Java内省之Introspector
在JavaBean规范中有如下描述:
大意是java默认情况下jdk使用低级的反射机制来分析Bean,为了方便其他人分析bean,java提供了一个内省类Introspector,使用Introspector的getBeanInfo方法可以获取一个封装了bean信息(包括属性和方法)的BeanInfo对象。
Introspector使用不当导致内存泄露的风险
框架几乎都使用了Introspector类来实现灵活性,但是Introspector在获取beanInfo对象时,为了提高性能使用了缓存保存beanInfo:
public static BeanInfo getBeanInfo(Class> beanClass) throws IntrospectionException { if (!ReflectUtil.isPackageAccessible(beanClass)) { return (new Introspector(beanClass, null, USE_ALL_BEANINFO)).getBeanInfo(); } ThreadGroupContext context = ThreadGroupContext.getContext(); BeanInfo beanInfo; synchronized (declaredMethodCache) { beanInfo = context.getBeanInfo(beanClass); } if (beanInfo == null) { beanInfo = new Introspector(beanClass, null, USE_ALL_BEANINFO).getBeanInfo(); synchronized (declaredMethodCache) { context.putBeanInfo(beanClass, beanInfo); } } return beanInfo; }
缓存使用ThreadGroupContext——线程组级别共享,类似与ThreadLocal。
内部使用WeakHashMap——key为弱引用来保存beanInfo,其中使用class作为key,beanInfo作为value。
同时使用WeakIdentityMap保存ThreadGroupContext对象(应该是ThreadGroupContext对象的hash值)与WeakHashMap的映射关系,也就是说不同线程组相互隔离:
final class ThreadGroupContext { //WeakIdentityMap 判断key是否重复只判断hash是否相等,不调用equals private static final WeakIdentityMapcontexts = new WeakIdentityMap () { protected ThreadGroupContext create(Object key) { return new ThreadGroupContext(); } }; /** * Returns the appropriate {@code ThreadGroupContext} for the caller, * as determined by its {@code ThreadGroup}. * * @return the application-dependent context */ static ThreadGroupContext getContext() { return contexts.get(Thread.currentThread().getThreadGroup()); } 。 。 。 }
但是beanInfo中持有class对象,因此WeakHashMap的弱引用失效,Introspector提供了清除缓存的方法flushCaches。
但有些框架在使用Introspector之后并没有清除缓存
在spring中有如下描述:
/** * Listener that flushes the JDK's {@link java.beans.Introspector JavaBeans Introspector} * cache on web app shutdown. Register this listener in your {@code web.xml} to * guarantee proper release of the web application class loader and its loaded classes. * *If the JavaBeans Introspector has been used to analyze application classes, * the system-level Introspector cache will hold a hard reference to those classes. * Consequently, those classes and the web application class loader will not be * garbage-collected on web app shutdown! This listener performs proper cleanup, * to allow for garbage collection to take effect. * *
Unfortunately, the only way to clean up the Introspector is to flush * the entire cache, as there is no way to specifically determine the * application's classes referenced there. This will remove cached * introspection results for all other applications in the server too. * *
Note that this listener is not necessary when using Spring's beans * infrastructure within the application, as Spring's own introspection results * cache will immediately flush an analyzed class from the JavaBeans Introspector * cache and only hold a cache within the application's own ClassLoader. * * Although Spring itself does not create JDK Introspector leaks, note that this * listener should nevertheless be used in scenarios where the Spring framework classes * themselves reside in a 'common' ClassLoader (such as the system ClassLoader). * In such a scenario, this listener will properly clean up Spring's introspection cache. * *
Application classes hardly ever need to use the JavaBeans Introspector * directly, so are normally not the cause of Introspector resource leaks. * Rather, many libraries and frameworks do not clean up the Introspector: * e.g. Struts and Quartz. * *
Note that a single such Introspector leak will cause the entire web * app class loader to not get garbage collected! This has the consequence that * you will see all the application's static class resources (like singletons) * around after web app shutdown, which is not the fault of those classes! * *
This listener should be registered as the first one in {@code web.xml}, * before any application listeners such as Spring's ContextLoaderListener. * This allows the listener to take full effect at the right time of the lifecycle.
