Struts2架构图
ActionMapper取得了ActionMapping后，在Dispatcher的serviceAction方法里创建ActionProxy，ActionProxy创建ActionInvocation，然后ActionInvocation调用Interceptors，执行Action本身，创建Result并返回，当然，如果要在返回之前做些什么，可以实现PreResultListener。
Struts2部分类介绍
这部分从Struts2参考文档中翻译就可以了。
ActionMapper
ActionMapper其实是HttpServletRequest和Action调用请求的一个映射，它屏蔽了Action对于Request等java Servlet类的依赖。Struts2中它的默认实现类是DefaultActionMapper，ActionMapper很大的用处可以根据自己的需要来设计url格式，它自己也有Restful的实现，具体可以参考文档的docs\actionmapper.html。
ActionProxy&ActionInvocation
Action的一个代理，由ActionProxyFactory创建，它本身不包括Action实例，默认实现DefaultActionProxy是由ActionInvocation持有Action实例。ActionProxy作用是如何取得Action，无论是本地还是远程。而ActionInvocation的作用是如何执行Action，拦截器的功能就是在ActionInvocation中实现的。
ConfigurationProvider&Configuration
ConfigurationProvider就是Struts2中配置文件的解析器，Struts2中的配置文件主要是尤其实现类XmlConfigurationProvider及其子类StrutsXmlConfigurationProvider来解析，
Struts2请求流程
1、客户端发送请求
2、请求先通过ActionContextCleanUp-->FilterDispatcher
3、FilterDispatcher通过ActionMapper来决定这个Request需要调用哪个Action
4、如果ActionMapper决定调用某个Action，FilterDispatcher把请求的处理交给ActionProxy，这儿已经转到它的Delegate--Dispatcher来执行
5、ActionProxy根据ActionMapping和ConfigurationManager找到需要调用的Action类
6、ActionProxy创建一个ActionInvocation的实例
7、ActionInvocation调用真正的Action，当然这涉及到相关拦截器的调用
8、Action执行完毕，ActionInvocation创建Result并返回，当然，如果要在返回之前做些什么，可以实现PreResultListener。添加PreResultListener可以在Interceptor中实现，不知道其它还有什么方式？
Struts2(2.1.2)部分源码阅读
从org.apache.struts2.dispatcher.FilterDispatcher开始
//创建Dispatcher，此类是一个Delegate，它是真正完成根据url解析，读取对应Action的地方
public void init(FilterConfig filterConfig) throws ServletException {
try {
this.filterConfig = filterConfig;
initLogging();
dispatcher = createDispatcher(filterConfig);
dispatcher.init();
dispatcher.getContainer().inject(this);
//读取初始参数pakages，调用parse()，解析成类似/org/apache/struts2/static,/template的数组
String param = filterConfig.getInitParameter("packages");
String packages = "org.apache.struts2.static template org.apache.struts2.interceptor.debugging";
if (param != null) {
packages = param + " " + packages;
}
this.pathPrefixes = parse(packages);
} finally {
ActionContext.setContext(null);
}
}
顺着流程我们看Disptcher的init方法。init方法里就是初始读取一些配置文件等，先看init_DefaultProperties，主要是读取properties配置文件。
private void init_DefaultProperties() {
configurationManager.addConfigurationProvider(new DefaultPropertiesProvider());
}
打开DefaultPropertiesProvider
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);
}
//PropertiesSettings
//读取org/apache/struts2/default.properties的配置信息，如果项目中需要覆盖，可以在classpath里的struts.properties里覆写
public PropertiesSettings(String name) {
URL settingsUrl = ClassLoaderUtils.getResource(name + ".properties", getClass());
if (settingsUrl == null) {
LOG.debug(name + ".properties missing");
settings = new LocatableProperties();
return;
}
settings = new LocatableProperties(new LocationImpl(null, settingsUrl.toString()));
// Load settings
InputStream in = null;
try {
in = settingsUrl.openStream();
settings.load(in);
} catch (IOException e) {
throw new StrutsException("Could not load " + name + ".properties:" + e, e);
} finally {
if(in != null) {
try {
in.close();
} catch(IOException io) {
LOG.warn("Unable to close input stream", io);
}
}
}
}
再来看init_TraditionalXmlConfigurations方法，这个是读取struts-default.xml和Struts.xml的方法。
private void init_TraditionalXmlConfigurations() {
