Error page jsf

Учебное пособие по проверке JSF: обработка ошибок в JSF

В продолжение моей предыдущей статьи о создании JSF-приложения JavaServer Faces в Eclipse я публикую следующую статью из этой серии. На этот раз мы расскажем подробнее о модели проверки JSF и обработке ошибок в JSF.

Технология JavaServer Faces поддерживает механизм проверки данных редактируемых компонентов. Каждый компонент в JavaServer Faces создает сообщения об ошибках в течение жизненного цикла JSF и присоединяет их к объекту FacesContext. Следовательно, каждое сообщение присоединяется к компоненту в дереве компонентов, и сообщение отображается в представлении в конце жизненного цикла JSF.

Отображение сообщений об ошибках при просмотре

JSF предоставляет различные теги для обработки и отображения сообщений в представлении. В эталонной реализации Sun библиотеки JSF HTML есть два тега сообщения:

h: messages используется для отображения всех сообщений одновременно. Вы можете разместить тег h: messages в начале вашей формы. Вам может потребоваться отображать только глобальные сообщения, используя тег h: messages. Для отображения только глобальных сообщений установите для атрибута globleOnly значения h: messages значение true.

Используйте h: message для отображения сообщения, прикрепленного к одному компоненту. Атрибут at = «» может использоваться для указания идентификатора компонента, чьи сообщения об ошибках нам нужно отобразить. H: сообщение используется для отображения сообщения об ошибке рядом с компонентом, который сгенерировал ошибку. Если к сообщению прикреплено более одного сообщения этот компонент, h: message будет отображать последнее сообщение.

Каждое сообщение может иметь краткое описание и подробное описание. При использовании тега h: message по умолчанию показывается подробное сообщение. При использовании тега h: messages по умолчанию отображаются сводные описания. Чтобы изменить значения по умолчанию, измените логические атрибуты showSummary и showDetail .

Вы также можете включить отображение подробного сообщения компонента в виде всплывающей подсказки. Для этого установите для атрибута подсказки тега сообщения значение true. Обратите внимание, что для включения этой опции для showDetail и showSummary должно быть установлено значение true.

Существует четыре формы проверки JSF:
1. Встроенные компоненты проверки
2. Проверка уровня приложения
3. Пользовательские компоненты проверки с использованием интерфейса Validator
4. Методы проверки в компонентах поддержки

Встроенные компоненты проверки

Эталонная реализация Sun JSF предоставляет некоторые компоненты проверки по умолчанию, которые можно использовать для проверки любого пользовательского ввода. Базовая библиотека JSF предоставляет теги для проверки ввода. Ниже приведены несколько тегов, которые можно использовать для проверки ввода.

f: validateDoubleRange : этот тег проверяет значение компонента в указанном диапазоне. Значение должно быть преобразовано в тип с плавающей точкой или саму плавающую точку.

f: validateLength : этот тег проверяет длину значения и ограничивает его в пределах указанного диапазона. Значение должно иметь тип java.lang.String.

f: validateLongRange : Проверяет, находится ли значение компонента в указанном диапазоне. Значение должно иметь числовой тип или строку, конвертируемую в long.
Пример:

Проверка с использованием методов Backing Bean

Метод базового компонента может использоваться для проверки любого поля ввода. Сначала нам нужно создать метод вспомогательного компонента, который будет вызываться в процессе проверки. Подпись метода может быть:

Как только метод вспомогательного компонента будет готов, мы можем связать его с компонентом, используя тег f: validator .

В приведенном выше фрагменте мы связали метод userBean checkUsername () с компонентом inputText. Можно связать более одного валидатора с одним компонентом.

Метод проверки базового компонента прост в реализации, но этот метод специфичен для одного приложения и не может быть повторно использован для другого приложения. Для создания универсальных валидаторов, которые можно использовать в разных приложениях, можно использовать интерфейс Validator.

Пользовательские компоненты проверки с использованием интерфейса Validator

Интерфейс Validator может быть расширен, и может быть создан специальный валидатор, который можно повторно использовать в различных приложениях в JSF. Чтобы создать собственный валидатор, нам нужно создать класс Java, который реализует интерфейс javax.faces.validator.Validator . Интерфейс Validator предоставляет метод validate (), который необходимо реализовать. Ниже приведена подпись метода validate ().

