singletonObjects (一级缓存)它是我们最熟悉的朋友,俗称“单例池”“容器”,缓存创建完成单例Bean的地方。
earlySingletonObjects(二级缓存)映射Bean的早期引用,也就是说在这个Map里的Bean不是完整的,甚至还不能称之为“Bean”,只是一个Instance.
singletonFactories(三级缓存) 映射创建Bean的原始工厂
这样做有什么好处呢?让我们来分析一下“A的某个field或者setter依赖了B的实例对象,同时B的某个field或者setter依赖了A的实例对象”这种循环依赖的情况:
spring初始化bean的过程如下:
1>首先尝试从一级缓存中获取serviceA实例,发现不存在并且serviceA不在创建过程中;
2>serviceA完成了初始化的第一步(实例化:调用createBeanInstance方法,即调用默认构造方法实例化);
3>将自己(serviceA)提前曝光到singletonFactories中;
4>此时进行初始化的第二步(注入属性serviceB),发现自己依赖对象serviceB,此时就尝试去get(B),发现B还没有被实create,所以走create流程;
5>serviceB完成了初始化的第一步(实例化:调用createBeanInstance方法,即调用默认构造方法实例化);
6>将自己(serviceB)提前曝光到singletonFactories中;
7>此时进行初始化的第二步(注入属性serviceA);
8>于是尝试get(A),尝试一级缓存singletonObjects(肯定没有,因为A还没初始化完全),尝试二级缓存earlySingletonObjects(也没有),尝试三级缓存singletonFactories,由于A通过ObjectFactory将自己提前曝光了,所以B能够通过ObjectFactory.getObject拿到A对象(虽然A还没有初始化完全,但是总比没有好呀);
9>B拿到A对象后顺利完成了初始化阶段1、2、3,完全初始化之后将自己放入到一级缓存singletonObjects中;
10>此时返回A中,A此时能拿到B的对象顺利完成自己的初始化阶段2、3,最终A也完成了初始化,进去了一级缓存singletonObjects中;
知道了这个原理时候,肯定就知道为啥Spring不能解决“A的构造方法中依赖了B的实例对象,同时B的构造方法中依赖了A的实例对象”这类问题了!因为加入singletonFactories三级缓存的前提是执行了构造器,所以构造器的循环依赖没法解决。
4、循环依赖的N种场景
这里主要讲解spring是如何解决单例setter注入时的循环依赖问题,其他四种循环依赖场景后文再一一分析讲解
3.1、单例的setter注入
这种注入方式应该是spring用的最多的,代码如下:
@Service
public class ServiceA {
@Autowired
private ServiceB serviceB;
}
@Service
public class ServiceB {
@Autowired
private ServiceA serviceA;
}
spring内部有三级缓存:
singletonObjects 一级缓存,用于保存实例化、注入、初始化完成的bean实例
earlySingletonObjects 二级缓存,用于保存实例化完成的bean实例
singletonFactories 三级缓存,用于保存bean创建工厂,以便于后面扩展有机会创建代理对象。
下面用一张图告诉你,spring是如何解决循环依赖的:
上面(3、Spring怎么解决循环依赖)中已经讲解过spring解决单例循环依赖的过程,这里细心的朋友可能会发现在这种场景中第二级缓存(earlySingletonObjects)作用不大。
那么问题来了,为什么要用第二级缓存呢?
试想一下,如果出现以下这种情况,我们要如何处理?
@Service
public class ServiceA {
@Autowired
private ServiceB serviceB;
@Autowired
private ServiceB serviceC;
}
@Service
public class ServiceB {
@Autowired
private ServiceA serviceA;
}
@Service
public class ServiceC {
@Autowired
private ServiceA serviceA;
}
serviceA依赖于serviceB和serviceC,而serviceB依赖于serviceA,同时serviceC也依赖于serviceA。按照上图的流程可以把serviceA注入到serviceB,并且sServiceA的实例是从第三级缓存中获取的。假设不用第二级缓存,serviceA注入到serviceC的流程如图:
图2
serviceA注入到serviceC又需要从第三级缓存中获取实例,而第三级缓存里保存的并非真正的实例对象,而是ObjectFactory对象。说白了,两次从三级缓存中获取都是ObjectFactory对象,而通过它创建的实例对象每次可能都不一样的。这样不是有问题?
为了解决这个问题,spring引入的第二级缓存。上面图1其实serviceA对象的实例已经被添加到第二级缓存中了,而在serviceA注入到TestService3时,只用从第二级缓存中获取该对象即可。
图3
还有个问题,第三级缓存中为什么要添加ObjectFactory对象,直接保存实例对象不行吗?