//首先读取web.xml中的config初始参数值
//如果没有配置就使用默认的"struts-default.xml,struts-plugin.xml,struts.xml"，
//这儿就可以看出为什么默认的配置文件必须取名为这三个名称了
//如果不想使用默认的名称，直接在web.xml中配置config初始参数即可
String configPaths = initParams.get("config");
if (configPaths == null) {
configPaths = DEFAULT_CONFIGURATION_PATHS;
}
String[] files = configPaths.split("\\s*[,]\\s*");
//依次解析配置文件，xwork.xml单独解析
for (String file : files) {
if (file.endsWith(".xml")) {
if ("xwork.xml".equals(file)) {
configurationManager.addConfigurationProvider(new XmlConfigurationProvider(file, false));
} else {
configurationManager.addConfigurationProvider(new StrutsXmlConfigurationProvider(file, false, servletContext));
}
} else {
throw new IllegalArgumentException("Invalid configuration file name");
}
}
}
对于其它配置文件只用StrutsXmlConfigurationProvider，此类继承XmlConfigurationProvider，而XmlConfigurationProvider又实现ConfigurationProvider接口。类XmlConfigurationProvider负责配置文件的读取和解析，addAction()方法负责读取<action>标签，并将数据保存在ActionConfig中；addResultTypes()方法负责将<result-type>标签转化为ResultTypeConfig对象；loadInterceptors()方法负责将<interceptor>标签转化为InterceptorConfi对象；loadInterceptorStack()方法负责将<interceptor-ref>标签转化为InterceptorStackConfig对象；loadInterceptorStacks()方法负责将<interceptor-stack>标签转化成InterceptorStackConfig对象。而上面的方法最终会被addPackage()方法调用，将所读取到的数据汇集到PackageConfig对象中。来看XmlConfigurationProvider的源代码，详细的我自己也就大体浏览了一下，各位可以自己研读。
protected PackageConfig addPackage(Element packageElement) throws ConfigurationException {
PackageConfig.Builder newPackage = buildPackageContext(packageElement);
if (newPackage.isNeedsRefresh()) {
return newPackage.build();
}
.
addResultTypes(newPackage, packageElement);
loadInterceptors(newPackage, packageElement);
loadDefaultInterceptorRef(newPackage, packageElement);
loadDefaultClassRef(newPackage, packageElement);
loadGlobalResults(newPackage, packageElement);
loadGobalExceptionMappings(newPackage, packageElement);
NodeList actionList = packageElement.getElementsByTagName("action");
for (int i = 0; i < actionList.getLength(); i++) {
Element actionElement = (Element) actionList.item(i);
addAction(actionElement, newPackage);
}
loadDefaultActionRef(newPackage, packageElement);
PackageConfig cfg = newPackage.build();
configuration.addPackageConfig(cfg.getName(), cfg);
return cfg;
}
这儿发现一个配置上的小技巧，我的xwork2.0.*是没有的，但是看源码是看到xwork2.1.*是可以的。继续看XmlConfigurationProvider的源代码：
private List loadConfigurationFiles(String fileName, Element includeElement) {
List<Document> docs = new ArrayList<Document>();
if (!includedFileNames.contains(fileName)) {
Element rootElement = doc.getDocumentElement();
NodeList children = rootElement.getChildNodes();
int childSize = children.getLength();
for (int i = 0; i < childSize; i++) {
Node childNode = children.item(i);
if (childNode instanceof Element) {
Element child = (Element) childNode;
final String nodeName = child.getNodeName();
//解析每个action配置是，对于include文件可以使用通配符*来进行配置
//如Struts.xml中可配置成<include file="actions_*.xml"/>
if (nodeName.equals("include")) {
String includeFileName = child.getAttribute("file");
if(includeFileName.indexOf('*') != -1 ) {
ClassPathFinder wildcardFinder = new ClassPathFinder();
wildcardFinder.setPattern(includeFileName);
Vector<String> wildcardMatches = wildcardFinder.findMatches();
for (String match : wildcardMatches) {
docs.addAll(loadConfigurationFiles(match, child));
