探究Struts2运行机制：StrutsPrepareAndExecuteFilter 源码剖析

探究Struts2运行机制：StrutsPrepareAndExecuteFilter 源码剖析

  作者：niumd

  blog：http://ari.iteye.com

 一、概述

     Struts2的核心是一个Filter，Action可以脱离web容器，那么是什么让http请求和action关联在一起的，下面我们深入源码来分析下Struts2是如何工作的。

FilterDispatcher API 写道

Deprecated. Since Struts 2.1.3, use StrutsPrepareAndExecuteFilter instead or StrutsPrepareFilter and StrutsExecuteFilter if needing using the ActionContextCleanUp filter in addition to this one

     鉴于常规情况官方推荐使用StrutsPrepareAndExecuteFilter替代FilterDispatcher，我们此文将剖析StrutsPrepareAndExecuteFilter，其在工程中作为一个Filter配置在web.xml中，配置如下：

<filter>
<filter-name>struts2</filter-name>
<filter-class>org.apache.struts2.dispatcher.ng.filter.StrutsPrepareAndExecuteFilter</filter-class>
</filter>
<filter-mapping>
<filter-name>struts2</filter-name>
<url-pattern>/*</url-pattern>
</filter-mapping>

二、源码属性方法简介

    下面我们研究下StrutsPrepareAndExecuteFilter源码，类的主要信息如下：

属性摘要
protected  List<Pattern>	excludedPatterns
protected  ExecuteOperations	execute
protected  PrepareOperations	prepare

    StrutsPrepareAndExecuteFilter与普通的Filter并无区别，方法除继承自Filter外，仅有一个回调方法，第三部分我们将按照Filter方法调用顺序，由init—>doFilter—>destroy顺序地分析源码。

方法摘要
void	destroy()             继承自Filter，用于资源释放
void	doFilter(ServletRequest req, ServletResponse res, FilterChain chain)             继承自Filter，执行方法
void	init(FilterConfig filterConfig)             继承自Filter，初始化参数
protected  void	postInit(Dispatcher dispatcher, FilterConfig filterConfig)            Callback for post initialization（一个空的方法，用于方法回调初始化）

三、源码剖析    

    1、init方法

         init是Filter第一个运行的方法，我们看下struts2的核心Filter在调用init方法初始化时做哪些工作：

 public void init(FilterConfig filterConfig) throws ServletException {
InitOperations init = new InitOperations();
try {
//封装filterConfig，其中有个主要方法getInitParameterNames将参数名字以String格式存储在List中
FilterHostConfig config = new FilterHostConfig(filterConfig);
// 初始化struts内部日志
init.initLogging(config);
//创建dispatcher ，并初始化，这部分下面我们重点分析，初始化时加载那些资源
Dispatcher dispatcher = init.initDispatcher(config);
init.initStaticContentLoader(config, dispatcher);
//初始化类属性：prepare 、execute
prepare = new PrepareOperations(filterConfig.getServletContext(), dispatcher);
execute = new ExecuteOperations(filterConfig.getServletContext(), dispatcher);
this.excludedPatterns = init.buildExcludedPatternsList(dispatcher);
//回调空的postInit方法
postInit(dispatcher, filterConfig);
} finally {
init.cleanup();
}
}

   首先看下FilterHostConfig ，源码如下：

public class FilterHostConfig implements HostConfig {
private FilterConfig config;
/**
*构造函数
*/
public FilterHostConfig(FilterConfig config) {
this.config = config;
}
/**
*  根据init-param配置的param-name获取param-value的值
*/
public String getInitParameter(String key) {
return config.getInitParameter(key);
}
/**
*  返回初始化参数名的List
*/
public Iterator<String> getInitParameterNames() {
return MakeIterator.convert(config.getInitParameterNames());
}
public ServletContext getServletContext() {
return config.getServletContext();
}
}

   只有短短的几行代码，getInitParameterNames是这个类的核心，将Filter初始化参数名称有枚举类型转为Iterator。此类的主要作为是对filterConfig 封装。

    重点来了，创建并初始化Dispatcher     

 public Dispatcher initDispatcher( HostConfig filterConfig ) {
Dispatcher dispatcher = createDispatcher(filterConfig);
dispatcher.init();
return dispatcher;
}

     创建Dispatcher，会读取 filterConfig 中的配置信息，将配置信息解析出来，封装成为一个Map，然后根绝servlet上下文和参数Map构造Dispatcher ：

private Dispatcher createDispatcher( HostConfig filterConfig ) {
Map<String, String> params = new HashMap<String, String>();
for ( Iterator e = filterConfig.getInitParameterNames(); e.hasNext(); ) {
String name = (String) e.next();
String value = filterConfig.getInitParameter(name);
params.put(name, value);
}
return new Dispatcher(filterConfig.getServletContext(), params);
}