答:不行,因为假如你想对添加到三级缓存中的实例对象进行增强,直接用实例对象是行不通的。
针对这种场景spring是怎么做的呢?
答案就在AbstractAutowireCapableBeanFactory类doCreateBean方法的这段代码中:
boolean earlySingletonExposure = (mbd.isSingleton() && this.allowCircularReferences &&
isSingletonCurrentlyInCreation(beanName));
if (earlySingletonExposure) {
if (logger.isDebugEnabled()) {
logger.debug("Eagerly caching bean '" + beanName +
"' to allow for resolving potential circular references");
}
addSingletonFactory(beanName, () -> getEarlyBeanReference(beanName, mbd, bean));
}
它定义了一个匿名内部类,通过getEarlyBeanReference方法获取代理对象,其实底层是通过AbstractAutoProxyCreator类的getEarlyBeanReference生成代理对象。
3.2、多例的setter注入
这种注入方法偶然会有,具体代码如下:
@Service
@Scope(ConfigurableBeanFactory.SCOPE_PROTOTYPE)
public class ServiceA {
@Autowired
private ServiceB serviceB;
}
@Service
@Scope(ConfigurableBeanFactory.SCOPE_PROTOTYPE)
public class ServiceB {
@Autowired
private ServiceA serviceA;
}
可能你会认为这种情况spring容器启动会报错,其实是不会,为什么呢?其实在AbstractApplicationContext类的refresh方法中告诉了我们答案,它会调用finishBeanFactoryInitialization方法,该方法的作用是为了spring容器启动的时候提前初始化一些bean。该方法的内部又调用了beanFactory.preInstantiateSingletons()方法:
public ConfigurableApplicationContext run(String... args) {
StopWatch stopWatch = new StopWatch();
stopWatch.start();
ConfigurableApplicationContext context = null;
Collection exceptionReporters = new ArrayList();
this.configureHeadlessProperty();
SpringApplicationRunListeners listeners = this.getRunListeners(args);
listeners.starting();
Collection exceptionReporters;
try {
ApplicationArguments applicationArguments = new DefaultApplicationArguments(args);
ConfigurableEnvironment environment = this.prepareEnvironment(listeners, applicationArguments);
this.configureIgnoreBeanInfo(environment);
Banner printedBanner = this.printBanner(environment);
context = this.createApplicationContext();
exceptionReporters = this.getSpringFactoriesInstances(SpringBootExceptionReporter.class, new Class[]{ConfigurableApplicationContext.class}, context);
this.prepareContext(context, environment, listeners, applicationArguments, printedBanner);
this.refreshContext(context);
this.afterRefresh(context, applicationArguments);
stopWatch.stop();
if (this.logStartupInfo) {
(new StartupInfoLogger(this.mainApplicationClass)).logStarted(this.getApplicationLog(), stopWatch);
}
listeners.started(context);
this.callRunners(context, applicationArguments);
} catch (Throwable var10) {
this.handleRunFailure(context, var10, exceptionReporters, listeners);
throw new IllegalStateException(var10);
}
try {
listeners.running(context);
return context;
} catch (Throwable var9) {
this.handleRunFailure(context, var9, exceptionReporters, (SpringApplicationRunListeners)null);
throw new IllegalStateException(var9);
}
}
refresh:
public void refresh() throws BeansException, IllegalStateException {
synchronized(this.startupShutdownMonitor) {
this.prepareRefresh();
ConfigurableListableBeanFactory beanFactory = this.obtainFreshBeanFactory();
this.prepareBeanFactory(beanFactory);
try {
this.postProcessBeanFactory(beanFactory);
this.invokeBeanFactoryPostProcessors(beanFactory);
this.registerBeanPostProcessors(beanFactory);
this.initMessageSource();
this.initApplicationEventMulticaster();
this.onRefresh();
this.registerListeners();
this.finishBeanFactoryInitialization(beanFactory);
this.finishRefresh();
} catch (BeansException var9) {
if (this.logger.isWarnEnabled()) {
this.logger.warn("Exception encountered during context initialization - cancelling refresh attempt: " + var9);