}
}
else {
docs.addAll(loadConfigurationFiles(includeFileName, child));
}
}
}
}
docs.add(doc);
loadedFileUrls.add(url.toString());
}
}
return docs;
}
init_CustomConfigurationProviders方式初始自定义的Provider，配置类全名和实现ConfigurationProvider接口，用逗号隔开即可。
private void init_CustomConfigurationProviders() {
String configProvs = initParams.get("configProviders");
if (configProvs != null) {
String[] classes = configProvs.split("\\s*[,]\\s*");
for (String cname : classes) {
try {
Class cls = ClassLoaderUtils.loadClass(cname, this.getClass());
ConfigurationProvider prov = (ConfigurationProvider)cls.newInstance();
configurationManager.addConfigurationProvider(prov);
}
}
}
}
好了，现在再回到FilterDispatcher，每次发送一个Request，FilterDispatcher都会调用doFilter方法。
public void doFilter(ServletRequest req, ServletResponse res, FilterChain chain) throws IOException, ServletException {
HttpServletRequest request = (HttpServletRequest) req;
HttpServletResponse response = (HttpServletResponse) res;
ServletContext servletContext = getServletContext();
String timerKey = "FilterDispatcher_doFilter: ";
try {
ValueStack stack = dispatcher.getContainer().getInstance(ValueStackFactory.class).createValueStack();
ActionContext ctx = new ActionContext(stack.getContext());
ActionContext.setContext(ctx);
UtilTimerStack.push(timerKey);
//根据content type来使用不同的Request封装，可以参见Dispatcher的wrapRequest
request = prepareDispatcherAndWrapRequest(request, response);
ActionMapping mapping;
try {
//根据url取得对应的Action的配置信息--ActionMapping，actionMapper是通过Container的inject注入的
mapping = actionMapper.getMapping(request, dispatcher.getConfigurationManager());
} catch (Exception ex) {
log.error("error getting ActionMapping", ex);
dispatcher.sendError(request, response, servletContext, HttpServletResponse.SC_INTERNAL_SERVER_ERROR, ex);
return;
}
//如果找不到对应的action配置，则直接返回。比如你输入***.jsp等等
//这儿有个例外，就是如果path是以“/struts”开头，则到初始参数packages配置的包路径去查找对应的静态资源并输出到页面流中，当然.class文件除外。如果再没有则跳转到404
if (mapping == null) {
// there is no action in this request, should we look for a static resource?
String resourcePath = RequestUtils.getServletPath(request);
if ("".equals(resourcePath) && null != request.getPathInfo()) {
resourcePath = request.getPathInfo();
}
if (serveStatic && resourcePath.startsWith("/struts")) {
String name = resourcePath.substring("/struts".length());
findStaticResource(name, request, response);
} else {
chain.doFilter(request, response);
}
return;
}
//正式开始Action的方法了
dispatcher.serviceAction(request, response, servletContext, mapping);
} finally {
try {
ActionContextCleanUp.cleanUp(req);
} finally {
UtilTimerStack.pop(timerKey);
}
}
}
Dispatcher类的serviceAction方法：
public void serviceAction(HttpServletRequest request, HttpServletResponse response, ServletContext context,ActionMapping mapping) throws ServletException {
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);
if (stack != null) {
extraContext.put(ActionContext.VALUE_STACK, valueStackFactory.createValueStack(stack));
}
String timerKey = "Handling request from Dispatcher";
try {
UtilTimerStack.push(timerKey);
String namespace = mapping.getNamespace();
String name = mapping.getName();
String method = mapping.getMethod();
Configuration config = configurationManager.getConfiguration();
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!
