javaweb技能之filter的执⾏顺序解密

1.引⾔

我们在编写javaweb程序的时候,时常会⽤filter这个组件,它能将我们⼀些通⽤逻辑抽取出来,在servlet执⾏业务逻辑之前运⾏,
达到简化代码和复⽤的⽬的.⽐如最常⽤的场景全站编码和登录验证功能.
servlet3.0以前我们只能通过web.xml的⽅式配置filter,并且多个filter的执⾏顺序是根据你web.xml中书写顺序来决定的.
servlet3.0以后,提供了注解的⽅式注⼊filter,只需要在filter类上加上@WebFilter()注解即可,⼤⼤的简化了开发复杂度.

2.抛出问题

注解的⽅式书写的filter的执⾏顺序⼜是如何的呢?
⽹上的很多资料都说是根据filter的类名来决定,也有说是根据filter的注解的name属性值的字⺟顺序来决定的.
对不对呢?

3.验证问题

我们创建了三个filter 来验证此问题
filter1号

package com.jk1123.web.filter.demo01;
import javax.servlet.*;
import javax.servlet.annotation.WebFilter;
import java.io.IOException;
@WebFilter("/*")
public class OrderFilter1 implements Filter {
 public void destroy() {
 }
 public void doFilter(ServletRequest req, ServletResponse resp, FilterChain chain)
throws ServletException, IOException {
 System.out.println("orderFilter1执⾏了..");
 chain.doFilter(req, resp);
 }
 public void init(FilterConfig config) throws ServletException {
}
}

filter2号

package com.jk1123.web.filter.demo02;
import javax.servlet.*;
import javax.servlet.annotation.WebFilter;
import java.io.IOException;
@WebFilter("/*")
public class OrderFilter2 implements Filter {
 public void destroy() {
 }
 public void doFilter(ServletRequest req, ServletResponse resp, FilterChain chain)
throws ServletException, IOException {
 System.out.println("orderFilter2执⾏了..");
 //这是放⾏
 chain.doFilter(req, resp);
 }
 public void init(FilterConfig config) throws ServletException {
 }
}

filter3号

package com.jk1123.web.filter.demo03;
import javax.servlet.*;
import javax.servlet.annotation.WebFilter;
import java.io.IOException;
@WebFilter("/*")
public class OrderFilter3 implements Filter {
 public void destroy() {
 }
 public void doFilter(ServletRequest req, ServletResponse resp, FilterChain chain)
throws ServletException, IOException {
System.out.println("orderFilter3执⾏了..");
 //这是放⾏
 chain.doFilter(req, resp);
 }
 public void init(FilterConfig config) throws ServletException {
 }
}

配上⼀个servlet 来访问试试

package com.jk1123.web.servlet;
import javax.servlet.ServletException;
import javax.servlet.annotation.WebServlet;
import javax.servlet.http.HttpServlet;
import javax.servlet.http.HttpServletRequest;
import javax.servlet.http.HttpServletResponse;
import java.io.IOException;
@WebServlet("/foo")
public class FooServlet extends HttpServlet {
 protected void doPost(HttpServletRequest request, HttpServletResponse response)
throws ServletException, IOException {
 this.doGet(request, response);
 }
 protected void doGet(HttpServletRequest request, HttpServletResponse response)
throws ServletException, IOException {
 response.getWriter().print("foo servlet");
 }
}

验证结果:

3.查看源码
可以从上⾯验证看出 好像并不是根据类名或者filter的name属性的字⺟排序执⾏,那到底是根据什么执⾏的呢?
点开源码,我们⼀点点探寻它的秘密.
需要搞清楚如下问题
1.filterChain是什么时候执⾏的呢?
2.filterChain中的filter来源何处?
3.standardContext什么时候开始收集的过滤器集合
我们先查询第⼀段源码 解密filterChain是什么时候执⾏的

//在org.apache.catalina.core.StandardWrapperValve 类中有如下⼀个⽅法
public final void invoke(Request request, Response response)
 throws IOException, ServletException {
 //省略掉⼤段⽆关代码
 //0.⽣命servlet对象
 Servlet servlet = null;
 /**
 * 省略⼀⼤段⽆关代码
 */
 try {
 if (!unavailable) {
 //1.这⾥创建正在访问的servlet对象
 servlet = wrapper.allocate();
 }
 } catch (UnavailableException e) {
 //省略掉⼤段⽆关代码
 }