}
this.destroyBeans();
this.cancelRefresh(var9);
throw var9;
} finally {
this.resetCommonCaches();
}
}
}
protected void finishBeanFactoryInitialization(ConfigurableListableBeanFactory beanFactory) {
if (beanFactory.containsBean("conversionService") && beanFactory.isTypeMatch("conversionService", ConversionService.class)) {
beanFactory.setConversionService((ConversionService)beanFactory.getBean("conversionService", ConversionService.class));
}
if (!beanFactory.hasEmbeddedValueResolver()) {
beanFactory.addEmbeddedValueResolver((strVal) -> {
return this.getEnvironment().resolvePlaceholders(strVal);
});
}
String[] weaverAwareNames = beanFactory.getBeanNamesForType(LoadTimeWeaverAware.class, false, false);
String[] var3 = weaverAwareNames;
int var4 = weaverAwareNames.length;
for(int var5 = 0; var5 < var4; ++var5) {
String weaverAwareName = var3[var5];
this.getBean(weaverAwareName);
}
beanFactory.setTempClassLoader((ClassLoader)null);
beanFactory.freezeConfiguration();
beanFactory.preInstantiateSingletons();
}
红框的地方明显能够看出:非抽象、单例 并且非懒加载的类才能被提前初始bean。而多例即SCOPE_PROTOTYPE类型的类,非单例,不会被提前初始化bean,所以程序能够正常启动。如何让它提前初始化bean呢?只需要再定义一个单例的类,在它里面注入serviceA:
@Service
public class ServiceC {
@Autowired
private ServiceA serviceA;
}
重新启动程序,执行结果:
Requested bean is currently in creation: Is there an unresolvable circular reference?
果然出现了循环依赖。
注意:这种循环依赖问题是无法解决的,因为它没有用缓存,每次都会生成一个新对象。
3.3、构造器注入
这种注入方式现在其实用的已经非常少了,但是我们还是有必要了解一下,看看如下代码:
@Service
public class ServiceA {
public ServiceA(ServiceB serviceB) {
}
}
@Service
public class ServiceB {
public ServiceB(ServiceA serviceA) {
}
}
运行结果:
Requested bean is currently in creation: Is there an unresolvable circular reference?
出现了循环依赖,为什么呢?
从图中的流程看出构造器注入没能添加到三级缓存,也没有使用缓存,所以也无法解决循环依赖问题。
3.4、单例的代理对象setter注入
这种注入方式其实也比较常用,比如平时使用:@Async注解的场景,会通过AOP自动生成代理对象。
@Service
@EnableAsync
public class ServiceA {
@Autowired
private ServiceB serviceB;
@Async
public void test1() {
System.out.println("async");
}
}
@Service
public class ServiceB {
@Autowired
private ServiceA serviceA;
}
程序启动会报错,出现了循环依赖:
org.springframework.beans.factory.BeanCurrentlyInCreationException: Error creating bean with name ‘serviceA’: Bean with name ‘serviceA’ has been injected into other beans [serviceB] in its raw version as part of a circular reference
为什么会循环依赖呢?答案就在下面这张图中:
doCreateBean
protected Object doCreateBean(String beanName, RootBeanDefinition mbd, @Nullable Object[] args) throws BeanCreationException {
BeanWrapper instanceWrapper = null;
if (mbd.isSingleton()) {
instanceWrapper = (BeanWrapper)this.factoryBeanInstanceCache.remove(beanName);
}
if (instanceWrapper == null) {
instanceWrapper = this.createBeanInstance(beanName, mbd, args);
}
Object bean = instanceWrapper.getWrappedInstance();
Class> beanType = instanceWrapper.getWrappedClass();
if (beanType != NullBean.class) {
mbd.resolvedTargetType = beanType;
}
synchronized(mbd.postProcessingLock) {
if (!mbd.postProcessed) {
try {
this.applyMergedBeanDefinitionPostProcessors(mbd, beanType, beanName);
} catch (Throwable var17) {
throw new BeanCreationException(mbd.getResourceDescription(), beanName, "Post-processing of merged bean definition failed", var17);
}
mbd.postProcessed = true;
}
}
boolean earlySingletonExposure = mbd.isSingleton() && this.allowCircularReferences && this.isSingletonCurrentlyInCreation(beanName);
if (earlySingletonExposure) {
if (this.logger.isTraceEnabled()) {
