Building a Web Store with Struts & ADF Frameworks

Abstract

By exploring the details of a sample application built using two popular off-the-shelf J2EE frameworks, Apache Struts and Oracle ADF, this paper illustrates how developers can build J2EE-compliant applications with maximum developer productivity using a framework-based approach. In the process, it highlights the full lifecycle support that the Oracle JDeveloper 10g IDE provides for framework-based development using Struts and ADF.

Contents

Lessons from the Past

The initial release of Sun's Java Pet Store Demo was a watershed event. Thirsty for guidance on implementing real-world J2EE applications, Java developers dove into its cool pools of code like parched creatures of the Kalahari. But after exploring the depths of its refreshing routines, many returned to the surface wondering why application infrastructure code dominated the demo. Obscured by repetitive implementations of J2EE design patterns, the more interesting business functionality of the web storefront was hard to find.

On further analysis, one point was clear to developers: for their own applications they would need to reimplement the same design pattern drudgery. Common sense dictated a framework approach, but developers would have to decide whether to build their own or leverage existing ones. To make a more informed decision, they read books like Core J2EE Patterns: Best Practices and Design Strategies which gave the design patterns names, organized them into functional layers, and explained how a typical J2EE application should use fifteen key patterns together. Another book, EJB Design Patterns: Advanced Patterns, Processes, and Idioms, came with a handy poster in back, detailing twenty-one design pattern tips and diagrams for easy cubicle-wall reference. These and other resources clarified that correctly and efficiently coding all these patterns from scratch would be no trivial task.

While their instincts undoubtedly warned them otherwise, many developers opted anyway for the "do-it-yourself" approach on their first J2EE application projects. A year later, many were still struggling to deliver feature-complete, well-performing applications.

Advice for the Future

On the opening page of his book Expert One-on-One: J2EE Design and Development (Wrox Press), Rod Johnson offers an observation on this phenomenon:

The return on investment for many J2EE projects is disappointing. Delivered systems are too often slow and unduly complex. Development time is often disproportionate to the complexity of business requirements.

Why? Not so much because of the shortcomings of J2EE as because J2EE is often used badly. This often results from approaches to architecture and development that ignore real world problems. A major contributing factor is the emphasis in many J2EE publications on the J2EE specifications rather than the real world problems people use them to address. Many issues that commonly arise in real applications are simply ignored.

Throughout the rest of his book, Rod debunks many myths about J2EE development and offers pragmatic guidance about which J2EE technologies to use under what circumstances. On page 166, he begins a section on frameworks and how they can help:

Many common problems (beyond those addressed by J2EE application servers) have been solved well by open source or commercial packages and frameworks. In such cases, designing and implementing a proprietary solution may be wasted effort. By adopting an existing solution, we are free to devote all our effort to meeting business requirements.

After commenting that existing frameworks can mean a slightly steeper learning curve, Rod later motivates why this trade-off is worthwhile to gain a strong application infrastructure. On page 395, he clearly explains the benefits:

Using a strong standard infrastructure can deliver better applications, faster. A strong infrastructure makes this possible by achieving the following goals:

• Allowing application code to concentrate on implementing business logic and other application functionality with a minimum of distraction. This reduces time to market by reducing development effort, and reduces costs throughout the project lifecycle by making application code more maintainable (because it is simpler and focused on the problem domain). This is the ultimate goal, which many of the following goals help us to achieve.
• Separating configuration from Java code
• Facilitating the use of OO design by eliminating the need for common compromises.
• Eliminating code duplication, by solving each problem only once. Once we have a good solution for a problem such as a complex API we should always use that solution, in whatever components or classes that encounter the problem
• Concealing the complexity of J2EE APIs. We've already seen this with JDBC; other APIs that are candidate for a higher-level of abstraction include JNDI and EJB access
• Ensuring correct error handling. We saw the importance of this when working with JDBC in Chapter 9.
• Facilitating internationalization if required.
• Enhancing productivity without compromising architectural principles. Without adequate infrastructure, it is tempting to cut corners by adopting quick, hacky solutions that will cause ongoing problems. Appropriate infrastructure should encourage and facilitate the application of sound design principles.
• Achieving consistency between applications within an organization. If all applications use the same infrastructure as well as the same application server and underlying technologies, productivity will be maximized, teamwork more effective, and risk reduced.
• Ensuring that applications are easy to test. Where possible, a framework should allow application code to be tested without deployment on an application server.

Several existing application frameworks provide ready-to-use implementations of the kind of strong application infrastructure that Rod recommends. If you use these frameworks, you won't have to design, code, debug, and maintain your own infrastructure code.

In this whitepaper, we examine two existing J2EE frameworks by studying a working sample application. By patterning the sample application after the "classic" Java Pet Store Demo, we've made it easier for readers familiar with the original demo to compare the developer productivity that a framework-based J2EE development approach can provide.

Rebuilding a Web Storefront with Struts and ADF

The ADF Toy Store demo is a simple web storefront application adhering to the Model/View/Controller (MVC) design pattern. It is implemented using two existing J2EE application frameworks: Apache Struts and Oracle Application Development Framework (ADF). Both the Struts and ADF frameworks have been iteratively developed to support the requirements of communities of application developers building real-world applications. Many aspects of their design and implementation echo the pragmatic suggestions that Rod Johnson details throughout his book.

As with all MVC-style web applications, the ADF Toy Store has the basic architecture illustrated in Figure 1:

• The model layer represents the business information needed by the application,
• The controller layer handles user input, interfaces with the model layer, and picks the presentation
• The view layer presents the model data to the end-user.

The model layer consists of one or more business services that expose application functionality and access to model data through a business service interface that is easy to test. These business services, in turn, rely on query components to retrieve that data and on business objects to validate and persist any new or modified data. Code implementing the business delegate design pattern abstracts the details of locating and using the business services. When JavaServer pages are used for the view layer along with a cleanly separated controller layer, many J2EE books refer to the architecture, shown in Figure 1, as a best practices "JSP Model 2" architecture. The number "2" is used because this MVC-based architecture for JSP is an evolution over first-generation JSP-based approaches.

Figure 1: Best Practices "JSP Model 2" MVC Web Application Architecture

By dissecting the framework-based implementation of our ADF Toy Store demo, we'll learn how ADF simplifies building all aspects of the model layer, and how the Struts and ADF frameworks cooperate to simplify implementing the view and controller layers. In the process, we'll also see plenty of evidence for how Oracle JDeveloper 10g provides a productive environment covering the full development lifecycle for building these kinds of MVC-style business applications.

Before diving into the explanation of the demo, let's make sure you can open and run the demo in JDeveloper 10g. The next section details the steps to get the demo setup correctly on your system.

Demo Installation and Setup

These instructions assume that you are running Oracle JDeveloper 10g production, version 9.0.5.1 or 9.0.5.2. The demo will not work with earlier versions of JDeveloper.

We also assume that you have access to an Oracle database, and privileges to create new user accounts to setup the sample data.

1. Download the adftoystore.zip file if you haven't already.

2. Extract the contents of the adftoystore.zip file with the standard JDK jar utility into a convenient directory.

   jar xvf adftoystore.zip

   This will create a directory adftoystore, and subdirectories. These instructions assume you've extracted the adftoystore.zip file into the root directory C:\ thus ending up with a demo "root" directory of C:\adftoystore.

   
   

   NOTE:

   If the jar command does not work on your system, double-check that you have included the JDKHOME/bin subdirectory in your system path. If you downloaded the full version of Oracle JDeveloper 10g, then it comes with a 1.4.2 JDK in the JDEVHOME/jdk directory.

   
   

3. Create the TOYSTORE and TOYSTORE_STATEMGMT user accounts in the database using the provided SQL script.

   Run the SQL script ./adftoystore/DatabaseSetup/CreateToyStoreUsers.sql like this:

   cd C:\adftoystore\DatabaseSetup
sqlplus /nolog @CreateToyStoreUsers.sql

   After entering your SYS account's password, the script will create the TOYSTORE and TOYSTORE_STATEMGMT user accounts, and then GRANT EXECUTE ON DBMS_LOCK TO TOYSTORE. The TOYSTORE schema will contain the ADF Toy Store application tables, while the TOYSTORE_STATEMGMT schema will be used by the ADF state management facility (described later in this whitepaper) to store pending data across web pages. The DBMS_LOCK.SLEEP packaged procedure is used by the LOCK_ALL_ITEMS_ORDERDED stored procedure explained in the Example of Accessing Stored Procedures section later in the paper.

4. Create the application tables for the ADF Toy Store demo, along with some sample data.

   Run the SQL script ./adftoystore/DatabaseSetup/ToyStore.sql like this:

   sqlplus toystore/toystore @ToyStore.sql

   
   

   NOTE:

   If you have a version of the Oracle database prior to Oracle 10g, the command purge recyclebin at the end of this script will give an error. It's harmless and you can just ignore it.

   
   

5. Setup two database connections in the JDeveloper 10g IDE corresponding to the two database accounts we created above.

   Define connections in the JDeveloper 10g IDE named...

   • toystore, corresponding to the TOYSTORE user (password TOYSTORE) and
   • toystore_statemgmt corresponding to the TOYSTORE_STATEMGMT user (password TOYSTORE).
   
   

   NOTE:

   The two connection names are case-sensitive and should be typed in lowercase as shown.

   
   

   To save some typing, you can import these two connections from the supplied jdev_toystore_connections.xml file in the ./adftoystore/DatabaseSetup directory. To do so, select the Database category folder in the Connections Navigator and choose Import Connections... from the right-mouse menu. Supply the jdev_toystore_connections.xml file name as the file to import from. After importing the two named connections, you should test each connection by selecting it, double-clicking to bring up the Connection Wizard, and visiting the Test tab. If clicking on the Test Connection button does not yield a "Success!" message, then correct the connection details on the Connection tab to work for the database to which you want to connect. By default, the connections are defined against a database on your local machine listening on port 1521 with a SID of ORCL.

6. Insure that the JUnit Extension for JDeveloper is installed.

   JUnit is the defacto standard tool for building regression tests for Java applications. Oracle JDeveloper 10g features native support for creating and running JUnit tests, but this feature is installed as a separately-downloadable IDE extension. You can tell if you already have the JUnit Extension installed by selecting File | New... from the JDeveloper main menu, and verifying that you have a Unit Tests (JUnit) subcategory under the General top-level category in the New Gallery.

   If you do not already have the JUnit extension installed, then download it from here. You'll find it along with all the other extensions available for JDeveloper in the JDeveloper Extension Exchange on OTN. To complete the installation of the extension, first exit from JDeveloper if you are currently running it. With JDeveloper not running, extract the contents of the downloaded zip file into the ./jdev/lib/ext subdirectory under your JDeveloper installation home directory. Then, restart JDeveloper.

   Finally, you should verify that the junit3.8.1 subdirectory exists in your JDeveloper installation home. This directory will automatically get created the first time you run any JUnit wizard from the Unit Tests (JUnit) category of the New Gallery. However if you don't plan on creating any JUnit tests yourself yet, you can do the following steps to make sure the directory gets setup correctly. Assuming your current directory is the JDeveloper installation home directory...

   • jar xvf jdev/lib/ext/junit_addin.jar junit3.8.1.zip

     This extracts the junit3.8.1.zip file from the junit_addin.jar archive. This zip file contains the distribution of JUnit with which JDeveloper has been tested.

   • jar xvf junit3.8.1.zip

     This extracts the contents of the junit3.8.1.zip file into the JDeveloper installation home directory.

7. Open the ./adftoystore/ADFToyStore.jws workspace in JDeveloper 10g

8. Run the application inside the JDeveloper 10g IDE by running the index.jsp page in the ToyStoreViewController.jpr project as shown in Figure 2.

   
   Figure 2: Running the ADF Toy Store Application Inside JDeveloper 10g
   

   NOTE:

   Since index.jsp is configured as the Default Run Target on the Runner panel of the project properties for the ToyStoreViewController project, you can also simply click the Run icon in the IDE toolbar when this project is active to run the application, or pick the Run menu item on the ToyStoreViewController project's right-mouse menu. You can see the project properties by clicking on it in the Navigator and selecting Property Properties... from the right-mouse menu.

   
   

   Running the index.jsp page from inside JDeveloper will startup the embedded Oracle Application Server 10g Oracle Containers for J2EE (OC4J) server, launch your default browser, and cause it to request the URL:

   http://yourmachine:8988/ADFToyStore/index.jsp

   If everything is working correctly, you will see the home page of the ADF Toy Store demo, as shown in Figure 3.

   
   

   NOTE:

   If following the steps above didn't produce the above demo home page as expected, see Appendix 1 for a list of known issues and troubleshooting tips.

   
   
   Figure 3: ADF Toy Store Demo Home Page

Quick Tour Through the Demo

Before we dive into explaining how the demo was built, let's begin with a quick overview of the end-user functionality of our web storefront application.

Browsing Products and Adding Them to Your Cart

The ADF Toy Store is a fictitious online store that sells toys. The products for sale are organized into five categories: Accessories, Games, Party Supplies, Toys, and Models. From the home page, you can browse products in the store in two ways:

1. Selecting a category name to see the products in that category, or
2. Using the What are you looking for? search box in the banner to find products by name, regardless of what category they belong to.

If the list contains more than three products, they are presented a page at a time. You can use the Next or Previous links that appear above the item list to browse through the complete list.

Clicking on the name of a product shows you a list of the different product items for sale. For example, clicking on the name of a product like Pinata, you will see a list of the different kinds of piñatas that are available as shown in Figure 4.

Figure 4: Browsing Different Kinds of Items for a Product Type

To see a detailed description and a picture of any product, just click on its name.

On any page where the button appears, you can click on it to add one of those items to your shopping cart.

You can see what items you have in your shopping cart at any time by clicking on the button, which shows a page listing the items and quantities you have selected so far, as shown in Figure 5.

Figure 5: Shopping Cart Display

To adjust the quantities of the items in the cart, just type over the current value in the Quantity field for one or more items, and click the button to see the recalculated shopping cart total. You can remove an item from your cart either by clicking the button, or by adjusting the item to have a zero quantity.

Checking Out and Signing In

From the Shopping Cart page, by clicking on the button you can proceed to the Review Checkout page. From there, you can review your purchase and if you are happy with it, press the button to continue.

If you have not already signed into the Toy Store as a registered user, you will be prompted to sign in at this point to continue with the checkout process. The sign in page looks like what you see in Figure 6. The user named j2ee is already registered, with a password of j2ee, so you can provide these credentials to continue.

Figure 6: ADF Toy Store Sign-in Page

After successfully signing in, you will proceed to the page where you can confirm your shipping and payment details. Here, to see some of the application validation logic that's implemented in the demo, you can try:

• Entering an invalid state abbreviation of ZA for the country USA
• Entering a credit card number that is not comprised of 16 digits

and pressing the button. You should see the multiple validation errors as shown in Figure 7 .

Figure 7: Shipping Information Validation Errors On Form Submission

After fixing those errors by entering a valid state abbreviation like CA, and filling out a full 16-digit credit card number, try causing some additional validation errors by:

• Entering a date in the past for the expiration date of your credit card
• Blanking out a required field like Last Name.

You should again see the relevant set of remaining validation errors that need to be corrected when you press the button again as shown in Figure 8.

Figure 8: Additional Shipping Information Validation Errors

After correcting these final validation problems and submitting again, your order will be placed, and you'll see the final "Thank You" page, with a reference to your order reference number. Clicking on the hyperlinked order reference number takes you to an order summary page which is implemented using the XML/XSLT-based Oracle XSQL Pages publishing framework instead of JSP Pages, to illustrate that multiple view-rendering technologies are possible.

Register a New User and Editing an Existing User's Profile

If you are not currently logged-in as a registered user of the web store, clicking the button brings you to the "Sign In" page as shown in Figure 6. From there, you can register as a new user by clicking on the Register as a New User link. This brings you to a form to complete with the necessary registration details.

This user registration page is another place in our application where it's easy to observe how business rules get enforced by the ADF framework. For example, if while filling out the form you try to:

• Enter a user name that has already been chosen by another user
• Forget to provide a password
• Enter an email address that is not properly formed

then when you submit the form, you'll see the full set of errors related to your registration as shown in Figure 9

Figure 9: Validation for New User Accounts In Action

If you are already logged into the site as a registered user, you will see the icon in the toolbar. Clicking on it brings you the page where you can edit your account details as shown in Figure 10. Of course, since we're working with the same underlying business object representing user accounts here in this "Update Account" form, the same validation will be enforced as above.

Figure 10: Editing Account Details

Trying Out the Demo in Another Language

The demo is built using the internationalization features supported by Struts and ADF Business Components, and it ships with support for three languages: English (the default), Italian, and German. The Struts and ADF frameworks automatically sense the language you want to see based on your browser settings. So, you can see what the demo looks like in Italian by simply setting your browser language preferences appropriately.

In Mozilla Firefox 0.9, you select your preferred languages on the General panel of the Tools | Options... menu by clicking the (Languages) button as shown in Figure 11. You can add "Italian [it]" and then press (Move Up) to move it to the top of the list.

Figure 11: Changing the Preferred Language in Mozilla Firefox 0.9

In Internet Explorer, you can do the same thing from the General tab of the Tools | Internet Options... dialog by clicking on the (Languages...) button.

Since Apache Struts caches the browser-user's preferred language at the servlet session level, you will need to close your browser window, and open a new one before you'll see the demo change into Italian. A quick way to relaunch your preferred browser with the right URL for the demo is to find the Target URL in the JDeveloper 10g log window as shown in Figure 12 and click on it.

Figure 12: Clicking on the Target URL in Log to Relaunch Browser

This will bring up the ADF Toy Store demo again, but this time in Italian. After adding the same items to your shopping cart again, it will look like what you see in Figure 13.

Figure 13: Shopping Cart With Preferred Browser Language Set to Italian

You can set your browser's preferred language back to English using similar steps, and close and relaunch the browser window to proceed in English again.

Experimenting with State Management and Failover Support

The last aspect of the demo we'll explore on our tour is a practical example of the ADF framework's state management and failover support for ADF Business Components. The feature sounds complicated, but it's easy to demonstrate. With the ADF Toy Store application running inside the embedded OC4J container that's part of the JDeveloper IDE, try the following:

1. Add several items to your shopping cart

2. Without closing your browser window, terminate the OC4J application server to simulate a hardware failure on your application server machine.

   To do this, select the View | Run Manager menu option to display the Run Manager. Find the Embedded OC4J Server process in the list, and select it. Finally, choose the Terminate menu option from the right-mouse menu item as shown in Figure 14.

   
   Figure 14: Terminating the OC4J Server to Simulate a Server Failure
3. Re-run the ADF Toy Store demo as we did in Figure 2.

After restarting the application server -- in this example, we've restarted the embedded OC4J application server in JDeveloper -- the browser window you left open in step 2 above will be able to continue where it left off, with all shopping cart items intact.

This failover capability works because the ADF framework offers automatic database-backed state management for pending data in your application when using the ADF Business Components. In the ADF Toy Store application, the pending shopping cart information is not stored in the HTTP session state the way most applications do. Instead, with a declarative checkbox on the ShoppingCart component at design time, we indicate that we'd like this component's pending data to be managed for us. And the framework takes care of the rest.

At this point we've seen the key functionality in the demo, so it's time to dive in to understand how it has all been built.

Dissecting the Demo

In this section we explain the demo in detail, highlighting the interesting details of how:

• The demo is architected into Model, View, and Controller layers
• The Model layer uses ADF's business service, data access, and business object components
• The business services can be tested using JUnit
• The Controller layer uses Struts actions to coordinate application flow
• The View layer uses standard Struts and JSTL tag libraries to simplify building the web UI
• The ADF features for seamless Struts integration work
• The features of Struts and ADF are used to deliver a multilingual application
• The your can use ADF with other view layer technologies like Oracle XSQL Pages
• The default framework behavior can be customized fit your needs.

How the Application is Organized Into Packages and Projects

Like all applications built in Java, the ADF Toy Store demo is comprised of a set of classes, organized hierarchically into packages. Figure 15 illustrates the key packages in the demo. We have used the package naming to make it clear how the application classes break down into model, view, and controller layers, as well as to clarify which classes are part of our framework customizations and regression testing suite.