 //省略⼤段⽆关代码

 //2.创建过滤器链对象
 ApplicationFilterChain filterChain =
 ApplicationFilterFactory.createFilterChain(request, wrapper, servlet);
 try {
 if ((servlet != null) && (filterChain != null)) {
 // Swallow output if needed
 if (context.getSwallowOutput()) {
 //省略掉⼤段⽆关代码
} else {
 if (request.isAsyncDispatching()) {
 request.getAsyncContextInternal().doInternalDispatch();
 } else {
 //3.执⾏过滤链对象doFilter⽅法
 filterChain.doFilter
 (request.getRequest(), response.getResponse());
 }
 }
 }
 } catch (ClientAbortException | CloseNowException e) {
 //省略掉⼤段⽆关代码
 }
 }

我们⼀会⼉回过头来看它是如何创建过滤器链对象的代码,我们先来看他是如何执⾏过滤器链的,过滤器链对象的实现为:

package org.apache.catalina.core;
//我们只保留 跟执⾏序列有关的代码
public final class ApplicationFilterChain implements FilterChain {
 //当前正在执⾏的filter索引
 private int pos = 0;
 //总共有多少个filter匹配上了
 private int n = 0;
 //关联的要执⾏的servlet对象
 private Servlet servlet = null;
 //匹配上的filter数组
 private ApplicationFilterConfig[] filters = new ApplicationFilterConfig[0];
 @Override
 public void doFilter(ServletRequest request, ServletResponse response)
 throws IOException, ServletException {
 if( Globals.IS_SECURITY_ENABLED ) {
 //删除⽆关代码
 } else {
 //0.执⾏内容的doFilter⽅法
 internalDoFilter(request,response);
 }
 }
 private void internalDoFilter(ServletRequest request,
 ServletResponse response)
throws IOException, ServletException {
 // Call the next filter if there is one
 if (pos < n) {
 //这个地⽅主义有个pos++ 进来⼀次 ++⼀次
 ApplicationFilterConfig filterConfig = filters[pos++];
 try {
 Filter filter = filterConfig.getFilter();
 //删除⼤段⽆关代码

 if( Globals.IS_SECURITY_ENABLED ) {
 //删除⼤段⽆关代码
 } else {
 //执⾏过滤器链中的过滤器的doFilter⽅法
 //⽽我们的过滤器中满⾜条件后 放⾏ 放⾏就会跳转回来执⾏过滤器链的
 //的doFilter 也就是⼜回来执⾏第⼆个
 filter.doFilter(request, response, this);
 }
 } catch (IOException | ServletException | RuntimeException e) {
 throw e;
 } catch (Throwable e) {
 e = ExceptionUtils.unwrapInvocationTargetException(e);
 ExceptionUtils.handleThrowable(e);
 throw new ServletException(sm.getString("filterChain.filter"), e);
 }
 return;
 }
 try {
 //删除⼤段⽆关代码
 // Use potentially wrapped request from this point
 if ((request instanceof HttpServletRequest) &&
 (response instanceof HttpServletResponse) &&
 Globals.IS_SECURITY_ENABLED ) {
 //删除⼤段⽆关代码
 } else {
 //如果没有需要执⾏的filter就会执⾏ servlet的service⽅法 也就是我们写的业务
逻辑
 servlet.service(request, response);
 }
 } catch (IOException | ServletException | RuntimeException e) {
 throw e;
 } catch (Throwable e) {
 e = ExceptionUtils.unwrapInvocationTargetException(e);
 ExceptionUtils.handleThrowable(e);
 throw new ServletException(sm.getString("filterChain.servlet"), e);
 } finally {
 if (ApplicationDispatcher.WRAP_SAME_OBJECT) {
lastServicedRequest.set(null);
 lastServicedResponse.set(null);
 }
 }
 }
}

从filterChain类的源码可以看出底层是包含了 所匹配上的filter数组 也就是添加进去匹配上过滤器对象是有序的 添加的时候就决定了!!!
那么它是什么时候添加的呢?
2.filterChain中的filter来源何处?
其实我们在在org.apache.catalina.core.StandardWrapperValve 类的invoke⽅法中