После того, как Валидатор реализован, нам нужно зарегистрировать этот валидатор в файлеface-config.xml. Для этого скопируйте следующий код вface-config.xml, предполагая, что наше имя класса валидатора — net.viralpatel.jsf.helloworld.EmailValidator.

Мы можем связать этот валидатор с любым компонентом, используя тег f: validator.

Обратите внимание, что в приведенном выше фрагменте кода атрибут validatorId тега f: validator указывает на идентификатор валидатора, который зарегистрирован в файлеface-config.xml.
Для проверки адреса электронной почты мы можем создать класс Validator как:

Таким образом, используя платформу JSF Validation, можно легко проверить вводимые пользователем данные. Мы видели разные способы проверки в JSF: валидаторы по умолчанию, методы поддержки bean-компонентов и валидация через интерфейс валидатора.

Дайте мне знать ваши комментарии об этом уроке.

Источник

JSF Validation Tutorial: Error Handling in JSF

Join the DZone community and get the full member experience.

In a continuation of my previous article on Creating JavaServer Faces JSF application in Eclipse , I am posting the next article in the series. This time we will cover details about JSF Validation Model and Handling Errors in JSF.

JavaServer Faces technology supports a mechanism for validating the data of editable components. Each component in JavaServer Faces creates Error Messages during the life cycle of JSF and attaches them to FacesContext object. Hence each message is attached to a component in the component tree and the message is displayed into the view at the end of JSF life cycle.

Showing Error Messages on View

JSF provides different tags to handle and display messages on the view. There are two message tags in Sun’s reference implementation of JSF HTML library:

h:messages is used to display all messages at once. You can place h:messages tag in start of your form. You may need to display only global messages using h:messages tag. For displaying only global messages set globleOnly attribute of h:messages to true.

Use h:message to display message attached to one component. An attibute for=”» can be used to specify the id of a component whose error messages we need to display. h:message is used to display error message next to the component that generated the error. If more then one message is attached to that component, h:message will display the last message.

Each message can have a summary description and a detailed description. When using the h:message tag, the default is to show the detail message. When using the h:messages tag, the default is to display the summary descriptions. To change the defaults, modify the boolean showSummary and showDetail attributes.

You can also enable a component’s detail message to appear as a tooltip. To do so, set tooltip attribute of message tag to true. Note that for enabling this option, showDetail and showSummary must be set to true.

There are four forms of JSF validation:
1. Built-in validation components
2. Application level validations
3. Custom validation components using Validator interface
4. Validation methods in backing beans

Built-in validation components

Sun’s reference implementation of JSF provides some default validation components that can be leveraged to implement validation of any user input. The JSF core library provides tags to validate input. The following are a few tags that can be used to validate the input.

f:validateDoubleRange : This tag checks the value of a component within specified range. The value must be convertible to floating-point type or a floating-point itself.

f:validateLength : This tag checks the length of a value and restricts it within a specified range. The value must be of type java.lang.String.

f:validateLongRange : Checks if a component value is within a specified range. The value must be of numeric type or string convertible to a long.
Example:

Validation using Backing Bean methods

A backing bean method can be used for doing validation of any input field. First we need to create the backing bean method that will get called during validation process. The signature of the method can be:

Once a backing bean method is ready we can bind it with a component using f:validator tag.

In above snippet, we have bound the checkUsername() method of userBean to the inputText component. It is possible to bind more than one validator to one component.

THe backing bean method of validation is easy to implement, but this method is specific for one application and may not be reused for different application. To create generic validators which can be used in different application, the Validator interface can be used.

Custom validation components using Validator interface

The Validator interface can be extended and a custom validator can be created which can be reused across different applications in JSF. To create a custom validator, we need to create a Java class that implements javax.faces.validator.Validator interface. The Validator interface provides a validate () method that needs to be implemented. The following is the signature of validate() method.

Once the Validator is implemented, we need to register this validator in faces-config.xml file. To do so copy following code in faces-config.xml assuming that our validator class name is net.viralpatel.jsf.helloworld.EmailValidator.

We can bind this validator to any component using f:validator tag.

Note that in the above code snippet, the validatorId attribute of f:validator tag points to the validator’s ID that is registered in faces-config.xml file.
For validating email address we can create a Validator class as:

Thus by using the JSF Validation framework, it is possible to validate user input easily. We saw different ways of validation in JSF: Default validators, backing bean methods and validation through validator interface.

Let me know your comments about this tutorial.

Opinions expressed by DZone contributors are their own.