  Dispatcher初始化，加载struts2的相关配置文件，将按照顺序逐一加载：default.properties，struts-default.xml,struts-plugin.xml,struts.xml，……

/**
*初始化过程中依次加载如下配置文件
*/
public void init() {
if (configurationManager == null) {
configurationManager = new ConfigurationManager(BeanSelectionProvider.DEFAULT_BEAN_NAME);
}
try {
//加载org/apache/struts2/default.properties
init_DefaultProperties(); // [1]
//加载struts-default.xml,struts-plugin.xml,struts.xml
init_TraditionalXmlConfigurations(); // [2]
init_LegacyStrutsProperties(); // [3]
//用户自己实现的ConfigurationProviders类
init_CustomConfigurationProviders(); // [5]
//Filter的初始化参数
init_FilterInitParameters() ; // [6]
init_AliasStandardObjects() ; // [7]
Container container = init_PreloadConfiguration();
container.inject(this);
init_CheckConfigurationReloading(container);
init_CheckWebLogicWorkaround(container);
if (!dispatcherListeners.isEmpty()) {
for (DispatcherListener l : dispatcherListeners) {
l.dispatcherInitialized(this);
}
}
} catch (Exception ex) {
if (LOG.isErrorEnabled())
LOG.error("Dispatcher initialization failed", ex);
throw new StrutsException(ex);
}
}

   初始化default.properties，具体的初始化操作在DefaultPropertiesProvider类中

  

 private void init_DefaultProperties() {
configurationManager.addConfigurationProvider(new DefaultPropertiesProvider());
}

    

   下面我们看下DefaultPropertiesProvider类源码：

public void register(ContainerBuilder builder, LocatableProperties props)
throws ConfigurationException {
Settings defaultSettings = null;
try {
defaultSettings = new PropertiesSettings("org/apache/struts2/default");
} catch (Exception e) {
throw new ConfigurationException("Could not find or error in org/apache/struts2/default.properties", e);
}
loadSettings(props, defaultSettings);
}

   其他的我们再次省略，大家可以浏览下各个初始化操作都加载了那些文件

3、doFilter方法

     doFilter是过滤器的执行方法，它拦截提交的HttpServletRequest请求，HttpServletResponse响应，作为strtus2的核心拦截器，在doFilter里面到底做了哪些工作，我们将逐行解读其源码，源码如下：

    public void doFilter(ServletRequest req, ServletResponse res, FilterChain chain) throws IOException, ServletException {
//父类向子类转：强转为http请求、响应
HttpServletRequest request = (HttpServletRequest) req;
HttpServletResponse response = (HttpServletResponse) res;
try {
//设置编码和国际化
prepare.setEncodingAndLocale(request, response);
//创建Action上下文（重点）
prepare.createActionContext(request, response);
prepare.assignDispatcherToThread();
if ( excludedPatterns != null && prepare.isUrlExcluded(request, excludedPatterns)) {
chain.doFilter(request, response);
} else {
request = prepare.wrapRequest(request);
ActionMapping mapping = prepare.findActionMapping(request, response, true);
if (mapping == null) {
boolean handled = execute.executeStaticResourceRequest(request, response);
if (!handled) {
chain.doFilter(request, response);
}
} else {
execute.executeAction(request, response, mapping);
}
}
} finally {
prepare.cleanupRequest(request);
}
}

    setEncodingAndLocale调用了dispatcher方法的prepare方法：

/**
* Sets the request encoding and locale on the response
*/
public void setEncodingAndLocale(HttpServletRequest request, HttpServletResponse response) {
dispatcher.prepare(request, response);
}

   下面我们看下prepare方法，这个方法很简单只是设置了encoding 、locale ，做的只是一些辅助的工作：

public void prepare(HttpServletRequest request, HttpServletResponse response) {
String encoding = null;
if (defaultEncoding != null) {
encoding = defaultEncoding;
}
Locale locale = null;
if (defaultLocale != null) {
locale = LocalizedTextUtil.localeFromString(defaultLocale, request.getLocale());
}
if (encoding != null) {
try {
request.setCharacterEncoding(encoding);
} catch (Exception e) {
LOG.error("Error setting character encoding to '" + encoding + "' - ignoring.", e);
}
}
if (locale != null) {
response.setLocale(locale);
}
if (paramsWorkaroundEnabled) {
request.getParameter("foo"); // simply read any parameter (existing or not) to "prime" the request
}
}