Figure 15: Java Package Hierarchy for ADF Toy Store demo

When building applications that leverage existing frameworks, your application-specific classes inherit default functionality from an appropriate framework base class. They inherit core behavior from their superclass, and add application-specific logic and metadata. Typically, the only code needed in your classes is the code that is specific to your application's business functionality. Figure 16 shows some representative examples of classes in the ADF Toy Store demo that inherit their behavior from a framework:

• The main business service component toystore.model.services.ToyStoreService extends the ADF framework base class oracle.jbo.server.ApplicationModuleImpl, adding custom business service methods and an application-specific "data model" of named collections of data transfer objects (also known as value objects) exposed to the client.
• An example query component toystore.model.dataaccess.ProductsInCategory extends the ADF framework base class oracle.jbo.server.ViewObjectImpl, adding an application-specific SQL query for products in a particular category and providing custom methods to encapsulate the setting of its bind parameters.
• An example business object toystore.model.businessobjects.Account extends the ADF framework base class oracle.jbo.server.EntityImpl, adding application-specific attributes for user accounts and specifying several business rules that govern an account's validity.
• The data transfer object toystore.model.dataaccess.common.ShoppingCartRow extends the ADF framework base interface oracle.jbo.Row, adding typesafe access to the application-specific attributes in the row of shopping cart information.
• The form bean toystore.controller.strutsformbeans.LoginForm extends the Struts framework base class org.apache.struts.action.ActionForm, adding application-specific bean attributes like Username and Password that will be entered by the user to sign in to the Toy Store web site.
• The action toystore.controller.strutsactions.PlaceOrderAction extends the Struts framework base class org.apache.struts.action.Action, adding application-specific controller logic needed before rendering the HTML form to collect shipping information for the order being placed.
• The test case toystore.test.unittests.CreateAnOrderTest extends the JUnit framework base class junit.framework.TestCase, adding application-specific testing logic that exercises the ToyStoreService business service by simulating the creation of an order after adding items to the shopping cart.

Figure 16: Example of Demo Classes that Extend Frameworks

JDeveloper 10g provides two constructs to organize our work: workspaces and projects. Projects contain a set of files that get compiled (and perhaps deployed) as a unit, and workspaces are a list of projects that go together to comprise a complete application. Theoretically, we can build any application with all of the files in a single project, but typically we organize our work into a number of separate projects to divide up the work into more logical groupings.

As shown in Figure 17, the ADF Toy Store application is comprised of a ADFToyStore workspace containing the following six projects:

1. ToyStoreModel.jpr

   This project contains the components in the toystore.model.* package tree, including the main business service toystore.model.services.ToyStoreService and all the business object and data access components on which it relies to provide its application functionality and model data to the client. It also contains the translated resources (in English, Italian, and German) related to these components.

2. ToyStoreViewController.jpr

   This project contains

   • the JSP pages that comprise the user interface of the web store, and the view-layer resource files in the toystore.view package of the translatable text that appears in all the pages.
   • the Struts configuration file struts-config.xml and the source code for all of the classes in the toystore.controller.* package tree. This includes the Struts actions that coordinate the interaction between the business service and the view-layer pages.

3. FwkExtensions.jpr

   This project contains the classes in the toystore.fwk.* package tree that extend the base ADF and Struts framework facilities to augment and/or customize the default framework behavior. These customizations are not specific to the web storefront and could be easily reused in another Struts/ADF application.

4. Testing.jpr

   This project contains the classes in the toystore.test.* package tree, including a JUnit regression test suite, test fixture, and unit tests for various aspects of the ToyStoreService component.

5. DatabaseSetup.jpr

   This project contains the two SQL scripts used to setup the database schema for the demo, as well as a JDeveloper database diagram of the Toy Store database design.

6. Documentation.jpr

   This project contains a copy of this whitepaper in the readme.html file.

Figure 17: The ADFToyStore Workspace in the JDeveloper Application Navigator

Advantages of a Model/View/Controller Architecture

First generation JSP applications freely mixed code "scriptlets" into the page among the HTML presentation tags. The code for parameter evaluation, data access, business rules enforcement, transaction management, error handling, and page flow was simply typed right into the same JSP file that would also eventually format the data for the end-user to see. Having everything in one file and being able to see compilation errors by refreshing the browser lent an immediacy to development that enticed many developers to follow this approach. However, this hybrid approach more often than not produced pages that were impossible to read. Attempts to alter the look and feel of the pages, unless performed by the original developer, could lead to hours of staring at the file, hunting for the unintended typographical error.

Code scriptlets in JSP pages began to fall out of favor as JSP 1.1's tag libraries allowed many common tasks to be performed using easier-to-read elements and attributes. However, the popularity of tag libraries that performed SQL data access or EJB component interaction directly from the JSP page was still an indication that developers were not correctly separating the presentation layer from the application layer. In these first generation JSP applications, the model, view, and controller layers were hopelessly intertwined.

As these applications evolved, attempts to respond quickly to new business needs requiring an updated look and feel or modified web page flow were greatly complicated by this "heavy page" approach. Developers bitten by the maintenance nightmares of the first-generation approach immediately understood the benefits that the Model, View, and Controller separation has to offer. In a nutshell, with an MVC architecture:

• Application look and feel can change without affecting core application logic
• Page flow and error handling are centralized and removed from individual pages
• Simpler-looking web pages can be understood and modified by less technical team members

With its advantages now clear, let's begin to look at how our Toy Store demo implements the Model, View, and Controller layers of its architecture.

Implementing the Model Layer Using ADF Business Components

The model layer is comprised of business services, query components, business objects, and collections of data transfer objects that the business service exposes to the controller and view layers. In this section we'll highlight some examples of these model layer components from the ADF Toy Store demo and briefly explain how they leverage the ADF framework for their implementation.

Considering Model Layer Approaches: EJB-Centric or Web-Tier-Centric?

Before exploring the ADF Toy Store model layer implementation in detail, we should first stop to consider the important choice of whether the model layer will be implemented using:

• EJB Technology deployed to the J2EE EJB Tier, or
• JavaBeans Technology deployed to the J2EE Web Tier.

As illustrated by the two separate sample applications provided by Sun's "J2EE Blueprints" demo team, the approach you choose for your model layer can have a major impact on the application's underlying implementation. The architecture documentation that accompanies the more recent Adventure Builder demo explains:

The Java Pet Store application illustrates how to write a Website application in an EJB-centric manner. The Adventure Builder application illustrates the other option: how to write a Website application in a Web-centric manner. EJB is a key technology in the J2EE platform, but not all J2EE applications need to use it.

The document goes on to explain some of the motivations behind making the choice:

One important design consideration is mapping application modules and functionality to the different tiers and technologies on the J2EE platform. Some choices are obvious, such as having a web tier when a web browser client is required. Other choices may depend on several factors. Issues such as data access and transactional needs, security, portability and modularity of design, lead to deciding how to optimally map the application modules to the client, Web, EJB, and EIS (data storage) tiers. An important question is whether to use an EJB tier. Based on the application's needs, one might choose not to use enterprise beans and the EJB container and tier. The expertise of the development team also affects this decision. For example, a team with strong Web-tier and SQL skills may find it easier to write a Web-only application especially when they are new to the EJB technology and are pressed for time to learn it.

Using the ADF framework, you build your J2EE application using a consistent development approach that is independent of your choice of deployment tier for your model layer. You develop, test, and debug the application using a model layer built from high-performance, well-architected, XML-configured JavaBeans. At any time during the development process, you can choose to deploy your model layer as JavaBeans to the J2EE Web Tier, or as an EJB Session Bean to the EJB tier. Some business requirements that might nudge you in the direction of an EJB tier deployment include the need to:

• Coordinate ADF-powered services with other Session Beans in the same transaction
• Leverage method-level security on your ADF-backed services.

Since the ADF framework provides an implementation of the best-practices Business Delegate design pattern, your model and view layers are isolated from these deployment details. Even if you change your mind mid-project on your preferred model-layer deployment architecture, none of your application code needs to change. In fact, you can try out both deployment options and pick the one that delivers best performance for your particular application scenario. In other words, using the ADF framework, you don't have to decide up front on an EJB-Centric or Web-Tier-Centric approach, and you can change your mind at any time, without rearchitecting your system.

For the purposes of this demo, we have selected to deploy the ADF Toy Store demo's model layer to the J2EE web tier to keep the demo as easy to follow as possible for the widest audience of Java developers. For the reasons we've just mentioned, redeploying the model layer to the EJB Tier would be a painless step for those wanting to take an EJB-centric approach.

Implementing Business Services with ADF Application Module Components

While Oracle ADF supports using virtually any kind of Java class as a business service, the ADF Business Components option we provide gives you the highest level of built-in application-building functionality and developer productivity. Business services built using the ADF business components are called application modules. These service components are:

• Cleanly architected with a client-side business service interface and server-tier implementation
• Efficiently implemented as JavaBeans, but deployable as EJB Session Beans as necessary, with support for container-managed transactions
• Automatically configured at runtime from XML metadata and created through framework-supplied factories
• Easily used by clients through ADF's implementation of the Business Delegate design pattern
• Cleverly designed to expose "active" collections of updateable data transfer objects that interact with your business objects without code

All of these features can be summarized by saying that ADF-powered service components make the J2EE developer's life a lot simpler. The key ADF framework components that cooperate to provide the business service implementation are:

• Application Modules to build transactional business services
• View Objects and View Links to build collections of updateable data transfer objects based on SQL queries
• Entity Objects and Associations to encapsulate business rules and persistence details of domain business objects and express the relationships between them
• Domains - to build custom datatypes, where necessary

Our toystore.model.services.ToyStoreService application module is the heart of our application. It is a JavaBean component that implements the business service interface shown in Example 1.

As shown in Figure 18, the ToyStoreService component is implemented as a set of files. The Application Navigator presents the component as a single, logical icon, while the Structure Window's Sources folder shows the detailed implementation files comprising that component:

• ToyStoreService.xml - Service definition file
• ToyStoreService.java - Service interface
• ToyStoreServiceImpl.java - Service implementation
• ToyStoreServiceClient.java - Service client proxy

The client proxy class is created when you've exposed custom methods to be accessed by clients, alongside the custom service interface that the ADF design time creates for your component. The client proxy class implements your custom component interface, and is used automatically at runtime when client layer is deployed in a separate tier from the business service layer or when using Batch Mode.

The application module editor in JDeveloper automatically keeps the XML component definition file and business service interface in sync with the declarative options you set using the editor. For example, the business service interface automatically appears in your project as soon as you mark any custom methods as part of the service interface on the Client Methods panel in the editor. As we'll see in the other sections below, all ADF business components follow this basic pattern for the names of the files that comprise their definition, implementation, and interface.

Figure 18: Structure Window Showing Sources Comprising ToyStoreService Application Module

If you have a look inside the ToyStoreServiceImpl.java file, you'll see it contains the code implementing the business service interface methods, and some JDeveloper-generated convenience methods to access collections of data transfer objects. In order to more clearly identify custom code from JDeveloper-generated code, we've surrounded all of the developer-written, application-specific code with marking comments like:

//--[Begin Custom Code]--

and

//--[End Custom Code]--

You'll see these same marking comments in all of the ADF-based JavaBean components in the application.

Exposing Model Data to the View and Controller Layers

When implementing Model/View/Controller (MVC) applications by hand, best practice techniques steer developers to expose model data to the controller and view layers using a HashMap object. This "model data map" gives the client layers a single object that represents the entire "data model" needed for the current application task. Using the model data map, the controller and view layers can easily find any collections of data transfer objects using a convenient string key name.

For example, the model data required to display the summary of an order might include:

• Account information for the customer placing the order
• Order information
• Order line item information to show the items and quantities the customer purchases
• Shipping option information to drive a poplist of delivery choices

Figure 19 illustrates what the model data map object would look like for such a task.

Figure 19: Find Named Collections of Data Transfer Objects Using a Model Data Map

Example 2 shows the typical code required to find the collection of data transfer objects for line items from a model data map.

ADF Application Modules Implement Your Model Data Map For You

The model data map discussed above is a necessary feature of any MVC application. Business services implemented as ADF application modules inherit a built-in model data map implementation. The application module cooperates with ADF view object components to allow you to build your model data map declaratively.

You create view object components to encapsulate SQL statements that will produce the required collections of data transfer objects. Then you define your model data map at design-time by adding instances of these view object components to your application module using the appropriate panel of the Application Module Editor shown in Figure 20. The names that appear in the Data Model list on the right are the string key names that you'll use at runtime to find the collection of data transfer objects produced by this view object instance. Of course, you can pick any names you like.

Figure 20: Declaratively Define Your Model Data Map Using the Application Module Editor

The names appear in a tree control to illustrate visually any master/detail coordination that the ADF framework is doing on your behalf among the collections of data transfer objects. For example, the indentation in Figure 20 shows that the collection named Orders will automatically contain only those orders for the current account data transfer object in the Accounts collection, and the LineItems collection will contain the line items for the current order data transfer object.

Figure 21 shows the same ToyStoreService application module in a visual way using the UML diagramming support for ADF components in JDeveloper 10g. Each of the named collections in its data model map appears, and master/detail relationships are indicated with their cardinality using lines and arrows between the collections.

Figure 21: Visual Diagram of the ToyStoreService's Data Model

Since the application module component implements the model data map for you, at runtime the view or controller layer can lookup a particular collection of data transfer objects by name using the instance of the application module service component that it is working with using syntax as shown in Example 3.

If you do not want to work with typesafe collections of data transfer objects, you can opt to work with ADF's generic collection implementation (oracle.jbo.RowSet) and generic data transfer object implementation (oracle.jbo.Row) by writing code like this instead:

RowSet lineItems = (RowSet)yourAppModule.findViewObject("LineItems");
while (lineItems.hasNext()) {
  Row line = lineItems.next();
  Long quantity = (Long)line.getAttribute("Quantity");
  /* etc. */
}

In addition to the useful findViewObject() method to access collections of data transfer objects from the built-in model data map, business services like our ToyStoreService inherit several other useful methods related to flexibly working with application data. They are beyond the scope of this article since we didn't require their use in the ADF Toy Store demo, but JDeveloper's Online Help system covers all of the framework API's in its reference documentation if you are curious for more details.

Implementing Domain Business Objects Using ADF Entity Object Components

Business objects built using ADF are called entity objects. Like application module components, your entity objects are JavaBeans that extend a framework base class, are configured from XML metadata, and are created by factories. They cooperate automatically with other ADF framework components to help make application building easier. The distinguishing role of entity objects is to be the software implementation of the domain business entities in your real-world business object model.

Since developers typically use the Unified Modeling Language (UML) to visualize their business object model, we can use JDeveloper 10g's UML modeling features for ADF to do just this. Figure 22 shows where to find the Business Objects UML model in the ToyStoreModel project.

Figure 22: Finding the "Business Objects" UML Model in ToyStoreModel Project

You can see the UML diagram named "Business Objects" in the toystore.uml package by double-clicking on it. Figure 23 shows what you will see when you open the diagram. It's the business object model for the ADF Toy Store demo. Here, I've set the display options for the diagram shapes to only show the object attributes to keep things simple, but you can easily turn on additional options to show methods and other UML artifacts on the diagram, too.

Figure 23: UML Diagram of the ADF Toy Store Domain Business Objects

For each real-world business entity in the application domain, an ADF entity object:

• Defines the names and datatypes of the attributes required to model its business data
• Declares how it is associated to and/or composed of other entities in the model
• Encapsulates the business rules governing the entity and any composed entities
• Handles the persistence of changes made to business objects

By default, each entity object inherits high-performance, relational-database persistence functionality from the ADF framework, too. Custom persistence schemes can be implemented by overriding one framework method in your domain-specific entity subclass. For example, some ADF framework users are doing this to adapt their entity objects to use an existing PL/SQL package API for updating information in their base tables.

One built-in feature we can notice from the UML model in Figure 23 is that the Orders entity object uses a DBSequence type for its Orderid attribute. By configuring an entity object to have a datatype of DBSequence, the ADF framework automatically handles the common case of primary key values assigned from a database trigger, without having to write code.

Our UML business model in Figure 23 visualizes several other interesting things about the relationships between our entity object components. In particular, it shows:

• The associations and compositions between entities
• The cardinality of the association (one-to-one, one-to-many, etc.)
• The programmatic navigation possible between entities (the arrowheads)

For example, Orders are composed of one or more Lineitem and each Lineitem is associated with an Item. Each Item is associated many-to-1 with Supplier, and one-to-one with Inventory. The arrowheads imply, for example, that code in Lineitem can call getItem().getInventory() to access the instance of the Inventory object that tracks the items quantity in stock. In fact, if you look at the finalizeOrder() method in the ToyStoreServiceImpl.java class, you'll see programmatic association traversal at work as shown in the code snippet below, accessing the inventory object for the item being ordered on the current order line item, and setting the inventory quantity on hand to the adjusted quantity.

// Decrement Inventory Quantity for current line item amount
InventoryImpl inv = newLine.getLineitem().getItem().getInventory();
double currentQty = inv.getQty().doubleValue();
double newQuantity = currentQty - (newLine.getQuantity().doubleValue());
inv.setQty(new Number(newQuantity));

As you can see in Figure 24, which shows the toystore.model.businessobject package in the JDeveloper Application Navigator, entity object components like Account are comprised of a number of constituent files:

• Account.xml - Entity definition file
• AccountImpl.java - Entity implementation
• AccountImplMsgBundle.java - Entity message bundle

Figure 24: Domain Business Objects and Associations in the ToyStoreModel Project

All of the declarative aspects of the Account entity object definition are kept in the Account.xml file. This includes attribute definitions, declarative business rules, and database table/column mapping information. As we saw with application module components, you never have to hand-modify the declarative XML yourself. The multi-panel Entity Object Editor in JDeveloper shows you all of your entity component's settings and lets you easily configure its declarative behavior. An important thing to notice is that entity objects do not have a client-accessible interface as application module components do. No client interface is required since entity objects are not meant to be directly accessed by the controller or view layers. Entity objects are private to the model layer by design.

Underlying Database Schema for the Toy Store Demo

To get an overall view of the database design that our business object layer maps onto for its persistence, visit the DatabaseSetup project in the ADFToyStore workspace, and double-click on the "Toystore Database Design" diagram under the toystore.db package in the Application Sources folder. This will bring up the JDeveloper 10g database diagrammer showing the ADF Toy Store table design as shown in Figure 25. With the exception of the STATES_FOR_COUNTRY table that is used by our State/Country validation rule, all of the other tables are identical to those used in the original Java Pet Store Demo.

Figure 25: ADF Toy Store Database Design

Supported Approaches for Implementing Business Rules

Figure 26 shows the Validation panel in the Entity Object Editor for the Account object, illustrating the object-level and attribute level business rules that we've defined for this component.

Figure 26: Validation Rules Panel in the Entity Object Editor for "Account"

Table 1 shows the declarative business rules that have been enabled for the Account and Orders entity objects.

Rules like the UniquePKValidationBean, ListValidationBean, and Method are supplied with the framework. As we'll see more in detail later in this paper, the VerifyStateForCountry rule is a declarative rule whose implementation we've written ourselves. Once a custom rule is written, other developers can use it declaratively just like any of the supplied rules by setting specific usage-specific parameter values that will drive the rule evaluation. In general, validation rules can be set at object-level and attribute-level with the following three implementation choices:

1. Use one or more pre-supplied rules

   This only requires picking the rule type and setting any properties that govern its behavior. For example, a RangeValidationRule might have LowValue and HighValue parameters that must be supplied to define the range.

2. Use one or more custom method validation rules

   Defining method validation rules causes the framework to evaluate various validateSomething() methods that you've written in your entity object's implementation class. These rules are appropriate for complex validation that don't make sense to generalize into a custom rule to be reused by other entities.

3. Using custom business rules

   Custom rules are JavaBeans that implement the JbiValidator interface in the oracle.jbo.server.rules package. Once defined, they can be packaged as reusable rule libraries and put to work declaratively by other developers on the team.

The validateCreditCardExpiration() method on the Orders entity object illustrates a custom validation method. Figure 27 shows the Method validation rule that we've defined on Orders to engage the validateCreditCardExpiration() method.

Figure 27: Method Validation Rule in Use for "Orders" Entity

Since this custom method-based validation rule depends on two different attributes (Creditcard and Exprdate) we implement it as a validation rule at the business object level, instead of at the attribute-level. Here we're illustrating how to use code-based validation as an alternative to declarative business rules like the VerifyStateForCountryRule that is also associated to this Orders business object.

The expected signature of validation methods invokeable by the Method business rule is:

public boolean validateSomeNameYouChoose()

This means that our custom validation methods should return true if the validation succeeds. For validation failures, we can throw an oracle.jbo.ValidationException for an object-level exception, or an oracle.jbo.AttrValException for an attribute-level exception, accompanied by a custom error message. Of course, if the framework's generic exception message is adequate, we can just return false to indicate failure as well.

By throwing an attribute validation exception, the error is attributed to a specific attribute instead of to the object as a whole. This attribute name information is then available to the controller-layer code that handles errors so it can decide where to present the error to the user. Example 4 shows what the code for the validateCreditCardExpiration() method looks like.

