Session beans act as agents to the client, controlling workflow (the business process) and filling the gaps between the representation of data by entity beans and the business logic that interacts with that data. Session beans are often used to manage interactions between entity beans and can perform complex manipulations of beans to accomplish some task. Since we have only defined one entity bean so far, we will focus on a complex manipulation of the Cabin bean rather than the interactions of the Cabin bean with other entity beans. In Chapter 7, "Session Beans", after we have had the opportunity to develop other entity beans, interactions of entity beans within session beans will be explored in greater detail.
Client applications and other beans use the Cabin bean in a variety of ways. Some of these uses were predictable when the Cabin bean was defined, but most were not. After all, an entity bean represents data--in this case, data describing a cabin. The uses to which we put that data will change over time--hence the importance of separating the data itself from the workflow. In Titan's business system, for example, we will need to list and report on cabins in ways that were not predictable when the Cabin bean was defined. Rather than change the Cabin bean every time we need to look at it differently, we will obtain the information we need using a session bean. Changing the definition of an entity bean should only be done within the context of a larger process--for example, a major redesign of the business system.
In Chapter 1, "Introduction" and Chapter 2, "Architectural Overview", we talked hypothetically about a TravelAgent bean that was responsible for the workflow of booking a passage on a cruise. This session bean will be used in client applications accessed by travel agents throughout the world. In addition to booking tickets, the TravelAgent bean also provides information about which cabins are available on the cruise. In this chapter, we will develop the first implementation of this listing behavior in the TravelAgent bean. The listing method we develop in this example is admittedly very crude and far from optimal. However, this example is useful for demonstrating how to develop a very simple stateless session bean and how these session beans can manage other beans. In Chapter 7, "Session Beans", we will rewrite the listing method. This "list cabins" behavior is used by travel agents to provide customers with a list of cabins that can accommodate the customer's needs. The Cabin bean does not directly support the kind of list, nor should it. The list we need is specific to the TravelAgent bean, so it's the Travel- Agent bean's responsibility to query the Cabin beans and produce the list.
Before we get started, we will need to create a development directory for the TravelAgent bean, as we did for the Cabin bean. We name this directory travelagent and nest it below the com/titan directory, which also contains the cabin directory (see Chapter 4, "Developing Your First Enterprise Beans").
As before, we start by defining the remote interface so that our focus is on the business purpose of the bean, rather than its implementation. Starting small, we know that the TravelAgent will need to provide a method for listing all the cabins available with a specified bed count for a specific ship. We'll call that method listCabins(). Since we only need a list of cabin names and deck levels, we'll define listCabins() to return an array of Strings. Here's the remote interface for TravelAgent:
package com.titan.travelagent;
import java.rmi.RemoteException;
import javax.ejb.FinderException;
public interface TravelAgent extends javax.ejb.EJBObject {
// String elements follow the format "id, name, deck level"
public String [] listCabins(int shipID, int bedCount)
throws RemoteException;
}
Copy the TravelAgent interface definition into your IDE, and save it to the travel- agent directory. Compile the class to ensure that it is correct.
The second step in the development of any bean is to create the home interface. The home interface for a session bean defines the create methods that initialize a new session bean for use by a client.
Find methods are not used in session beans; they are used with entity beans to locate persistent entities for use on a client. Unlike entity beans, session beans are not persistent and do not represent data in the database, so a find method would not be meaningful; there is no specific session to locate. A session bean is dedicated to a client for the life of that client (or less). For the same reason, we don't need to worry about primary keys; since session beans don't represent persistent data, we don't need a key to access that data.
package com.titan.travelagent;
import java.rmi.RemoteException;
import javax.ejb.CreateException;
public interface TravelAgentHome extends javax.ejb.EJBHome {
public TravelAgent create()
throws RemoteException, CreateException;
}
In the case of the TravelAgent bean, we only need a simple create() method to get a reference to the bean. Invoking this create() method returns a Travel-Agent remote reference that the client can use for the reservation process. Copy the TravelAgentHome definition into your IDE and save it to the travelagent directory. Compile the class to ensure that it is correct.
