lastMatch(doService);
			if(lastMatch){
				System.out.println("第二次筛选通过");
			}
			else{
				System.out.println("第二次筛选未通过");
			}
		}
		else {
			System.out.println("第一次筛选未通过");
		}
	}
}

结果(就不截图了,怀疑的小伙伴可以自己试试~)

第一次筛选通过
第二次筛选通过

当我们新建一个类Test,把切点表达式换成

execution(* com.my.spring.test.aop.Test.*(..))

测试结果为

第一次筛选未通过

再把切点表达式换成指定的方法

execution(* com.my.spring.test.aop.*.matchMethod(..))

结果

第一次筛选通过
第二次筛选未通过

到这里,小伙伴们应该明白了AspectJ的使用方法吧

代理对象创建过程

接下来,我们就来看看Spring是如何使用AspectJ匹配出相应的advisor并创建代理对象的吧,以下为创建代理对象的大致路程图

创建代理对象.png
创建代理对象是在bean初始化后完成的,所以对应的beanPostProcessor调用时机为postProcessAfterInitialization

AbstractAutoProxyCreator#postProcessAfterInitialization

public Object postProcessAfterInitialization(@Nullable Object bean, String beanName) {
		if (bean != null) {
			// 获取缓存key值,其实就是beanName
			Object cacheKey = getCacheKey(bean.getClass(), beanName);
			// 判断缓存中是否有该对象,有则说明该对象已被动态代理,跳过
			if (this.earlyProxyReferences.remove(cacheKey) != bean) {
				return wrapIfNecessary(bean, beanName, cacheKey);
			}
		}
		return bean;
	}

wrapIfNecessary

protected Object wrapIfNecessary(Object bean, String beanName, Object cacheKey) {
	// 根据bean获取到匹配的advisor
  Object[] specificInterceptors = getAdvicesAndAdvisorsForBean(bean.getClass(), beanName, null);
  if (specificInterceptors != DO_NOT_PROXY) {
    // 创建代理对象
    Object proxy = createProxy(
      bean.getClass(), beanName, specificInterceptors, new SingletonTargetSource(bean));
    return proxy;
  }
  return bean;
}

getAdvicesAndAdvisorsForBean

protected Object[] getAdvicesAndAdvisorsForBean(
			Class<?> beanClass, String beanName, @Nullable TargetSource targetSource) {
  // 获取合适的advisor
  List<Advisor> advisors = findEligibleAdvisors(beanClass, beanName);
  return advisors.toArray();
}

findEligibleAdvisors

protected List<Advisor> findEligibleAdvisors(Class<?> beanClass, String beanName) {
  // 先获取到所有的advisor, 这里和解析过程相同,由于已经解析好,所以会直接从缓存中取出
  List<Advisor> candidateAdvisors = findCandidateAdvisors();
  // 筛选出匹配的advisor
  List<Advisor> eligibleAdvisors = findAdvisorsThatCanApply(candidateAdvisors, beanClass, beanName);
  // 增加一个默认的advisor
  extendAdvisors(eligibleAdvisors);
  if (!eligibleAdvisors.isEmpty()) {
    // 排序
    eligibleAdvisors = sortAdvisors(eligibleAdvisors);
  }
  return eligibleAdvisors;
}

findAdvisorsThatCanApply

protected List<Advisor> findAdvisorsThatCanApply(
			List<Advisor> candidateAdvisors, Class<?> beanClass, String beanName) {
  // 查找匹配的advisor
  return AopUtils.findAdvisorsThatCanApply(candidateAdvisors, beanClass);
}

findAdvisorsThatCanApply

public static List<Advisor> findAdvisorsThatCanApply(List<Advisor> candidateAdvisors, Class<?> clazz){
  List<Advisor> eligibleAdvisors = new ArrayList<>();
  for (Advisor candidate : candidateAdvisors) {
    // 判断是否匹配
    if (canApply(candidate, clazz, hasIntroductions)) {
      // 加入到合适的advisors集合中
      eligibleAdvisors.add(candidate);
    }
  }
  return eligibleAdvisors;
}

canApply

public static boolean canApply(Advisor advisor, Class<?> targetClass, boolean hasIntroductions) {
  if (advisor instanceof PointcutAdvisor) {
    PointcutAdvisor pca = (PointcutAdvisor) advisor;
    // 判断是否匹配
    return canApply(pca.getPointcut(), targetClass, hasIntroductions);
  }
  else {
    // It doesn't have a pointcut so we assume it applies.
    return true;
  }
}

canApply

public static boolean canApply(Pointcut pc, Class<?> targetClass, boolean hasIntroductions) {
	// 第一次筛选,对class筛选判断是否满足匹配条件
  // 这里将会初始化切点表达式
  if (!pc.getClassFilter().matches(targetClass)) {
    return false;
  }
  