Using View Object Components to Simplify All Aspects of Data Access

One of the most frequent and fundamental tasks that business application developers do is access business information for iteration, presentation, and modification. As simple proof of this fact, we need only note that virtually every page of the ADF Toy Store demo relies on displaying or editing business information. For example:

• while browsing, the user sees pages showing store categories, products, and items
• the Register as a New User page captures new data from the user
• the Edit Profile page allows the user to update existing account information
• the Your Cart page shows the items the user has added to their shopping cart.

Unfortunately, in the J2EE world without frameworks, this omnipresent data-access task is fraught with implementation complexity in the name of adhering to J2EE best practices design patterns. Just look at what developers are encouraged to do by most J2EE design pattern books:

1. Write code using the JDBC API to implement the Fast Lane Reader pattern to query database data,
2. Encapsulate that JDBC code in a Data Access Object to isolate the retrieval method in case the datasource changes later
3. Write code to implement the Transfer Object pattern. These data transfer objects (or value objects) reflect the structure of the JDBC result rows.
4. Write code to implement the Value List Handler pattern to iterate over the rows in the JDBC ResultSet and construct instances of row-like data transfer objects holding copies of the queried row data.
5. Implement the Transfer Object Assembler pattern to aggregate multiple collections of data transfer objects into a single object for further minimizing network round trips.

Of course, these suggestions are not bad in and of themselves. All of their functionality is interesting and valid. However, their hand-coded implementation implies a ton of uninteresting, infrastructure-level programming tasks that should not be required by application developers. This observation becomes even more painfully clear when we consider that only really interesting things changing in each usage scenario of the above patterns is:

• What SQL query is required to retrieve the data I need?
• What are the attribute names and datatypes of the data transfer object needed to transport a row of data from the result set of this query to the client?

As with many hand-coded J2EE development techniques, there is a better way! Rod Johnson devotes all of chapter 9 in Expert One-on-One: J2EE Design and Development to "Practical Data Access", and he recommends developing a generic JavaBean component that can handle all of these gory details of data access. Specific query components then extends this generic "query bean" to add the usage-specific details like the SQL statement.

This is precisely what the ADF framework provides with its View Object component. Like the other ADF framework components we've seen, your view objects are JavaBeans that extend a framework base class, are configured from XML metadata, and are created by factories. Once configured with a SQL statement you want it to execute, each view object component automatically implements all of the following design patterns for you: Data Access Object, Fast Lane Reader, and Transfer Object. They also cooperate with the application module component to provide an implementation of the Transfer Object Assembler pattern, too.

While defining your view object, in addition to setting up your query (or letting the editor build it for you), you also configure the attribute names and datatypes of the data transfer object that will carry the query's result row data to the client. Figure 28 illustrates this relationship between the view object and its "view object row" data transfer object. When the view object produces collections of data transfer objects after executing a query, each object in the collection is an instance of the view object row class.

Figure 28: View Object and Its View Object Row "Data Transfer Object"

Of course, the ADF design time editor defaults the names and datatypes for you based on metadata it obtains from your query statement, but you can change the names and datatypes of the attributes in the view object row as required. The declarative information about the query and the view-specific data transfer object attributes is saved in the view object's XML file.

In addition to the ViewName.xml file that all ADF components have, as shown in Figure 29, the view object component can comprise a number of optional additional files as well:

1. View Object implementation file (ViewNameImpl.java) used to customize the behavior of the view object component or implement custom methods.
2. View Object interface (ViewName.java) appears automatically when you expose one or more of your custom view object methods on the Client Methods panel of the View Object Editor.
3. View Row implementation file (ViewNameRowImpl.java) used to customize the behavior of the data transfer object associated with this view object.
4. View Row interface (ViewNameRow.java) appears automatically when you expose one or more of your custom view row (data transfer object) methods on the Client Row Methods panel of the View Object Editor.
5. View Row message bundle (ViewNameRowImplMsgBundle.java) appears automatically when you define any built-in prompts, tooltips, format masks, etc., for your view object.

Client proxy classes can be generated for view object and/or view rows as well if you have exposed custom methods on either or both of these.

For example, toystore.model.dataaccess.Accounts is implemented by the following set of files:

• Accounts.xml - View Object definition file
• AccountsImpl.java - View Object implementation
• AccountsRowImpl.java - View Row implementation
• AccountsRowImplMsgBundle.java - View Row message bundle

Figure 29: Data Access Objects in the ToyStoreModel Project

Looking at the files associated with other view objects in the same package like ItemsForSale, LineItems, and ReviewOrder, we see that the additional files only appear as needed. In the case of ReviewOrder, we didn't need to customize the view or view row code, and we didn't need a view row message bundle, so only the ReviewOrder.xml file exists. Everything necessary for this particular query is captured in metadata. The LineItems view object illustrates using only code in the view row class. The ItemsForSale view object is an example that uses custom code in the ItemsForSaleImpl.java class to encapsulate bind variable parameter setting, and it exposes these custom methods through the automatically-maintained ItemsForSale.java client interface.

Other view objects in the demo that contain custom methods to encapsulate their bind variable details are ProductsInCategory, FindProducts, and ProductList. By exposing custom view object interfaces, your controller layer can contain code like Example 5 to set query parameters without being aware that the query is being performed using SQL.

By default, ADF creates instances of a generic data transfer object (oracle.jbo.Row), but by checking the Expose Accessors to the Client checkbox on the Java panel of the View Object Editor, you can expose a typesafe data transfer object interface that lets you access the data using compile-time-checked accessor methods like getQuantity() instead of using generic, runtime-evaluated accessor methods like getAttribute("Quantity").

At runtime, your view object components handle:

1. Encapsulating the data access code
2. Querying the data using optimized, best-practices JDBC techniques
3. Constructing instances of the data transfer objects for query result rows
4. Browsing data a page at a time
5. Exposing one or more collections of data transfer objects based on the view object's query

While a view object can be used to produce multiple collections of data transfer objects, and can iterate each collection using one or more iterators, in 90% of the cases, developers need to just iterate through a single collection of results. To cater to this frequent use case, the base ViewObject framework component implements the RowSet interface and delegates that interface's methods to an aggregated instance called the "default rowset". Similarly, all RowSet objects implement the RowSetIterator interface, and delegate its methods to an aggregated instance called the "default iterator". Figure 30 illustrates how the pieces fit together. This setup allows developers in most common cases to work with a single view object instance at runtime, and easily iterate its default collection of data transfer objects.

Figure 30: View Object Aggregates a Default RowSet for Ease of Use

Since view objects encapsulate how data is accessed, it's possible to retrieve data from datasources other than the normal SQL database query without changing other parts of your application. As a concrete example of this, CountryList, CreditCardList, ExpirationYearList, and ShippingOptionsList view objects in the ADF Toy Store demo "query" their data from a standard Java *.properties file.

View Object queries can automatically be related master/detail by creating additional components called View Links. A view link connects a source and a target view object, and defines (in XML, of course!) which attributes in the source view object row need to correspond to the attributes in the target view object row. By creating view links between view objects, you make it possible for the framework to automate the coordination of the correlated queries on your behalf. You also get the programmatic benefit of being able to call "view link accessor" methods to traverse from one row to its detail collection of correlated rows very easily.

How Framework Components Cooperate to Simplify Data Modification

Easily querying and iterating collections of data transfer objects illustrates only part of the view object's value proposition for application developers. View objects cooperate with entity objects and application modules to dramatically simplify creating, updating, and deleting business information as well. In order to appreciate the work this powerful trio is saving us, we'll again study the manual steps that J2EE developers do to modify data when working without a framework.

The controller layer provides data from the end-user to the business service in the form of one or more instances of data transfer objects. The task at hand might involve creating, updating or deleting instances of business objects in the model layer. The steps for creating new data typically require business service code to:

1. Create a new instance of the appropriate back-end business object
2. Populate the attributes of the business object based on corresponding data in the data transfer object(s)
3. Commit the transaction

Steps for updating existing data typically require business service code to:

1. Lookup the existing instance of the correct back-end business object by primary key, using appropriate attribute value(s) from the data transfer object
2. Update the attributes of the business object instance based on appropriate, corresponding data in the data transfer objects, ideally only updating the attributes that the client actually changed in case attribute-level auditing is being done inside the business object layer
3. Commit the transaction

Steps for removing existing data typically involve:

1. Looking up the existing instance of the correct back-end business object by primary key, using appropriate attribute value(s) from the data transfer object.
2. Deleting the instance by invoking an appropriate delete method on it.
3. Committing the transaction

The above steps are those required for the business service to handle a single instance of an incoming a data transfer object from the controller layer. In reality, the incoming data might be a collection of data transfer objects, or a set of collections of one or more different kinds of data transfer objects. In that case, you would combine the steps above with iteration and conditional logic to lookup the correct back-end business object based on the kind of data transfer object coming in.

As with much of the uninteresting application infrastructure code that too many J2EE developers are used to writing themselves, this code is not mind-bendingly difficult to write. However, the unmistakable déjà vu you feel each time you write the code above tells any good programmer that this is a task that can be generalized and driven through metadata to cut down on hand-written code.

The only pieces of information that are unique to each situation are:

• Which back-end business objects are related to the data in this data transfer object?
• How do data transfer object attributes correspond to back-end business object attributes?

Using this information, our implementation of the Value List Handler pattern could be enhanced to support updating the data transfer objects and automatically coordinating them with the back-end business objects.

Luckily, this information is precisely what ADF view objects store in their XML metadata file, along with the other metadata about the SQL query and the attribute names and datatypes of the associated data transfer object. This means that view objects can be related to back-end entity objects in a declarative way, and they store this "attribute map" information for use at runtime to completely automate the coordination between client-modified data transfer objects and the back-end business objects to which they relate. This makes creating, updating, and deleting back-end business data very easy because the controller layer can simply work with the collections of data transfer objects (DTO's) to:

• Add new DTO instances to any collection
• Update the attributes of existing DTO instances
• Remove existing DTO instances from any collection
• Find a DTO instance in a collection by primary key

All of the changes made on the data transfer objects in a view object's RowSet collection are delegated automatically back to the appropriate entity objects without user-written code to orchestrate that cooperation. Since the entity objects encapsulate the business rules as we explored in a previous section, this delegation of DTO attribute setting to appropriate entity object attributes causes any relevant business rules to be evaluated. Of course, any failures are thrown as exceptions that can be trapped and presented to the user.

Developers using the ADF framework for their model layer have control over how "eagerly" this delegation between DTO attribute modifications and entity object attribute modifications is performed. For rich-client applications with highly-interactive graphical user interfaces built in Swing, the delegation can be set to be "immediate". This causes errors raised by business rules in the underlying entity objects to be thrown immediately for interactive user feedback as expected.

For web-style applications, the entity-level delegation of changes made to the collections of DTO's can be done in more of a "batch" style. To avoid frustrating the end user by presenting one error at a time to be corrected, ADF offers additional facilities for bundling a maximal set of errors raised by entity-level validation failures. This "bundled exception mode" allows your web application to present all errors to the user in a single round-trip to the browser. The built-in ADF/Struts integration leverages this bundled exception mode.

Recall from above the entity objects representing the domain business objects are not directly accessible to clients. The controller layer can modify entity objects only indirectly in one of these three ways. Two of the three ways involve invoking a business service method. That method's private implementation code can:

• directly create, lookup and modify, or lookup and delete entity instances, or
• update view object row DTO's in a view object's rowset.

Alternatively, the controller layer can directly update any data transfer object "rows" itself, in any collection in the application module's model data map.

In all three cases, any changes effected successfully to entity object instances are held in the entity object cache until the transaction is committed. The application module contains a reference to a transaction object that coordinates the entity object caches during commit or rollback. This coordination insures that any new, modified, or removed entity instances are validated if required, and then persisted appropriately. In effect, the application module component represents the "unit of work" and any entities modified by interaction with the application module business service are automatically "enrolled" in the unit of work, without requiring developer-written code. Of course, when you choose to deploy your application module component as an EJB Session Bean, the ADF transaction object automatically becomes a "slave" to the container-managed transaction so that it participates correctly in transactions that can span multiple session beans.

Figure 31 summarizes in a single picture how the ADF framework components interact. All of the components can be accessed freely by Business Service implementation code inside your application module component. The component interfaces on the left side of the double-dotted line can be used by the view and controller layers as well.

Figure 31: How ADF Business Components Relate And Which Are Client-Accessible

Creating Smarter Custom Data Types Using Domains

The Java language comes with a number of basic datatypes like String, Long, Float, etc. The ADF framework adds a few additional types called "domains" that live in the oracle.jbo.domain.* package. These additional types include:

• Number
• Date
• DBSequence
• ClobDomain
• BlobDomain
• Array

These are designed to improve performance and ease-of-use when working with corresponding types in the Oracle database.

In addition, there are times when you need to create your own custom data types to capture some frequently used concept like an email address or a credit card number that require custom value-level validation whenever they are constructed. The ADF framework provides the simple feature of domains to support this use case.

Figure 32 shows the three custom datatypes we've created for the ADF Toy Store demo. The toystore.model.datatypes.CreditCardNumber is a String-based datatype with construction-time validation that the credit card number has 16 digits. The toystore.model.datatypes.Email domain is another String-based datatype that checks to make sure its value has the format of a valid email address.

Figure 32: Custom DataTypes in the ToyStoreModel Project

The toystore.model.datatypes.ExpirationDate type models string-based data in the database stored in the format MM/YY, and internally treats it as a date for comparing the expiration date with the current date for detecting if a credit card expiration date is in the future or the past.

Once you've created custom domain types, your entity objects and view objects can use these new types in the same way that they can use any of the built-in domains or the base Java types.

Testing Business Tier Components with JUnit

Testing is an important consideration when implementing your J2EE model layer. J2EE experts like Rod Johnson recommend building the business services that support your model layer in a way that allows them to be easily tested outside of the J2EE runtime container environment where they will ultimately be deployed. Being able to run regression tests on business services from the command line simplifies the testing process and allows it to be more easily automated. Simplifying and automating regression tests leads to higher quality software since it encourages engineers to develop more tests and run them more often after making changes to the system.

The open source JUnit testing framework is the defacto standard that most Java development teams use to write and run their regression tests. Oracle JDeveloper 10g features native support for working with the JUnit testing framework (available as a small, separate download explained in the Demo Installation and Setup section above). One of the wizards available in the Unit Tests (JUnit) category of the New... Gallery allows you to create a skeleton Business Components Test Suite. This wizard allows you to pick an application module component, and a particular configuration that you'd like to use for testing, and then it generates you:

• A JUnit test fixture that encapsulates getting an instance of the desired application module
• An example JUnit test case class which uses this fixture and asserts that all expected view object instances exist in the application module's model data map.
• A JUnit Test class that runs all of the test cases (initially just one).

By using the above wizard and the built-in JDeveloper refactoring tools, I renamed the packages containing the test runner, the unit test cases, and the test fixture into separate packages, and then created several additional tests to exercise the ToyStoreService application module component.

Running the RunAllTests.java class in the Testing.jpr project launches the Swing-based JUnit test runner as shown in Figure 33. This result is what you'll see if all of the seventeen regression tests pass. If any failures occur, the progress bar turns red and error messages in the log window alert you to the failing test.

Figure 33: Successful Run of ADF Toy Store JUnit Test Suite

By developing this suite of business service regression tests earlier during the development cycle, I was able to rerun my JUnit test suite after any major new feature additions I made as well as after the many project refactorings that I performed as I continually improved the demonstration. Any problems that my changes might have produced inadvertently were instantly uncovered by the regression tests so that I could easily fix them before moving on to the next task.

Implementing the Controller Layer with the Apache Struts Framework

In the MVC architecture, all of the logic that processes user input, handles interaction with business services, and decides what page the user should see lives in the controller layer. The best practice approach for your web application's controller layer is to implement the Front Controller design pattern. Based on the type of incoming request, the front controller delegates the handling of the request to more task-specific controller logic. The front controller is typically implemented as a servlet and configured in web.xml to handle all requests for URL's related to your application. While you could implement this controller servlet by hand, the popular Apache Struts framework provides a ready-made implementation that handles the job well.

Handling Model Interaction and Page Flow With Struts Actions

Using Struts, you describe the "page flow" of your web application by:

• assigning logical names to the actions your application needs to perform, and
• indicating the possible "routes" the user can follow between the actions.

For example, in a simple web store application, the user might inspect the contents of her shopping cart by using an action named /yourcart. Using Struts, actions are executed when the end-user browses a URL matching the *.do suffix, so the URL for the /yourcart action might be:

http://yourcompany.com/ADFToyStore/yourcart.do

While looking at her shopping cart, the user might begin the checkout procedure by clicking on a link to the /reviewcheckout action, letting her confirm what she is about to purchase.

A visual representation of these actions and the links that connect them might look like Figure 34. The /yourcart and /reviewcheckout actions show as gear icons, while the JSP pages that represent the view layer are shown as page icons. The links between the actions, known as "forwards" in Struts parlance, are shown here as solid lines pointing from the source action to the target page or action.

Figure 34: Struts Actions and Forwards

The dotted lines represent web page hyperlinks or HTML forms which target the action that their corresponding arrows are pointing to. While the above is a trivial example, you can imagine that a visual diagram of your Struts actions and pages for a more complex application would form an easy-to-understand "page flow" showing where the user can go and how they get can get from one page to another.

As you might imagine, the controller-layer behavior associated with each action is written in an associated Java class. These action classes orchestrate what happens in the web tier in response to each request. Figure 35 illustrates the typical steps that occur each time the user clicks on a link representing a Struts action.

1. The controller receives a page request and routes it to the appropriate action class
2. The action acquires an instance of a business service, via the business delegate, and invokes a method on it
3. The business delegate implementation delegates the method call to the business service
4. The business service uses one or more query components to retrieve business data from SQL queries, and exposes the results as one or more collections of data transfer objects in its "model data map"
5. The business delegate exposes the "model data map" to the view layer
6. The controller selects an appropriate view and forwards the request to the view layer for rendering
7. The view layer finds the collections of data transfer objects from the "model data map" and iterates over them to render the data to the end-user.

Figure 35: Typical MVC Processing Steps with Struts and JSP

Notice back in Figure 34 how the /yourcart action and the yourcart.jsp page are in a kind of symbiotic relationship. The /yourcart action prepares the data to be displayed by yourcart.jsp, and the yourcart.jsp page targets any links and submitted HTML forms back to the /yourcart action so it can handle the controller-layer details and decide what to do next. When an action and a page are mutually dependent like this, it's often said that they implement the "Postback Pattern"; the action prepares the data for the page to render, and the page posts-back to the action to handle its events.

Figure 36 shows the two different ways that JDeveloper 10g's Struts Page Flow Diagrammer supports implementing the "Postback Pattern" using Struts and ADF. We've already seen the first approach above. You see a discrete icon for both the action and the page, as well as connecting lines in both directions between the two. The alternative is to use the ADF "DataPage" which represents the combination of the action and the page involved in implementing the "Postback Pattern" with a single icon.

Figure 36: Pages and Actions Implementing a "Postback Pattern"

As you might expect, using the "DataPage" approach makes for much easier to read page flow diagrams for applications of any non-trivial size. As we'll see shortly, the ADF Toy Store controller layer is implemented using this DataPage approach.

Along with other configuration information, Struts keeps the mapping of action names to action handler classes in the struts-config.xml file. As shown in the ToyStoreViewController project in Figure 37, this configuration file lives in the standard WEB-INF directory of your application, among the other elements that comprise your Web Content.

Figure 37: Locating the Struts Configuration File

When you double-click on the struts-config.xml file, by default, JDeveloper 10g shows you a visual page flow diagram. Let's take a look at what one of these diagrams looks like by studying the page flow of the ADF Toy Store demo.

Figure 38 shows the page flow diagram for the ADF Toy Store Demo's controller layer. Each DataPage is implemented by the combination of a JSP page and an Struts action that prepares that page's model data to render and handles its post-back events. Since these page/action pairs render as a single icon, the page flow diagram is tidy and easy to understand.

Figure 38: ADF Toy Store demo Struts Action Flow

By clicking on the Source tab of the editor, you can see the XML source code for the Struts action mappings for the ADF Toy Store demo. Searching for /yourcart, you will find the XML tags shown in Example 6. Under the Search main menu, you can use either the Find... or Incremental Find Forward (Ctrl+E) options to locate it.

Among other pieces of configuration information, the struts-config.xml file contains one <action> tag for each action in your application. The <forward> element nested inside Struts action tags provides a logical name like "reviewcheckout" for the available page navigation "routes" the user can take from the "/yourcart" page. In this case, there is only a single forward route that the user can follow to navigate to review the items for checkout.