Источник

JSF Custom Error Pages

When you run an application in the development project stage and you encounter an error, you get an error message in an undesirable form. You probably don’t want your users to see that message in such that ugly way.

To substitute a better error page, use error-page tag in the web.xml file, in that you can specify either a Java Exception or an HTTP error code. So in case the type of thrown exception has matched that type mentioned in the web.xml exception-type or the error code that generated by the server has matched error-code that mentioned in the web.xml, the JSF framework will handle it by forwarding the user into the desired view that you’ve defined for such those errors or exceptions.

1. The Deployment Descriptor

web.xml

[code lang=”xml”]

404
/faces/error.xhtml

500
/faces/error.xhtml

java.lang.Exception
/faces/error.xhtml

State saving method: ‘client’ or ‘server’
(=default). See JSF Specification 2.5.2

Faces Servlet
javax.faces.webapp.FacesServlet
1

Faces Servlet
/faces/*

Faces Servlet
*.xhtml

{scrollbar}

Error handling for MyFaces Core 2.0 and later versions

Since JSF 2.0, it is possible to provide a custom javax.faces.context.ExceptionHandler or javax.faces.context.ExceptionHandlerWrapper implementation to deal with exceptions. To do that, just create your custom class, an factory that wrap/override it and add the following into your faces-config.xml:

faces-config.xml<faces-config xmlns=»http://java.sun.com/xml/ns/javaee»
xmlns:xsi=»http://www.w3.org/2001/XMLSchema-instance»
xsi:schemaLocation=»http://java.sun.com/xml/ns/javaee http://java.sun.com/xml/ns/javaee/web-facesconfig_2_0.xsd»
version=»2.0″>

<!— … —>
<factory>
<!— … —>
<exception-handler-factory>org.apache.myfaces.context.ExceptionHandlerFactoryImpl</exception-handler-factory>
<!— … —>
</factory>
<!— … —>
</faces-config>

This is an example of an ExceptionHandlerFactory, from myfaces code:

ExceptionHandlerFactoryImpl.javapublic class ExceptionHandlerFactoryImpl extends ExceptionHandlerFactory
{

@Override
public ExceptionHandler getExceptionHandler()
{
return new SwitchAjaxExceptionHandlerWrapperImpl(
new MyFacesExceptionHandlerWrapperImpl(new ExceptionHandlerImpl()) ,
new AjaxExceptionHandlerImpl());
}
}

If you need a wrapper:

ExceptionHandlerFactoryImpl.javapublic class ExceptionHandlerFactoryImpl extends ExceptionHandlerFactory
{

@Override
public ExceptionHandler getExceptionHandler()
{
return new CustomExceptionHandlerWrapper(getWrapped().getExceptionHandler());
}
}

The most important method to override is ExceptionHandler.handle().

MyFacesExceptionHandlerWrapperImpl.javapublic class MyFacesExceptionHandlerWrapperImpl extends ExceptionHandlerWrapper
{
//…
public void handle() throws FacesException
{
//… some custom code goes here …
}
}

Take a look at MyFaces Core source code, to know in detail how ExceptionHandler implementations works.

MyFaces ExceptionHandler

MyFaces Core provide a custom ExceptionHandler to deal with exceptions and provide detailed information about it. Since 2.0.8/2.1.2 this is disabled on Production environments unless it enabled on web.xml file. Don’t forget to provide your custom error page in this scenario, to prevent show more information than necessary.

xml <context-param>
<param-name>org.apache.myfaces.ERROR_HANDLING</param-name>
<param-value>true</param-value>
</context-param>

<!— if you want to use a different resource template file than
«META-INF/rsc/myfaces-dev-error.xml» this param let you configure
it. (since 1.2.4-SNAPSHOT and 1.1.6-SNAPSHOT)—>
<context-param>
<param-name>org.apache.myfaces.ERROR_TEMPLATE_RESOURCE</param-name>
<param-value>META-INF/rsc/custom-dev-error.xml</param-value>
</context-param>

Provide an error page

If org.apache.myfaces.ERROR_HANDLING is set to «false», or if it is undefined and the project stage is «Development», the default ExceptionHandler just throws an exception. So you can still setup an error page by adding something like this in your web.xml file:

xml <error-page>
<error-code>500</error-code>
<location>/somepage.jsp</location>
</error-page>

Error handling for MyFaces Core 1.2 and earlier versions

MyFaces, from version 1.2.1 and 1.1.6, includes automatic error-handling for the full JSF-Lifecycle (taken over mostly from Facelets, with a few adoptions and additions). So for most projects during development, you will have exactly what you want with these new error-handling possibilities.