  IntroductionAwareMethodMatcher introductionAwareMethodMatcher = null;
  if (methodMatcher instanceof IntroductionAwareMethodMatcher) {
    introductionAwareMethodMatcher = (IntroductionAwareMethodMatcher) methodMatcher;
  }
  
  for (Class<?> clazz : classes) {
    Method[] methods = ReflectionUtils.getAllDeclaredMethods(clazz);
    // 循环所有方法进行第二次筛选,判断是否有方法满足匹配条件
    for (Method method : methods) {
      if (introductionAwareMethodMatcher != null ?
          introductionAwareMethodMatcher.matches(method, targetClass, hasIntroductions) :
          methodMatcher.matches(method, targetClass)) {
        return true;
      }
    }
  }
  return false;
}

pc.getClassFilter()

public ClassFilter getClassFilter() {
  obtainPointcutExpression();
  return this;
}

obtainPointcutExpression

private PointcutExpression obtainPointcutExpression() {
  if (this.pointcutExpression == null) {
    // 确认类加载器
    this.pointcutClassLoader = determinePointcutClassLoader();
    // 创建切点表达式
    this.pointcutExpression = buildPointcutExpression(this.pointcutClassLoader);
  }
  return this.pointcutExpression;
}

buildPointcutExpression

private PointcutExpression buildPointcutExpression(@Nullable ClassLoader classLoader) {
  // 初始化切点解析器
  PointcutParser parser = initializePointcutParser(classLoader);
  PointcutParameter[] pointcutParameters = new PointcutParameter[this.pointcutParameterNames.length];
  for (int i = 0; i < pointcutParameters.length; i++) {
    pointcutParameters[i] = parser.createPointcutParameter(
      this.pointcutParameterNames[i], this.pointcutParameterTypes[i]);
  }
  // 使用切点解析器进行解析表达式获取切点表达式
  return parser.parsePointcutExpression(replaceBooleanOperators(resolveExpression()),
                                        this.pointcutDeclarationScope, pointcutParameters);
}

initializePointcutParser

private PointcutParser initializePointcutParser(@Nullable ClassLoader classLoader) {
  // 获得切点解析器
  PointcutParser parser = PointcutParser
    .getPointcutParserSupportingSpecifiedPrimitivesAndUsingSpecifiedClassLoaderForResolution(
    SUPPORTED_PRIMITIVES, classLoader);
  parser.registerPointcutDesignatorHandler(new BeanPointcutDesignatorHandler());
  return parser;
}

pc.getClassFilter便是完成了以上事情,此时再进行调用matchs方法

public boolean matches(Class<?> targetClass) {
  PointcutExpression pointcutExpression = obtainPointcutExpression();
  // 使用切点表达式进行粗筛
  return pointcutExpression.couldMatchJoinPointsInType(targetClass);
}

introductionAwareMethodMatcher.matches 同样如此

以上便是寻找合适的advisor的过程,下面,就是通过这些advisor进行创建动态代理了

createProxy

protected Object createProxy(Class<?> beanClass, @Nullable String beanName,
			@Nullable Object[] specificInterceptors, TargetSource targetSource) {
  ProxyFactory proxyFactory = new ProxyFactory();
  proxyFactory.copyFrom(this);
	// 将specificInterceptors(现在是Object)转化为Advisor返回
  Advisor[] advisors = buildAdvisors(beanName, specificInterceptors);
  // 赋值到proxyFactory的advisors属性中
  proxyFactory.addAdvisors(advisors);
  proxyFactory.setTargetSource(targetSource);
  customizeProxyFactory(proxyFactory);
  // 创建动态代理
  return proxyFactory.getProxy(getProxyClassLoader());
}

proxyFactory.getProxy

public Object getProxy(@Nullable ClassLoader classLoader) {
  // 创建代理对象
  return createAopProxy().getProxy(classLoader);
}