ApplicationFilterChain filterChain =
 ApplicationFilterFactory.createFilterChain(request, wrapper, servlet);

点开这段代码查询把!

public static ApplicationFilterChain createFilterChain(ServletRequest request,
 Wrapper wrapper, Servlet servlet) {
 // If there is no servlet to execute, return null
 if (servlet == null)
 return null;
 // 在这⾥创建ApplicationFilterChain 对象
 //但是对象⾥还没有filter对象
 ApplicationFilterChain filterChain = null;
 if (request instanceof Request) {
 Request req = (Request) request;
 if (Globals.IS_SECURITY_ENABLED) {
 // Security: Do not recycle
 filterChain = new ApplicationFilterChain();
 } else {
 filterChain = (ApplicationFilterChain) req.getFilterChain();
 if (filterChain == null) {
 filterChain = new ApplicationFilterChain();
 req.setFilterChain(filterChain);
 }
 }
 } else {
 // Request dispatcher in use
 filterChain = new ApplicationFilterChain();
 }
filterChain.setServlet(servlet);
 filterChain.setServletSupportsAsync(wrapper.isAsyncSupported());
 // Acquire the filter mappings for this Context
 StandardContext context = (StandardContext) wrapper.getParent();
 //获取ServletContext对象 注册的所有的filter数组
 FilterMap filterMaps[] = context.findFilterMaps();
 // If there are no filter mappings, we are done
 if ((filterMaps == null) || (filterMaps.length == 0))
 return filterChain;
 // Acquire the information we will need to match filter mappings
 DispatcherType dispatcher =
 (DispatcherType) request.getAttribute(Globals.DISPATCHER_TYPE_ATTR);
 String requestPath = null;
 Object attribute = request.getAttribute(Globals.DISPATCHER_REQUEST_PATH_ATTR);
 if (attribute != null){
 requestPath = attribute.toString();
 }
 String servletName = wrapper.getName();
 //这⾥开始遍历 filterMaps数组根据请求路径匹配添加
 for (int i = 0; i < filterMaps.length; i++) {
 if (!matchDispatcher(filterMaps[i] ,dispatcher)) {
 continue;
 }
 if (!matchFiltersURL(filterMaps[i], requestPath))
 continue;
 ApplicationFilterConfig filterConfig = (ApplicationFilterConfig)
 context.findFilterConfig(filterMaps[i].getFilterName());
 if (filterConfig == null) {
 // FIXME - log configuration problem
 continue;
 }
 filterChain.addFilter(filterConfig);
 }
 // Add filters that match on servlet name second
 for (int i = 0; i < filterMaps.length; i++) {
 if (!matchDispatcher(filterMaps[i] ,dispatcher)) {
 continue;
 }
 if (!matchFiltersServlet(filterMaps[i], servletName))
 continue;
 ApplicationFilterConfig filterConfig = (ApplicationFilterConfig)
context.findFilterConfig(filterMaps[i].getFilterName());
 if (filterConfig == null) {
 // FIXME - log configuration problem
 continue;
 }
 filterChain.addFilter(filterConfig);
 }
 // Return the completed filter chain
 return filterChain;
 }

可以看出在创建filterChain对象时候,从ServletContext获取所有注册的filter的数组 取出需要的添加到这次请求创建的filterChain对象中
⽽且servletContext对象的注册的所有的过滤器本身就是⼀个数组 本身就是有序的,所以遍历匹配的时候,也就是有序的!
3.ServletContext什么时候开始收集的数组,从哪来的呢?
这个要从tomcat启动的时候来看了!

⽽StandardContext创建完成以后就要开始初始化操作了!
StandardContext.startInternal()⽅法---->fireLifecycleEvent()⽅法-->ContextConfig.lifecycleEvent()--
>ContextConfig.lifecycleEventcon-->ContextConfig.con.figureStart()-->ContextConfig.webConfig()
好了我们现在查看该⽅法:

