先说一下什么是循环依赖,Spring在初始化A的时候需要注入B,而初始化B的时候需要注入A,在Spring启动后这2个Bean都要被初始化完成
Spring的循环依赖有两种场景
构造器的循环依赖,可以在构造函数中使用@Lazy注解延迟加载。在注入依赖时,先注入代理对象,当首次使用时再创建对象完成注入
属性的循环依赖主要是通过3个map来解决的
@Component
public class ConstructorA {
private ConstructorB constructorB;
@Autowired
public ConstructorA(ConstructorB constructorB) {
this.constructorB = constructorB;
}
}
@Component
public class ConstructorB {
private ConstructorA constructorA;
@Autowired
public ConstructorB(ConstructorA constructorA) {
this.constructorA = constructorA;
}
}
@Configuration
@ComponentScan("com.javashitang.dependency.constructor")
public class ConstructorConfig {
}
public class ConstructorMain {
public static void main(String[] args) {
AnnotationConfigApplicationContext context =
new AnnotationConfigApplicationContext(ConstructorConfig.class);
System.out.println(context.getBean(ConstructorA.class));
System.out.println(context.getBean(ConstructorB.class));
}
}
运行ConstructorMain的main方法的时候会在第一行就报异常,说明Spring没办法初始化所有的Bean,即上面这种形式的循环依赖Spring无法解决。
我们可以在ConstructorA或者ConstructorB构造函数的参数上加上@Lazy注解就可以解决
@Autowired
public ConstructorB(@Lazy ConstructorA constructorA) {
this.constructorA = constructorA;
}
因为我们主要关注属性的循环依赖,构造器的循环依赖就不做过多分析了
先演示一下什么是属性的循环依赖
@Component
public class FieldA {
@Autowired
private FieldB fieldB;
}
@Component
public class FieldB {
@Autowired
private FieldA fieldA;
}
@Configuration
@ComponentScan("com.javashitang.dependency.field")
public class FieldConfig {
}
public class FieldMain {
public static void main(String[] args) {
AnnotationConfigApplicationContext context =
new AnnotationConfigApplicationContext(FieldConfig.class);
// com.javashitang.dependency.field.FieldA@3aa9e816
System.out.println(context.getBean(FieldA.class));
// com.javashitang.dependency.field.FieldB@17d99928
System.out.println(context.getBean(FieldB.class));
}
}
Spring容器正常启动,能获取到FieldA和FieldB这2个Bean
属性的循环依赖在面试中还是经常被问到的。总体来说也不复杂,但是涉及到Spring Bean的初始化过程,所以感觉比较复杂,我写个demo演示一下整个过程
Spring的Bean的初始化过程其实比较复杂,为了方便理解Demo,我就把Spring Bean的初始化过程分为2部分
bean初始化过程完毕,则bean就能被正常创建出来了
下面开始写Demo,ObjectFactory接口用来生产Bean,和Spring中定义的接口一样
public interface ObjectFactory<T> {
T getObject();
}
public class DependencyDemo {
// 初始化完毕的Bean
private final Map<String, Object> singletonObjects =
new ConcurrentHashMap<>(256);
// 正在初始化的Bean对应的工厂,此时对象已经被实例化
private final Map<String, ObjectFactory<?>> singletonFactories =
new HashMap<>(16);
// 存放正在初始化的Bean,对象还没有被实例化之前就放进来了
private final Set<String> singletonsCurrentlyInCreation =
Collections.newSetFromMap(new ConcurrentHashMap<>(16));
public <T> T getBean(Class<T> beanClass) throws Exception {
// 类名为Bean的名字
String beanName = beanClass.getSimpleName();
// 已经初始化好了,或者正在初始化
Object initObj = getSingleton(beanName, true);
if (initObj != null) {
return (T) initObj;
}
// bean正在被初始化
singletonsCurrentlyInCreation.add(beanName);
// 实例化bean
Object object = beanClass.getDeclaredConstructor().newInstance();
singletonFactories.put(beanName, () -> {
return object;
});
// 开始初始化bean,即填充属性
Field[] fields = object.getClass().getDeclaredFields();
for (Field field : fields) {
field.setAccessible(true);
// 获取需要注入字段的class
Class<?> fieldClass = field.getType();
field.set(object, getBean(fieldClass));
}
// 初始化完毕
singletonObjects.put(beanName, object);
singletonsCurrentlyInCreation.remove(beanName);
return (T) object;
}
/**
* allowEarlyReference参数的含义是Spring是否允许循环依赖,默认为true
* 所以当allowEarlyReference设置为false的时候,当项目存在循环依赖,会启动失败
*/