The value of the parameter attribute tells us that our /yourcart action is related to the yourcart.jsp page that lives in the /WEB-INF/jsp directory. We can also see that it is mapped to the YourCartAction class in the toystore.controller.strutsaction package. This action contains the controller logic to handle the "UpdateCart" and "RemoveItem" events, generated by the user's clicking on appropriate buttons in the page, by invoking the adjustQuantitiesInCartAsStringArrays() on the ToyStoreService to adjust shopping cart quantities. It also contains a line of code to retrieve the shopping cart total from the business service for display in the page.

The YourCartAction class extends the ToyStoreDataForwardAction class in the FwkExtensions project. This framework extension class, in turn, inherits a goodly amount of built-in behavior from the ADF DataForwardAction, adding a few handy helper methods in the process that are useful to many of the actions in the Toy Store Demo. The value of the modelReference property that we see points to some declarative metadata that ADF uses to simplify data binding.

Table 2 gives a description of what each of the actions does.

Whenever the struts-config.xml file is active, the JDeveloper Structure Window and Property Inspector provide additional design time support. The Structure Window shows a tree reflecting the internal structure of the configuration file. You can click on elements in the tree to see or edit their properties in the Property Inspector. In Figure 39 you can see all of the ADF Toy Store Demo's Struts action mappings under the Action Mappings folder in the tree and the properties of the selected "/yourcart" action in the Property Inspector.

Figure 39: Structure Window and Property Inspector Support for Struts

By selecting an appropriate container node, like Form Beans or Action Mappings for example, you can use the New option in the right-mouse menu to create a new entry of that kind. You can then use the Property Inspector to configure its various properties as you would expect. Alternatively, you can click Edit in the right-mouse menu to edit any of the configuration file items through the Struts Configuration Editor dialog shown in Figure 40. Other right-mouse options certain nodes allow quick navigation to related files. For example, to quickly jump to the code of any of your Struts actions, you can select an action and pick Go to Code on its right-mouse menu.

Figure 40: Oracle JDeveloper 10g Struts Configuration File Editor

As we've seen, for working with Struts configuration information you can achieve everything you need to do using a combination of the Struts Page Flow Diagram, the Structure Window, and the Property Inspector. Keep in mind that if you do need (or prefer!) to edit the XML source of the file directory, JDeveloper 10g is happy to oblige and will keep the page flow diagram in sync with your manual edits automatically. The choice is yours.

Using Struts Form Beans to Process HTML Form Input

To complement its features for managing controller-layer action classes, Struts also helps developers process HTML form input in their actions. The Struts controller servlet populates the attributes of a so-called Form Bean with the parameters from the posted HTML form. Then, the developer writing the action class can work with the HTML form values using simple bean getter and setter methods. Since the data entered by the user in an HTML form arrives in string form, and since any errors made to data entered should be redisplayed verbatim so the user can correct them, typically all of the form bean's properties are typed as String.

Example 7 shows that the /signin action is mapped to the SignInAction action class in the toystore.controller.strutsactions package. However, additional attributes on the <action> tag indicate to the Struts controller to use the Form Bean named loginform to encapsulate posted HTML form parameters and to return to the signin.jsp page in case of any validation errors.

Note that loginform is a logical name defined in the <form-beans> section which corresponds to the toystore.controller.strutsformbeans.LoginForm JavaBean class.

With these additional attributes set, at runtime the struts controller will populate the LoginForm bean with the values of the username and password HTML form fields from the signin page when the user clicks to submit the form. Those values are accessible by the onVerifySignin() event-handler method of the SignInAction class which will handle the sign-in form's postback when the user submits the form.

Example 8 shows the onVerifySignin() method in the form bean. It verifies that neither the username nor password properties are blank, and then calls the:

public boolean validSignon(String username, String password)

method on the ToyStoreService business service interface to verify whether the username/password combination represents a valid webstore user. If any validation check fails, the method adds a Struts ActionError object to the ActionErrors collection so the view layer can present appropriate error messages to the user. We'll dive into more details about how the error string keys, like those represented by the INVALIDLOGIN constant, are translated into user-readable messages in the appropriate language in the Struts and ADF Features for Building Multilingual Applications section below.

If the login form submission validation above fails, then the user is returned to the signin page to try again. If it succeeds, then the code above calls a helper method to flag the current user as signed in, and then returns the appropriate page to forward the request to. Since several different actions in the demo can cause the user to login, this action uses the target parameter to return the correct "next" page in the flow, based on which action caused the user to need to log in.

Understanding ADF/Struts Integration

As we've seen above, Struts is a handy framework for implementing the controller layer of your web application. As we'll see in more detail in the Building the View Layer with JSP Pages and JSTL section, Struts also provides some JSP tag libraries to simplify building your view layer. However, it effectively provides virtually no built-in functionality to assist with implementing the model layer, arguably the most important and time-consuming part of your application development. The model layer is where all of your business logic and data access functionality resides. To complete the strong MVC application infrastructure that J2EE developers require, you need a productive approach to building the model layer that integrates nicely with Struts. This section explains how the Oracle ADF framework is the perfect companion to Struts in this venture.

In addition to the business component building-blocks that ADF provides to accelerate model layer development, it also provides specific additional functionality to dovetail with Struts' controller and view layer features. For example, ADF provides:

• Comprehensive declarative data binding support for "wiring" your view layer pages and your model layer data together in a bidirectional way without writing code.
• High-performance Business Delegate implementation that finds and pools business service components
• Declarative, database-backed state management to easily support complex transactions spanning multiple pages

Of course, JDeveloper 10g provides a productive design time environment for putting all of these feature to work in your own J2EE applications.

Brief Overview of ADF Binding and Controller Concepts

Oracle ADF includes a "data control" abstraction layer for back-end business services and a binding layer for declaratively connecting front-end user interface controls to data supplied by any data control. These ingredients, based on JSR 227, provide you a consistent and pluggable model layer for your J2EE applications that is on its way to becoming a J2EE standard.

Figure 41 illustrates where the ADF Data Control and ADF Bindings fit into the overall ADF Model, View, Controller, and Business Services architecture.

Figure 41: Oracle ADF Architecture for J2EE Applications

The key data binding concepts in Oracle ADF are the following:

• Data Controls

  A data control abstracts the implementation of a business service, allowing the binding layer to access the data from all services in a consistent way.

• Iterator Bindings and Control Bindings

  Bindings are lightweight objects that decouple back-end data and front-end UI display. An iterator binding provides a consistent way to work with a collection of data objects supplied by a data control. Control bindings provide a standard interface for UI components to interact with an iterator's data or to invoke "action" methods for preparing model data and handling events. Bindings also expose key metadata to simplify building dynamic, multi-lingual user interfaces.

• Binding Containers

  A binding container is a named group of related iterator and control bindings that you use together for a particular page (or panel) of your application. A binding container is also known as a "UI Model" since it provides the appropriate subset of model data for a specific UI.

• Binding Context

  The binding context provides the data environment for your application. It contains all of the data controls and binding containers that your application can access.

JDeveloper 10g supports these data binding concepts at design time in a rich way. Once you've created a business service, you can expose it as a Data Control and it will appear in the Data Control palette. Figure 42 shows what our ToyStoreService data control, based on our ToyStoreServiceImpl application module component, looks like in the Data Control Palette. We can see all of the data collections in its model data map, as well as the custom methods on the ToyStoreService service interface under the Operations folder.

Figure 42: ToyStoreService Data Control in the Data Control Palette

As you are creating pages with databound content -- through drag and drop from the Data Control Palette or by explicit creation -- JDeveloper maintains metadata about your application's Binding Context. This information describes the data controls your application is using, the different binding containers that you've created, and the binding objects they contain. The DataBindings.cpx file is the place JDeveloper stores this Binding Context metadata. After clicking on this file in the Navigator, as shown in Figure 43, the Structure Window displays the Data Controls in use by your application, as well as the various binding containers (also known as "UI Models") that support the different pages in your application. Clicking on something in the Structure Window like the ToyStoreService, the Property Inspector shows you its properties.

Figure 43: Binding Context Info in Navigator, Structure Window, and Property Inspector

While looking at a given JSP page, you can click on UI Model tab of the Structure Window as shown in Figure 44 to see the details of the binding objects it contains. By clicking on a particular binding object, you can see and edit its properties in the Property Inspector, as well as through the more structured object editor by selecting Edit... on the right-mouse menu.

Figure 44: UI Model Tab Showing Binding Container for the Search Page

To simplify using the ADF binding layer in Apache Struts applications, Oracle ADF also provides Struts-specific controller-layer components to seamlessly integrate Struts with the ADF Binding layer. These components include:

• A servlet filter named ADFBindingFilter that coordinates the use of the business delegate, business service pooling, and state-management.
• A multi-purpose Struts DynaActionForm called BindingContainerActionForm that eliminates the need to create Struts form beans for each page you build.
• A powerful Struts Action class named DataAction that implements an easy-to-customize request-handling lifecycle with automatic support for the ADF data binding functionality, and
• A standard Struts ActionMapping class called DataActionMapping that captures addition declarative metadata in standard action mapping extension properties used to automate the data binding.

Rounding off the JDeveloper 10g IDE support for using ADF and Struts together is the Visual Struts Page Flow diagrammer and the concept of the ADF "DataPage". The DataPage simplifies working with a page and a Struts action that want to cooperate to implement the "Postback Pattern." JDeveloper 10g represents the combination of a web page and a DataForwardAction that prepares its data and handled its events into a single node in the Struts diagram. The DataForwardAction combines the ADF DataAction's lifecycle features and DispatchAction-like event-handling with the implicit page forwarding behavior and event-handling of the basic Struts ForwardAction.

The DataForwardAction inherits event-handling functionality (from DataAction) that is similar to that offered by the Struts DispatchAction. Your HTML forms or hyperlinks can include a special parameter named event in the request, whose value represents the name of an event to be "fired" and handled by the DataAction being targeted. So, sending a parameter event=YourName as a request parameter for form field will "fire" the YourName event.

For HTML buttons, whose value attribute serves as both the user-visible button label and the value of the control, we support the alternative event-signalling approach of naming the button event_YourName so that the button can have any value, and hence any user-visible label, that is appropriate to the current user's language.

The ADF DataAction supports three key features around the handling of named events.

When an event named YourEvent fires, then...

1. If you have a public void onYourEvent(DataActionContext ctx) method in the data action class handling the request, it will be invoked to handle the event with custom code.

2. If you have an action binding in the current binding container named YourEvent, it will be invoked.

   When used in combination with 1, your event-handler code needs to explicitly invoke the default action for the current event by using code like:

   if (ctx.getEventActionBinding() != null) {
  PageLifecycle p = (PageLifecycle)getPageLifecycle(ctx);
  p.invokeActionBinding(ctx,ctx.getEventActionBinding().getName());
}

3. If you have a Struts forward named YourEvent, it will be used to determine the next page to forward control to.

   When used in combination with 1, if your event-handling code invokes ctx.setActionForward(), then your programmatically set forward takes precedence.

With these basics fresh in our minds, we can now dive deeper into understanding the ADF/Struts integration and how it's put to work in the ADF Toy Store Demo.

Lifecycle of a Web Page Request Using Struts and ADF

Figure 45 shows all of the "moving parts" together in a single sequence diagram to show the lifecycle of a web page request using the Struts and ADF frameworks in tandem.

Figure 45: Lifecycle of a Web Page Request Using Struts and ADF

1. A web request for http://yourserver/yourapp/some.do arrives

2. The ADFBindingFilter finds the ADF Binding Context in the HTTP Session, and if not yet present, initializes it for the first time.

   During Binding Context initialization, the ADFBindingFilter:

   • Consults the servlet context initialization parameter named CpxFileName and appends the *.cpx file extension to its value to determine the name of the Binding Context metadata file. By default the parameter value will be "DataBindings", so it will look for a file named DataBindings.cpx.
   • Reads the Binding Context metadata file to discover the DataControl definitions and BindingContainer names.
   • Constructs an instance of each DataControl, and a reference to each BindingContainer. The contents of each binding container are loaded lazily on their first use by some page.
3. The ADFBindingFilter invokes the beginRequest() method on each DataControl in the binding context. This gives every data control a notification at the start of every request where they can perform any necessary setup.
4. A DataControl based on an ADF ApplicationModule uses this beginRequest notification to acquire an instance of the ApplicationModule from the ApplicationModule pool.

5. The Struts RequestProcessor forwards control to the appropriate action class for the "/some" action mapping as configured in struts-config.xml.

   
   

   NOTE:

   The Struts ActionServlet is what initially forwards control to the RequestProcessor but since RequestProessor does all the work, we don't bother showing the ActionServlet in the diagram.

   
   

6. If the Struts action class is or extends the ADF DataAction class, then it goes through a set of processing steps (known collectively as the request-processing "lifecycle") that include:

   • Finding the BindingContainer...

     ...or initializing it if it's the first time it has been used. During initialization, the binding objects are created based on the binding container corresponding *UIModel.xml metadata file.

   • Executing the iterator bindings if they haven't been yet, in the prepareModel() phase
   • Updating posted data (if present) into corresponding binding objects, during the processUpdateModel() phase
   • Executing custom controller layer logic in the invokeCustomMethod() phase, and
   • Deciding on what page to display next during the findForward() phase.

   Each phase of the lifecycle can be overridden by the developer and each method is passed the DataActionContext object that gives you access to the BindingContext, the BindingContainer, and all Struts-related elements like the servlet request and response as well as the action mapping and action form. The action returns an appropriate Struts ActionForward indicating which action or page we want to navigate to next.

   
   

   NOTE:

   If the Struts action class does not extend DataAction, then it must find (and if necessary, initialize) the binding container manually and manage working with the bindings itself. The ADF Data Binding Primer whitepaper has example code illustrating how to do this.

   
   
7. The Struts RequestProcessor forwards control to the action or page returned by the action as the "next" one to forward control to.
8. Assuming it's a JSP page, the page accesses and iterates over collection of data transfer objects from the control bindings, to write out the formatted web page destined to appear in the browser.
9. The Struts RequestProcessor returns control to the ActionServlet which returns control back to the servlet container.
10. The ADFBindingFilter invokes the endRequest() method on each DataControl in the binding context. This gives every data control a notification at the end of every request where they can perform any necessary resource cleanup.
11. A DataControl based on an ADF ApplicationModule uses this endRequest notification to release the instance of the ApplicationModule back to the ApplicationModule pool.
12. The user sees the resulting page in the browser.

The sequence diagram illustrates that the binding container is available both for model data manipulation by the controller layer as well as model data iteration and rendering by the view layer.

Configuring Your J2EE Web Application to Use Struts and ADF

When you use JDeveloper to build your ADF-based Struts application, your web application is automatically configured for you by the design time tools. However, it's useful to understand what the tool is setting up for you so that you understand how the setup pieces fit together. Being a J2EE web application, it's understandable that several of the configuration steps involve the standard web.xml file. We'll highlight the interesting aspects of the setup steps in this section.

First, in order to use Struts you need to have the Struts front-controller servlet named ActionServlet configured and mapped to a URL pattern. In the ADF Toy Store we're using the conventional *.do URL pattern to map requests to the ActionServlet. The relevant portions of the web.xml file, located under the WEB-INF directory in the Web Content folder of the ToyStoreViewController project, look like this:

<web-app>
  :
  <servlet>
    <servlet-name>action</servlet-name>
    <servlet-class>org.apache.struts.action.ActionServlet</servlet-class>
    <init-param>
      <param-name>config</param-name>
      <param-value>/WEB-INF/struts-config.xml</param-value>
    </init-param>
    <load-on-startup>1</load-on-startup>
  </servlet>
   :
  <servlet-mapping>
    <servlet-name>action</servlet-name>
    <url-pattern>*.do</url-pattern>
  </servlet-mapping>
   :
</web-app>

The second step is setting up the ADFBindingFilter and configuring it to be engaged when *.jsp or Struts ActionServlet-related URL's are processed. The relevant portions of the web.xml file look like this:

<web-app>
  :
  <filter>
    <filter-name>ADFBindingFilter</filter-name>
    <filter-class>oracle.adf.model.servlet.ADFBindingFilter</filter-class>
  </filter>
  <filter-mapping>
    <filter-name>ADFBindingFilter</filter-name>
    <servlet-name>action</servlet-name>
  </filter-mapping>
  <filter-mapping>
    <filter-name>ADFBindingFilter</filter-name>
    <url-pattern>/*.jsp</url-pattern>
  </filter-mapping>
   :
</web-app>

The ADFBindingFilter needs to know where it should look for the ADF metadata file that describes the binding context for your application. It reads a servlet context initialization parameter named CpxFileName to determine the file name to look for. We see in the ADF Toy Store's web.xml file, that this parameter is using the default value of DataBindings. This means that the ADFBindingFilter will use the DataBindings.cpx file to read this information.

<!-- web.xml file -->
<web-app>
  <description>web.xml file for the ADF Toy Store demo application</description>
  <context-param>
    <param-name>CpxFileName</param-name>
    <param-value>DataBindings</param-value>
  </context-param>
    :
</web-app>

You can see the DataBindings.cpx file under the Application Sources folder in the ToyStoreViewController project. It is an XML file that describes all of the ADF data controls and binding containers used by the ADF Toy Store application. Peeking inside it reveals that the ADF Toy Store demo has just a single ToyStoreService data control. You can also see the same information in the Structure Window tree display when the DataBindings.cpx file is active in the navigator or code editor.

<!-- DataBindings.cpx File -->
<JboProject id="DataBindings" ... >
  <Contents>
      <DataControl
         id="ToyStoreService"
         SubType="DCBC4J"
         SupportsFindMode="true"
         SupportsTransactions="true"
         Package="toystore.model.services"
         FactoryClass="oracle.adf.model.bc4j.DataControlFactoryImpl"
         Configuration="ToyStoreServiceLocal" >
         <Parameters >
            <Parameter
               name="Sync"
               value="Immediate" >
            </Parameter>
         </Parameters>
      </DataControl>
       :
   </Contents>
</JboProject>

Notice that this ADF Business Components-based data control has properties named Package and Configuration. The values of these attributes are used by the ADF implementation of the Business Delegate pattern to lookup an appropriate instance of the business service whose configuration properties are defined by the configuration named ToyStoreServiceLocal. Given the package name, the ADF runtime first looks for a bc4j.xcfg file in the common subdirectory under the directory corresponding to the package name. In our example above, it will open and read the resource named ./toystore/model/services/common/bc4j.xcfg from the runtime classpath since the package name provided is toystore.model.services. Inside this file, the ToyStoreServiceLocal configuration contains a property indicating the fully-qualified name of the application module that will be used as our business service implementation.

This best-practice abstraction of the business service lookup lets us easily change model layer deployment strategy without affecting the rest of our application. As long as you're using ADF Business Components to build your model layer, then to change an ADF/Struts application from using a simple JavaBean-based business service to using an application module deployed as an EJB Session Bean instead, the only change required is changing the value of the Configuration attribute above in the data control metadata. This alternative configuration would contain the lookup details for the EJB session bean, cleanly factored out into XML configuration information instead of being lost somewhere in your code. Of course, the ADF design time sets up these multiple configurations for you automatically and gives you productive dialogs to edit the configuration parameters at design time.

Overview of the ADF DataAction and BindingContainerActionForm

When you build HTML forms to allow the user to enter or modify application data, as we discussed above, the Struts infrastructure abstracts the form data into an instance of a FormBean for easier processing by your actions. Three very frequent tasks that actions need to perform in service of HTML forms are:

1. Populating a FormBean with appropriate default values to display a data entry form
2. Populating a FormBean with existing row data to display an data editing form
3. Updating the model layer from a FormBean containing user-submitted changes

Oracle ADF provides two key ingredients that collaborate to make implementing these scenarios easy:

• A multi-purpose Struts DynaActionForm called BindingContainerActionForm that eliminates the need to create separate Struts form beans for each page you build.
• A powerful Struts Action class named DataAction that implements an easy-to-customize request-handling lifecycle with automatic support for the ADF data binding functionality

If you use the /register DataPage to register a new user in the ADF Toy Store site as shown back in Figure 9, you'll notice that the country is defaulted to "USA". This defaulting happens automatically because the Accounts view object being used by this data entry form is related to the underlying Account entity object, and that Account entity object has a default value of "US" defined for its Country attribute.

When the user clicks the (Save Changes) button on the "Register as a New User" form, the form is submitted back to the /register.do action. The Struts framework populates the contents of the form parameters into the FormBean configured for this action mapping, "DataForm". As we can see in the form-beans section of struts-config.xml, the "DataForm" form bean is using the ADF BindingContainerActionForm class.