createAopProxy

protected final synchronized AopProxy createAopProxy() {
  // 创建AOP代理对象
  return getAopProxyFactory().createAopProxy(this);
}
public AopProxy createAopProxy(AdvisedSupport config) throws AopConfigException {
  // @EnableAspectJAutoProxy的proxyTargetClass是否配置为true
  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.");
    }
    // 如何是接口则创建jdk动态代理
    if (targetClass.isInterface() || Proxy.isProxyClass(targetClass)) {
      return new JdkDynamicAopProxy(config);
    }
    // cglib动态代理
    return new ObjenesisCglibAopProxy(config);
  }
  // 默认是jdk动态代理
  else {
    return new JdkDynamicAopProxy(config);
  }
}
public Object getProxy(@Nullable ClassLoader classLoader) {
  // 获取到代理的接口
  Class<?>[] proxiedInterfaces = AopProxyUtils.completeProxiedInterfaces(this.advised, true);
  findDefinedEqualsAndHashCodeMethods(proxiedInterfaces);
  // 创建jdk代理,传入的为JdkDynamicAopProxy对象,里面包含了被代理的bean以及匹配的advisor
  return Proxy.newProxyInstance(classLoader, proxiedInterfaces, this);
}

动态代理创建完成~

代理对象调用过程

对象都给你创建好了,接下当然是开..发起调用咯

以下是调用的大致流程图

代理对象调用过程

代理对象被调用的是invoke方法,我们所创建的代理对象为JdkDynamicAopProxy,所以

JdkDynamicAopProxy#invoke

public Object invoke(Object proxy, Method method, Object[] args) throws Throwable {
  Object oldProxy = null;
  boolean setProxyContext = false;
  // 取出包装了被代理bean的对象->创建代理对象时的SingletonTargetSource, advised为ProxyFactory
  TargetSource targetSource = this.advised.targetSource;
  Object target = null;
  // 拿到bean
  target = targetSource.getTarget();
  Class<?> targetClass = (target != null ? target.getClass() : null);
  // 将所有advisor中的advice取出,并转化为对应的interceptor
  List<Object> chain = this.advised.getInterceptorsAndDynamicInterceptionAdvice(method, targetClass);
  // 创建一个最外层的MethodInvocation用于发起调用
  MethodInvocation invocation =
    new ReflectiveMethodInvocation(proxy, target, method, args, targetClass, chain);
  // 发起链式调用
  Object retVal = invocation.proceed();
  return retVal;
}

我们先看获取interceptor的过程

getInterceptorsAndDynamicInterceptionAdvice

public List<Object> getInterceptorsAndDynamicInterceptionAdvice(Method method, @Nullable Class<?> targetClass) {
  // 将所有advisor中的advice取出并封装成intercept
  return this.advisorChainFactory.getInterceptorsAndDynamicInterceptionAdvice(this, method, targetClass);
}
public List<Object> getInterceptorsAndDynamicInterceptionAdvice(
  Advised config, Method method, @Nullable Class<?> targetClass) {
	// 创建一个advisor适配器的注册器用于转化advice,创建时将默认注册三个适配器
  AdvisorAdapterRegistry registry = GlobalAdvisorAdapterRegistry.getInstance();
  Advisor[] advisors = config.getAdvisors();
  // 循环遍历所有advisor
  for (Advisor advisor : advisors) {
  	// 将advisor中的advice转化为interceptor
    MethodInterceptor[] interceptors = registry.getInterceptors(advisor);
    interceptorList.addAll(Arrays.asList(interceptors));
    return interceptorList;
  }
}

GlobalAdvisorAdapterRegistry.getInstance() 类初始化时调用静态方法

private static AdvisorAdapterRegistry instance = new DefaultAdvisorAdapterRegistry()
public static AdvisorAdapterRegistry getInstance() {
		return instance;
}
public DefaultAdvisorAdapterRegistry() {
  // 注册三个适配器
  registerAdvisorAdapter(new MethodBeforeAdviceAdapter());
  registerAdvisorAdapter(new AfterReturningAdviceAdapter());
  registerAdvisorAdapter(new ThrowsAdviceAdapter());
}
public void registerAdvisorAdapter(AdvisorAdapter adapter) {
  // 将适配器加入集合
  this.adapters.add(adapter);
}

registry.getInterceptors 这里面包含了advice转化成interceptor的过程

public MethodInterceptor[] getInterceptors(Advisor advisor) throws UnknownAdviceTypeException {
  List<MethodInterceptor> interceptors = new ArrayList<>(3);
  Advice advice = advisor.getAdvice();
  // advice本身是否就是MethodInterceptor
  if (advice instanceof MethodInterceptor) {
    interceptors.add((MethodInterceptor) advice);
  }
  for (AdvisorAdapter adapter : this.adapters) {
    // 判断advice是哪个advice 如:(advice instanceof MethodBeforeAdvice)
    if (adapter.supportsAdvice(advice)) {
      // 将advice封装到对应的interceptor
      interceptors.add(adapter.getInterceptor(advisor));
    }
  }
  return interceptors.toArray(new MethodInterceptor[0]);
}