Using the remote interface as a guide, we can define the TravelAgentBean class that implements the listCabins() method. The following code contains the complete definition of TravelAgentBean for this example. Copy the TravelAgentBean definition into your IDE and save it to the travelagent directory. Compile the class to ensure that it is correct. EJB 1.1 and EJB 1.0 differ significantly in how one bean locates another, so I have provided separate TravelAgentBean listings for each version.
Here's the code for the EJB 1.1 version of the TravelAgentBean:
package com.titan.travelagent;
import com.titan.cabin.Cabin;
import com.titan.cabin.CabinHome;
import com.titan.cabin.CabinPK;
import java.rmi.RemoteException;
import javax.naming.InitialContext;
import javax.naming.Context;
import java.util.Properties;
import java.util.Vector;
public class TravelAgentBean implements javax.ejb.SessionBean {
public void ejbCreate() {
// Do nothing.
}
public String [] listCabins(int shipID, int bedCount) {
try {
javax.naming.Context jndiContext = new InitialContext();
Object obj = jndiContext.lookup("java:comp/env/ejb/CabinHome");
CabinHome home = (CabinHome)
javax.rmi.PortableRemoteObject.narrow(obj, CabinHome.class);
Vector vect = new Vector();
CabinPK pk = new CabinPK();
Cabin cabin;
for (int i = 1; ; i++) {
pk.id = i;
try {
cabin = home.findByPrimaryKey(pk);
} catch(javax.ejb.FinderException fe) {
break;
}
// Check to see if the bed count and ship ID match.
if (cabin.getShip() == shipID &&
cabin.getBedCount() == bedCount) {
String details =
i+","+cabin.getName()+","+cabin.getDeckLevel();
vect.addElement(details);
}
}
String [] list = new String[vect.size()];
vect.copyInto(list);
return list;
} catch(Exception e) {throw new EJBException(e);}
}
private javax.naming.Context getInitialContext()
throws javax.naming.NamingException {
Properties p = new Properties();
// ... Specify the JNDI properties specific to the vendor.
return new javax.naming.InitialContext(p);
}
public void ejbRemove(){}
public void ejbActivate(){}
public void ejbPassivate(){}
public void setSessionContext(javax.ejb.SessionContext cntx){}
}
Examining the listCabins() method in detail, we can address the implementation in pieces, starting with the use of JNDI to locate the CabinHome:
javax.naming.Context jndiContext = new InitialContext();
Object obj = jndiContext.lookup("java:comp/env/ejb/CabinHome");
CabinHome home = (CabinHome)
javax.rmi.PortableRemoteObject.narrow(obj, CabinHome.class);
Beans are clients to other beans, just like client applications. This means that they must interact with other beans in the same way that client applications interact with beans. In order for one bean to locate and use another bean, it must first locate and obtain a reference to the bean's EJB home. This is accomplished using JNDI, in exactly the same way we used JNDI to obtain a reference to the CabinHome in the client application we developed earlier. In EJB 1.1, all beans have a default JNDI context called the environment context, which was discussed a little in Chapter 3, "Resource Management and the Primary Services". The default context exists in the name space (directory) called "java:comp/env" and its subdirectories. When the bean is deployed, any beans it uses are mapped into the subdirectory "java:comp/env/ejb", so that bean references can be obtained at runtime through a simple and consistent use of the JNDI default context. We'll come back to this when we take a look at the deployment descriptor for the TravelAgent bean below.
Once the EJB home of the Cabin bean is obtained, we can use it to produce a list of cabins that match the parameters passed. The following code loops through all the Cabin beans and produces a list that includes only those cabins with the ship and bed count specified:
Vector vect = new Vector();
CabinPK pk = new CabinPK();
Cabin cabin;
for (int i = 1; ; i++) {
pk.id = i;
try {
cabin = home.findByPrimaryKey(pk);
} catch(javax.ejb.FinderException fe){
break;
}
// Check to see if the bed count and ship ID match.
if (cabin.getShip() == shipID && cabin.getBedCount() == bedCount) {
String details = i+","+cabin.getName()+","+cabin.getDeckLevel();
vect.addElement(details);
}
}
This method simply iterates through all the primary keys, obtaining a remote reference to each Cabin bean in the system and checking whether its ship and bedCount match the parameters passed in. The for loop continues until a FinderException is thrown, which would probably occur when a primary key is used that isn't associated with a bean. (This isn't the most robust code possible, but it will do for now.) Following this block of code, we simply copy the Vector's contents into an array and return it to the client.