   Action上下文创建（重点）

       ActionContext是一个容器，这个容易主要存储request、session、application、parameters等相关信息.ActionContext是一个线程的本地变量，这意味着不同的action之间不会共享ActionContext，所以也不用考虑线程安全问题。其实质是一个Map，key是标示request、session、……的字符串，值是其对应的对象：

static ThreadLocal actionContext = new ThreadLocal();
Map<String, Object> context;

 
   下面我们看下如何创建action上下文的，代码如下：

/**
*创建Action上下文，初始化thread local
*/
public ActionContext createActionContext(HttpServletRequest request, HttpServletResponse response) {
ActionContext ctx;
Integer counter = 1;
Integer oldCounter = (Integer) request.getAttribute(CLEANUP_RECURSION_COUNTER);
if (oldCounter != null) {
counter = oldCounter + 1;
}
//注意此处是从ThreadLocal中获取此ActionContext变量
ActionContext oldContext = ActionContext.getContext();
if (oldContext != null) {
// detected existing context, so we are probably in a forward
ctx = new ActionContext(new HashMap<String, Object>(oldContext.getContextMap()));
} else {
ValueStack stack = dispatcher.getContainer().getInstance(ValueStackFactory.class).createValueStack();
stack.getContext().putAll(dispatcher.createContextMap(request, response, null, servletContext));
//stack.getContext()返回的是一个Map<String，Object>，根据此Map构造一个ActionContext
ctx = new ActionContext(stack.getContext());
}
request.setAttribute(CLEANUP_RECURSION_COUNTER, counter);
//将ActionContext存如ThreadLocal
ActionContext.setContext(ctx);
return ctx;
}

    上面代码中dispatcher.createContextMap，如何封装相关参数：

public Map<String,Object> createContextMap(HttpServletRequest request, HttpServletResponse response,
ActionMapping mapping, ServletContext context) {
// request map wrapping the http request objects
Map requestMap = new RequestMap(request);
// parameters map wrapping the http parameters.  ActionMapping parameters are now handled and applied separately
Map params = new HashMap(request.getParameterMap());
// session map wrapping the http session
Map session = new SessionMap(request);
// application map wrapping the ServletContext
Map application = new ApplicationMap(context);
//requestMap、params、session等Map封装成为一个上下文Map，逐个调用了map.put(Map p).
Map<String,Object> extraContext = createContextMap(requestMap, params, session, application, request, response, context);
if (mapping != null) {
extraContext.put(ServletActionContext.ACTION_MAPPING, mapping);
}
return extraContext;
}

 我们简单看下RequestMap，其他的省略。RequestMap类实现了抽象Map，故其本身是一个Map，主要方法实现：

//map的get实现
public Object get(Object key) {
return request.getAttribute(key.toString());
}
//map的put实现
public Object put(Object key, Object value) {
Object oldValue = get(key);
entries = null;
request.setAttribute(key.toString(), value);
return oldValue;
}

   下面是源码展示了如何执行Action控制器：

public void executeAction(HttpServletRequest request, HttpServletResponse response, ActionMapping mapping) throws ServletException {
dispatcher.serviceAction(request, response, servletContext, mapping);
}
public void serviceAction(HttpServletRequest request, HttpServletResponse response, ServletContext context,
ActionMapping mapping) throws ServletException {
//封装执行的上下文环境，主要讲相关信息存储入map
Map<String, Object> extraContext = createContextMap(request, response, mapping, context);
// If there was a previous value stack, then create a new copy and pass it in to be used by the new Action
ValueStack stack = (ValueStack) request.getAttribute(ServletActionContext.STRUTS_VALUESTACK_KEY);
boolean nullStack = stack == null;
if (nullStack) {
ActionContext ctx = ActionContext.getContext();
if (ctx != null) {
stack = ctx.getValueStack();
}
}
if (stack != null) {
extraContext.put(ActionContext.VALUE_STACK, valueStackFactory.createValueStack(stack));
}
String timerKey = "Handling request from Dispatcher";
try {
UtilTimerStack.push(timerKey);
//获取命名空间
String namespace = mapping.getNamespace();
//获取action配置的name属性
String name = mapping.getName();
//获取action配置的method属性
String method = mapping.getMethod();
Configuration config = configurationManager.getConfiguration();
//根据执行上下文参数，命名空间，名称等创建用户自定义Action的代理对象
ActionProxy proxy = config.getContainer().getInstance(ActionProxyFactory.class).createActionProxy(
namespace, name, method, extraContext, true, false);
request.setAttribute(ServletActionContext.STRUTS_VALUESTACK_KEY, proxy.getInvocation().getStack());
// if the ActionMapping says to go straight to a result, do it!
//执行execute方法，并转向结果
if (mapping.getResult() != null) {
Result result = mapping.getResult();
result.execute(proxy.getInvocation());
} else {
proxy.execute();
}
// If there was a previous value stack then set it back onto the request
if (!nullStack) {
request.setAttribute(ServletActionContext.STRUTS_VALUESTACK_KEY, stack);
}
} catch (ConfigurationException e) {
// WW-2874 Only log error if in devMode
if(devMode) {
String reqStr = request.getRequestURI();
if (request.getQueryString() != null) {
reqStr = reqStr + "?" + request.getQueryString();
}
LOG.error("Could not find action or result\n" + reqStr, e);
}
else {
LOG.warn("Could not find action or result", e);
}
sendError(request, response, context, HttpServletResponse.SC_NOT_FOUND, e);
} catch (Exception e) {
sendError(request, response, context, HttpServletResponse.SC_INTERNAL_SERVER_ERROR, e);
} finally {
UtilTimerStack.pop(timerKey);
}
}