<form-beans>
  <form-bean name="loginform"
    type="toystore.controller.strutsformbeans.LoginForm"/>
  <form-bean name="DataForm"
    type="oracle.adf.controller.struts.forms.BindingContainerActionForm"/>
</form-beans>

This clever form bean is a Struts DynaActionForm that takes on the properties corresponding to names of the bindings in the current BindingContainer. The /register action is configured to use the RegisterAction class in the ToyStoreViewController project. This action class extends the default ADF DataForwardAction that applies the data in the form bean to the appropriate bindings during the "update model" phase of the its request-processing lifecycle. The bindings turn around and carry out the out the work of populating the target row in the default rowset of the Accounts view object. Figure 46 illustrates this, and helps explain what happens next.

Figure 46: How New Account Information Gets To the Database

Specifically the steps involved in posting the changes from the browser to the database go like this:

1. The Struts RequestProcessor populates the form bean properties with the values from the submitted form parameters, and forwards control to the RegisterAction.
2. The RegisterAction, which inherits its lifecycle handling functionality from the ADF DataForwardAction, finds its binding container and updates the model bindings with corresponding values from the BindingContainerActionForm form bean.
3. The control value bindings are bound to the current row in an iterator binding based on the Accounts view object, so setting the values on the bindings sets these values on the underlying view row attributes.
4. The view row delegates the attribute setter calls to the correct, underlying entity object attributes, causing business rules to be evaluated
5. Since the (Save Changes) button is named event_Save, the RegisterAction invokes the onSave() event handling method. It inherits this event-handling functionality from the ADF DataForwardAction. This method calls the invokeEventAction() helper method to invoke the action named "Save" in the binding container. This action binding is bound to the built-in "Commit" operation on the ToyStoreService data control, so the transaction is committed.
6. On transaction commit, if no validation errors have occurred, the entity objects persist their changes into their underlying tables.

When the ADF view object row is populated, since the view object was associated to underlying entity objects at design time, the framework delegates the attribute setting down to the entity objects. These entity objects contain business rules that get evaluated and which succeed or fail based on the values the user has entered. If any business rules are violated, all exceptions are collected into a single, wrapping "bundled exception" by the ADF framework which is saved in the DataActionContext so that later when the ADF DataAction lifecycle method reportErrors() is called, they may be translated into Struts Action Errors for reporting to the user.

Building Multi-Page Units of Work with Automatic State Management

When your applications need to support an end-user task requiring data entry on many different web pages to complete, you have to invent a scheme for "carrying" the pending data between pages. For example, consider a web application that lets employees file expense reports. The task requires entering many expense line items, potentially requiring multiple pages to complete the task. As the user is entering these details, where do you store the pending expense report data as the end-user enters it bit by bit?

The most common approaches in use today are to store the pending data in:

1. HTTP Session Objects
2. Hidden Form Fields
3. Temporary database "staging tables"

All of these approaches have downsides, though:

1. For some J2EE application servers, data stored in the HTTP Session cannot be reliably accessed in a "server farm". Even using servers like Oracle Application Server that do support clustering and replication of session state, the more and larger the objects you add to your HTTP Session, the more traffic you generate between servers to keep each other's HTTP Session objects synchronized.
2. Storing pending data in hidden form fields, in all but the simplest of cases, requires inventing a scheme to reflect the objects' field structure in the names of the hidden fields, and increases the size of every page downloaded.
3. Since pending data might not yet be complete, it often is impractical to commit it directly into the underlying application tables, which might raise database constraint violations. So, pending data could be stored in some staging tables that allow in-progress work to be saved and loaded later. However, either a special set of tables need to be created for each kind of pending data you need to store, or you need to invent a generic scheme for serializing the pending data, which is not fun.

The ADF framework provides a built-in state management facility for your business services that addresses all of these issues out of the box. Any application module instance can "snapshot" its pending state to the database and "reactivate" that state from the database at a future point in time. The pending state is serialized by generic framework code into a single XML document, containing only the pending changes, and saved in a single round-trip to the database in a generic table named PS_TXN, managed by ADF. This approach has the following benefits:

• Works on any J2EE-compliant application server
• Works well in application server farms of any size
• Has low overhead: only a single database round-trip per request
• Avoids application server HTTP Session synchronization traffic
• Doesn't increase the size of web pages delivered to the user
• Uses the same, generic mechanism for all applications you build

The database-backed session management occurs on a different JDBC connection from the application data access and persistence. Typically you will want to have the state management table in a separate schema from the application tables. You can control this by setting the value of the jbo.server.internal_connection configuration property to either the name of a JDBC datasource name, or a JDBC URL connection string. The ADF Toy Store demo sets the value of jbo.server.internal_connection to use the jdbc/toystore_statemgmtDS JDBC datasource name so that the PS_TXN table is created and managed in the TOYSTORE_STATEMGMT schema, instead of the TOYSTORE schema where the application tables live.

The ADF application module pool supports three different usage patterns for how application module components are used by incoming browser-user requests. The three modes are called:

• Stateless Mode

  This retains no business service component state, forcing pending changes to commit/rollback in the current request.

• Stateful Mode

  This allows pending state to span multiple page requests, using the built-in state management mechanism just described.

• Reserved Mode

  This dedicates an instance of the application module component to the current browser session until later released in stateless mode, or until the browser session times out.

When we picked the name "Stateful Mode", we honestly made a really unfortunate choice of terms. We've since come to learn that the word "stateful" conjures up ideas of terrible scalability in web developers' minds. Luckily, it's just the name that's unfortunate; the actual functionality is great. In hindsight, we should have called "Stateful Mode" something like "Managed State with Stateless Performance Mode".

Using "Stateful Mode", you can build web applications with performance very near that of a completely stateless application, but with the programming simplicity of a stateful application. To deliver on this promise, the ADF application module pool implements an algorithm known as "Stateless with Session Affinity". Across multiple page requests in a "stateful mode" page flow, the pool attempts to return the same component instance used by the current session on its most recent request in the current page flow. If this instance is still available in the pool, and it has not been used by another session in the meantime, this scenario gives the highest performance.

If instead the pool needed to hand out that instance to service another browser user session in the meantime, or if the request has come into a different server in the server farm, then the pool grabs any available business service component instance and "reactivates" the current browser user's pending state from the database. This means that if the user had created 5 new rows, updated 2 rows, and deleted 1 row, that these same pending changes would be present after reactivating the state from the database. Also, view object iterator position information is included in the "state snapshot", so that if on the last request your iterator was pointing at the third row in a collection of twenty, your current row will still be row number three after state reactivation.

The executive summary is that if you size the application module pool for the load your application needs to handle, then you get excellent performance and scalability.

Stateful Mode is the ADF default mode of operation. Since the task of filling the user's shopping cart and buying the items is all part of a logical transaction, this default works well for the Toy Store Demo.

To look at one example, let's take the ADF Toy Store Demo's shopping cart. It is implemented as an ADF view object named toystore.model.dataaccess.ShoppingCart in the ToyStoreModel project, This view object has all transient attributes and no SQL query. The YourCartAction Struts action class calls the ToyStoreService business interface method adjustQuantitiesInCartAsStringArrays() to add, change, or remove items from the cart. Since we're using "Stateful Mode", the application code does not have to worry about how to store the pending shopping cart data. By visiting the Tuning panel of the View Object editor for the ShoppingCart view object, we indicate declaratively that we want all transient attributes of this view object to be passivated by the framework's state management mechanism. Just checking the checkbox there is the only work required to leverage this feature.

On each subsequent request, our actions can access the ShoppingCart view object to work with its collection of data transfer objects, and programmatically adjust the contents of the view object's default rowset.

The final action in the demo's page flow, the "/thankyou" action uses the "Stateless" release mode to indicate that we no longer need any more pending state to be managed. In order to release in stateless mode, it overrides the main handleLifecycle() method of the DataForwardAction and adds a call to the helper method releaseToyStoreServiceStateless(), which it inherits from the ToyStoreServiceDataForwardAction:

/* From ToyStoreServiceDataForwardAction */
  protected void releaseToyStoreServiceStateless(DataActionContext ctx) {
    releaseDataControlStateless(DATACONTROLNAME, ctx);
  }

This method turns around and invokes a helper method on the ToyStoreDataForwardAction, passing in the name of the data control to release in stateless mode. As you can see in the method below, it finds this data control and invokes its releaseState() method, so that all pending statement management info is released at the end of the request.

/* From ToyStoreDataForwardAction */
  protected void releaseDataControlStateless(String dcName,
    DataActionContext ctx) {
    DCDataControl dc = ctx.getBindingContext().findDataControl(dcName);
    if (dc != null) {
      dc.resetState();
    }
  }

Releasing an application module to the pool in stateless mode automatically cleans up any database-backed pending state that was being managed by the framework on its behalf.

Building the View Layer with JSP Pages and JSTL

As Figure 47 shows, the majority of the view layer in the ADF Toy Store demo is implemented using JSP pages, with the exception of one example illustrating how to use an XML/XSLT-based approach as an alternative. We've consciously placed the pages in subdirectories of the WEB-INF directory so that users cannot directly browse the pages. We want to force them to go through our controller layer for every request, and not be able to bookmark and return to URL's that short-circuit the controller layer by going straight to the JSP page. Our controller layer can forward control to pages under WEB-INF, but they cannot be browsed directly based on rules laid out in the J2EE specifications. So this is a safe, best-practice approach for guaranteeing that our controller layer is in total control.

Figure 47: Display-Related Resources and Pages in the ToyStoreView Project

The toystore.view package contains the ToyStoreResources.properties file and the GlobalErrors.properties file that store translatable strings used by our view layer pages, for the default language (English). The additional versions of these two files with the "_it" and "_de" in the name contain the Italian and German translations of the same strings, respectively.

DataBindings.cpx is the ADF Binding Context file that store metadata about data controls and binding container names. The *UIModel.xml files contain the detailed binding container metadata, describing the bindings each one contains. Clicking on DataBindings.cpx in JDeveloper's Application Navigator, and looking at its details in the Structure Window, you can see that our demo only has a single data control named ToyStoreService defined.

Accessing and Iterating Over Model Data

Let's start by looking at one of the view layer JSP pages to study how it's built, the yourcart.jsp page. Recall from the Lifecycle of a Web Page Request Using Struts and ADF section that the ADFBindingFilter, DataAction, and DataControl collaborate to handle the details of acquiring an instance of your application module, making it available to the controller and view layers as a data control, and then releasing it at the end of the request. Using the DataActionContext object that the DataAction creates at the beginning of each request, your controller layer code can access:

• The BindingContainer, to work with any of the bindings
• The BindingContext, to work with any of the data controls

Once you have the data control in "hand", you can access its data provider and cast it to the application module's business service interface and can directly or indirectly access the collections of data transfer objects in its model data map. Or, by working with the bindings, you can avoid the need directly work with the business service if you choose. Remember that given the roles that the controller layer and the view layer play in the MVC architecture, the controller can modify the model data, whereas the view layer can only iterate over it to present it.

The popular JSP Standard Tag Library (JSTL) provides a set of handy tags that virtually all JSP web page developers need. Using the JSTL tags, your JSP pages can read and write attributes from page, request, session, and application scope, easily iterate over their values, output their values into your page where needed, and conditionalize the rendering of the page. You have always been able to perform these tasks in a JSP page: it's just that with JSTL, you can do them without resorting to pesky JSP scriptlets that make your page harder to manage.

Using the JSTL tags, you identify which objects and properties you want to work with by using a simple dot notation called the JSTL Expression Language, known as "EL" for short. To distinguish themselves from literal values, these EL expressions appear in the attribute values of JSTL tags surrounded by the ${ and } like this: ${expression}.

For example, to refer to the value of an attribute named bindings, you would use the expression ${bindings}. If the object returned is a JavaBean with properties of its own, or a Map with named members, you can use a dot notation to refer to its members. For example, if the object named bindings is a map that contains a member named ShoppingCart, then you can refer to it using the expression ${bindings.ShoppingCart}.

ADF binding objects are easy to use with any client technology that can interact with JavaBean's and the Java Collections Framework, which includes the JSTL tag library. On each request, the ADF framework's DataAction makes the current binding container available as an attribute named bindings and the current Binding Context available as an attribute named data. You can use EL expressions in JSTL tags to refer to any information in these contexts.

When working with JSTL tags, in addition to the standard context-sensitive "Code Insight" provided for the tag names and tag attributes, as shown in Figure 48, JDeveloper 10g also provides EL-specific Code Insight for helping you create your EL expressions.

Figure 48: Code Insight in JDeveloper 10g for EL Expressions

We'll see in the next sections that all the JSP pages in the ADF Toy Store demo make use of JSTL and EL expressions for presenting the data exposed by the ADF binding layer.

Let's start by taking a peek at how the shopping cart page works. The /yourcart Data Page, which used the related /WEB-INF/jsp/yourcart.jsp page for its view-layer rendering, is configured like this:

    <action path="yourcart"
            className="oracle.adf.controller.struts.actions.DataActionMapping"
            type="toystore.controller.strutsactions.YourCartAction"
            name="DataForm"
            parameter="WEB-INF/jsp/yourcart.jsp"
            unknown="false">
      <set-property property="modelReference"
                       value="WEB_INF_jsp_yourcartUIModel"/>
      <forward name="reviewcheckout" path="reviewcheckout.do"/>
    </action>

It is mapped to the YourCartAction action class, which performs the setup necessary to retrieve the appropriate model data. We can see from the value of the modelReference property above that the binding container name for the /yourcart DataPage is WEB_INF_jsp_yourcartUIModel (a name generated for us by the ADF design time tools). With the yourcart.jsp page active, clicking on the UI Model tab of the Structure Window shows the contents of the binding container for the page as shown in Figure 49. There is one iterator binding (ShoppingCartIterator), and one range binding (ShoppingCart).

Figure 49: UI Model Tab Showing Binding Container for yourcart.jsp

At runtime, the default behavior of the DataForwardAction is to forward control for rendering to its "companion" JSP page. In this case, our ShowProductDetails action is configured to forward to the yourcart.jsp page.

The page starts by declaring four tag libraries:

1. The Struts Bean tag library, whose tags will be prefixed by bean:
2. The Struts HTML tag library, whose tags will be prefixed by html:
3. The JSTL Core tag library, whose tags will be prefixed by c:
4. The ADF tag library, whose tags will be prefixed by adf:

We're using the Struts Bean tag library to access the handy <bean:message> tag, which makes it easy to include translatable text strings into our pages, based on string keys like "details.title", "cart.addItem", and "images.buttons.addtocart".

We're using the Struts HTML tag library to access the <html:errors> tag, which makes it easy to displays any errors that occur during runtime processing at the top of the page in a standard way.

We're using the JSTL Core tag library to iterate our data collections and include values of their attributes in the page. The JSTL tags in use in this page are:

• <c:choose>, <c:when>, and <c:otherwise> which work like an if/then/else statement to conditionalize the display based on whether your shopping cart is empty or not.

    <c:choose>
    <c:when test="${not empty bindings.ShoppingCart.rangeSet}">
     <form action="yourcart.do" method="post">
        <!-- etc. -->
     </form>
    </c:when>
    <c:otherwise>
      <br><br><bean:message key="cart.empty"/>    
    </c:otherwise>
  </c:choose>


• <c:forEach> to iterate over the items in your shopping cart, and <c:out> to display attribute values of those items.

  <c:forEach var="Row" items="${bindings.ShoppingCart.rangeSet}" >
  <tr bgcolor="#f3f3f3">
    <td>
      <a href="yourcart.do?event=removeItem&id=<c:out value='${Row.Itemid}'/>"
      ><img src="<bean:message key='images.buttons.removefromcart'/>"
      border="0" alt="<bean:message key="cart.removeItem"/>"></a>
    </td>
    <td><c:out value="${Row.Itemid}"/></td>
    <!-- etc. -->
  </tr>
</c:forEach>

We're using the ADF tag library to use the <adf:render> tag that outputs formatted attribute data based on language-sensitive format masks that you can define in your business components.

The output that appears in the browser is shown in Figure 50.

Figure 50: Your Cart Page

Recall that the contents of the shopping cart was kept in transient view object rows and populated programmatically as the user adds and removes items from the cart. So, in the /yourcart example from the previous section, there was no database query involved in rendering the page.

Next we'll take a look at a page like the /showproductdetails DataPage that presents queried database information. The /showproductdetails Data Page, which used the related showproductdetails.jsp page for its view-layer rendering, is configured like this:

<action path="showproductdetails"
        className="oracle.adf.controller.struts.actions.DataActionMapping"
        type="toystore.controller.strutsactions.ShowProductDetailsAction"
        name="DataForm"
        parameter="WEB-INF/jsp/showproductdetails.jsp"
        unknown="false">
  <set-property property="modelReference"
                   value="WEB_INF_jsp_showproductdetailsUIModel"/>
  <forward name="addToCart" path="yourcart.do"/>
</action>

It is mapped to the ShowProductDetailsAction action class, which performs the setup necessary to retrieve the appropriate model data. That action contains the code shown in Example 9.

We customized the DataAction lifecycle in a framework extension class to add the additional initializeModelForPage() method. We override this method in our actions set bind variable values in our view object's queries before the DataAction's prepareModel() lifecycle phase will cause the queries to be executed. This is a common use case that our additional lifecycle method makes easier to handle.

If you look in the ToyStoreDataForwardAction, you'll see where we've introduced this new lifecycle method by overriding the existing prepareModel() method and augmenting its default behavior to call initializeModelForPage() first if we are not handling any events:

/* From: toystore.fwk.controller.ToyStoreDataAction */
  /**
   * Overridden method.
   *
   * Add two new overrideable methods into the ADF DataAction lifecycle as
   * part of the prepareModel() processing to make handling the page
   * initialization use case easier.
   */
  protected void prepareModel(DataActionContext ctx) throws Exception {
    if (!handlingEvents(ctx)) {
      initializeModelForPage(ctx);
    }
    super.prepareModel(ctx);
    if (!handlingEvents(ctx)) {
      initializeBindingsForPage(ctx);
    }
  }

The body of the initializeModelForPage() method back in Example 9 uses the DataActionContext to retrieve the id parameter from the HttpServletRequest and then pass it as an argument to the prepareToShowProductDetails() method on our ToyStoreService business service interface. Behind the tier-independent ToyStoreService interface is the ToyStoreServiceImpl implementation class, whose prepareToShowProductDetails() method looks like this.

/* From: toystore.model.services.ToyStoreServiceImpl */
  public void prepareToShowProductDetails(String id) {
    getFindItems().setItemToFind(id);
    getFindItems().executeQuery();
  }

It simply encapsulates the setting of the item id to find, and re-execution of the view object's query. The getFindItems() method is a view object instance getter method that the ADF design time tools generate for you when you add a view object instance named "FindItems" to your application module's data model. The view layer will be able to access these data transfer objects representing the view object query's results by using appropriate control value binding objects. Under the covers, these control value bindings are related to an iterator binding that knows how to obtain the data from the "FindItems" collection from the model data map.

Notice again back in Example 9 that we're using the getToyStoreService() helper method from the ToyStoreServiceDataForwardAction superclass that returns our business service's custom service interface named ToyStoreService.

/* From toystore.controller.strutsactions.ToyStoreServiceDataForwardAction */
  protected ToyStoreService getToyStoreService(DataActionContext ctx) {
    return (ToyStoreService) getApplicationModule(DATACONTROLNAME, ctx);
  }

This code, in turn, calls the getApplicationModule() helper method from its superclass: ToyStoreDataForwardAction. As you can see below, that method finds the data control by name, then if it is of the expected type for a data control based on an ADF Application Module, it returns the instance of the service as an ApplicationModule interface to work with.

/* From toystore.fwk.controller.ToyStoreDataForwardAction */
  protected ApplicationModule getApplicationModule(String dataControlName,
    DataActionContext ctx) {
    DCDataControl dc = ctx.getBindingContext().findDataControl(dataControlName);
    if ((dc != null) && dc instanceof DCJboDataControl) {
      return (ApplicationModule) dc.getDataProvider();
    }
    return null;
  }

At runtime, after the initializeModelForPage() method has set the appropriate bind variable values, the default behavior of the DataForwardAction is to forward control for rendering to its "companion" JSP page. In this case, our ShowProductDetails action is configured to forward to the showproductdetails.jsp page that you see in Example 10. The tag libraries in use are the same ones as for the yourcart.jsp page. The only difference here is that the page displays data a single row from FindItemsIterator, so we don't need a range binding in the binding container and don't need a <c:forEach> loop. We also don't have any conditional display logic going on, so this page has no <c:choose> or <c:if> tags.

The output that appears in the browser is shown in Figure 51.

Figure 51: Show Product Details Page

All of the JSP pages in the ADF Toy Store demo follow this basic approach for iterating and formatting data.

