Spring代码分析一:加载与初始化

转载地址：http://www.cnblogs.com/bobzeng/articles/1877140.html

 http://www.360doc.com/content/10/1223/08/1720440_80574231.shtml

通常的Web项目都会在web.xml中加入Spring监听器，内容以下：

?

1 2 3 4 5 6 7 8

< listener >          < listener-class >org.springframework.web.context.ContextLoaderListener</ listener-class > </ listener >   < context-param >          < param-name >contextConfigLocation</ param-name >          < param-value >classpath*:applicationContext-struts.xml,classpath*:spring/applicationContext.xml</ param-value > </ context-param >

咱们的问题是，Spring是什么时候以及如何加载咱们的配置文件来初始化Bean工厂的，带着这些问题，咱们展开研究：

咱们先来看看web.xml中配置的监听器的类，来回答咱们的问题，Spring是什么时候来加载咱们的配置文件的：

org.springframework.web.context.ContextLoaderListener

它继承了javax.servlet.ServletContextListener接口。

ServletContextListener是J2EE Servlet API中的一个标准接口，

它可以监听ServletContext对象的生命周期，实际上就是监听Web应用的生命周期。

当Servlet容器启动或终止Web应用时，会触发ServletContextEvent事件，该事件由ServletContextListener来处理。

 

这里面有两个方法咱们比较感兴趣：

?

1 2 3 4 5 6 7

/**   * Create the ContextLoader to use. Can be overridden in subclasses.   * @return the new ContextLoader   */ protected  ContextLoader createContextLoader() {        return  new  ContextLoader(); }

这个方法构造一个默认的ContextLoader，ContextLoader能够理解为Spring上下文的加载器。之因此这样去定义这样一个类，是为了开发人员进行重写此方法来使用一个自定义的Spring上下文的加载器。

?

1 2 3 4 5 6 7

/**   * Initialize the root web application context.   */ public  void  contextInitialized(ServletContextEvent event) {       this .contextLoader = createContextLoader();       this .contextLoader.initWebApplicationContext(event.getServletContext()); }

这个方法很简单，仅仅只是调用了createContextLoader()构造了ContextLoader，并调用其初始化方法。

由此，咱们能够得出结论，Spring是在Web项目启动时，经过ServletContextListener机制，来加载以及初始化Spring上下文的。

 

下面，咱们好好研究一下Spring是如何加载其上下文的：

咱们先定位ContextLoader类。

看看此类的initWebApplicationContext()方法（省略了不重要的语句）

?

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39

/**   * Initialize Spring's web application context for the given servlet context,   * according to the "{@link #CONTEXT_CLASS_PARAM contextClass}" and   * "{@link #CONFIG_LOCATION_PARAM contextConfigLocation}" context-params.   * @param servletContext current servlet context   * @return the new WebApplicationContext   * @throws IllegalStateException if there is already a root application context present   * @throws BeansException if the context failed to initialize   * @see #CONTEXT_CLASS_PARAM   * @see #CONFIG_LOCATION_PARAM   */ public  WebApplicationContext initWebApplicationContext(ServletContext servletContext)          throws  IllegalStateException, BeansException {      if  (servletContext.getAttribute(WebApplicationContext.ROOT_WEB_APPLICATION_CONTEXT_ATTRIBUTE) != null ) {              throw  new  IllegalStateException(                      "Cannot initialize context because there is already a root application context present - "  +                      "check whether you have multiple ContextLoader* definitions in your web.xml!" );      }      try  {          // Determine parent for root web application context, if any.          ApplicationContext parent = loadParentContext(servletContext);            // Store context in local instance variable, to guarantee that          // it is available on ServletContext shutdown.          this .context = createWebApplicationContext(servletContext, parent);          servletContext.setAttribute(WebApplicationContext.ROOT_WEB_APPLICATION_CONTEXT_ATTRIBUTE, this .context);          currentContextPerThread.put(Thread.currentThread().getContextClassLoader(), this .context);            return  this .context;      } catch  (RuntimeException ex) {          logger.error( "Context initialization failed" , ex);          servletContext.setAttribute(WebApplicationContext.ROOT_WEB_APPLICATION_CONTEXT_ATTRIBUTE, ex);          throw  ex;      } catch  (Error err) {          logger.error( "Context initialization failed" , err);          servletContext.setAttribute(WebApplicationContext.ROOT_WEB_APPLICATION_CONTEXT_ATTRIBUTE, err);          throw  err;      } }

其中的有两句比较重要，咱们来看看：

ApplicationContext parent = loadParentContext(servletContext);

这个方法的用途主要是用来解决Spring共享环境的，即，若是咱们有多个WAR包部署在同一个服务器上，并且这些WAR都共享某一套业务逻辑层。如何共享一套业务逻辑包配置而不要每一个WAR都单独配置，这时咱们就可能须要Spring的共享环境了。

?