protected void webConfig() {

 WebXmlParser webXmlParser = new WebXmlParser(context.getXmlNamespaceAware(),
 context.getXmlValidation(), context.getXmlBlockExternal());
 Set defaults = new HashSet<>();
 defaults.add(getDefaultWebXmlFragment(webXmlParser));
 //创建了web.xml 配置⽂件对象
 //也就是它就代表我们项⽬的配置相关的信息
 WebXml webXml = createWebXml();
 // Parse context level web.xml
 InputSource contextWebXml = getContextWebXmlSource();
 //解析web.xml 配置⽂件
 //发现配置⽂件中的 filter servlet listener等配置
 //⽽xml的解析是从上到下 所以你在web.xml 配置filter
 //得到filter集合就是有序的
 if (!webXmlParser.parseWebXml(contextWebXml, webXml, false)) {
 ok = false;
 }
 ServletContext sContext = context.getServletContext();
 //省略⼤段⽆关代码
 if (!webXml.isMetadataComplete() || typeInitializerMap.size() > 0) {
 // Steps 4 & 5.
 //扫描编译的类⽂件 寻找注解⽅式书写的servlet filter listener
 processClasses(webXml, orderedFragments);
 }
 //省略⼤段⽆关代码
 }

从上⾯代码 可以看出web,xml配置的filter肯定是有序的了 解析的时候 就会收集到webXml对象的

//采⽤的是linkedHashset来存储的 是有序的
private final Set filterMaps = new LinkedHashSet<>();

解析xml过程我们就不看了 ⼈家才是的digester的xml解析框架来做.
我们来查看processClasses(webXml, orderedFragments); 这个⽅法是解析注解⽤的

protected void processClasses(WebXml webXml, Set orderedFragments) {
 // Step 4. Process /WEB-INF/classes for annotations and
 // @HandlesTypes matches
 Map javaClassCache = new HashMap<>();
 if (ok) {
//获取项⽬下的类路径
 WebResource[] webResources =
 context.getResources().listResources("/WEB-INF/classes");
 for (WebResource webResource : webResources) {
 // Skip the META-INF directory from any JARs that have been
 // expanded in to WEB-INF/classes (sometimes IDEs do this).
 if ("META-INF".equals(webResource.getName())) {
 continue;
 }
 //开始根据注解解析了
 processAnnotationsWebResource(webResource, webXml,
 webXml.isMetadataComplete(), javaClassCache);
 }
 }
 // Step 5. Process JARs for annotations and
 // @HandlesTypes matches - only need to process those fragments we
 // are going to use (remember orderedFragments includes any
 // container fragments)
 if (ok) {
 processAnnotations(
 orderedFragments, webXml.isMetadataComplete(), javaClassCache);
 }
 // Cache, if used, is no longer required so clear it
 javaClassCache.clear();
 }

查看processAnnotationsWebResource⽅法

protected void processAnnotationsWebResource(WebResource webResource,
 WebXml fragment, boolean handlesTypesOnly,
 Map javaClassCache) {
 //看看是否是个⽬录
 if (webResource.isDirectory()) {
 WebResource[] webResources =
 webResource.getWebResourceRoot().listResources(
 webResource.getWebappPath());
 if (webResources.length > 0) {
 if (log.isDebugEnabled()) {
 log.debug(sm.getString(
 "contextConfig.processAnnotationsWebDir.debug",
webResource.getURL()));
 }
 //遍历⽬录
 for (WebResource r : webResources) {
 //递归处理
processAnnotationsWebResource(r, fragment, handlesTypesOnly,
javaClassCache);
 }
 }
 } else if (webResource.isFile() &&
 webResource.getName().endsWith(".class")) {
 try (InputStream is = webResource.getInputStream()) {
 //如果是类⽂件的话 开始处理
 processAnnotationsStream(is, fragment, handlesTypesOnly,
javaClassCache);
 } catch (IOException e) {
 log.error(sm.getString("contextConfig.inputStreamWebResource",
 webResource.getWebappPath()),e);
 } catch (ClassFormatException e) {
 log.error(sm.getString("contextConfig.inputStreamWebResource",
 webResource.getWebappPath()),e);
 }
 }
 }