Three different pages in the demo offer the ability to view results a page at a time:

• /search - showing results from a storewide product search
• /showcategory - showing the products in a category
• /showproduct - showing available items of given product type

We would like to create a reusable paging control that displays feedback like 1 - 3 of 18 and conditionally shows Previous and Next links when it makes sense. The pieces of information we need to do the job are:

• Total Number of Rows
• First Row Number Shown on the Current Page
• Number of Rows Shown per Page

Each of these pages includes an appropriate RangeBinding in their respective binding container. The range binding is an ADF binding object designed for presenting rows of data in a grid, optionally navigating them a page (or "range") at a time. You set the size of the range on the related iterator binding to be the number of rows that you want to appear on a page at a time. If you set the range size to the value -1, then all rows in the data set will display at once.

Figure 52 shows the binding container for the /search DataPage, with its FindProducts range binding, and the underlying FindProductsIterator iterator binding and the range binding is related to.

Figure 52: UI Model Tab Showing Binding Container for Search Page

At runtime, the range binding exposes properties that we can access via JSTL expressions to retrieve interesting information about the collection to which its related iterator bound. Some of these properties include:

• estimatedRowCount - an estimate of the rows in the collection
• rangeStart - the zero-based row number that appears at the top the current range of rows
• rangeSize - the number of rows to show per page

Using these three pieces of information we can calculate all of the bits of information we need. Example 11 shows the code of our pagingControl.jsp that you'll find in the ToyStoreViewController project. You can see that it leverages the following different JSTL tags, using EL expressions to access the binding objects it needs:

• <c:set> - to set page-local variable values to use later in the page

• <c:choose> - to perform if/then/else logic without writing scriptlet code

  (along with its nested <c:when> and <c:otherwise> tags).

• <c:out> - to output the value of an EL-expression

The Previous and Next links that are conditionally created are generated to have the event=Previous and event=Next parameters in their respective URL's. This will cause the action bindings of the matching Next and Previous names to be declaratively executed in the current page's binding container if they exist.

If you look inside the search.jsp page's source code, you'll see that it makes use of this reusable pagingControl.jsp page using <jsp:include>, passing a couple of parameters needed by the page with nested <jsp:param> tags. Notice that it uses a <c:choose> tag that uses the EL not empty operator to test whether any products have been found or not. If no products were found, it outputs a specific "No matching products" message instead of showing an empty table of results.

    <c:choose>
      <c:when test="${not empty bindings.FindProducts.rangeSet}">
        <jsp:include page="pagingControl.jsp">
          <jsp:param name="rangeBindingName" value="FindProducts"/>
          <jsp:param name="targetPageName" value="search.do"/>
        </jsp:include>
        <table border="0" bgcolor="#003399">
          <!-- etc. -->
        </table>
      </c:when>
      <c:otherwise>
        <br><br><bean:message key="search.nomatchingproducts"/>
      </c:otherwise>
    </c:choose>

The first thing that the pagingControl.jsp does is take the string-valued rangeBindingName parameter that we pass in, and lookup the range binding object having that name using the syntax:

<c:set var="rangeBinding" value="${bindings[param.rangeBindingName]}"/>

Inside the pagingControl.jsp, it can use EL expressions to refer to properties of that range binding to do its job in rendering the paging control display. Notice that instead of using the EL dot notation like ${bindings.SomeRangeBinding}, this expression uses array-index-style notation using square brackets. This allows the member name your are trying to access, in this case on the bindings object, to be provided by another expression value, instead of providing the name as a literal.

One detail you might be asking yourself is, "Where do I set the number of rows per page to display?" Excellent question! Clicking on the FindProductsIterator in the UI Model tab shown in Figure 52 and looking at the Property Inspector, we can see that the Range Size property is set to the value 3. This means that at runtime, this iterator will present its results in pages of up to 3 rows at a time. To change to show five at a time, we would only need to modify this declarative iterator property to the value 5 instead of changing code. If you need to support functionality allowing the user to change the number of rows displayed per page, keep in mind that you can also call setRangeSize() on your iterator binding.

A Data Entry Form Using Traditional Form Layout Approach

The /editaccount DataPage illustrates an example of a page that displays a data entry form to edit user profile information. It uses a traditional JSP page layout approach of placing each control to use inside an HTML form.

The EditAccountAction sets up the model layer in its initializeModelForPage() method as we've seen above. It calls the custom service method prepareToEditAccountInfoFor() on the ToyStoreService interface, passing in the name of the current user as an argument.

The implementation of this method in the ToyStoreServiceImpl class looks what you see in Example 12. It performs the following three basic steps:

1. Creates a oracle.jbo.Key object based on the current user's name passed in
2. Looks up an existing row in the Accounts view object by passing this key to the findByKey() method on the view object.
3. Sets that row as the current row in the view object.

In the corresponding JSP page, named editexistingaccount.jsp, we use the <html:form> tag from the Struts HTML tag library to implement the "Postback Pattern" by having its action post back to the DataPage like this:

<html:form action="editaccount.do" method="post">

At runtime, the Struts <html:form> tag sees the action attribute value of /updateaccount.do and uses it, along with its action mapping information, to determine that the FormBean named DataForm is the one that should be used to render this form. The DataForm form bean is defined in struts-config.xml to use our ADF BindingContainerActionForm.

Since we're rendering the data entry form for just a single "row" of user account information, we don't need to use the use the JSTL <c:forEach> in this page and don't need a range binding in our binding container. We simply format the individual fields in the form, using normal HTML table tags to get the prompts and controls to line up nicely. As this page shows off several different techniques in use, we'll try to highlight each of the important ones in turn.

Figure 53 shows the binding container for the EditAcount page. Notice that we have basic attribute bindings for all of the Accounts attributes except for Country, which is a list binding (its icon shows a poplist in it). We have two iterator bindings: AccountsIterator for the main Accounts information we're editing, and CountryListIterator that will supply a poplist with the valid country names the user can choose for the Country attribute.

Figure 53: UI Model Tab Showing Binding Container for EditAccount Page

We also have an action binding named save that is bound to the built-in Commit operation on the ToyStoreService data control.

Example 13 shows the tags in the page that output the HTML table row containing the prompt and data for the Username property. Since the username in this application is not updateable once it's been created, we don't need to render an HTML form control for the data. Using the <c:out> tag, we can easily just output the value of the field for display using its corresponding binding object. The <bean:message> tags are outputting translatable strings from the default ToyStoreResources.properties properties file to display the tooltip and the label for the username.

When data needs to be entered or edited, you can use a number of other tags in the Struts HTML library to render databound controls. Example 14 shows using the <html:password> tag to show the Password property.

<html:password property="Password" size="25" maxlength="30"/>

Recall that the ADF BindingContainerActionForm presents Struts (and here in particular, the Struts HTML tag library's tags) with a DynaActionForm bean having properties that are named after, and "wired to", the bindings in your current binding container, so when the <html:password> tag gets and sets the value of the Password property on this form bean, behind the scenes ADF is coordinating the properties of that form bean with the corresponding binding objects.

The example also illustrates using the Struts HTML tag <html:errors> to display any validation errors that are specific to the Password attribute. Of course, when the form is first rendered there won't be any validation errors, so this table cell will be empty. However, if the user submits the form and validation errors in the model layer are thrown, when this page is rendered again, any errors related to Password will show up next to the Password field on the screen. Also, since we know this field is mandatory, we've included a <bean:write> tag to show the string corresponding to the key "dataentryform.mandatory" as a visual marker to the user that the field is required. By default, we render an asterisk.

ADF entity object and view object components have a number of built-in features that allow developers to define control hints like locale-sensitive labels, tooltips, and format masks. The ADF binding layer exposes this metadata directly on the binding objects for convenient access by your view layer pages. The <c:out> tags shown in Example 15 illustrate the EL expressions for the tooltip and label information that have been associated with the business object attributes or the view object attributes. If an entity object has defined a tooltip for one of its attributes named Firstname, for example, then this tooltip is inherited by any view objects that include Firstname. Of course, the view object can also override these control hints if necessary.

Each binding object exposes runtime metadata about the objects to which it is bound that you can access at runtime using EL expressions. For example, the control value binding for an attribute exposes information about the underlying attribute in the model layer. In Example 15 we see an example of using this metadata to detect at runtime whether a given attribute, like Firstname, is mandatory or not. Combined with the JSTL tag <c:if>, we can use this information to conditionally output the mandatory marker on a required field.

<c:if test="${bindings.Firstname.mandatory}">
  <bean:message key="dataentryform.mandatory"/>
</c:if>

One helpful tip to remember is that to get a quick review of all the available properties on a binding object, you can just click on the binding in the UI Model tab of the Structure Window and press the F1 key. The online help topic for the appropriate binding object appears in an IDE window for your reference.

Finally, we look at an example of a data-bound form control like a poplist showing the country where a user resides. As shown in Figure 54 their are two dimensions to the control:

1. The value of the underlying Country binding, reflected by the selection in the list, and
2. The list of all available country names to chose from.

Figure 54: Poplist Has Both Current Value and List of Valid Values

ADF provides more sophisticated binding objects to handle controls list this that have multiple facets to their data binding requirements. The ADF List Binding caters specifically to poplist-type controls that need to manage both a current bound attribute value, as well as a list of valid choices to present to the user. For hierarchical data, ADF supplies a Tree Binding object that can come in handy on occasion as well.

By clicking on the UI Model tab of the Structure Window while the editexistingaccount.jsp is active, you'll see the bindings we saw back in Figure 53. If you select the CountryListIterator and look in the Property Inspector, you'll see that it has a range size of -1. This value indicates that you want all rows to appear in the list of countries, instead of only a partial set.

Clicking on the Country binding and selecting Edit... from the right-mouse menu, you will see the List Binding Editor shown in Figure 55. It allows you to see the binding metadata required to support the Country poplist:

• The datasource for the list of available choices comes from CountryListIterator
• The iterator whose current row will be used to determine the current value of the binding and to update in case of selecting a new item from the list is AccountsIterator.
• The (source,target) attribute pairs show that the value of the Code property from the selected row in the CountryList will be set into the Country property on the current row of the target AccountsIterator.

If you click on the LOV Display Attributes tab, you can observe that the Description attribute from the CountryListIterator is indicated as the value to display to the user in the list.

Figure 55: List Binding Editor for the Country List Binding

Example 16 shows how to use the <html:select> and <html:optionsCollection> to leverage this Country list binding to put the poplist onto our page. The <html:select> is bound to the Country property of the form bean, which is non other than our list binding object. The <html:optionsCollection> gets its data from the nested, List-valued displayData property of that same Country binding. The beans in this display data collection each have a prompt and an index property, so we indicate to use those as the label and value (respectively) for each option in the list.

Rendering Data Entry Forms in a More Generic Way Using Metadata

In contrast to the more "traditional" technique explained above, the Toy Store demo also includes another page that renders a data entry form in a more generic, metadata-driven way. Both forms render the same set of controls for Accounts data, so it's even easier to compare the two approaches and pick the one that will suit your applications best.

Example 17 shows the registernewuser.jsp page (used by the /register DataPage) which renders the data entry form allowing users to register on the site for the first time. The results produced in the browser of this page are nearly identical to the editexistingaccount.jsp page we looked at above, but as you can see from the example, the whole form is rendered by the single <jsp:include page="formControl.jsp"> tag. This tag works like a reusable component, including the contents of the formControl.jsp page. The nested <jsp:param> tags pass three parameters to the reusable component page:

1. dataPage - The name of the current datapage
2. saveButtonLabelKey - The message bundle key to the label to display on the (Save) button
3. saveButtonEvent - The name of the event to association with the pressing of the (Save) button.

So the actual work being done lies in the formControl.jsp "component" page. The page builds a data entry form with one data-bound control for each control value binding in the current binding container.

The page begins with some examples of using the <c:choose>, <c:if>, and <c:set> tags to conditionally set up the values of page local variables named eventName, buttonLabel, and buttonLabelKey based on whether and which of the expected input parameters were provided. We'll use these variables later in the page as part of constructing the (Save) button at the bottom of the generated form.

<c:choose>
  <c:when test="${not empty param.saveButtonEvent}">
    <c:set var="eventName" value="${param.saveButtonEvent}"/>
  </c:when>
  <c:otherwise>
    <c:set var="eventName" value="Commit"/>
  </c:otherwise>
</c:choose>
<c:if test="${not empty param.saveButtonLabel}">
  <c:set var="buttonLabel" value="${param.saveButtonLabel}"/>
</c:if>    
<c:if test="${not empty param.saveButtonLabelKey}">
  <c:set var="buttonLabelKey" value="${param.saveButtonLabelKey}"/>
</c:if> 

The formControl.jsp page goes on to use <html:errors> tag as part of a "global errors" section of the input form, where any errors that are not attribute-specific will show up:

<center>
  <table border="0">
    <tr>
      <td><html:errors bundle="GlobalErrors"
                     property="<%= ActionErrors.GLOBAL_ERROR %>"/></td>
    </tr>
  </table>
</center>

Next the form uses the value of the dataPage parameter passed in by the jsp:include as part of opening the <html:form> tag. Notice that since we cannot use EL expressions directly in the <html:form> tag's action attribute, we use <c:set> to first set a page local variable named name with the EL-expression value we want, then we use a JSP scriptlet to pass the value of this name variable to the action attribute:

  <c:set var="name" value="${param.dataPage}.do"/>
  <html:form action='<%= pageContext.getAttribute("name")%>'>
    <!-- etc. -->
  </html:form>

The form includes the standard hidden field that the ADF controller layer uses to detect whether the user has tried to submit the same form multiple times in rapid succession:

    <input type="hidden" name="<c:out value='${bindings.statetokenid}'/>"
           value="<c:out value='${bindings.statetoken}'/>"/>


Next we begin the loop that will create an HTML form field for each control value binding in the binding container. Inside the <table> tag, we have the following <c:forEach> iteration:

<c:forEach var="curBinding" items="${bindings.ctrlBindingList}">
  <% JUControlBinding cb =
       (JUControlBinding)pageContext.getAttribute("curBinding");
     if (cb instanceof JUCtrlValueBinding &&
       !(cb instanceof JUCtrlRangeBinding) &&
       !(cb instanceof JUCtrlHierNodeBinding)) { %>
   <!-- Build control for current control value binding in here -->
 <% } %>
</c:forEach>

The <c:forEach> loop iterates over the list of control value bindings from the binding container. Since this list might include control action bindings, we need to skip over those when rendering the input controls. Since we want to keep things simple, we'll also skip over the RangBindings and TreeBindings, too. The EL expression language doesn't have a built-in instanceof operator, so we're using a JSP scriptlet to use a regular Java-language if statement to perform the combination of instanceof checks.

Since we specified the var="curBinding" attribute on the <c:forEach> tag, inside the loop we can refer to this curBinding loop variable to access the current control value binding as part of our generic form input control generation. Notice how we're making use of the binding properties in our EL expressions like tooltip, mandatory, and label to access this metadata from the current control binding.

<c:forEach var="curBinding" items="${bindings.ctrlBindingList}">
  <% JUControlBinding cb =
       (JUControlBinding)pageContext.getAttribute("curBinding");
     if (cb instanceof JUCtrlValueBinding &&
       !(cb instanceof JUCtrlRangeBinding) &&
       !(cb instanceof JUCtrlHierNodeBinding)) { %>
   <tr>
     <th align="right" title="<c:out value='${curBinding.tooltip}'/>">
       <c:if test="${curBinding.mandatory}">*&nbsp;</c:if>
       <c:out value="${curBinding.label}"/>
     </th>
     <td>
        <c:set var="name" value="bindings.${curBinding.name}"/>
        <adf:inputrender model='<%= pageContext.getAttribute("name")%>'/>
     </td>
     <c:set var="name" value="${curBinding.name}"/>
     <td>
       <html:errors property='<%= pageContext.getAttribute("name") %>'/>
     </td>
   </tr>
 <% } %>
</c:forEach>

To actually render the HTML form control, we use the <adf:inputrender> tag which is setup to render an appropriate tag based on the datatype of the current binding's attribute value. As we'll see, we can also use some additional attribute metadata to customize the way that the <adf:inputrender> tag does its job. We repeat our trick of using <c:set> to set a local page variable named name to the concatenation of the string "bindings." with the name of the current binding, which is what the <adf:inputrender> tag wants to "see" as the value of its model attribute.

Finally, we use the <html:errors> tag to show any attribute-level validation errors that might occur next to the control to which they are relevant. We again use the <c:set> trick to get the value of the <html:errors> tag's property attribute to be the name of the current binding.

Last, but not least, we use a <c:choose> to put the appropriately-labeled (Save) button at the bottom of the form. We're based on whether the user specified a button label or a button label key, we either use the literal label string, or employ the bean:message tag to lookup the label key for us. We're using our page local variable eventName that we setup at the top of the page to fill-in the right name for the button to generate that event when clicked by the user.

<c:choose>
  <c:when test="${not empty buttonLabel}">
    <input name="event_<c:out value="${eventName}"/>" type="submit"
            value='<c:out value="${buttonLabel}"/>'/>
  </c:when>
  <c:when test="${not empty buttonLabelKey}">
    <input name="event_<c:out value="${eventName}"/>" type="submit"
            value='<bean:message name="buttonLabelKey"/>'>
  </c:when>
  <c:otherwise>
     <input name="event_<c:out value="${eventName}"/>" type="submit"
            value='Submit'/>
  </c:otherwise>                      
</c:choose>

Each ADF Business Component supports setting custom properties that can be read at runtime and be used to drive metadata-driven behavior. For ADF entity objects and view objects which have attributes, you can also set custom properties on individual attributes as well.

To edit attribute properties, expand the Attributes heading in the respective object editor, and click on the name of the attribute whose properties you want to modify. Then, click on the Attribute Properties tab on the editor panel on the right. Figure 56 shows what this looks like for editing the custom attribute properties of the Country attribute of the Accounts view object.

The <adf:inputrender> tag implementation consults the value of the EditRenderer attribute property to see if the attribute has specified a custom renderer. If none is specified, a default heuristic is used to pick an appropriate control.

Figure 56: Custom Attribute Properties for Country Attribute in Accounts View Object

In this example, we've specified the class name toystore.fwk.view.ListBindingPoplistRenderer which implements a customized poplist renderer for ADF list bindings. This custom field renderer extends the default oracle.jdeveloper.html.StaticPickList renderer to populate some of its properties based on information it can retrieve from the list binding object. The source code for the custom renderer (from the FwkExtensions project in the demo) is shown in Example 18. You can see that the code accesses the JUControlBinding object from the datasource, and after checking that its a JUCtrlListBinding, calls the getDisplayData() method on the list binding to access the list display data. In order to populate the String[] variables for the labels and the values, it iterates over the display data collection and adds the prompt attribute from each bean in the collection to the label array. Since the ADF binding layer will expect the value coming back from the page to be the numerical row number, we populate the values array by converting the loop variable z to a string on each iteration. The net effect is that when our generic formControl.jsp "component" page renders an HTML form for the bindings in the current binding container, the Country binding will render as a data-bound poplist populated from the display data collection configured of value objects in the model data map named "CountryList".

If you adopt a generic data form rendering technique like this in your applications, you can more easily insure that all data entry forms in your application look and act similarly since they are all rendered by the same generic code.

Struts and ADF Features for Building Multilingual Applications

Above we've seen the use of translatable strings in a few example JSP pages, as well as references to translatable attribute labels, tooltips, and format masks. In this section we'll quickly cover the features provided by the Struts and ADF frameworks for building applications that need to support user interfaces in multiple languages.

Struts Message Resource File Support

Struts provides a basic facility for using translated messages stored in standard Java *.properties files. There is a default message resource file, but you can define secondary message resources as well. The ADF Toy Store demo uses:

• The default Struts resource message file

  In ./ToyStoreView/src/toystore/view/ToyStoreResources.properties

• A secondary resource message file for global errors identified by the key "GlobalErrors"

  In ./ToyStoreView/src/toystore/view/GlobalErrors.properties

You make Struts aware of the names of your message resource files in struts-config.xml as shown in Example 19. The <message-resources> element with no key attribute defines the location of the default message resource. Secondary message resources identify their key by specifying a value for the key attribute.

The messages are stored in a properties file that pairs a string key with a text message. For example, two lines that appear in ToyStoreResources.properties are:

  :
cart.addItem=Add Item to Your Shopping Cart
cart.removeItem=Remove Item from Your Cart
  :

You provide translations of the messages in the message resource file by creating a properties file in the same directory with the same name, except with the locale suffix appended to it. For example, the Italian translations of the messages in ToyStoreResources.properties are in the ToyStoreResources_it.properties file in the same directory. The translated message resource files contain the same string key as the default language, but with the translated version of the default language's message string. For example, the above two messages in English look like this in the Italian version of the message resource file:

  :
cart.addItem=Aggiungi al carrello
cart.removeItem=Togli dal carrello
  :

If there are messages that do not need to be translated, there is no need to repeat them in the translated message resource files. If a message cannot be found in the locale-specific message bundle, the one from the default message resource file will be used as a fallback.