   文中对如何解析Struts.xml，如何将URL与action映射匹配为分析，有需要的我后续补全，因为StrutsXmlConfigurationProvider继承XmlConfigurationProvider，并在register方法回调父类的register，有兴趣的可以深入阅读下下XmlConfigurationProvider源码：

 public void register(ContainerBuilder containerBuilder, LocatableProperties props) throws ConfigurationException {
if (servletContext != null && !containerBuilder.contains(ServletContext.class)) {
containerBuilder.factory(ServletContext.class, new Factory<ServletContext>() {
public ServletContext create(Context context) throws Exception {
return servletContext;
}
});
}
//调用父类的register，关键点所在
super.register(containerBuilder, props);
}

     struts2-core-2.2.1.jar包中struts-2.1.7.dtd对于Action的定义如下：

<!ELEMENT action (param|result|interceptor-ref|exception-mapping)*>
<!ATTLIST action
name CDATA #REQUIRED
class CDATA #IMPLIED
method CDATA #IMPLIED
converter CDATA #IMPLIED
>

    从上述DTD中可见Action元素可以含有name 、class 、method 、converter 属性。

   XmlConfigurationProvider解析struts.xml配置的Action元素：

   protected void addAction(Element actionElement, PackageConfig.Builder packageContext) throws ConfigurationException {
String name = actionElement.getAttribute("name");
String className = actionElement.getAttribute("class");
String methodName = actionElement.getAttribute("method");
Location location = DomHelper.getLocationObject(actionElement);
if (location == null) {
LOG.warn("location null for " + className);
}
//methodName should be null if it's not set
methodName = (methodName.trim().length() > 0) ? methodName.trim() : null;
// if there isnt a class name specified for an <action/> then try to
// use the default-class-ref from the <package/>
if (StringUtils.isEmpty(className)) {
// if there is a package default-class-ref use that, otherwise use action support
/* if (StringUtils.isNotEmpty(packageContext.getDefaultClassRef())) {
className = packageContext.getDefaultClassRef();
} else {
className = ActionSupport.class.getName();
}*/
} else {
if (!verifyAction(className, name, location)) {
if (LOG.isErrorEnabled())
LOG.error("Unable to verify action [#0] with class [#1], from [#2]", name, className, location.toString());
return;
}
}
Map<String, ResultConfig> results;
try {
results = buildResults(actionElement, packageContext);
} catch (ConfigurationException e) {
throw new ConfigurationException("Error building results for action " + name + " in namespace " + packageContext.getNamespace(), e, actionElement);
}
List<InterceptorMapping> interceptorList = buildInterceptorList(actionElement, packageContext);
List<ExceptionMappingConfig> exceptionMappings = buildExceptionMappings(actionElement, packageContext);
ActionConfig actionConfig = new ActionConfig.Builder(packageContext.getName(), name, className)
.methodName(methodName)
.addResultConfigs(results)
.addInterceptors(interceptorList)
.addExceptionMappings(exceptionMappings)
.addParams(XmlHelper.getParams(actionElement))
.location(location)
.build();
packageContext.addActionConfig(name, actionConfig);
if (LOG.isDebugEnabled()) {
LOG.debug("Loaded " + (StringUtils.isNotEmpty(packageContext.getNamespace()) ? (packageContext.getNamespace() + "/") : "") + name + " in '" + packageContext.getName() + "' package:" + actionConfig);
}
}

     工作中不涉及Struts2，本周工作有个2天的空档期，稍微看了下struts2的文档，写了个demo，从源码的角度研究了下运行原理，如有分析不当请指出，我后续逐步完善更正，大家共同提高。