SpringIOC——DI循环依赖
要弄清楚循环依赖
1、需要知道Bean初始化的两个阶段
① Bean实例化创建实例对象(new Bean())
② Bean实例对象初始化(DI:注解自动注入)
2、DefaultSingletonBeanRegistry类中的5个容器(补充:很多地方说的三级缓存就是下面的 1/2/3容器)
/** 记录已将创建的单例<beanName,singletonBean> */ private final Map<String, Object> singletonObjects = new ConcurrentHashMap<>(256); /** 记录singletonFactory<beanName,singletonFactory> singeletonFactory中存放beanName和上面的①阶段的bean:Bean实例化实例对象(还未初始化DI)*/ private final Map<String, ObjectFactory<?>> singletonFactories = new HashMap<>(16); /** 记录早期的singletonBean 存放的也是① */ private final Map<String, Object> earlySingletonObjects = new HashMap<>(16); /** 存放已经初始化后的beanName,有序 */ private final Set<String> registeredSingletons = new LinkedHashSet<>(256); /** 记录正在初始化的bean的beanName */ private final Set<String> singletonsCurrentlyInCreation = Collections.newSetFromMap(new ConcurrentHashMap<>(16));
其中跟循环依赖相关的是singletonFactories、singeletonsCurrentlyInCreation、earlysingletonObjects.
3、循环依赖实现
①bean初始化前后会打标,加入到singletonsCurrentlyInCreation容器中,这个打标会在核心方法getSingleton()中起作用
/* org.springframework.beans.factory.support.DefaultSingletonBeanRegistry#getSingleton(java.lang.String, org.springframework.beans.factory.ObjectFactory<?>) */ public Object getSingleton(String beanName, ObjectFactory<?> singletonFactory) { ... //循环依赖相关:初始化前先singletonsCurrentlyInCreation.add(beanName) beforeSingletonCreation(beanName);
... //lamda表达式:其实是调用createBean(beanName, mbd, args):Bean初始化 singletonObject = singletonFactory.getObject(); ... //循环依赖相关:初始化后singletonsCurrentlyInCreation.remove(beanName) afterSingletonCreation(beanName); ...//初始化完后 //this.singletonObjects.put(beanName, singletonObject);放入到单例容器中 //this.singletonFactories.remove(beanName);清空循环依赖的两个打标 //this.earlySingletonObjects.remove(beanName); //this.registeredSingletons.add(beanName);放入单例beanName容器中 addSingleton(beanName, singletonObject); ... } }
② 上面singletonObject = singletonFactory.getObject()时向singletonFactories中记录了(new Bean()),singletonFactories也会在核心方法getSingleton()中起作用
/* org.springframework.beans.factory.support.AbstractAutowireCapableBeanFactory#doCreateBean */ protected Object doCreateBean(final String beanName, final RootBeanDefinition mbd, final @Nullable Object[] args) throws BeanCreationException { ... //循环依赖相关逻辑: //this.singletonFactories.put(beanName, singletonFactory); //将实例化bean(①阶段)、beanName组装成singletonFactory装入singletonFactories容器 //this.earlySingletonObjects.remove(beanName); //删除earlySingletonObjects中beanName addSingletonFactory(beanName, () -> getEarlyBeanReference(beanName, mbd, bean)); ... //实例初始化 就是在这里面实现依赖注入DI的:反射实现 //调用AutowiredAnnotationBeanPostProcessor.postProcessProperties populateBean(beanName, mbd, instanceWrapper); ... }
③ 核心方法getSingleton
/* org.springframework.beans.factory.support.DefaultSingletonBeanRegistry#getSingleton(java.lang.String, boolean) */ protected Object getSingleton(String beanName, boolean allowEarlyReference) { Object singletonObject = this.singletonObjects.get(beanName); //循环依赖核心就在于这个判断,由于打标+记录了①阶段的bean,
//循环依赖第二次调用getBean("a")时,这里会直接返回第一次调用getBean("a")创建的①阶段的bean
//而不会调用createBean("a")再次bean初始化(造成两个bean的循环创建) if (singletonObject == null && isSingletonCurrentlyInCreation(beanName)) { synchronized (this.singletonObjects) { singletonObject = this.earlySingletonObjects.get(beanName); if (singletonObject == null && allowEarlyReference) { ObjectFactory<?> singletonFactory = this.singletonFactories.get(beanName); if (singletonFactory != null) { singletonObject = singletonFactory.getObject(); this.earlySingletonObjects.put(beanName, singletonObject); this.singletonFactories.remove(beanName); } } } } return singletonObject; }
④ 循环依赖流程
/* org.springframework.beans.factory.support.AbstractBeanFactory#doGetBean */ protected <T> T doGetBean(final String name, @Nullable final Class<T> requiredType, @Nullable final Object[] args, boolean typeCheckOnly) throws BeansException { ... //假设A、B互相依赖 //第一次getBean(A),sharedInstance == null(一级缓存singletonObjects中没有),bean还未被初始化,走else,createBean初始化bean //A打标:正在创建中,代码在①中beforeSingletonCreation()中,,主要逻辑:放入到singetonsCurrentlyInCreation这个Set中
//A实例化(早期Bean)后保存到singletonFactories中②中addSingletonFactory(beanName,singletonFactory)(即三级缓存)
//DI依赖注入:②中populateBean(beanName,mbd,instanceWrapper),发现依赖B,调用getBean(B)初始化B的单例
//调用getBean(B)重复上面步骤,DI依赖注入发现依赖A,调用getBean(A)
//第二次getBean(A),一级缓存singletonObjects中依然没有初始化完成的Bean,singleObject == null
//③中if(singletonObject == null && isSingletonCurrentlyInCreation(A))由于打标了所以返回上面创建的早期Bean-A(具体是先访问二级缓存earlySingletonObjects,再访问三级缓存singletonFactories)
//下面if条件直接返回bean,没有走else破坏了循环 Object sharedInstance = getSingleton(beanName); if (sharedInstance != null && args == null) { // bean = getObjectForBeanInstance(sharedInstance, name, beanName, null); } else { ... // createBean(beanName, mbd, args); bean = getObjectForBeanInstance(sharedInstance, name, beanName, null); } return bean; }
四、总结
未看源码之前,其实对循环依赖有一个想法:循环依赖可以看做是一个死锁。
预防死锁的方法:打破死锁的四个必要条件(互斥、请求并等待、不可剥夺、循环等待),由于循环依赖的资源是对象自身,所以常用破坏循环等待条件方法:编号顺序执行,不适用
选择破坏请求并等待条件:先创建对象,再赋值,模型
A a = new A(); B b = new B(); A.b = b; B.a = a;
研究源码之后发现:想法差不多,但是代码实现非常精彩。模型(打标没想到过)
A a = new A(); B b = new B(); b.a = a; a.b = b;