While this is a very crude approach to locating the right Cabin beans--we will define a better method in Chapter 7, "Session Beans"--it is adequate for our current purposes. The purpose of this example is to illustrate that the workflow associated with this listing behavior is not included in the Cabin bean nor is it embedded in a client application. Workflow logic, whether it's a process like booking a reservation or obtaining a list, is placed in a session bean.
Here's the code for the EJB 1.0 version of the TravelAgentBean:
package com.titan.travelagent;
import com.titan.cabin.Cabin;
import com.titan.cabin.CabinHome;
import com.titan.cabin.CabinPK;
import java.rmi.RemoteException;
import javax.naming.InitialContext;
import javax.naming.Context;
import java.util.Properties;
import java.util.Vector;
public class TravelAgentBean implements javax.ejb.SessionBean {
public void ejbCreate() {
// Do nothing.
}
public String [] listCabins(int shipID, int bedCount)
throws RemoteException {
try {
Context jndiContext = getInitialContext();
CabinHome home = (CabinHome)jndiContext.lookup("CabinHome");
Vector vect = new Vector();
CabinPK pk = new CabinPK();
Cabin cabin;
for (int i = 1; ; i++) {
pk.id = i;
try {
cabin = home.findByPrimaryKey(pk);
} catch(javax.ejb.FinderException fe) {
break;
}
// Check to see if the bed count and ship ID match.
if (cabin.getShip() == shipID &&
cabin.getBedCount() == bedCount) {
String details =
i+","+cabin.getName()+","+cabin.getDeckLevel();
vect.addElement(details);
}
}
String [] list = new String[vect.size()];
vect.copyInto(list);
return list;
} catch (javax.naming.NamingException ne) {
throw new RemoteException("Unable to locate CabinHome",ne);
}
}
private javax.naming.Context getInitialContext()
throws javax.naming.NamingException {
Properties p = new Properties();
// ... Specify the JNDI properties specific to the vendor.
return new javax.naming.InitialContext(p);
}
public void ejbRemove(){}
public void ejbActivate(){}
public void ejbPassivate(){}
public void setSessionContext(javax.ejb.SessionContext cntx){}
}
The most significant difference between this code and the EJB 1.1 code is the use of JNDI to locate the CabinHome:
Context jndiContext = getInitialContext();
CabinHome cabinHome = (CabinHome)jndiContext.lookup("CabinHome");
Beans interact with other beans in the same way that clients interact with beans. In order for one bean to locate and use another bean, it must first locate and obtain a reference to the bean's EJB home. This is accomplished using JNDI, in exactly the same way we used JNDI to obtain a reference to the CabinHome in the client application we developed earlier. If you take a close look at the method getInitialContext(), you will discover that it is exactly the same as the getInitialContext() method in the client classes defined earlier. The only difference is that the method is not static. You will need to change this code to match the correct settings for your EJB server. Once the EJB home of the Cabin bean is obtained, we can use it to produce our list of cabins that match the parameters passed.
The logic for finding beans with cabins that match the desired parameters is the same in EJB 1.1 and EJB 1.0. Again, it's a crude approach: we will define a better method in Chapter 7, "Session Beans". Our purpose here is to demonstrate that the workflow associated with this listing behavior is not included in the Cabin bean nor is it embedded in a client application. Workflow logic, whether it's a process like booking a reservation or obtaining a list, is placed in a session bean.
The TravelAgent bean uses an XML deployment descriptor similar to the one used for the Cabin entity bean. Here is the ejb-jar.xml file used to deploy the TravelAgent. In Chapter 10, "XML Deployment Descriptors", you will learn how to deploy several beans in one deployment descriptor, but for now the TravelAgent and Cabin beans are deployed separately.