If this is not what you want, though, you can always disable or modify this error-handling with the following parameters:

xml <!— if you want to disable the behaviour completely —>
<context-param>
<param-name>org.apache.myfaces.ERROR_HANDLING</param-name>
<param-value>false</param-value>
</context-param>
<!— if you are using myfaces + facelets don’t forget to do this —>
<context-param>
<param-name>facelets.DEVELOPMENT</param-name>
<param-value>false</param-value>
</context-param>

<!— if you want to use a different resource template file than
«META-INF/rsc/myfaces-dev-error.xml» this param let you configure
it. (since 1.2.4-SNAPSHOT and 1.1.6-SNAPSHOT)—>
<context-param>
<param-name>org.apache.myfaces.ERROR_TEMPLATE_RESOURCE</param-name>
<param-value>META-INF/rsc/custom-dev-error.xml</param-value>
</context-param>

<!— if you want to choose a different class for handling the exception — the error-handler needs to include a method handleException(FacesContext fc, Exception ex)—>
<context-param>
<param-name>org.apache.myfaces.ERROR_HANDLER</param-name>
<param-value>my.project.ErrorHandler</param-value>
</context-param>

If you do this, you can now read on to get to general ways of handling server-errors.

Server errors such as HTTP 500 can occur for a number of reasons such as uncaught exceptions, missing JSFs or backing beans, bad URL and the list goes on. While we hope these only occur during development it is important to plan to catch and deal with these errors gracefully when running live with multiple users.

Several approaches have been discussed on the mailing list:

Use default handler

Myfaces has a default error handler (class javax.faces.webapp._ErrorPageWriter) that uses a jsp template file (META-INF/rsc/myfaces-dev-error.xml and META-INF/rsc/myfaces-dev-debug.xml) to handle errors.

For define a custom template file:

xml <context-param>
<param-name>org.apache.myfaces.ERROR_HANDLING</param-name>
<param-value>true</param-value>
</context-param>

<context-param>
<param-name>org.apache.myfaces.ERROR_TEMPLATE_RESOURCE</param-name>
<param-value>META-INF/rsc/mycustom-template-error.xml</param-value>
</context-param>

Use sandbox org.apache.myfaces.tomahawk.util.ErrorRedirectJSFPageHandler

This handler uses myfaces error handling feature, redirecting to a jsf page when an error occurs.

An example jsf page for redirect can be found at http://issues.apache.org/jira/browse/TOMAHAWK-1297

This class is set as a config-parameter org.apache.myfaces.ERROR_HANDLER
available on myfaces core jsf. (This does not work with RI)

The idea is extends myfaces error handling feature, making possible to redirect
to a jsf page when an error occur, using navigation rules.

If this handler is not able to handle the error, an alternate error handler
could be set in the config-parameter org.apache.myfaces.ERROR_REDIRECT_ALTERNATE_HANDLER

The info of the error in the jsf page can be found using:

#{exceptionContext.cause} : Cause retrieved from the exception
#{exceptionContext.stackTrace} : Stack trace of the exception
#{exceptionContext.exception} : Exception handled by this page
#{exceptionContext.tree} : Print the component tree of the page that cause the error
#{exceptionContext.vars} : Enviroment variables from the request

Using servlets

Mert Caliskan (http://www.jroller.com/page/mert?entry=handling_errors_with_an_errror) describes an approach which wraps the JSF servlet with a new servlet which delegates to the faces servlet but handles uncaught exceptions allowing the developer to redirect to a custom error page.

Andrea Paternesi has refined this technique for MyFaces as described here:
[http://patton-prog-tips.blogspot.com/2008/10/myfaces-handling-viewexpiredexception.html]

With JSF

Another approach is described in the book ‘Core Server Faces’ which uses the servlet engine to catch the error and delegate to a JSF error page. While not as elegant as the first approach this method has several advantages for some users

1 It uses a standard JSP approach and framework<<BR>>
2 It does not require custom servlets<<BR>>
3 It can easily be customized/changed by simply changing the error handler definition in the web.xml<<BR>>

To implement this method perform the following steps

1 define an error handling web page in web.xml

xml <error-page>
<error-code>500</error-code>
<location>/ErrorDisplay.jsf</location>
</error-page>