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 (stack != null) {
request.setAttribute(ServletActionContext.STRUTS_VALUESTACK_KEY, stack);
}
} catch (ConfigurationException e) {
LOG.error("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);
}
}
第一句createContextMap()方法，该方法主要把Application、Session、Request的key value值拷贝到Map中，并放在HashMap<String,Object>中，可以参见createContextMap方法：
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);
Map<String,Object> extraContext = createContextMap(requestMap, params, session, application, request, response, context);
extraContext.put(ServletActionContext.ACTION_MAPPING, mapping);
return extraContext;
}
后面才是最主要的--ActionProxy，ActionInvocation。ActionProxy是Action的一个代理类，也就是说Action的调用是通过ActionProxy实现的，其实就是调用了ActionProxy.execute()方法，而该方法又调用了ActionInvocation.invoke()方法。归根到底，最后调用的是DefaultActionInvocation.invokeAction()方法。先看DefaultActionInvocation的init方法。
public void init(ActionProxy proxy) {
this.proxy = proxy;
Map contextMap = createContextMap();
// Setting this so that other classes, like object factories, can use the ActionProxy and other
// contextual information to operate
ActionContext actionContext = ActionContext.getContext();
if(actionContext != null) {
actionContext.setActionInvocation(this);
}
//创建Action，可Struts2里是每次请求都新建一个Action
createAction(contextMap);
if (pushAction) {
stack.push(action);
contextMap.put("action", action);
}
invocationContext = new ActionContext(contextMap);
invocationContext.setName(proxy.getActionName());
// get a new List so we don't get problems with the iterator if someone changes the list
List interceptorList = new ArrayList(proxy.getConfig().getInterceptors());
interceptors = interceptorList.iterator();
}
protected void createAction(Map contextMap) {
// load action
String timerKey = "actionCreate: "+proxy.getActionName();
try {
UtilTimerStack.push(timerKey);
//这儿默认建立Action是StrutsObjectFactory，实际中我使用的时候都是使用Spring创建的Action，这个时候使用的是SpringObjectFactory
action = objectFactory.buildAction(proxy.getActionName(), proxy.getNamespace(), proxy.getConfig(), contextMap);
}
..
} finally {
UtilTimerStack.pop(timerKey);
}
if (actionEventListener != null) {
action = actionEventListener.prepare(action, stack);
}
}
接下来看看DefaultActionInvocation 的invoke方法。
public String invoke() throws Exception {
String profileKey = "invoke: ";
try {
UtilTimerStack.push(profileKey);
if (executed) {
throw new IllegalStateException("Action has already executed");
}
//先执行interceptors
if (interceptors.hasNext()) {
final InterceptorMapping interceptor = (InterceptorMapping) interceptors.next();
UtilTimerStack.profile("interceptor: "+interceptor.getName(),
new UtilTimerStack.ProfilingBlock<String>() {
public String doProfiling() throws Exception {
resultCode = interceptor.getInterceptor().intercept(DefaultActionInvocation.this);
return null;
}
});
} else {
//interceptor执行完了之后执行action
resultCode = invokeActionOnly();
}
// this is needed because the result will be executed, then control will return to the Interceptor, which will
// return above and flow through again
if (!executed) {
//在Result返回之前调用preResultListeners
if (preResultListeners != null) {
for (Iterator iterator = preResultListeners.iterator();
iterator.hasNext();) {
PreResultListener listener = (PreResultListener) iterator.next();
String _profileKey="preResultListener: ";
try {
UtilTimerStack.push(_profileKey);
listener.beforeResult(this, resultCode);
}
finally {
UtilTimerStack.pop(_profileKey);
}
}
}
// now execute the result, if we're supposed to
if (proxy.getExecuteResult()) {