this.logger.trace("Eagerly caching bean '" + beanName + "' to allow for resolving potential circular references");
}
this.addSingletonFactory(beanName, () -> {
return this.getEarlyBeanReference(beanName, mbd, bean);
});
}
Object exposedObject = bean;
try {
this.populateBean(beanName, mbd, instanceWrapper);
exposedObject = this.initializeBean(beanName, exposedObject, mbd);
} catch (Throwable var18) {
if (var18 instanceof BeanCreationException && beanName.equals(((BeanCreationException)var18).getBeanName())) {
throw (BeanCreationException)var18;
}
throw new BeanCreationException(mbd.getResourceDescription(), beanName, "Initialization of bean failed", var18);
}
if (earlySingletonExposure) {
Object earlySingletonReference = this.getSingleton(beanName, false);
if (earlySingletonReference != null) {
if (exposedObject == bean) {
exposedObject = earlySingletonReference;
} else if (!this.allowRawInjectionDespiteWrapping && this.hasDependentBean(beanName)) {
String[] dependentBeans = this.getDependentBeans(beanName);
Set actualDependentBeans = new LinkedHashSet(dependentBeans.length);
String[] var12 = dependentBeans;
int var13 = dependentBeans.length;
for(int var14 = 0; var14 < var13; ++var14) {
String dependentBean = var12[var14];
if (!this.removeSingletonIfCreatedForTypeCheckOnly(dependentBean)) {
actualDependentBeans.add(dependentBean);
}
}
if (!actualDependentBeans.isEmpty()) {
throw new BeanCurrentlyInCreationException(beanName, "Bean with name '" + beanName + "' has been injected into other beans [" + StringUtils.collectionToCommaDelimitedString(actualDependentBeans) + "] in its raw version as part of a circular reference, but has eventually been wrapped. This means that said other beans do not use the final version of the bean. This is often the result of over-eager type matching - consider using 'getBeanNamesOfType' with the 'allowEagerInit' flag turned off, for example.");
}
}
}
}
try {
this.registerDisposableBeanIfNecessary(beanName, bean, mbd);
return exposedObject;
} catch (BeanDefinitionValidationException var16) {
throw new BeanCreationException(mbd.getResourceDescription(), beanName, "Invalid destruction signature", var16);
}
}
说白了,bean初始化完成之后,后面还有一步去检查:第二级缓存是否存在, 代理对象和原始对象是否相等。
到此发现第二级缓存存在并且代理对象和原始对象是不相等,因此抛出上面异常。
如果这时候把ServiceA改个名字,改成:ServiceC,其他的都不变。
@Service
public class ServiceB {
@Autowired
private ServiceC serviceC;
}
@Service
@EnableAsync
public class ServiceC {
@Autowired
private ServiceB serviceB;
@Async
public void test1() {
System.out.println("async");
}
}
再重新启动一下程序,成功了,没有报错了!!!
这又是为什么?
这就要从spring的bean加载顺序说起了,默认情况下,spring是按照文件完整路径递归查找的,按路径+文件名排序,排在前面的先加载。所以ServiceA比ServiceB先加载,而改了文件名称之后,ServiceB比ServiceC先加载。为什么ServiceB比ServiceC先加载就没问题呢?答案在下面这张图中
这种情况ServiceC第二级缓存是空的,不需要跟原始对象判断,所以不会抛出循环依赖。
3.5、DependsOn循环依赖
@Service
@DependsOn("serviceB")
public class ServiceA {
}
@Service
@DependsOn("serviceA")
public class ServiceB {
}
程序启动之后,执行结果:
Circular depends-on relationship between ‘serviceB’ and ‘serviceA’
这个例子中本来如果ServiceA和ServiceB都没有加@DependsOn注解是没问题的,反而加了这个注解会出现循环依赖问题。
这又是为什么?
答案在AbstractBeanFactory类的doGetBean方法的这段代码中:
初始化bean前,它会检查dependsOn的实例有没有循环依赖,如果有循环依赖则抛异常。
项目中如果出现循环依赖问题,说明是spring默认无法解决的循环依赖,要看项目的打印日志,属于哪种循环依赖。目前包含下面几种情况:
5.1、生成代理对象产生的循环依赖
生成代理对象产生的循环依赖这类循环依赖问题解决方法很多,主要有:
1、使用@Lazy注解,延迟加载
2、使用@DependsOn注解,指定加载先后关系
3、修改文件名称,改变循环依赖类的加载顺序
5.2、DependsOn循环依赖
使用@DependsOn产生的循环依赖这类循环依赖问题要找到@DependsOn注解循环依赖的地方,迫使它不循环依赖就可以解决问题。
5.3、多例循环依赖
多例循环依赖这类循环依赖问题可以通过把bean改成单例的解决。
5.3、构造器循环依赖
构造器循环依赖这类循环依赖问题可以通过使用@Lazy注解解决。