2 Create the error handler display. Due to a problem with the JSF 1.1 specification, the error handling page cannot use a <f:view> but must use a subview.

xml<!DOCTYPE html PUBLIC «-//W3C//DTD XHTML 1.0 Transitional//EN» «http://www.w3.org/TR/xhtml1/DTD/xhtml1-transitional.dtd»>
<f:subview id=»error»
xmlns:f=»http://java.sun.com/jsf/core»
xmlns:t=»http://myfaces.apache.org/tomahawk»
xmlns:h=»http://java.sun.com/jsf/html»>

<html>
<head>
<meta content=»no-cache» http-equiv=»Cache-Control» />
<meta content=»no-cache» http-equiv=»Pragma» />
<title>CMS — Error</title>
<t:stylesheet path=»#{SessionBean.styleSheet}» />
</head>
<body>
<h:form>
:
: set up the normal view
:
<h:outputText styleClass=»infoMessage» escape=»false» value=»#{ErrorDisplay.infoMessage}» />
<t:htmlTag value=»br» />
<h:inputTextarea style=»width: 99%;» rows=»10″ readonly=»true» value=»#{ErrorDisplay.stackTrace}» />
:
: more view stuff
:
</h:form>
</body>
</html>
</f:subview>

3 Create a backing bean in request scope which can process the error. (don’t forget to register it on faces-config.xml)

ErrorDisplay.javaimport cms.beans.framework.AbstractUIBean;

import com.c2gl.jsf.framework.ApplicationResource;

import java.io.PrintWriter;
import java.io.StringWriter;

import java.util.Map;

import javax.faces.context.FacesContext;

import javax.servlet.ServletException;

public class ErrorDisplay extends AbstractUIBean {
private static final long serialVersionUID = 3123969847287207137L;
private static final String BEAN_NAME = ErrorDisplay.class.getName();

public String getInfoMessage() {
return «An unexpected processing error has occurred. Please cut and paste the following information» + » into an email and send it to <b>» +
some email address + «</b>. If this error » + «continues to occur please contact our technical support staff at <b>» +
some phone number etc + «</b>.»;
}

public String getStackTrace() {
FacesContext context = FacesContext.getCurrentInstance();
Map requestMap = context.getExternalContext().getRequestMap();
Throwable ex = (Throwable) requestMap.get(«javax.servlet.error.exception»);

StringWriter writer = new StringWriter();
PrintWriter pw = new PrintWriter(writer);
fillStackTrace(ex, pw);

return writer.toString();
}

private void fillStackTrace(Throwable ex, PrintWriter pw) {
if (null == ex) {
return;
}

ex.printStackTrace(pw);

if (ex instanceof ServletException) {
Throwable cause = ((ServletException) ex).getRootCause();

if (null != cause) {
pw.println(«Root Cause:»);
fillStackTrace(cause, pw);
}
} else {
Throwable cause = ex.getCause();

if (null != cause) {
pw.println(«Cause:»);
fillStackTrace(cause, pw);
}
}
}
}

Also have a look at our ExceptionUtils class. It encapsulates the way how to get the real root cause

java [1] List exceptions = ExceptionUtils.getExceptions(exception);
[2] Throwable throwable = (Throwable) exceptions.get(exceptions.size()-1);
[3] String exceptionMessage = ExceptionUtils.getExceptionMessage(exceptions);

1 get a list of all exceptions — using getRootCause if available or getCause<<BR>>
2 get the initial exception<<BR>>
3 get the first exception with an message starting with the initial exception

So the new fillStackTrace become

java private void fillStackTrace(Throwable ex, PrintWriter pw)
{
if (null == ex) {
return;
}

List exceptions = ExceptionUtils.getExceptions(exception);
Throwable throwable = (Throwable) exceptions.get(exceptions.size()-1);

for (int i = 0; i<exceptions.size(); i++)
{
if (i > 0)
{
pw.println(«Cause:»);
}
throwable.printStackTrace(pw);
}
}

In the backing bean we construct a message which informs the user that something we didnt’t plan for has happened with directions on who to call, what to do etc. We don’t really care if they actually do cut-and-paste the error and email it to us as it is also in Tomcat’s logs but giving the user something to do and become part of resolving the problem is always a good idea

ViewExpiredException: No saved view state could be found for the view identifier

In some configurations, you might run into this problem exception with the error handling method shown above. This will happen if an error results in a forward, rather than redirect. The »ViewHandler» will call »response.sendError()» in case of an error, which will lookup your »<error-page>» declarations in »web.xml» and forward to the error url. If the error url is picked up by the »FacesServlet» (i.e. it’s a JSF url), a new JSF lifecycle will be started. However, since this is a forward, the request object will still contain all of the request parameters, including »»javax.faces.ViewState»», which makes the request look like a postback to the »ViewHandler». Since it’s a postback, the »ViewHandler» will expect a saved view, which is clearly not going to be there, since our »viewId» is now referencing the error page.