executeResult();
}
executed = true;
}
return resultCode;
}
finally {
UtilTimerStack.pop(profileKey);
}
}
看程序中的if(interceptors.hasNext())语句，当然，interceptors里存储的是interceptorMapping列表（它包括一个Interceptor和一个name），所有的截拦器必须实现Interceptor的intercept方法，而该方法的参数恰恰又是ActionInvocation，在intercept方法中还是调用invocation.invoke()，从而实现了一个Interceptor链的调用。当所有的Interceptor执行完，最后调用invokeActionOnly方法来执行Action相应的方法。
protected String invokeAction(Object action, ActionConfig actionConfig) throws Exception {
String methodName = proxy.getMethod();
String timerKey = "invokeAction: "+proxy.getActionName();
try {
UtilTimerStack.push(timerKey);
boolean methodCalled = false;
Object methodResult = null;
Method method = null;
try {
//获得需要执行的方法
method = getAction().getClass().getMethod(methodName, new Class[0]);
} catch (NoSuchMethodException e) {
//如果没有对应的方法，则使用do+Xxxx来再次获得方法
try {
String altMethodName = "do" + methodName.substring(0, 1).toUpperCase() + methodName.substring(1);
method = getAction().getClass().getMethod(altMethodName, new Class[0]);
} catch (NoSuchMethodException e1) {
// well, give the unknown handler a shot
if (unknownHandler != null) {
try {
methodResult = unknownHandler.handleUnknownActionMethod(action, methodName);
methodCalled = true;
} catch (NoSuchMethodException e2) {
// throw the original one
throw e;
}
} else {
throw e;
}
}
}
if (!methodCalled) {
methodResult = method.invoke(action, new Object[0]);
}
//根据不同的Result类型返回不同值
//如输出流Result
if (methodResult instanceof Result) {
this.explicitResult = (Result) methodResult;
return null;
} else {
return (String) methodResult;
}
} catch (NoSuchMethodException e) {
throw new IllegalArgumentException("The " + methodName + "() is not defined in action " + getAction().getClass() + "");
} catch (InvocationTargetException e) {
// We try to return the source exception.
Throwable t = e.getTargetException();
if (actionEventListener != null) {
String result = actionEventListener.handleException(t, getStack());
if (result != null) {
return result;
}
}
if (t instanceof Exception) {
throw(Exception) t;
} else {
throw e;
}
} finally {
UtilTimerStack.pop(timerKey);
}
}
好了，action执行完了，还要根据ResultConfig返回到view，也就是在invoke方法中调用executeResult方法。
private void executeResult() throws Exception {
//根据ResultConfig创建Result
result = createResult();
String timerKey = "executeResult: "+getResultCode();
try {
UtilTimerStack.push(timerKey);
if (result != null) {
//这儿正式执行:)
//可以参考Result的实现，如用了比较多的ServletDispatcherResult,ServletActionRedirectResult,ServletRedirectResult
result.execute(this);
} else if (resultCode != null && !Action.NONE.equals(resultCode)) {
throw new ConfigurationException("No result defined for action " + getAction().getClass().getName()
+ " and result " + getResultCode(), proxy.getConfig());
} else {
if (LOG.isDebugEnabled()) {
LOG.debug("No result returned for action "+getAction().getClass().getName()+" at "+proxy.getConfig().getLocation());
}
}
} finally {
UtilTimerStack.pop(timerKey);
}
}
public Result createResult() throws Exception {
if (explicitResult != null) {
Result ret = explicitResult;
explicitResult = null;;
return ret;
}
ActionConfig config = proxy.getConfig();
Map results = config.getResults();
ResultConfig resultConfig = null;
synchronized (config) {
try {
//根据result名称获得ResultConfig，resultCode就是result的name
resultConfig = (ResultConfig) results.get(resultCode);
} catch (NullPointerException e) {
}
if (resultConfig == null) {
//如果找不到对应name的ResultConfig，则使用name为*的Result
resultConfig = (ResultConfig) results.get("*");
}
}
