As we have discussed in previous months, the servlet container loads only a single copy of every servlet into memory at a given time. It can get away with this because Java is multithreaded, allowing a particular servlet to be executing simultaneously for several HTTP requests. Since JSPs are actually servlets in disguise, they are also subject to issues of multithreading.
This raises the question of what happens to our JavaBeans: how many times are they loaded; what is their scope? If a JSP sets a property that happens to modify a class' fields, does this affect all of the other instances of the same bean?
The answer is: It depends. There are four different types of bean scopes, and the type of scope that you choose will profoundly affect the way your bean is used. Application scope means a single copy of the bean for all JSPs is running within the servlet container. Session scope is for a single user, from the time they enter the site to the time they exit. If a user opens two browser windows onto your system, they will have identical sessions. Request scope extends through the end of an HTTP request. This is useful if you want to set a bean's properties in one JSP, use the <jsp:forward/> tag to perform an internal redirect to a second JSP and then continue to use that same bean from the second JSP. Page scope is for a single JSP page. When the page exits, so does the scope.
By default, beans are placed in the session scope. You can change this by adding a scope parameter to the <jsp:useBean/>:
<jsp:useBean id="simple" scope="application" class="il.co.lerner.SimpleBean"/>
Once we have done the above, properties set by one JSP will be visible to other JSPs. Of course, because beans in the application and server scopes might be executed by more than one JSP simultaneously, they must be threadsafe. Consider the scopes in which your beans are meant to be used, and make them threadsafe as necessary. If a bean is not threadsafe, be sure to indicate as much in its documentation so that others will not mistakenly use it in the wrong way.
Last month, we continued our simple investigations of web logs with a JSP that directly accessed a relational database to get the latest contents of a web log. While such a JSP is certainly legal, it looks awkward, is difficult to debug and fails to separate code from logic as elegantly as we might have hoped.
By putting the database logic into a JavaBean, we can achieve several of our goals: the code is reusable, available even to nonprogrammers and allows us to change the source of information or the application logic without rewriting our JSP. Listing 5 (available at ftp.linuxjournal.com/pub/lj/listings/issue86) contains the source code for our bean, which I have made threadsafe (using the “synchronized” keyword within the getBlog method) so we can create a single instance of application scope. This JavaBean connects to the web log database and retrieves the latest information. Listing 6 contains a small JSP that uses this JavaBean to display the web log.
There really isn't anything special about the bean in Listing 5, but it does bring together a number of things that we have been discussing in the last few months. We now have a way for nonprogrammers to access information in our web log without having to write a single line of code on their own! Using a few simple JSP tags, we can place our current blog contents in an HTML page quickly and easily.
JavaBeans are a wonderful way in which to reduce the amount of code we put in our JSPs, while making it easier to use. However, complex JSPs will still contain some code, such as when they have to iterate through loops or work with complex data.
Next month, we will see how we can write our own tags similar to the jsp: tags we saw this month, using JSP's custom actions facility. Therefore, we can create our own new tags, associating them with any code we wish. In this way, JSPs allow us to create our own new formatting language, as well as use the tags that have been provided.