大意是在web应用中使用Introspector分析bean,当web应用停止时(这里应该指的是正常销毁,而非杀死进程暴力销毁),由于Introspector持有被分析bean的强引用,导致bean以及加载bean的classload无法被gc,造成内存泄露。
个人猜想,如果web应用停止之后,main方法运行结束,jvm退出应该不存在内存泄露的情况。但是,当web服务销毁之后main方法还在执行,那么就出现内存泄露。例如在一个tomcat中部署多个应用,在tomcat的manager App 页面关闭应用就会导致内存泄露。
大部分框架在创建线程池的时候都继承parentThreadGroup,因此即使使用WeakIdentityMap保存ThreadGroup对象的软引用与WeakHashMap的映射关系,但其他未关闭的web应用仍然持有ThreadGroup的强引用,因此WeakIdentityMap中的beanInfo缓存不会被回收——内存泄露。
IntrospectorCleanupListener
为了解决其他框架如:
Struts和Quartz(大部分博客均指出这两个框架使用Introspector后没有flushCaches,但我没有考证),一心为我们考虑的spring提供了解决方案 ——IntrospectorCleanupListener:
public class IntrospectorCleanupListener implements ServletContextListener { @Override public void contextInitialized(ServletContextEvent event) { CachedIntrospectionResults.acceptClassLoader(Thread.currentThread().getContextClassLoader()); } @Override public void contextDestroyed(ServletContextEvent event) { CachedIntrospectionResults.clearClassLoader(Thread.currentThread().getContextClassLoader()); Introspector.flushCaches(); } }
IntrospectorCleanupListener是servletContext的监听器,在servletContext销毁时,会执行contextDestroyed方法,调用Introspector.flushCaches(),防止内存泄露。
spring同时说明spring框架没有使用Introspector的缓存,而是使用Introspector分析bean之后,随即清理了Introspector缓存,并使用自己的缓存逻辑进行缓存,应该就是
CachedIntrospectionResults.acceptClassLoader(Thread.currentThread().getContextClassLoader())
这行代码实现——未考证,因此spring声明在只使用spring框架时不需要考虑introspector导致内存泄露的问题。
但个人认为,如果某个框架在创建自己的线程池时,传入了新的ThreadGroup对象,那么IntrospectorCleanupListener 可能也无法工作。
Java内省Introspector应用
IntroSpecor介绍
内省(IntroSpector)是Java语言对JavaBean 类属性、事件的一种缺省处理方法。
例如类A中有属性name, 那我们可以通过getName,setName 来得到其值或者设置新的值。
通过getName/setName 来访问name属性,这就是默认的规则。
Java中提供了一套API 用来访问某个属性的getter/setter方法,通过这些API 可以使你不需要了解这个规则,这些API存放于包java.beans 中。
Class Diagram
一般的做法是通过类Introspector的getBeanInfo方法获取某个对象的BeanInfo 信息,然后通过BeanInfo来获取属性的描述器(PropertyDescriptor),通过这个属性描述器就可以获取某个属性对应的getter/setter方法,然后我们就可以通过反射机制来调用这些方法。
我们又通常把javabean的实例对象称之为值对象(Value Object),因为这些bean中通常只有一些信息字段和存储方法,没有功能性方法。
一个JavaBean类可以不当JavaBean用,而当成普通类用。JavaBean实际就是一种规范,当一个类满足这个规范,这个类就能被其它特定的类调用。一个类被当作javaBean使用时,JavaBean的属性是根据方法名推断出来的,它根本看不到java类内部的成员变量。去掉set前缀,然后取剩余部分,如果剩余部分的第二个字母是小写的,则把剩余部分的首字母改成小的。
除了反射用到的类需要引入外,内省需要引入的类如下所示,它们都属于java.beans包中的类,自己写程序的时候也不能忘了引入相应的包或者类。
简单示例1
下面代码片断是设置某个JavaBean类某个属性的关键代码:
package com.jasun.test; import java.beans.BeanInfo; import java.beans.IntrospectionException; import java.beans.Introspector; import java.beans.PropertyDescriptor; import java.lang.reflect.InvocationTargetException; import java.lang.reflect.Method; import org.apache.commons.beanutils.BeanUtils; publicclass IntrospectorTest { publi static void main(String[] args) throws IllegalArgumentException, IntrospectionException, IllegalAccessException, InvocationTargetException, NoSuchMethodException { UserInfo userInfo=new UserInfo("zhangsan", "123456"); String propertyName="userName"; Object retVal=getProperty(userInfo, propertyName); System.out.println("retVal="+retVal); //retVal=zhangsan Object value="abc"; setProperty(userInfo, propertyName, value); retVal=getProperty(userInfo, propertyName); System.out.println("retVal="+retVal); //retVal=abc //使用BeanUtils工具包操作JavaBean String userName=BeanUtils.getProperty(userInfo, propertyName); System.out.println("userName="+userName); BeanUtils.setProperty(userInfo, propertyName, "linjiqin"); userName=BeanUtils.getProperty(userInfo, propertyName); System.out.println("userName="+userName); } /** * 设置属性 * * @param clazz 对象名 * @param propertyName 属性名 * @param value 属性值 */ private static void setProperty(Object clazz, String propertyName, Object value) throws IntrospectionException,IllegalAccessException, InvocationTargetException{ //方法一 /*PropertyDescriptor pd=new PropertyDescriptor(propertyName, clazz.getClass()); Method