若adapter为MethodBeforeAdviceAdapter,则

public MethodInterceptor getInterceptor(Advisor advisor) {
  MethodBeforeAdvice advice = (MethodBeforeAdvice) advisor.getAdvice();
  return new MethodBeforeAdviceInterceptor(advice);
}

其他advice转化过程相同

以上,便将所有的advice转化成了interceptor,接下来,则是经典的链式递归调用过程

以下过程小伙伴们可以对照流程图阅读,毕竟递归还是有些复杂,需要一定的功底

ReflectiveMethodInvocation#proceed

public Object proceed() throws Throwable {
  // currentInterceptorIndex 初始值为-1
  // 当currentInterceptorIndex等于advice的数量减一时,则调用目标方法
  // 由于advice已排好序,所以调用顺序为before, after, afterReturn, afterThrowing
  // 注意,并非调用到相应的advice就会执行advice方法,这里是类似递归调用的方式,会存在一个归过程
  // 有些是递的时候发起调用,如beforeAdvice, 但有些则是归的时候发起调用,如afterAdvice
  // 递归的终止条件则是这下面这个return invokeJoinpoint();
  if (this.currentInterceptorIndex == this.interceptorsAndDynamicMethodMatchers.size() - 1) {
    return invokeJoinpoint();
  }
	// currentInterceptorIndex自增并获取到interceptor
  Object interceptorOrInterceptionAdvice =
    this.interceptorsAndDynamicMethodMatchers.get(++this.currentInterceptorIndex);
  // 将interceptro强转为MethodInterceptor发起调用
  return ((MethodInterceptor) interceptorOrInterceptionAdvice).invoke(this);
}

此时currentInterceptorIndex值为0,而我们的advice为4个(去除了默认的),所以当currentInterceptorIndex为3时便会调用我们的实际方法

首先调用的是MethodBeforeAdviceInterceptor

public Object invoke(MethodInvocation mi) throws Throwable {
  // 调用前置通知
  this.advice.before(mi.getMethod(), mi.getArguments(), mi.getThis());
  return mi.proceed();
}

mi为传入的this,所有mi.proceed()将会回到最开始的方法

再次循环,此时currentInterceptorIndex值为1

调用的是AspectJAfterAdvice

public Object invoke(MethodInvocation mi) throws Throwable {
  try {
    return mi.proceed();
  }
  finally {
    // finally意味着不管怎样都会被调用
    invokeAdviceMethod(getJoinPointMatch(), null, null);
  }
}

继续,此时currentInterceptorIndex值为2

调用的是AfterReturningAdviceInterceptor

public Object invoke(MethodInvocation mi) throws Throwable {
  Object retVal = mi.proceed();
  this.advice.afterReturning(retVal, mi.getMethod(), mi.getArguments(), mi.getThis());
  return retVal;
}

继续,此时currentInterceptorIndex值为3

调用的是AspectJAfterThrowingAdvice

public Object invoke(MethodInvocation mi) throws Throwable {
  try {
    return mi.proceed();
  }
  catch (Throwable ex) {
    if (shouldInvokeOnThrowing(ex)) {
      // 调用异常通知
      invokeAdviceMethod(getJoinPointMatch(), null, ex);
    }
    // 往外抛出异常
    throw ex;
  }
}

所以如果我们的业务方法发生了异常,会调用到异常通知,而这里又把异常往外抛,所以afterReturn就会被跳过直接到after的finally方法

现在currentInterceptorIndex值为3了,再回调最初的方法中时,就会调用到我们的业务方法了。调用完毕则进行归的过程,调用过程便结束了。

以上,便是整个AOP的过程了,下一篇,事务相关源码解析~

本篇文章中涉及到图片的矢量图地址为:https://www.processon.com/view/link/5fa8afdae401fd45d109f257,有需要的小伙伴可自取

# spring 

评论

Your browser is out-of-date!

Update your browser to view this website correctly. Update my browser now

×