As we've seen in the sections above, you include a message resource string in a JSP page using the Struts "Bean" tag library's <bean:write> tag, providing the string key to lookup like this:

<bean:write key="cart.addItem"/>

At runtime, the Struts <bean:write> tag attempts to return the string for the most specifically matching message resource file that is available. For example, if the user's locale is Swiss German (de_CH) then it will:

1. Return the string from ToyStoreResources_de_CH.properties file if it exists, then
2. Try finding it in the ToyStoreResources_de.properties file if it exists, then
3. Return the string matching the key from the default message resource file

Struts infers the locale of the current browser user by looking at HTTP request header properties that browsers send with each request, indicating an ordered list of the user's preferred languages. Specifically, on each request through the Struts RequestProcessor, the default implementation of processLocale() method checks to see if the HTTP Session attribute whose name is defined by the Action.LOCALE_KEY constant is present. If the session attribute exists, processing continues. If it does not exist, Struts infers the locale and saves it in that same session attribute. The net result is that the browser user's language is determined once per session by Struts.

Custom ADF Component Message Bundles

ADF entity objects and view objects can have an optional, associated Java message bundle. This component-specific message bundle stores locale-sensitive values of attribute-level control hints like label, tooltip, format mask, and others, as well as component-specific exception messages. The ADF design time wizards automatically handle the maintenance of the component-specific message bundle Java class for the default language. You are responsible for creating translated message bundles with a locale-specific suffix on the end of the name.

For example, for an entity object named toystore.model.businessobjects.Account, the message bundle will be named AccountImplMsgBundle.java and will look like Example 20. Notice that for a component in package toystore.model.businessobjects, the message bundles (as well as the custom client interfaces) are created automatically in the toystore.model.businessobjects.common subpackage. This package naming scheme emphasizes the fact that message bundle classes and client interfaces are common to both the client tier and the server tier. All of the *Impl.java classes and XML files in the toystore.model.* packages are not shared by both tiers. Their use is restricted to the server tier. Allowing the client or web tiers work only with interfaces (and a minimal number of classes like message bundles) is a best-practice technique that the ADF design time automatically encourages through consistent naming patterns and built-in deployment packaging support.

In the Toy Store Demo, we've also added by hand an extra entity-specific exception message to the two-dimensional array of strings like this:

static final Object[][] sMessageStrings = {
    :
  {ErrorMessages.ENTITY_ALREADY_EXISTS,
   "Another user has already chosen this name. Please try another."},
    :
}

Just as the order of the string keys in the Struts message resource properties files was not meaningful, the order here of the {String,String} elements in the Object array is not meaningful either.

To create a translated version of this resource bundle, for example for Italian, you would create the AccountImplMsgBundle_it class in the toystore.model.businessobjects.common package. As a shortcut, you can:

1. Copy AccountImplMsgBundle.java to AccountImplMsgBundle_it.java

2. Rename the class declaration in the new AccountImplMsgBundle_it.java and make it extend the default language message bundle class, changing:

   public class AccountImplMsgBundle extends JboResourceBundle

   to:

   public class AccountImplMsgBundle_it extends AccountImplMsgBundle

3. Rename the default constructor in the new AccountImplMsgBundle_it.java file from:

   public AccountImplMsgBundle(){}

   to:

   public AccountImplMsgBundle_it(){}

4. Edit the user-visible strings to be in Italian.

This will give you a translated message bundle like Example 21

The message bundle classes extend the oracle.jbo.common.JboResourceBundle framework base class so the messages can inherit the message "merging" functionality. The merging occurs when your message bundle's getContents() method invokes super.getMergedArray(). This allows the translated bundles to only include messages that need translating while other messages are inherited from the superclass bundle. To insure that the message merging works correctly, translated resource bundles should also include this overridden getContents() method.

View objects also support custom message bundles for storing locale-sensitive values of control hints. For a view object named some.pkg.ViewObjectName, the custom message bundle will be named some.pkg.common.ViewObjectNameRowImplMsgBundle.java. Table 3 lists the components in the ADF Toy Store demo that have associated custom message bundles.

Since the business service supporting the model layer could be deployed remotely from the web tier that is accessing it, the model layer tracks its own per-user-session notion of the current user's locale. The ADFBindingFilter also performs a similar preferred language inference to automatically set the current language context on the ADF binding context. Each ADF data control then picks up its locale off that binding context.

Using ADF, XSQL Pages, XSLT, and XML Schema Together

Oracle XSQL Pages is a flexible publishing platform for XML-based information that comes with the free Oracle XML Developers Kit for Java. You create server-side "datapage" templates out of XML documents mixed with "action handler" tags which can assist in building all or parts of your XML-based datapage. It comes with a number of built-in action handlers for pulling XML content from SQL queries, stored procedures, web services, and other sources, and easily combining them with XSLT transformations to produce any kind data format required by the requesting user. The most typical formats are HTML, XML, or Text, but it also supports producing,PDF (when used in combination with Apache FOP), SVG, WML, and others.

Custom XSQL Action Handler for ADF Iterator Bindings

When one of the built-in action handlers doesn't fit the bill, you can write your own custom action handlers to assist the publishing engine in getting XML data or performing other programmatic tasks as part of the XSQL page template processing.

Since Oracle ADF business components feature automatic, bidirectional support for working with XML messages, we can easily conjure up a custom XSQL action handler to allow an Oracle XSQL Page to include XML data from an ADF view object. Example 22 shows the code required to accomplish this.

This custom action handler can then be used inside an XSQL page template using the syntax:

<xsql:action handler="toystore.fwk.xsql.ADFViewObject" iterator="YourIteratorName"/>

We'll see a couple of interesting uses of this handler in the next couple of sections.

Designing and Serving Schema-Compliant XML Datagrams

Let's say we wanted to allow a user to request a review of their order in XML format. This might be used as part of some workflow automation process so that the order that has been placed can be audited automatically. Generally when two programs need to exchange XML, they first agree on an XML Schema that describes the structure of the XML to be exchanged, and then at runtime they exchange documents that comply with that schema.

Using the JDeveloper 10g XML Schema Designer, I designed the "Toystore Orders" XML Schema that you see in Figure 57. This allowed me to create the XML schema with very little knowledge of the low-level details of XML Schema itself.

Each of the rounded boxes represents an XML element in the schema. Attributes, like the id attribute of an Order, appear nested within their element's box. When an element needs to contain a sequence of subelements, I dragged and dropped a sequence connector from the component palette onto the diagram. I defined some custom types like PersonType and ItemType which then I can reuse in other parts of the schema as I've done with the OrderedBy element (reusing PersonType) and the Line element (reusing ItemType).

Figure 57: Toystore Order Schema in JDeveloper XML Schema Designer

To serve up an XML datagram for my order review that complies with this XML schema, I did the following:

1. Created a /revieworderxml DataPage in the Struts Page Flow Diagram
2. Double-clicked on this new DataPage node and specified revieworderAsXML.xsql as the name of the view-layer page.
3. Clicked again on the /revieworderxml DataPage in the Struts Page Flow Diagram and then click on the UI Model tab of the Structure Pane to see <No Bindings>
4. Right-moused on <No Bindings> and selected Create UI Model.
5. Right-moused on the binding container node just created and selected Create Binding > Data > Iterator to create an iterator binding for the ReviewOrder data collection in the ToyStoreService data control.

6. Added an <xsql:action> tag to the XSQL page template to use the custom action handler we created in the previous section, indicating the name of the iterator I want to use like this:

   <Page xmlns:xsql="urn:oracle-xsql">
  <xsql:action handler="toystore.fwk.xsql.ADFViewObject"
               iterator="ReviewOrderIterator"/>
</Page>

The underlying ReviewOrder view object component has a query that is defined to use a bind variable. We need to set the bind variable using an overridden initializeModelForPage() method as we've done in previous examples. Therefore, I customized the DataForwardAction class for the /revieworderxml to use the ReviewOrderAction class that you see in the demo. It looks like this:

package toystore.controller.strutsactions;
import oracle.adf.controller.struts.actions.DataActionContext;
/**
 * ADF DataAction for the "/revieworderxml" 
 * and "/revieworder" action mapping (data page).
 * 
 * @author Steve Muench
 */
public class ReviewOrderAction extends ToyStoreServiceDataForwardAction {
  /**
   * Model initialization logic for this page.
   */
  protected void initializeModelForPage(DataActionContext ctx) {
    String id = ctx.getHttpServletRequest().getParameter("id");
    getToyStoreService(ctx).prepareToShowReviewOrder(id);
  }
  /**
   * Illustrate using an alternative mechanism, implemented in the
   * base ToyStoreServiceDataForwardAction class, to release all
   * data controls in use by the current binding container in 
   * stateless mode.
   */
  protected boolean releaseStateless() {
    return true;
  }
}

If we were to access the /revieworderxml datapage at this stage using a URL like:

http://localhost:8988/ADFToyStore/revieworderxml.do?id=1001

we would see results like that you see in Example 23, reflecting the canonical XML format produced by the ADF view object.

To transform this into the format expected by our XML Schema "contract", we need to create an XSLT stylesheet like the one shown in Example 24 which transforms the above canonical XML format into the "Toystore Orders" XML Schema-compliant syntax.

Finally, we need to augment our XSQL page template to engage this XSLT stylesheet to transform the XSQL datapage before returning to the client. This entails adding one extra line to the top of the revieworderAsXML.xsql page that looks like this:

<?xml-stylesheet type="text/xsl" href="revieworderASXML.xsl"?>

With that additional line in the template, then requesting the XML order review produces the schema-compliant datagram like this:

<Order id="1001" xsi:schemaLocation="urn:oracle-toystore-order schemas/Order.xsd"
       xmlns="urn:oracle-toystore-order"
       xmlns:xsi="http://www.w3.org/2001/XMLSchema-instance">
  <OrderDate>2004-06-15</OrderDate>
  <OrderTotal>19.98</OrderTotal>
  <OrderedBy>
    <GivenName>Jay</GivenName>
    <FamilyName>Tooey</FamilyName>
  </OrderedBy>
  <Lines>
    <Line id="1">
      <ItemId>EST-27</ItemId>
      <Description>White Dice</Description>
      <Quantity>1</Quantity>
      <UnitPrice>3.99</UnitPrice>
      <LineTotal>3.99</LineTotal>
    </Line>
    <Line id="2">
      <ItemId>EST-18</ItemId>
      <Description>Apollo-13 Rocket</Description>
      <Quantity>1</Quantity>
      <UnitPrice>15.99</UnitPrice>
      <LineTotal>15.99</LineTotal>
    </Line>
  </Lines>
</Order>


Transforming the XML Using XSLT

After placing an order at the ADF Toy Store web site the user sees a link that they can bookmark showing their order review. We've designed this /revieworder page as a DataPage related to an XSQL template similar to the above /revieworderxml action. Since the model layer setup was identical to the XML-flavored order review above, we are reusing the same ReviewOrderAction class that we studied above. Rather than double-clicking on the /revieworder DataPage in the Struts Page Flow diagram to create a new action class, instead I just used the Property Inspector to set the value of its type property to toystore.controller.strutsactions.ReviewOrderAction, the same class we used above.

The revieworder.xsql template is the same as the one used above, with one exception: the name of the XSLT stylesheet is different.

<?xml-stylesheet type="text/xsl" href="revieworder.xsl"?>
<Page xmlns:xsql="urn:oracle-xsql">
  <xsql:set-stylesheet-param name="orderId" value="{@id}"/>
  <xsql:action handler="toystore.fwk.xsql.ADFViewObject"
               iterator="ReviewOrderIterator"/>
</Page>

With this in place, clicking on the link from the ADF Toy Store "Thank You" page after placing an order will show you the XML/XSLT-based Order Review that you see in Figure 58. You can study the revieworder.xsl stylesheet in the ./WEB-INF/xsql directory to see how a stylesheet that produces HTML is just as simple as using it to transform XML into a different XML format. Remember that the XML document that it is transforming is the canonical XML produced by the ADF view object that we saw back in Example 23.

Figure 58: Overview of Your Order Built Using ADF and XSQL

Customizing the Default Framework Behavior

Two of the biggest benefits of framework-based J2EE development are:

1. Your application components stand on the shoulders of the base framework functionality
2. When you need to make application-wide changes, you can extend the base framework

For example, if the base ADF framework EntityImpl class does not support a feature that you need all of your entity objects to have, not a problem. No need to file an enhancement request with Oracle Corporation and wait until the ADF Development team implements your desired feature for you. Just take the bull by the horns and add the feature yourself by making a framework customization.

Just create a Java class that extends oracle.jbo.server.EntityImpl and add your additional functionality into that class like this:

public class ToyStoreEntityImpl extends oracle.jbo.server.EntityImpl {
  /* 
   * Any new or customized entity object behavior goes here
   */
}

Then, when you create your entity objects for your application, just set up your components to extend from ToyStoreEntityImpl instead of from the default EntityImpl base class. Figure 59 illustrates how this looks for one of the entity objects in the ADF Toy Store demo like Account which does exactly this.

Figure 59: Application Components Can Extend a Customized Framework Base Class

The same opportunity that is available for customizing ADF framework base classes also exists for many aspects of the Struts framework, too. In this section we highlight the ADF framework customizations that were made to support the ADF Toy Store demo. They all live in the FwkExtensions project.

Struts Framework Customizations for the Controller Layer

In the toystore.fwk.controller package we have the ToyStoreDataForwardAction, which extends the base ADF DataForwardAction to make the following customizations:

1. Added two additional DataAction lifecycle methods named initializeModelForPage() and initializeBindingsForPage() which fire just before and just after the default prepareModel() lifecycle method in the situation when no "postback" events are being handled by the action. By default they do nothing: they live to be overridden by subclasses.

   • initializeModelForPage() is useful for calling custom methods to set bind variables in your business service queries before the prepareModel() phase goes about executing the iterators in your binding container. This setup could be done with a separate DataAction having a separate binding container and a custom method invocation association with it, however that meant that every page requiring bind variable setup would have required an extra action in the page flow model and I wanted to keep it as absolutely simple as possible.
   • initializeBindingsForPage() can be used to programmatically modify the values of bindings before the page has a chance to see them.
2. Added a helper method getApplicationModule() to retrieve an ApplicationModule based on its DataControl name.
3. Customized the default way that a "tree" of bundled ADF exceptions get translated into Struts ActionError objects for display to the user by overriding the reportErrors() method.
4. Added a boolean method releaseStateless() that can be overridden by a subclassing action to return true in order to indicate that all data controls in use by the current binding container should be released stateless at the end of the request. By default, an ADF ApplicationModule will have its pending state managed by the framework. There's also an alternative approach illustrated in this class with the releaseDataControlStateless() method that subclasses can explicitly call to release a data control by name in stateless mode at the end of the request.
5. Added an evalEL() helper method that subclassing actions can call to evaluate an EL expression.
6. Added an invokeEventAction() helper method that subclassing actions can call in their onEventName event-handler methods to carry out the default declarative before of invoking the action binding whose name matches the name of the event being handled before or after writing other custom code in the event handler method.

ADF Framework Customizations for the Model Layer Business Objects

In the toystore.fwk.model.businessobjects package, we created the ToyStoreEntityImpl class that extends the base ADF entity implementation base class to add the following features:

1. Declarative ability to force attribute values to UPPER or lower case.
2. Throwing of a specific custom exception EntityAlreadyExistsException when unique keys are violated

Both customizations rely on overriding the framework's setAttributeInternal() method. To perform the declarative case folding, we add some custom code before calling super.setAttributeInternal(). To support the custom exception handling, we write some custom code in a catch block around the call to super.setAttributeInternal().

The private foldCaseOfStringIfCasePropertySet() method illustrates how to check for a custom attribute-level property named "Case" and if provided, behave accordingly based on whether its value is "lower" or "upper". The State and Country attributes of the Account entity, as well as the Shipstate, Billstate, Shipcountry, and Billcountry attributes of the Orders entity have this custom property set to "upper" in their XML metadata.

If you peek in the Account.xml file or the Orders.xml file in the business objects package directory, you'll see that the custom attribute-level properties show up in the XML nested inside the <Attribute> element that they are related to as shown in Example 25.

Figure 60 shows how to customize the base class for an ADF component like an entity object. On the Java tab of the object editor, there is an (Extends...) button that brings up a Framework Base Classes dialog, where you can set any or all of the relevant base classes to use a custom class. The figure shows the Entity Object Wizard, but the process is similar for View Objects and Application Modules as well.

Figure 60: Setting ADF Framework Base Classes from the Java Panel

ADF Framework Customizations for the Model Layer Data Access Components

In the toystore.fwk.model.dataaccess package, we've implemented the following customizations:

• PropertyFileBasedLookupViewObjectImpl

  Customizes the base ViewObjectImpl class to support "fetching" data from a Java *.properties file instead of fetching from a JDBC RowSet resulting from a SQL query. The CountryList, CreditCardList, ExpirationYearList, and ShippingOptionsList view objects specify this customized class as their base view object class.

• ViewDefHelper

  Exposes utility methods to simplify creating dynamic view object definitions of many kinds, including ones with updatable, entity-mapped attributes.

ADF Framework Customizations for the Business Service Layer

In the toystore.fwk.model.service package, the ToyStoreDBTransactionImpl class provides a customized implementation of the framework's oracle.jbo.server.DBTransaction interface. To rewrite the minimum amount of code possible, it extends the oracle.jbo.server.DBTransactionImpl2 class, and overrides the postChanges() method. This customized postChanges() implementation catches any DMLConstraintException resulting from the posting attempt and throws a JboException with an application-specific error message that it looks up from the from the toystore.fwk.util.ErrorMessages message bundle based on the database constraint name that has been violated.

The companion ToyStoreDBTransactionFactory class in this package extends the framework oracle.jbo.server.DatabaseTransactionFactory class to return an instance of the customized ToyStoreDBTransactionImpl.

The ADF configuration property named TransactionFactory needs to be set to the fully-qualified class name of the custom DBTransactionFactory class, so we would have:

TransactionFactory=toystore.fwk.model.service.ToyStoreDBTransactionFactory

to use our custom DB transaction implementation.

The ToyStoreApplicationModuleImpl class extends the base ADF Application Module implementation class to add a helper method named getConfigurationProperty(). This illustrates how to retrieve a property from the ADF configuration at runtime, and if not present, falls back to check the value of the same property as a Java System property. This is used by two places in the demo implementation code (ShoppingCartImpl in toystore.model.dataaccess package and LockAllInventoryItemsHelper in the toystore.model.services package) to pick between alternate implementations based on a value of a configuration property.

Custom Validation Rule as Framework Extension

The toystore.fwk.rules package contains the VerifyStateForCountry class which implements the JbiValidator interface in the oracle.jbo.server.rules package to provide a custom, parameter-driven business rule. Both the Account entity and the Orders entity make declarative use of this reusable validation rule. For example, in the Account.xml file, you'll find the following snippet of XML that records the usage of this custom business rule and captures the parameter values that the generic rule uses to perform its validation. In this case we see that the countryAttributeName parameter is set to "Country" and the stateAttributeName is set to "State".

<ValidationBean
  OperandType="LITERAL"
  Name="VerifyStateForCountryRule"
  BeanClass="toystore.fwk.rules.VerifyStateForCountryRule" >
  <NamedData
    NDName="countryAttributeName"
    NDType="java.lang.String"
    NDValue="Country" >
  </NamedData>
  <NamedData
    NDName="stateAttributeName"
    NDType="java.lang.String"
    NDValue="State" >
  </NamedData>
</ValidationBean>

In the Orders.xml file, you'll find a similar block of XML tags that set the parameter values differently so that the countryAttributeName parameter is set to "Shipcountry" and the stateAttributeName is set to "Shipstate".

The code for the VerifyStateForCountry shows several interesting techniques in use:

1. On-demand creation of the toystore.fwk.rules.dataaccess.StatesForCountry view object
2. Reuse of the same view object instance for subsequent executions of the rule
3. Setting of Max Fetch Size to 1 to improve performance of view objects that are known to fetch a single row.

Also, this example illustrates that it's easy to build reusable rules that make use of ADF components like view objects that can be packaged together with the rule into a reusable library like the FwkExtensions.jar that is created by the deployment profile in the FwkExtensions project.

Custom XSQL Action Handler for ADF

The toystore.fwk.xsql.ADFViewObject class implements a custom Oracle XSQL Pages action handler for involving ADF View Objects from the current binding container in XML/XSLT based view-layer rendering.

Additional Points of Interest Around the Demo

In this section we'll study a few miscellaneous points of interest that we haven't already covered about the demo's implementation.