<?xml version="1.0"?>
<!DOCTYPE ejb-jar PUBLIC "-//Sun Microsystems, Inc.//DTD Enterprise
JavaBeans 1.1//EN" "http://java.sun.com/j2ee/dtds/ejb-jar_1_1.dtd">
<ejb-jar>
<enterprise-beans>
<session>
<ejb-name>TravelAgentBean</ejb-name>
<home>com.titan.travelagent.TravelAgentHome</home>
<remote>com.titan.travelagent.TravelAgent</remote>
<ejb-class>com.titan.travelagent.TravelAgentBean</ejb-class>
<session-type>Stateless</session-type>
<transaction-type>Container</transaction-type>
<ejb-ref>
<ejb-ref-name>ejb/CabinHome</ejb-ref-name>
<ejb-ref-type>Entity</ejb-ref-type>
<home>com.titan.cabin.CabinHome</home>
<remote>com.titan.cabin.Cabin</remote>
</ejb-ref>
</session>
</enterprise-beans>
<assembly-descriptor>
<security-role>
<description>
This role represents everyone who is allowed full access
to the cabin bean.
</description>
<role-name>everyone</role-name>
</security-role>
<method-permission>
<role-name>everyone</role-name>
<method>
<ejb-name>TravelAgentBean</ejb-name>
<method-name>*</method-name>
</method>
</method-permission>
<container-transaction>
<method>
<ejb-name>TravelAgentBean</ejb-name>
<method-name>*</method-name>
</method>
<trans-attribute>Required</trans-attribute>
</container-transaction>
</assembly-descriptor>
</ejb-jar>
Other than the <session-type>and <ejb-ref> elements, this XML deployment descriptor should make sense since it uses many of the same elements as the Cabin bean's. The <session-type> element can be Stateful or Stateless to indicate which type of session bean is used.
The <ejb-ref> element is used at deployment time to map the bean references used within the TravelAgent bean. In this case, the <ejb-ref> element describes the Cabin bean, which we already deployed. The <ejb-ref-name> element specifies the name that must be used by the TravelAgent bean to obtain a reference to the Cabin bean's home. The <ejb-ref-type> tells the container what kind of bean it is, Entity or Session. The <home> and <remote> elements specify the fully qualified interface names of the Cabin's home and remote bean interfaces.
When the bean is deployed, the <ejb-ref> will be mapped to the Cabin bean in the EJB server. This is a vendor-specific process, but the outcome should always be the same. When the TravelAgent does a JNDI lookup using the context name "java:comp/env/ejb/CabinHome" it will obtain a remote reference to the Cabin bean's home. The purpose of the <ejb-ref> element is to eliminate network specific and implementation specific use of JNDI to obtain bean references. This makes a bean more portable because the network location and JNDI service provider can change without impacting the bean code or even the XML deployment descriptor.
Deploying the TravelAgent bean is essentially the same as deploying the Cabin bean, except we use a SessionDescriptor instead of an EntityDescriptor. Here is the definition of the MakeDD for creating and serializing a SessionDescriptor for the TravelAgentBean:
package com.titan.travelagent;
import javax.ejb.deployment.*;
import javax.naming.CompoundName;
import java.util.*;
import java.io.*;
public class MakeDD {
public static void main(String [] args) {
try {
if (args.length <1) {
System.out.println("must specify target directory");
return;
}
SessionDescriptor sd = new SessionDescriptor();
sd.setEnterpriseBeanClassName(
"com.titan.travelagent.TravelAgentBean");
sd.setHomeInterfaceClassName(
"com.titan.travelagent.TravelAgentHome");
sd.setRemoteInterfaceClassName(
"com.titan.travelagent.TravelAgent");
sd.setSessionTimeout(300);
sd.setStateManagementType(SessionDescriptor.STATELESS_SESSION);
ControlDescriptor cd = new ControlDescriptor();
cd.setIsolationLevel(ControlDescriptor.TRANSACTION_READ_COMMITTED);
cd.setMethod(null);
cd.setRunAsMode(ControlDescriptor.CLIENT_IDENTITY);
cd.setTransactionAttribute(ControlDescriptor.TX_REQUIRED);
ControlDescriptor [] cdArray = {cd};
sd.setControlDescriptors(cdArray);
CompoundName jndiName =
new CompoundName("TravelAgentHome", new Properties());
sd.setBeanHomeName(jndiName);
String fileSeparator =
System.getProperties().getProperty("file.separator");
if(! args[0].endsWith(fileSeparator))
args[0] += fileSeparator;
FileOutputStream fis =
new FileOutputStream(args[0]+"TravelAgentDD.ser");
ObjectOutputStream oos = new ObjectOutputStream(fis);
oos.writeObject(sd);
oos.flush();
oos.close();
fis.close();
} catch(Throwable t) { t.printStackTrace(); }
}
}
The MakeDD definition for the TravelAgent bean is essentially the same as the one for the Cabin bean. The difference is that we are using a SessionDescriptor instead of an EntityDescriptor and the bean class names and JNDI name are different. We do not specify any container-managed fields because session beans are not persistent.