public Object getSingleton(String beanName, boolean allowEarlyReference) {
Object singletonObject = this.singletonObjects.get(beanName);
if (singletonObject == null
&& isSingletonCurrentlyInCreation(beanName)) {
synchronized (this.singletonObjects) {
if (singletonObject == null && allowEarlyReference) {
ObjectFactory<?> singletonFactory =
this.singletonFactories.get(beanName);
if (singletonFactory != null) {
singletonObject = singletonFactory.getObject();
}
}
}
}
return singletonObject;
}
/**
* 判断bean是否正在被初始化
*/
public boolean isSingletonCurrentlyInCreation(String beanName) {
return this.singletonsCurrentlyInCreation.contains(beanName);
}
}
测试一波
public static void main(String[] args) throws Exception {
DependencyDemo dependencyDemo = new DependencyDemo();
// 假装扫描出来的对象
Class[] classes = {A.class, B.class};
// 假装项目初始化所有bean
for (Class aClass : classes) {
dependencyDemo.getBean(aClass);
}
// true
System.out.println(
dependencyDemo.getBean(B.class).getA() == dependencyDemo.getBean(A.class));
// true
System.out.println(
dependencyDemo.getBean(A.class).getB() == dependencyDemo.getBean(B.class));
}
是不是很简单?我们只用了2个map就搞定了Spring的循环依赖
2个Map就能搞定循环依赖,那为什么Spring要用3个Map呢?
原因其实也很简单,当我们从singletonFactories中根据BeanName获取相应的ObjectFactory,然后调用getObject()这个方法返回对应的Bean。在我们的例子中
ObjectFactory的实现很简单哈,就是将实例化好的对象直接返回,但是在Spring中就没有这么简单了,执行过程比较复杂,为了避免每次拿到ObjectFactory然后调用getObject(),我们直接把ObjectFactory创建的对象缓存起来不就行了,这样就能提高效率了
比如A依赖B和C,B和C又依赖A,如果不做缓存那么初始化B和C都会调用A对应的ObjectFactory的getObject()方法。如果做缓存只需要B或者C调用一次即可。
知道了思路,我们把上面的代码改一波,加个缓存。
public class DependencyDemo {
// 初始化完毕的Bean
private final Map<String, Object> singletonObjects =
new ConcurrentHashMap<>(256);
// 正在初始化的Bean对应的工厂,此时对象已经被实例化
private final Map<String, ObjectFactory<?>> singletonFactories =
new HashMap<>(16);
// 缓存Bean对应的工厂生产好的Bean
private final Map<String, Object> earlySingletonObjects =
new HashMap<>(16);
// 存放正在初始化的Bean,对象还没有被实例化之前就放进来了
private final Set<String> singletonsCurrentlyInCreation =
Collections.newSetFromMap(new ConcurrentHashMap<>(16));
public <T> T getBean(Class<T> beanClass) throws Exception {
// 类名为Bean的名字
String beanName = beanClass.getSimpleName();
// 已经初始化好了,或者正在初始化
Object initObj = getSingleton(beanName, true);
if (initObj != null) {
return (T) initObj;
}
// bean正在被初始化
singletonsCurrentlyInCreation.add(beanName);
// 实例化bean
Object object = beanClass.getDeclaredConstructor().newInstance();
singletonFactories.put(beanName, () -> {
return object;
});
// 开始初始化bean,即填充属性
Field[] fields = object.getClass().getDeclaredFields();
for (Field field : fields) {
field.setAccessible(true);
// 获取需要注入字段的class
Class<?> fieldClass = field.getType();
field.set(object, getBean(fieldClass));
}
singletonObjects.put(beanName, object);
singletonsCurrentlyInCreation.remove(beanName);
earlySingletonObjects.remove(beanName);
return (T) object;
}
/**
* allowEarlyReference参数的含义是Spring是否允许循环依赖,默认为true
*/
public Object getSingleton(String beanName, boolean allowEarlyReference) {
Object singletonObject = this.singletonObjects.get(beanName);
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;
}
}
我们写的getSingleton的实现和org.springframework.beans.factory.support.DefaultSingletonBeanRegistry#getSingleton(java.lang.String, boolean)的实现一模一样,这个方法几乎所有分析Spring循环依赖的文章都会提到,这次你明白工作原理是什么了把
[1]https://mp.weixin.qq.com/s/gBr3UfC1HRcw4U-ZMmtRaQ
[2]https://mp.weixin.qq.com/s/5mwkgJB7GyLdKDgzijyvXw
比较详细
[1]https://zhuanlan.zhihu.com/p/84267654
[2]https://juejin.im/post/5c98a7b4f265da60ee12e9b2