查看processAnnotationsStream⽅法

protected void processAnnotationsStream(InputStream is, WebXml fragment,
 boolean handlesTypesOnly, Map javaClassCache)
 throws ClassFormatException, IOException {
 ClassParser parser = new ClassParser(is);
 JavaClass clazz = parser.parse();
 checkHandlesTypes(clazz, javaClassCache);
 if (handlesTypesOnly) {
 return;
 }
 //处理开始
 processClass(fragment, clazz);
 }
 protected void processClass(WebXml fragment, JavaClass clazz) {
 AnnotationEntry[] annotationsEntries = clazz.getAnnotationEntries();
 if (annotationsEntries != null) {
 String className = clazz.getClassName();
 for (AnnotationEntry ae : annotationsEntries) {
 String type = ae.getAnnotationType();
 if ("Ljavax/servlet/annotation/WebServlet;".equals(type)) {
 processAnnotationWebServlet(className, ae, fragment);
 //判断是否webFilter注解 如果是就添加到 webxml配置对象中
 }else if ("Ljavax/servlet/annotation/WebFilter;".equals(type)) {
 processAnnotationWebFilter(className, ae, fragment);
}else if ("Ljavax/servlet/annotation/WebListener;".equals(type)) {
 fragment.addListener(className);
 } else {
 // Unknown annotation - ignore
 }
 }
 }
 }

从上⾯可以看出原来扫描类路径的时候,就是先遍历⽂件夹 遍历⽂件夹下类⽂件 反射查看是否是⼀个带有WebFilter注解的类
如果是就添加到web.xml中set集合中,⽽那个set集合是有序的linkedset
所有顺序就是递归遍历⽂件夹的顺序 ⼀切就看 递归的时候如何获取下级⽂件夹的代码了 看它是否进⾏排序了?
也就是说由如下代码决定的

WebResource[] webResources =
 webResource.getWebResourceRoot().listResources(
 webResource.getWebappPath());

点开这段代码

protected WebResource[] listResources(String path, boolean validate) {
 if (validate) {
 path = validate(path);
 }
 String[] resources = list(path, false);
 WebResource[] result = new WebResource[resources.length];
 for (int i = 0; i < resources.length; i++) {
 if (path.charAt(path.length() - 1) == '/') {
 result[i] = getResource(path + resources[i], false, false);
 } else {
 result[i] = getResource(path + '/' + resources[i], false, false);
 }
 }
 return result;
 }
 //继续
private String[] list(String path, boolean validate) {
 if (validate) {
 path = validate(path);
 }
 // Set because we don't want duplicates
 // LinkedHashSet to retain the order. It is the order of the
// WebResourceSet that matters but it is simpler to retain the order
 // over all of the JARs.
 HashSet result = new LinkedHashSet<>();
 for (List list : allResources) {
 for (WebResourceSet webResourceSet : list) {
 if (!webResourceSet.getClassLoaderOnly()) {
 String[] entries = webResourceSet.list(path);
 for (String entry : entries) {
 result.add(entry);
 }
 }
 }
 }
 return result.toArray(new String[result.size()]);
 }
//继续
public String[] list(String path) {
 checkPath(path);
 String webAppMount = getWebAppMount();
 if (path.startsWith(webAppMount)) {
 File f = file(path.substring(webAppMount.length()), true);
 if (f == null) {
 return EMPTY_STRING_ARRAY;
 }
 //就到这⾥了 我们可以看到 它没有排序就是调⽤了
 //file类的list⽅法
 String[] result = f.list();
 if (result == null) {
 return EMPTY_STRING_ARRAY;
 } else {
 return result;
 }
 } else {
 if (!path.endsWith("/")) {
 path = path + "/";
 }
 if (webAppMount.startsWith(path)) {
 int i = webAppMount.indexOf('/', path.length());
 if (i == -1) {
 return new String[] {webAppMount.substring(path.length())};
 } else {
 return new String[] {
 webAppMount.substring(path.length(), i)};
 }
 }
 return EMPTY_STRING_ARRAY;
 }
 }

来来打开file类的list⽅法看看

所以最终 web.xml收集到所有filter的set集合 如果采⽤的是注解⽅式 没有任何顺序可⾔的.
然后接下来的代码我们就不看了⽆⾮就是将收集到的filter集合转换成数组 设置给StandardContext对象

4.得出结论

如果采⽤web.xml写的filter执⾏顺序跟书写顺序有关
⽽采⽤注解⽅式的是没有顺序可⾔的!!!!
⽽采⽤注解⽅式的是没有顺序可⾔的!!!!
⽽采⽤注解⽅式的是没有顺序可⾔的!!!!
实践出真知!切记⼈云亦云!有问题找源码!!!