After instantiating the javax.ejb.SessionDescriptor, the MakeDD application sets the remote interface and bean class names:
sd.setEnterpriseBeanClassName("com.titan.travelagent.TravelAgentBean");
sd.setHomeInterfaceClassName("com.titan.travelagent.TravelAgentHome");
sd.setRemoteInterfaceClassName("com.titan.travelagent.TravelAgent");
Next, we set two properties that control session timeouts (what happens if the bean is idle) and state management:
sd.setSessionTimeout(300); sd.setStateManagementType(SessionDescriptor.STATELESS_SESSION);
setSessionTimeout() specifies how many seconds the session should remain alive if it is not being used. In MakeDD we specify 300 seconds. This means that if no method is invoked on the session for over five minutes, it will be removed and will no longer be available for use.[3] If a method is invoked on a bean that has timed out, a javax.ejb.ObjectNotFoundException will be thrown. Once a stateful session bean has timed out, all of its accumulated state is lost. When a session bean times out, the client must create a new TravelAgent bean by invoking the TravelAgentHome.create() method. The setStateManagement() method determines whether the bean is stateful or stateless. At this point in it its development, the TravelAgentBean doesn't have any conversational state that needs to be maintained from one method to the next, so we make it a stateless session bean, which is more efficient. Both of these methods are unique to session descriptors; there are no corresponding methods in the EntityDescriptor class.
[3] Whether a session timeout is measured from creation time (the time the session bean is created) or from the time of last activity (when the last business method is invoked) is not clearly described in EJB 1.0. As a result, some vendors set the timeout relative to one of these two events (creation or last activity). Consult your vendor's documentation to determine your EJB server's timeout policy.
The next section specifies the default ControlDescriptor for the TravelAgentBean. These settings are the same as those used in the Cabin bean. The isolation level determines the visibility of the data being accessed. Chapter 8, "Transactions" explores isolation levels in more detail. The transactional attribute, TX_REQUIRED, tells the EJB server that this bean must be included in the transactional scope of the client invoking it; if the client is not in a transaction, a new transaction must be created for the method invocation, as follows:
ControlDescriptor cd = new ControlDescriptor();
cd.setIsolationLevel(ControlDescriptor.TRANSACTION_READ_COMMITTED);
cd.setMethod(null);
cd.setRunAsMode(ControlDescriptor.CLIENT_IDENTITY);
cd.setTransactionAttribute(ControlDescriptor.TX_REQUIRED);
ControlDescriptor [] cdArray = {cd};
sd.setControlDescriptors(cdArray);
The next section creates a JNDI name for TravelAgent's EJB home. When we use JNDI to look up the TravelAgentHome, this will be the name we specify:
CompoundName jndiName = new CompoundName("TravelAgentHome",new Properties());
Finally, the MakeDD serializes the SessionDescriptor to a file named TravelAgentDD.ser and saves it to the travelagent directory.