methodSet=pd.getWriteMethod(); methodSet.invoke(clazz, value);*/ //方法二 BeanInfo beanInfo=Introspector.getBeanInfo(clazz.getClass()); PropertyDescriptor[] pds=beanInfo.getPropertyDescriptors(); for(PropertyDescriptor pd:pds){ if(propertyName.equals(pd.getName())){ Method methodSet=pd.getWriteMethod(); methodSet.invoke(clazz, value); break; } } } /** * 获取属性 * * @param clazz 对象名 * @param propertyName 属性名 * @return * @throws IntrospectionException * @throws InvocationTargetException * @throws IllegalAccessException * @throws IllegalArgumentException */ private static Object getProperty(Object clazz, String propertyName) throws IntrospectionException, IllegalArgumentException, IllegalAccessException, InvocationTargetException{ //方法一 /*PropertyDescriptor pd=new PropertyDescriptor(propertyName, clazz.getClass()); Method methodGet=pd.getReadMethod(); return methodGet.invoke(clazz);*/ //方法二 Object retVal=null; BeanInfo beanInfo=Introspector.getBeanInfo(clazz.getClass()); PropertyDescriptor[] pds=beanInfo.getPropertyDescriptors(); for(PropertyDescriptor pd:pds){ if(propertyName.equals(pd.getName())){ Method methodGet=pd.getReadMethod(); retVal=methodGet.invoke(clazz); break; } } return retVal; } }
UserInfo类
package com.ljq.test; publicclass UserInfo { private String userName; private String pwd; public UserInfo(String userName, String pwd) { super(); this.userName = userName; this.pwd = pwd; } public String getUserName() { return userName; } publicvoid setUserName(String userName) { this.userName = userName; } public String getPwd() { return pwd; } publicvoid setPwd(String pwd) { this.pwd = pwd; } }
简单示例2::(仅作参考)
package com.siyuan.jdktest; import java.beans.BeanDescriptor; import java.beans.BeanInfo; import java.beans.IntrospectionException; import java.beans.Introspector; import java.beans.MethodDescriptor; import java.beans.PropertyDescriptor; import java.lang.reflect.Method; class Person { private String name; private int age; /** * @return the age */ public int getAge() { return age; } /** * @param age the age to set */ public void setAge(int age) { this.age = age; } /** * @return the name */ public String getName() { return name; } /** * @param name the name to set */ public void setName(String name) { this.name = name; } } public class IntrospectorTest { /** * @param args * @throws IntrospectionException */ public static void main(String[] args) throws IntrospectionException { // TODO Auto-generated method stub BeanInfo beanInfo = Introspector.getBeanInfo(Person.class); System.out.println("BeanDescriptor==========================================="); BeanDescriptor beanDesc = beanInfo.getBeanDescriptor(); Class cls = beanDesc.getBeanClass(); System.out.println(cls.getName()); System.out.println("MethodDescriptor==========================================="); MethodDescriptor[] methodDescs = beanInfo.getMethodDescriptors(); for (int i = 0; i < methodDescs.length; i++) { Method method = methodDescs[i].getMethod(); System.out.println(method.getName()); } System.out.println("PropertyDescriptor==========================================="); PropertyDescriptor[] propDescs = beanInfo.getPropertyDescriptors(); for (int i = 0; i < propDescs.length; i++) { Method methodR = propDescs[i].getReadMethod(); if (methodR != null) { System.out.println(methodR.getName()); } Method methodW = propDescs[i].getWriteMethod(); if (methodW != null) { System.out.println(methodW.getName()); } } } }
运行结果
BeanDescriptor===========================================
com.siyuan.jdktest.Person
MethodDescriptor===========================================
hashCode
setAge
equals
wait
wait
notify
getClass
toString
getAge
notifyAll
setName
wait
getName
PropertyDescriptor===========================================
getAge
setAge
getClass
getName
setName
以上为个人经验,希望能给大家一个参考,也希望大家多多支持脚本之家。