With the demise of the Sun Engineer's Sandbox, my content posted there has disappeared. Thankfully I saved a copy. Thanks to srcerer on the ##jsf chat room on freenode for reminding me to repost this.


To celebrate the release of Ruby on Rails (RoR) 1.0, I have taken the flashconcept from Rails's ActionController and brought it to JSF. For those unfamiliar with RoR or the flash, the flash is basically a map whose entries have a finite, container enforced lifetime. The concept is familiar; it's the same old "scope" concept used in the "request", "session" and "application" classes in the Servlet spec. The point of the flash is to hold information you discover on this request, that you want to use on thenext request, but want to be forgotten after the next request is finished

Figure 1 is a cheesy graphical depiction of the request, flash, dialog, session, and application scopes.

The nice thing about these scopes is that they all have a Map into which you can stick things. These Maps have the special property that they will be cleared when the scope ends. In the case of the flash, the map is cleared after exactly two runs through the JSF request processing lifecycle.

I included Shale's Dialog concept for good measure since I believe this concept is important to the future of JSF For example, it's so important that JBoss Seam also provides a similar concept. A Shale dialog or Seam conversation is a scope whose length is determined by the application developer.


Of course, I needed a release-vehicle for this thing so I created a new glassfish sub-project on java.net called jsf-extensions. My intent is to use this project for a number of extensions to the core jsf implementation. The flash is just the first the first that is ready for public use. Others in the works include:

  • A completion and implementation of Jacob Hookom's JSF Avatar proposal

  • A generalization of the Sun JSF implementation's JSP tag library generator. This tool is used in the Sun implementation to generate the TLD and tag handlers for the html_basic taglib from the existing standard JSF component metadata in thestandard-html-renderkit.xml file. The idea is to generalize this tool so it can be used to generate the taglib for any JSF component.

  • A generalization of the PhaseListener approach used in the Sun Blueprints Solutions Catalog AJAX components to serve up JavaScript files so that it can serve up any file out of a component jar. This problem is solved in MyFaces using the Extensions Filter but a PhaseListener is a more self-container approach. This problem is also solved by the Weblets Project but personally I don't think the problem of loading resources is big enough to warrant an entire top-level project. Also, weblets has extra XML configuration steps that simply are not necessary if you use the PhaseListener approach.

But enough about the future of jsf-extensions, let's quickly examine how to use the flash by examining the sample war included in the jsf-extensions download.

Getting Started

  • Download and unpack the jsf-extensions download.

  • Download a recent build of glassfish, I used Build 31.

  • Start it up by going to the bin directory and running:

    ./asadmin start-domain domain1
  • Then deploy the sample app by executing:

    ./asadmin deploy --user admin --password adminadmin ~/jsf-extensions-1.0alpha1/wars/run-time-test-0.1.war

    Of course, you need to adjust your path to the war accordingly. The app will then be accessible at http://localhost:8080/run-time-test-0.1/. This will take you to an index page where you should click on the "Ruby on Rails-style Flash" link.

App Traversal

This app illustrates using the Flash from JSP. It is also possible to use the flash directly from Java Code, or via the EL API. These latter two methods are described in the JavaDocs.

As you may know from Jennifer and Pierre's article on the Unified EL, JSF expressions can be accessed in a "get" context, or a "set context", otherwise known as "rvalue" or "lvalue", respectively. Thefore, the EL Expression #{flash.foo} will store a value into the hash under the key "foo" when evaluated as an lvalue, while it will retrieve the value under key "foo" from the flash when evaluated as an rvalue.

In JSP pages, most expressions act as rvalues, whereas the lvalue behavior does not occurr until the form is submitted and the values entered by the user are stored into the expressions. However, to make it easier to set expression values from JSP, including setting into the flash, the jsf-extensions library includes the jsfExt:set tag. A simple usage of this tag found on the first page in the sample app, flash.jsp, is shown below.


<!DOCTYPE HTML PUBLIC "-//W3C//DTD HTML 4.01 Transitional//EN">
    <title>RoR Flash Test Page 1</title>
    <%@ taglib uri="http://java.sun.com/jsf/core"  prefix="f" %>
    <%@ taglib uri="http://java.sun.com/jsf/html"  prefix="h" %>
    <%@ taglib prefix="jsfExt" uri="http://java.sun.com/jsf/extensions" %>


  <h:form id="form1">

  <h:panelGrid columns="2" border="1" width="600">

    Store into the flash when rendering this page      

    <jsfExt:set var="#{flash.foo}" value="fooValue" />

    Value of <code>\#{flash.foo}</code>, should be <code>null</code>.

    <h:outputText value="#{flash.foo}" />

    <h:commandButton value="reload" />

    <h:commandButton value="next" action="next" />




We're using the panelGrid tag (in way we not possible before JSF 1.2, due to the problems pointed out in Hans Bergsten challenge) to lay out the contents in a simple two column table. The first row shows the use of thejsfExt:set tag to store a value into the flash. The second row shows how to get things from the flash. When viewing this page the first time, you won't see anything in the "Value of #{flash.foo} row" until you press the "reload" button in the page. This is because the normal use-case of the flash is to store things in this request that will be accessed on the next request.

If you want to store something in the flash for use on this request, click on the "next" button in the sample app for an example. In this page we use the "flash.now" syntax, shown below in JSP.

<jsfExt:set var="#{flash.now.bar}" value="barValue" />

<h:outputText value="#{flash.now.bar}" />

By inserting the special keyword "now" we tell the flash that this store operation should be accessible on this request, rather than the next request. If, during processing, you decide you want to promote the value from flash.now to the real flash, use "flash.next" as shown on the next page of the sample app.

<jsfExt:set var="\#{flash.now.buckaroo}" value="banzai" />

Value of <code>\#{flash.now.buckaroo}</code>, should be

<h:outputText value="#{flash.now.buckaroo}" />

Promote buckaroo to stick around for the next request.

<jsfExt:set var="#{flash.keep.buckaroo}" value="#{flash.now.buckaroo}" />

When you click on the next button, you will be shown a page that shows that the value in the flash has indeed survived the postback.

Show that buckaroo is still here.

<h:outputText value="#{flash.buckaroo}" />

This is a brief introduction to using the flash, a more in-depth article would cover a common real-world use-case, such as doing master-detail, and also cover the other two entry points to using the flash, the EL API, and direct programmatic use.

Technorati Tags: edburns