if (resultConfig != null) {
try {
//参照StrutsObjectFactory的代码
Result result = objectFactory.buildResult(resultConfig, invocationContext.getContextMap());
return result;
} catch (Exception e) {
LOG.error("There was an exception while instantiating the result of type " + resultConfig.getClassName(), e);
throw new XWorkException(e, resultConfig);
}
} else if (resultCode != null && !Action.NONE.equals(resultCode) && unknownHandler != null) {
return unknownHandler.handleUnknownResult(invocationContext, proxy.getActionName(), proxy.getConfig(), resultCode);
}
return null;
}
//StrutsObjectFactory
public Result buildResult(ResultConfig resultConfig, Map extraContext) throws Exception {
String resultClassName = resultConfig.getClassName();
if (resultClassName == null)
return null;
//创建Result，因为Result是有状态的，所以每次请求都新建一个
Object result = buildBean(resultClassName, extraContext);
//这句很重要，后面将会谈到，reflectionProvider参见OgnlReflectionProvider；
//resultConfig.getParams()就是result配置文件里所配置的参数<param></param>
//setProperties方法最终调用的是Ognl类的setValue方法
//这句其实就是把param名值设置到根对象result上
reflectionProvider.setProperties(resultConfig.getParams(), result, extraContext);
if (result instanceof Result)
return (Result) result;
throw new ConfigurationException(result.getClass().getName() + " does not implement Result.");
}
最后补充一下，Struts2的查找值和设置值都是使用Ognl来实现的。关于Ognl的介绍可以到其官方网站查看http://www.ognl.org/，我在网上也找到另外一篇http://www.javaeye.com/topic/254684和http://www.javaeye.com/topic/223612。完了来看下面这段小测试程序（其它的Ognl的测试可以自己添加）。
public class TestOgnl {
private User user;
private Map context;
@Before
public void setUp() throws Exception {
}
@Test
public void ognlGetValue() throws Exception {
reset();
Assert.assertEquals("myyate", Ognl.getValue("name", user));
Assert.assertEquals("cares", Ognl.getValue("dept.name", user));
Assert.assertEquals("myyate", Ognl.getValue("name", context, user));
Assert.assertEquals("contextmap", Ognl.getValue("#name", context, user));
Assert.assertEquals("parker", Ognl.getValue("#pen", context, user));
}
@Test
public void ognlSetValue() throws Exception {
reset();
Ognl.setValue("name", user, "myyateC");
Assert.assertEquals("myyateC", Ognl.getValue("name", user));
Ognl.setValue("dept.name", user, "caresC");
Assert.assertEquals("caresC", Ognl.getValue("dept.name", user));
Assert.assertEquals("contextmap", Ognl.getValue("#name", context, user));
Ognl.setValue("#name", context, user, "contextmapC");
Assert.assertEquals("contextmapC", Ognl.getValue("#name", context, user));
Assert.assertEquals("parker", Ognl.getValue("#pen", context, user));
Ognl.setValue("#name", context, user, "parkerC");
Assert.assertEquals("parkerC", Ognl.getValue("#name", context, user));
}
public static void main(String[] args) throws Exception {
JUnitCore.runClasses(TestOgnl.class);
}
private void reset() {
user = new User("myyate", new Dept("cares"));
context = new OgnlContext();
context.put("pen", "parker");
context.put("name", "contextmap");
}
}
class User {
public User(String name, Dept dept) {
this.name = name;
this.dept = dept;
}
String name;
private Dept dept;
public Dept getDept() {
return dept;
}
public String getName() {
return name;
}
public void setDept(Dept dept) {
this.dept = dept;
}
public void setName(String name) {
this.name = name;
}
}
class Dept {
public Dept(String name) {
this.name = name;
}
private String name;
public String getName() {
return name;
}
public void setName(String name) {
this.name = name;
}
}
这样，一个Struts2的请求流程基本上就结束了。其实我觉得做项目把Struts2参考文档看两遍就可以了，呵呵！（写博客比看代码还累）
扫一扫
3560
被折叠的 条评论
为什么被折叠?
到【灌水乐园】发言