This problem can be solved by customizing the »ViewHandler» to use »response.sendRedirect()» instead of »response.sendError()», but that will mean that we can no longer use »web.xml» for specifying the error page mappings.

Another way to handle this would be to use an intermediate step by specifying a non-JSF URL as the error page and then somehow redirecting to the JSF error page. For example, for the 404 error code you could specify »/error/404_redirect.html»:

xml<html><head><meta http-equiv=»Refresh» content=»0; URL=/DPSV4/error/404.jsf»></head></html>

This works, but requires you to hard code the context path.

The solution I ended up with involves a »RedirectServlet»:

RedirectServlet.javapublic class RedirectServlet extends HttpServlet {
private static final String URL_PREFIX = «url=»;

@Override
protected void doGet(HttpServletRequest req, HttpServletResponse resp) throws ServletException, IOException {
String query = req.getQueryString();
if ( query.contains(URL_PREFIX) ) {
String url = query.replace(URL_PREFIX, «»);
if ( !url.startsWith(req.getContextPath()) ) {
url = req.getContextPath() + url;
}

resp.sendRedirect(url);
}
}

@Override
protected void doPost(HttpServletRequest req, HttpServletResponse resp) throws ServletException, IOException {
doGet(req, resp);
}

}

used like this in »web.xml»:

xml <servlet>
<servlet-name>Redirect Servlet</servlet-name>
<servlet-class>ca.gc.agr.ops.web.jsf.redirect.RedirectServlet</servlet-class>
</servlet>
<servlet-mapping>
<servlet-name>Redirect Servlet</servlet-name>
<url-pattern>/redirect</url-pattern>
</servlet-mapping>

<error-page>
<error-code>403</error-code>
<location>/redirect?url=/error/403.jsf</location>
</error-page>
<error-page>
<error-code>404</error-code>
<location>/redirect?url=/error/404.jsf</location>
</error-page>
<error-page>
<error-code>500</error-code>
<location>/redirect?url=/error/500.jsf</location>
</error-page>

With plain JSP

If you didn’t require any JSF functionality in your JSP page it might be worth to consider using the following error.jsp

[http://svn.apache.org/viewvc/myfaces/tomahawk/trunk/examples/simple/src/main/webapp/error.jsp?view=markup]

with this web.xml configuration

xml <error-page>
<error-code>500</error-code>
<location>/error.jsp</location>
</error-page>

This reduces the number of technologies required to show the error page which might improve the availablity of this page

Custom error handler under Apache Geronimo JavaEE server

If you would use your own error handler (org.apache.myfaces.ERROR_REDIRECT_ALTERNATE_HANDLER) under Apache Geronimo, you must set properly your deployment plan (geronimo-web.xml usualy), otherwise you will have classloading problem of your class. By default the MyFaces classes are loaded to your classpath through dependencies at org.apache.geronimo.framework.jee-specs/CAR. This is OK for common cases, but if you instruct MyFaces to use your own error handler class, you get the error because MyFaces cannot find your class in calling class.forName(). Avoid this situation is quite simple — in your deployment plan specify dependencies on myfaces-api and myfaces-impl and then modify classloading via hidden-classes setting.

A fragment of your dependency plan would be like this:

xml<dep:dependencies>
.
.
<dep:dependency>
<dep:groupId>org.apache.myfaces.core</dep:groupId>
<dep:artifactId>myfaces-api</dep:artifactId>
<dep:type>jar</dep:type>
</dep:dependency>
<dep:dependency>
<dep:groupId>org.apache.myfaces.core</dep:groupId>
<dep:artifactId>myfaces-impl</dep:artifactId>
<dep:type>jar</dep:type>
</dep:dependency>
.
.
</dep:dependencies>

<dep:hidden-classes>
<dep:filter>javax.faces</dep:filter>
<dep:filter>org.apache.myfaces</dep:filter>
</dep:hidden-classes>

This solution was tested under Apache Geronimo 2.1.3., but will probably be similar in other versions.{scrollbar}

JSF Error Pages That Actually Work

Here is an annoying problem using JSF error pages for JSF requests. Everything looks just fine, HttpServletResponse.sendError() sends the error page, but JSF continues processing and starts throwing exceptions after the response is complete. This happens even if you call FacesContext.responseComplete(), and also when the error page is sent at different stages of the JSF lifecycle.