You will need to compile and run the MakeDD class before continuing:
\dev % java com.titan.travelagent.MakeDD com/titan/travelagent F:\..\dev>java com.titan.travelagent.MakeDD com\titan\travelagent
To place the TravelAgent bean in a JAR file, we use the same process we used for the Cabin bean. We shrink-wrap the TravelAgent bean class and its deployment descriptor into a JAR file and save to the com/titan/travelagent directory:
\dev % jar cf cabin.jar com/titan/travelagent/*.class META-INF/ejb-jar.xml F:\..\dev>jar cf cabin.jar com\titan\travelagent\*.class META-INF\ejb-jar.xml
You might have to create the META-INF directory first, and copy ejb-jar.xml into that directory. The TravelAgent bean is now complete and ready to be deployed.
To place the TravelAgent bean in a JAR file, we use the same process we used for the Cabin bean. First, we have to create a manifest file, which we save in the com/titan/travelagent directory:
Name: com/titan/travelagent/TravelAgentDD.ser Enterprise-Bean: True
Now that the manifest is ready, we can shrink-wrap the TravelAgent bean so that it's ready for deployment:
\dev % jar cmf com/titan/travelagent/manifest \ TravelAgent.jar com/titan/travelagent/*.class com/titan/travelagent/*.ser F:\..\dev>jar cmf com\titan\travelagent\manifest TravelAgent.jar com\titan\travelagent\*.class com\titan\travelagent\*.ser
The TravelAgent bean is now complete and ready to be deployed.
To make your TravelAgent bean available to a client application, you need to use the deployment utility or wizard of your EJB server. The deployment utility reads the JAR file to add the TravelAgent bean to the EJB server environment. Unless your EJB server has special requirements, it is unlikely that you will need to change or add any new attributes to the bean. You will not need to create a database table for this example, since the TravelAgent bean is using only the Cabin bean and is not itself persistent. Deploy the TravelAgent bean and proceed to the next section.
To show that our session bean works, we'll create a simple client application that uses it. This client simply produces a list of cabins assigned to ship 1 with a bed count of 3. Its logic is similar to the client we created earlier to test the Cabin bean: it creates a context for looking up TravelAgentHome, creates a TravelAgent bean, and invokes listCabins() to generate a list of the cabins available. Here's the code:
package com.titan.travelagent;
import com.titan.cabin.CabinHome;
import com.titan.cabin.Cabin;
import com.titan.cabin.CabinPK;
import javax.naming.InitialContext;
import javax.naming.Context;
import javax.naming.NamingException;
import javax.ejb.CreateException;
import java.rmi.RemoteException;
import java.util.Properties;
public class Client_1 {
public static int SHIP_ID = 1;
public static int BED_COUNT = 3;
public static void main(String [] args) {
try {
Context jndiContext = getInitialContext();
Object ref = (TravelAgentHome)
jndiContext.lookup("TravelAgentHome");
TravelAgentHome home = (TravelAgentHome)
// EJB 1.0: Use Java cast instead of narrow()
PortableRemoteObject.narrow(ref,TravelAgentHome.class);
TravelAgent reserve = home.create();
// Get a list of all cabins on ship 1 with a bed count of 3.
String list [] = reserve.listCabins(SHIP_ID,BED_COUNT);
for(int i = 0; i < list.length; i++){
System.out.println(list[i]);
}
} catch(java.rmi.RemoteException re){re.printStackTrace();}
catch(Throwable t){t.printStackTrace();}
}
static public Context getInitialContext() throws Exception {
Properties p = new Properties();
// ... Specify the JNDI properties specific to the vendor.
return new InitialContext(p);
}
}
The output should look like this:
1,Master Suite ,1 3,Suite 101 ,1 5,Suite 103 ,1 7,Suite 105 ,1 9,Suite 107 ,1 12,Suite 201 ,2 14,Suite 203 ,2 16,Suite 205 ,2 18,Suite 207 ,2 20,Suite 209 ,2 22,Suite 301 ,3 24,Suite 303 ,3 26,Suite 305 ,3 28,Suite 307 ,3 30,Suite 309 ,3
You have now successfully created the first piece of the TravelAgent session bean: a method that obtains a list of cabins by manipulating the Cabin bean entity.
 |  |  |
| 4.2. Developing an Entity Bean |  | 5. The Client View |
Copyright © 2001 O'Reilly & Associates. All rights reserved.