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17

protected  ApplicationContext loadParentContext(ServletContext servletContext) throws  BeansException {      ApplicationContext parentContext = null ;      // 从web.xml中读取父工厂的配置文件，默认为："classpath*:beanRefContext.xml"      String locatorFactorySelector = servletContext.getInitParameter(LOCATOR_FACTORY_SELECTOR_PARAM);        // 从web.xml中读取父类工厂的名称      String parentContextKey = servletContext.getInitParameter(LOCATOR_FACTORY_KEY_PARAM);        if  (parentContextKey != null ) {          // locatorFactorySelector may be null, indicating the default "classpath*:beanRefContext.xml"          BeanFactoryLocator locator = ContextSingletonBeanFactoryLocator.getInstance(locatorFactorySelector);          this .parentContextRef = locator.useBeanFactory(parentContextKey);          parentContext = (ApplicationContext) this .parentContextRef.getFactory();      }        return  parentContext; }

如今咱们引入BeanFactoryLocator，它是Spring配置文件的一个定位器，Spring官方给它的定义是用来查找，使用和释放一个BeanFactory或其子类的接口。下面咱们看看此图：

ContextSingletonBeanFactoryLocator.getInstance(locatorFactorySelector);

是根据参数locatorFactorySelector去一个单例工厂中去拿一个对应的BeanFactoryLocator，也即，若是工厂中没有对应于locatorFactorySelector的BeanFactoryLocator对象，那就返回一个新的BeanFactoryLocator实例（这里是ContextSingletonBeanFactoryLocator的实例），不然，就从工厂里取现有的BeanFactoryLocator对象。

ContextSingletonBeanFactoryLocator里维护了一个静态的Map对象instances，每次须要新增BeanFactoryLocator实例时都会更新这个Map对象，这个Map对象是以配置文件名为KEY，BeanFactoryLocator对象为值。缘由很简单，就是但愿同一个配置文件只被初始化一次。

若是没有在web.xml中定义locatorFactorySelector这个参数，父环境的配置文件默认使用:"classpath*:beanRefContext.xml"

 

this.parentContextRef = locator.useBeanFactory(parentContextKey);

此方法定义在SingletonBeanFactoryLocator类中，一样是一个单例工厂模式，判断传入的参数parentContextKey对应的BeanFactory是否有被初始化，通过上面的ContextSingletonBeanFactoryLocator.getInstance(locatorFactorySelector)指定Spring父环境配置文件，这个方法判断指定的父环境是否被初始化，若是有则返回，没有就进行初始化。看看此方法的实现：

?

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49

public  BeanFactoryReference useBeanFactory(String factoryKey) throws  BeansException {      synchronized  ( this .bfgInstancesByKey) {          BeanFactoryGroup bfg = (BeanFactoryGroup) this .bfgInstancesByKey.get( this .resourceLocation);            if  (bfg != null ) {              bfg.refCount++;          } else  {              // Create the BeanFactory but don't initialize it.              BeanFactory groupContext = createDefinition( this .resourceLocation, factoryKey);                // Record its existence now, before instantiating any singletons.              bfg = new  BeanFactoryGroup();              bfg.definition = groupContext;              bfg.refCount = 1 ;              this .bfgInstancesByKey.put( this .resourceLocation, bfg);              this .bfgInstancesByObj.put(groupContext, bfg);                // Now initialize the BeanFactory. This may cause a re-entrant invocation              // of this method, but since we've already added the BeanFactory to our              // mappings, the next time it will be found and simply have its              // reference count incremented.              try  {                  initializeDefinition(groupContext);              } catch  (BeansException ex) {                  this .bfgInstancesByKey.remove( this .resourceLocation);                  this .bfgInstancesByObj.remove(groupContext);                  throw  new  BootstrapException( "Unable to initialize group definition. "  +                      "Group resource name ["  + this .resourceLocation + "], factory key ["  + factoryKey + "]" , ex);              }          }            try  {              BeanFactory beanFactory = null ;              if  (factoryKey != null ) {                  beanFactory = (BeanFactory) bfg.definition.getBean(factoryKey, BeanFactory. class );              } else  if  (bfg.definition instanceof  ListableBeanFactory) {                  beanFactory = (BeanFactory) BeanFactoryUtils.beanOfType((ListableBeanFactory) bfg.definition, BeanFactory. class );              } else  {                  throw  new  IllegalStateException(                      "Factory key is null, and underlying factory is not a ListableBeanFactory: "  + bfg.definition);              }              return  new  CountingBeanFactoryReference(beanFactory, bfg.definition);           } catch  (BeansException ex) {               throw  new  BootstrapException( "Unable to return specified BeanFactory instance: factory key ["  +                      factoryKey + "], from group with resource name ["  + this .resourceLocation + "]" , ex);           }        } }

此方法分为两做了两件事，

第一，初始化上下文，主意这里初始化的是从web.xml配置参数里的Spring配置文件，也是上面讲loadParentContext方法里的

BeanFactoryLocator locator = ContextSingletonBeanFactoryLocator.getInstance(locatorFactorySelector);

这句指定的参数。这里初始化的是这个配置文件全部Bean。咱们指定的factoryKey对应的Bean也是其中之一。

 

第二，从已经初始化的Spring上下文环境中获取Spring父环境。

?

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16

< beans >        < bean  id = "factoryBeanId"  class = "org.springframework.context.support.ClassPathXmlApplicationContext" >            < constructor-arg >                   < list >                        < value >sharebean.xml</ value >                   </ list >            </ constructor-arg >        </ bean >       < bean  id = "factoryBeanId2"  class = "org.springframework.context.support.ClassPathXmlApplicationContext" >            < constructor-arg >                   < list >                        < value >sharebean2.xml</ value >                   </ list >            </ constructor-arg >        </ bean > </ beans >

?

1 2 3 4 5 6 7 8 9 10

<!—========================= web.xml ========================= -->  < context-param >           < param-name >locatorFactorySelector</ param-name >           < param-value >beanRefFactory.xml</ param-value >  </ context-param >    < context-param >           < param-name >parentContextKey</ param-name >           < param-value >factoryBeanId</