It seems like invoking the FacesServlet for sendError() breaks the state of the original FacesContext.

When sending an error during view build I get this exception:

JSF Error Pages That Actually Work

java.lang.NullPointerException
	at com.sun.faces.facelets.util.Resource.getResourceUrl(Resource.java:105)
	at com.sun.faces.facelets.impl.DefaultResourceResolver.resolveUrl(DefaultResourceResolver.java:77)
	at com.sun.faces.facelets.impl.DefaultFaceletFactory.resolveURL(DefaultFaceletFactory.java:229)
	at com.sun.faces.facelets.impl.DefaultFacelet.getRelativePath(DefaultFacelet.java:273)
	at com.sun.faces.facelets.impl.DefaultFacelet.include(DefaultFacelet.java:341)
	at com.sun.faces.facelets.impl.DefaultFaceletContext.includeFacelet(DefaultFaceletContext.java:199)
	at com.sun.faces.facelets.tag.ui.DecorateHandler.apply(DecorateHandler.java:145)
	at com.sun.faces.facelets.compiler.NamespaceHandler.apply(NamespaceHandler.java:93)
	at com.sun.faces.facelets.compiler.EncodingHandler.apply(EncodingHandler.java:86)
	at com.sun.faces.facelets.impl.DefaultFacelet.apply(DefaultFacelet.java:149)
	at com.sun.faces.application.view.FaceletViewHandlingStrategy.buildView(FaceletViewHandlingStrategy.java:838)
	at com.sun.faces.lifecycle.RenderResponsePhase.execute(RenderResponsePhase.java:100)
	at com.sun.faces.lifecycle.Phase.doPhase(Phase.java:101)

and if renderView() has already started, it gets even more obscure:

java.lang.NullPointerException
	at org.richfaces.skin.SkinFactoryImpl.clearSkinCaches(SkinFactoryImpl.java:94)
	at org.richfaces.skin.SkinFactoryPreRenderViewListener.processEvent(SkinFactoryPreRenderViewListener.java:35)
	at javax.faces.event.SystemEvent.processListener(SystemEvent.java:106)
	at com.sun.faces.application.ApplicationImpl.processListeners(ApplicationImpl.java:2169)
	at com.sun.faces.application.ApplicationImpl.invokeListenersFor(ApplicationImpl.java:2145)
	at com.sun.faces.application.ApplicationImpl.publishEvent(ApplicationImpl.java:303)
	at com.sun.faces.application.ApplicationImpl.publishEvent(ApplicationImpl.java:247)
	at com.sun.faces.lifecycle.RenderResponsePhase.execute(RenderResponsePhase.java:108)
	at com.sun.faces.lifecycle.Phase.doPhase(Phase.java:101)
	at com.sun.faces.lifecycle.LifecycleImpl.render(LifecycleImpl.java:139)
	at javax.faces.webapp.FacesServlet.service(FacesServlet.java:594)P

JSF continues to RENDER phase in an all messed up drunken way.

• Calling FacesContext.responseComplete() from your managed bean’s @PostConstruct method doesn’t help because rendering has already started.
• Additionally, calling FaceContext.responseComplete() from a preRenderView listener just doesn’t work. It looks like the preRenderView event is added during view construction which happens in the Render View phase anyway. Could this be a regression bug?
• Finally, throwing an exception to be caught by an error filter or exception handler doesn’t resolve the problem because JSF swallows the exception from @PostConstruct and rethrows its own. 

I couldn’t believe something so basic should be so complicated.

Well it turns out there is a fairly simple solution. Calling reponse.setStatus() instead of response.sendError() does not interrupt the JSF lifecycle. This works nicely, except the original view is still rendered in spite of the error.

So all we have to do is manually render a new view (the error page) as soon as the error occurs. This doesn’t break JSF state and lets the lifecycle finish without all those random exceptions.

Here’s what I’m talking about.

    /**
     * The standard request.sendError() breaks JSF state if it is called 
     * too late in the lifecycle. This method does the same thing but 
     * copes better with interrupting the current request.
     */
    public void sendError(FacesContext faces, int code, String message) {
        try {
            faces.getExternalContext().setResponseStatus(code);
            faces.getExternalContext().getRequestMap().put
                    ("javax.servlet.error.message", message);
            ViewHandler views = faces.getApplication().getViewHandler();
            String template = "/error/" + code + ".xhtml";
            UIViewRoot view = views.createView(faces, template);
            faces.setViewRoot(view);
            views.getViewDeclarationLanguage(faces, template).
                    buildView(faces, view);
            views.renderView(faces, view);
            faces.responseComplete();
        } catch (IOException ioe) {
            throw new RuntimeException(ioe);
        }
    }

This method works any time before the view has started rendering. Normally it should be triggered during the view build by an event or managed bean @PostConstruct method. In fact it also works during the render phase but you get a mixed up response (see the comments below).

Hope you find that useful.

NB: if you use this method yourself, don’t forget to update the code with the correct path of your error templates.

About Roger Keays

Roger Keays is an artist, an engineer, and a student of life. He has no fixed address and has left footprints on 40-something different countries around the world. Roger is addicted to surfing. His other interests are music, psychology, languages, the proper use of semicolons, and finding good food.

For me, 3 lines was enough:

FacesContext fc = FacesContext.getCurrentInstance();
((HttpServletResponse) fc.getExternalContext().getResponse()).sendError(404);
fc.responseComplete();

Hello Roger,

I need your help, How do I implement this method in my app?

Thanks!

I just tried this little experiment:

<ui:composition xmlns:ui="http://java.sun.com/jsf/facelets"
                xmlns:f="http://java.sun.com/jsf/core"
                xmlns:h="http://java.sun.com/jsf/html">
   <h:outputText value="Hello World!"/>
   <h:outputText value="${ui.sendError(facesContext, 404, 'Woah')}"/>
   <h:outputText value="Error Sent"/>
</ui:composition> 

And it returned a 404 with «Hello World», my error page AND «Error Sent». It’s all combined because everything is running in the same FacesContext, but notice that the call to responseComplete() doesn’t terminate the rendering once it has already started.

You have to make sure it happens before (or while) the view is being built to prevent the render occurring. From most Facelets and JSTL tags it will work.

Thanks for that tip! Ran into the same issue recently, too. And thought about the same «I couldn’t believe something so basic should be so complicated»

I still have one question: Will this also work if JSF has already started rendering the page? I.e. not only using this approach to handle an error that occurs in a postConstruct or preRenderView method but for an error that comes from a method that is used e.g. in some rendered attribute? Have you tested that? Or would such an exception then be caught by a JSF ExceptionHandler (where you could use your sendError method)?

<web-app>  
   ...  
   <!-- A custom error page for error code == 500 -->  
   <error-page>   
     <error-code>500</error-code>   
     <location>/my-foo-bar-500-page.html</location>   
   </error-page>   
     
   <!-- A custom error page for exception type org.myapp.foo.bar.MyException -->  
   <error-page>   
     <exception-type>org.myapp.foo.bar.MyException</exception-type>   
     <location>/my-foo-bar-exception-page.html</location>   
   </error-page>   
   ...  
     
 </web-app>  

<web-app>  
   ...  
     
   <!-- A custom error page for exception type org.jboss.weld.context.NonexistentConversationException -->  
   <error-page>   
     <exception-type>org.jboss.weld.context.NonexistentConversationException</exception-type>   
     <location>/my-foo-bar-exception-page.html</location>   
   </error-page>   
   ...  
     
 </web-app>

<web-app>  
   ...  
     
   <!-- A custom error page for exception type org.jboss.weld.context.NonexistentConversationException -->  
   <error-page>   
     <exception-type>org.jboss.weld.context.NonexistentConversationException</exception-type>   
     <location>/my-foo-bar-exception-page.xhtml</location>   
   </error-page>   
   ...  
     
 </web-app> 

<web-app>  
   ...  
     
   <!-- A custom error page for exception type org.jboss.weld.context.NonexistentConversationException.  
     Notice the "nocid" parameter being passed to make sure that the non-existent conversation id  
     isn't passed to the error page  
   -->  
   <error-page>   
     <exception-type>org.jboss.weld.context.NonexistentConversationException</exception-type>   
     <location>/my-foo-bar-exception-page.xhtml?nocid=true</location>   
   </error-page>   
   ...  
     
 </web-app>