Understanding a View Object's Fetch-Related Tuning Parameters

ADF View Object components have two similarly-named properties, FetchSize and MaxFetchSize, that are important to understand for performance tuning reasons. Both of them can be set either declaratively in the View Object Editor on the Tuning panel, or programmatically at runtime with appropriate API's.

The FetchSize property determines the number of rows at a time that will be fetched from the database. For example, if a view object's query identifies 200 rows in the result, and its FetchSize is set to 50, it will make (200/50=4) four round-trips to the database to retrieve those 200 rows, fifty at a time. If the FetchSize were set to 1, which is the default, then the view object would make two hundred (200/1=200) round-trips to the database to fetch all those rows.

Sizing the FetchSize too large can mean using more memory than necessary while fetching if you don't fill up the JDBC row buffers each time. So unless there were very few attributes in each row, we probably wouldn't want to set the value to 200 or larger. Setting the FetchSize too small, or leaving at the default of 1 when that is not appropriate, can mean many additional round-trips to the database server to retrieve your view object's query results.

It is easy to understand why you should consider the value of FetchSize on each of your view objects to see whether the default of 1 is appropriate. If it's not, set a reasonable value as a function of the amount of rows you expect to retrieve in that view object's query. The ADF Toy Store Demo view objects that fetch mutiple rows like FindProducts, ItemsForSale, LineItems, ProductList, ProductsInCategory, and ReviewLineItems in the toystore.model.dataaccess package all set FetchSize to 10.

Next let's look at the other parameter MaxFetchSize whose name often makes customers confuse it with the FetchSize we've just discussed.

The MaxFetchSize property sets an upper limit on the number of rows that view object will attempt to fetch from the database. The default value is -1 which means to not apply any maximum limit. In other words, the default is to fetch all rows from the query. If the above view object's query identifies those same 200 rows in the result, but its MaxFetchSize is set to 10, then it will only fetch the first 10 of those 200 rows.

The value of the FetchSize property is orthogonal to the FetchSize property we saw in the last section. With MaxFetchSize set to 10 and FetchSize set to 50, for example, when querying our 200-row example view object its first 10 rows will be returned in a single round-trip to the database, because the FetchSize of 50 is enough to accommodate those first 10 rows. As you would expect, with a FetchSize of 1 (the default!), the first 10 rows would be fetched in ten round-trips to the database.

In practice, a non-default MaxFetchSize value is used to indicate:

1. That you expect a single row result (MaxFetchSize=1)
2. That you will use the view object for insert-only (MaxFetchSize=0).

Indicating a MaxFetchSize of 1 when you expect a maximum of one row is good for performance because it prevents the view object from fetching again to see if there are more rows in the result or not. Indicating a MaxFetchSize of zero causes the view object to never execute its query. Rows can be created in that view object or found via the combination of findByKey() plus setCurrentRow(), but they won't be fetched via its normal query.

The ADF Toy Store demo uses the Accounts view object in both of these ways. The code in that view object's findAccountByUsernamePassword() method sets the MaxFetchSize to 1 before performing the query using the username and password to lookup, then it sets it back to zero.

// From: toystore.model.dataaccess.AccountsImpl
  public boolean findAccountByUsernamePassword(String username, String password) {
    /*
     * We're expecting either zero or 1 row here, so indicate that
     * by setting the max fetch size to 1.
     */
    setMaxFetchSize(1);
    setWhereClause("username = :0 and password = :1");
    setWhereClauseParam(0, username);
    setWhereClauseParam(1, password);
    executeQuery();
    boolean found = first() != null;
    setWhereClause(null);
    setWhereClauseParams(null);
    setMaxFetchSize(0);
    return found;
  }

The prepareToEditAccountInfoFor() and prepareToCreateNewAccount() methods set the MaxFetchSize to zero before finding a row by key and before creating a new row, respectively.

For good measure, I also wrote code in the application module's prepareSession() method to force the MaxFetchSize to zero on Accounts.

// From toystore.model.services.ToyStoreServiceImpl
  protected void prepareSession(Session session) {
    super.prepareSession(session);
    getAccounts().setMaxFetchSize(0);
  }

This make sure that even if we don't get the same application module instance from the pool on each request, that this setting stays at the default value what we want. The prepareSession() method is always called by the framework each time an ApplicationModule instance is used from the pool, so it's a good place to write code like this.

Some Observations about Batch Mode

With the ADF Bindings and Data Controls layers in place, you are free to choose your front-end client technology and your back-end business service implementation. However, it's important to note that this consistency does not imply that all data controls are reduced to have the lowest common denominator functionality of a generic Java Bean. You work with the iterators, bindings, and data controls in a consistent way, but you also can take advantage of unique features offered by a particular data control provider.

For example, Figure 61 illustrates that the data control based on an ADF application module has a Sync property that the other data controls do not. In JDeveloper 10g, the value of Sync defaults to "Batch" to use the new batch mode.

Figure 61: Sync Parameter of a Data Control Based on an Application Module

Batch mode is a network roundtrip-reduction feature that, as its name implies, batches up data-related operations in order to perform them in more coarse-grained chunks.

The client layer works with its application module and its view objects to perform all of the setup operations on any view objects whose data is required by the current task at hand. These operations are batched up using a batch-mode specific client-side ApplicationModule implementation class, which ADF provides for you behind the normal ApplicationModule interface when Sync Mode is set to "Batch". Your application then makes a call to the following method to perform all of the data operations and retrieve all the data that you are expecting from the datasources in a single network round-trip:

yourBindingContainer.refreshControl();

When you call refreshControl() on the binding container, it calls the sync() method on the data provider of each data control in use in that binding container. In the case of an ADF Application Module running in batch mode, this data provider is the client-side ApplicationModule object. This sync() operation causes the pending operations and any data changes in the client-side cache to be sent to the server-side business service and get executed. Any data changes that are made by components in the middle-tier layer, either by executing queries or programmatically modifying data, will be returned back to the client cache in the same round-trip.

Eventually this batched mode of operation will allow improved application scalability even when client layer and business service are co-located in the same J2EE web container, by minimizing the span of time that any given client makes use of an application module from the pool. In our 9.0.5 release of JDeveloper 10g, when running co-located like this, our Immediate mode still gives better performance.

We are focusing a lot of research and development effort on making batch mode the most scalable choice in the future, although we are arriving at that goal in phases. Even if you plan to deploy your client layer (like your Struts action classes) and business service layer in the same web container -- where reducing network traffic would not be a worry you have in mind -- batch mode can still provide value for you during development time.

When you run and test your ADF application in batch mode, it will guarantee to insure the best practices approach of working exclusively with component interfaces on the client layer, and never with the underlying implementation classes. The guarantees comes from the ClassCastException errors that you will get in batch mode at runtime if your client-layer coding has gotten "sloppy" and contains downcasts to the business service tier's *Impl classes (ViewObjectImpl, ViewRowImpl, EntityImpl, or ApplicationModuleImpl, or your own classes extending these).

Maintaining this best practice approach insures that you can easily redeploy your application into physically separate client and server layer at some later point in time if you decide you need to. It also insures that you can take advantage of clever new optimizations that the ADF team implements in batch mode for scalability in future releases. This explains why we've made Batch mode the default for JDeveloper 10g.

For now, my recommendation for web-based applications like the ADF Toy Store demo is to use Sync Mode of Batch during development and testing, but to switch to use a Sync Mode of Immediate before performing final tests and deploying in production. Changing the Sync Mode is as simple as flipping the property as shown in Figure 61.

A final important point about Batch mode is that the Oracle ADF DataAction tries to shield you from having to remember to sync your batch-mode application module. In fact, the default DataAction page handling lifecycle automatically performs the refreshControl() operation on your binding container at the appropriate times. Once during the prepareModel() phase at the beginning of the lifecycle handling, and once at the end of the request in the refreshModel() phase. This is designed on the assumption that it will be the view layer to be the first to access the data returned by the business service, so the batch mode sync operation occurs at the end of the data action lifecycle, before the Struts RequestProcessor forwards control to the page. If you need to iterate the data retrieved from the business service within the Struts action itself, then you'll need to need to make an extra call to refreshModel() just before your action code that needs to iterate the data. Failure to do so might result in an error like:

JBO-25048: Operation YYYYY is invalid for a working set view object

or other InvalidOperException type errors.

In versions of ADF beyond 9.0.5 that are currently in development, we've already made the automatic batch mode sync behavior even smarter to further avoid network round trips in cases where we can detect that there is no need to go over the wire. We've also further improved the ability for the DataControl to checkout the application module instance from the pool for an even shorter amount of time.

Encapsulating Model Manipulation Code on the Server

In the ADF Toy Store Demo, I've taken a very service-centric approach to my application business logic. As a best practice, any logic which manipulates the business objects or governs setup of the queries over that business object data, I've put into service methods on my ToyStoreService component. This keeps my controller layer very thin and encapsulates the implementation details of my service to the maximum.

The other approach would have been to leave much of the data model manipulation code inside my struts actions. I opted against this approach because the service-centric approach made the application functionality much easier to test outside of a web environment. It also just made sense to me to try and accomplish as much as possible over inside the service implementation. Besides the obvious benefit of the additional encapsulation, this approach will maximize how my application benefits in the future to improvements in the ADF batch mode.

Example of Accessing Stored Procedures

While running stress tests on the ADF Toy Store, I discovered a potential issue when different web users completed their orders for similar items at almost exactly the same time. The problem would arise in the stress test if the random orders being placed by the load testing tool contained intersecting sets of item id's. In a very narrow window of time, it sometimes occurred that a second user would receive an error that one of the inventory items was locked by another user when the ToyStoreServiceImpl class' finalizeOrder() method was decrementing the inventory quantities for the items ordered.

I decided to implement a solution to pre-lock the inventory item rows for all items in the cart in an "all or nothing" fashion, so that in the rare case where users might be simultaneously ordering the same items, those orders would never clash. This approach seemed more scalable to me than simply marking the finalizeOrder() method as synchronized, because that would serialize all orders, even when they contained no overlapping items.

The helper method that is called by finalizeOrder() is a static method on the LockAllInventoryItemsHelper class that you'll find inside the ToyStoreServiceImpl.java file toystore.model.services package. It looks like this:

  static void lockAllInventoryItemsInCart(ToyStoreServiceImpl am) {
    if (useViewObjectObjectForLockingAllItems(am)) {
      ensureLockAllInventoryItemsViewObjectExists(am);
      lockAllItemsInArrayUsingViewObject(am);
    }
    else {
      lockAllItemsInArrayUsingStoredProcedure(am);
    }
  }

As you can see in the code, for educational purposes, I implemented the "lock all or nothing" solution in two different ways: firstly as a stored procedure, and secondly as Java code in the application module that uses a View Object to perform a similar task. In the useViewObjectObjectForLockingAllItems() method you can see that it uses the getConfigurationProperty() method that we added in our Application Module framework customization class to read the value of the toystore.lockinventoryitems configuration property. Valid values are ViewObject and StoredProcedure (the default). In both implementations, if an order that is about to be finalized cannot lock all of its items, it will wait for a second, and then try again.

The method that does the call to the stored procedure is the following one:

  private static void lockAllItemsInArrayUsingStoredProcedure(
                                     ToyStoreServiceImpl am) {
    PreparedStatement ps = null;
    try {
      ps = am.getDBTransaction().createPreparedStatement(STMT, 0);
      ps.setObject(1, arrayOfItemIdsInShoppingCart(am));
      ps.execute();
    }
    catch (SQLException s) {
      throw new AlreadyLockedException(s);
    }
    finally {
      if (ps != null) {
        try {
          ps.close();
        }
        catch (SQLException s) {}
      }
    }
  }

Using the DBTransaction object that it obtains from the application module, it creates a JDBC PreparedStatement to invoke the block of PL/SQL code stored in the private constant variable named STMT, and looks like this:

begin lock_all_items_ordered(?); end;

It makes sure to close the prepared statement at the end of the block. It's interesting to note that the bind variable that it passes is an array typed bind variable that is returned by the arrayOfItemIdsInShoppingCart() method. At runtime this array will contain the item id's of all the items being ordered in the shopping cart. The lock_all_items_ordered procedure uses this array-valued bind variable in a SQL statement to lock all or none of the rows in a single try:

-- from lock_all_items_ordered
  CURSOR c(cp_items table_of_itemid) IS
     SELECT null
       FROM inventory
      WHERE itemid IN (
        SELECT *
          FROM TABLE(CAST(cp_items AS table_of_itemid))
      ) FOR UPDATE NOWAIT;

The array parameter is defined to be of type table_of_itemid which is an Oracle object type defined as a TABLE OF VARCHAR(10). By combining the TABLE() operator and the CAST(ArrayExpr AS ArrayType) syntax, we can use a nested SELECT statement to select over the array element values as if it were a table.

Dynamically Constructing View Objects at Runtime

If you try setting the ToyStoreService's configuration property named toystore.lockinventoryitems to the value ViewObject, then our "lock all or nothing" helper method will use a View Object based implementation instead of the stored procedure based approach from above. You can explore the code of the lockAllItemsInArrayUsingViewObject() method to see how it's implemented.

I initially tried to create a design-time View Object to support the SELECT FOR UPDATE NOWAIT query that this implementation needs to perform, but I ran into trouble because the design time verification of the correctness of my query's SQL syntax kept failing due to the FOR UPDATE NOWAIT being in there. Not a problem! It gave me an excuse to come up with a helper class named ViewDefHelper in the toystore.fwk.model.dataaccess package (part of the FwkExtensions project) to simplify the runtime creation of view objects.

I wanted to avoid the usual runtime overhead associated with ApplicationModule's handy createViewObjectFromQueryStmt() method, which is forced to perform a round-trip to the database to "describe" the select list of the query to calculate the appropriate datatype of each view object attribute. So I dynamically create the ViewDefImpl view object definition class in ensureLockAllInventoryItemsViewObjectExists() and then proceed to use some ViewDefImpl member methods along with some helper methods in ViewDefHelper to create the view object along with the metadata about its attributes to avoid the runtime describe of the query. Notice that since the View definition object is shared by all components running in the VM, we use the synchronized modifier to make sure to avoid multi-threading issues for the definition creation and registering. Once the view object definition is resolved and registered, then I can use it to create a view object instance based on this definition. Since we always first try to find the view object instance before going about creating it, the same view object instance will get used over and over (with only the array-valued bind variable changing on each execution).

Changing Default Field Renders By Attribute Type

While developing the generic formControl.jsp page that we studied in the Rendering Data Entry Forms in a More Generic Way Using Metadata section above, I ran into a small problem. While entering her personal details in the Register New User form, if the user made a mistake like entering an invalid email address, the invalid value was getting reverted to the (blank) value it had before instead of continuing to show the value the user needed to correct. I diagnosed the issue to be a bug with the default HTMLFieldRendererImpl class in the oracle.jdeveloper.html package, which was incorrectly reading the attribute value to display in the form field from the underlying model-layer row instead of reading it from the binding object. It's the binding object that will cache an invalid value until it is successfully able to be set on the underlying model object, so by reading the value from the binding object, we would get the behavior we were expecting.

Until the bug (#3703925) gets fixed in a future JDeveloper release, I worked around the problem by creating a custom renderer that reads the value from the binding object instead of from the row's attribute directly. My ControlBindingTextFieldRenderer looks like what you see in Example 26. It extends the default TextField renderer and overrides its getHTMLValue() method to change the way it works when we're dealing with a BindingContainer-based datasource. If we're working with a BindingContainer datasource, we access the control binding for the current field being rendered (which has already been setup for us by the superclass) and get the value of the binding by calling its getInputValue() method. If we're not using a BindingContainer, then we return the value of what the superclass would have done before.

To illustrate another technique for perform more global overrides of the field renderer classes that the <adf:inputrender> tag will use, I added the helper method you see in Example 27 to my RegisterAction, which is the DataForwardAction that supports the /register DataPage to register new users.

We are able to set the default edit renderer class for any given Java type by setting a HTTP request attribute whose name looks like:

JavaTypeWithDotsConvertedToUnderscores_EditRenderer

to the fully-qualified name of the edit renderer class it should use. Accordingly, in order to have my ControlBindingTextFieldRenderer used as the default edit renderer class for edit controls rendered for attributes of type java.lang.String and toystore.model.datatypes.common.Email, I've set the values of the request attributes named:

• java_lang_String_EditRenderer
• toystore_model_datatypes_common_Email_EditRenderer

to the fully-qualified name of the ControlBindingTextFieldRenderer class.

I call the helper method from an overridden handleLifecycle() method in my RegisterAction like this:

// From: toystore.controller.strutsactions.RegisterAction
  protected void handleLifecycle(DataActionContext ctx) throws Exception {
    setupDefaultFieldRenderers(ctx);
    super.handleLifecycle(ctx);
  }


The handleLifecycle() method is the root method to override in a DataAction to perform custom code before the built-in lifecycle starts or after it is done. Once Bug 3703925 gets fixed, we'll be able to comment out this workaround code. Luckily, since we were using a framework, working around problems we encounter is typically just as easy as adding new behavior. Both involve customizing a framework base class and getting our customized class to be used instead of the default framework implementation class.

Different Approaches for Looking Up Item Info for the Shopping Cart

The fillInCartItemDetails() method in the ShoppingCartImpl view object class illustrates two different approaches for looking up information using the ADF business components. As in the example above, the implementation that is used at runtime depends on the value of a configuration property, in this case named toystore.shoppingcartlookup. When that parameter has the value "ViewObject", it uses an approach that makes use of an instance of the ShoppingCartItemLookup view object. On the other hand, if the parameter has the value "EntityObject", it uses an approach that finds the inventory item entity object instance by primary key from the entity cache.

Different Approaches for Decoding One-Character Flag Fields

Often your data will contain single-character "flag" fields. In the ADF Toy Store Demo, our example is the "In Stock" field that tells whether an item is in stock or not. In the ItemsForSale view object (in the toystore.model.dataaccess package) you can see that the expert-mode query uses a DECODE() statement to return a string like "Back Ordered" or "In Stock" depending on the value of the INVENTORY.QTY, the quantity currently in the inventory:

SELECT Item.ITEMID, 
       ITEM.ATTR1||' '||PRODUCT.NAME AS NAME, 
       Item.LISTPRICE, 
       Item.PRODUCTID, 
       DECODE(INVENTORY.QTY,NULL,'Back Ordered',
                               0,'Back Ordered',
                               'In Stock')
       AS IN_STOCK, 
       /* etc */

On the other hand, the ShoppingCart view object contains a simple transient field named InStock which takes on the values of either Y or N to indicate if the item is in stock or not. The value of this is determined by the fillInCartItemDetails() method we discussed in the previous section. When the yourcart.jsp page goes to display the InStock information, rather than showing the "raw" Y or N value, it uses the Y or N as part of the string key name for a translatable string in the Struts message resource file. If first uses the JSTL <c:set> tag to set a page-local variable named inStockMsgKey to the value of "cart.instock." concatenated to the value of the InStock field in the current Row of the <c:forEach> loop, and then it uses <bean:message> to display the translated string based on either the cart.instock.Y or cart.instock.N message key value in that inStockMsgKey object.

<td>
  <%--
   | NOTE: Here we are using the value of the "InStock"
   | ----  attribute, which will be "Y" or "N", as part
   |       of the key to lookup a translated value to 
   |       show to the user like "In Stock" or "Back Ordered"
   +--%>
  <c:set var="inStockMsgKey" value="cart.instock.${Row.InStock}"/>
  <bean:message name="inStockMsgKey"/>
</td>

Making Sure You See Errors that Occur

You might have noticed that every one of our web pages includes an <html:errors> tag at the top. By default, the ADF DataAction bundles up all exceptions that occur during the request processing lifecycle and translates them at the end of the request (during the lifecycle's reportErrors() phase) to the Struts layer as Struts ActionError objects. This means that business validation errors that occur during the processing of the lifecycle neatly appear on the page, wherever we've placed the <html:errors> tag(s). However, a failure to include any <html:errors> tag in your page will result in the errors being reported to the Struts layer, but never displayed. This means that even some kind of unexpected error will only show up if you explicitly render the Struts errors using the <html:errors> tag. The JDeveloper 10g Data Control Palette does its best to insure you always have an <html:errors> tag in your page, but in case you develop your pages in a more manual way, forewarned is forearmed.

Deployment and Packaging Considerations

JDeveloper 10g supports the ability to create deployment profiles that encapsulate all the details required to build and package Java archive files of various kinds for deployment. It also supports popular the build and packaging tool Apache Ant. The ADF Toy Store Demo contains examples of both approaches. This section describes the deployment profiles and Ant build file in use in the ADF Toy Store Demo.

JDeveloper Deployment Profiles In the Demo

As shown in Figure 62 the ADFToyStore.deploy is a WAR deployment profile in the ToyStoreViewController project that depends on all of the other deployment profiles. It packages up the JAR files produced by these other deployment profiles to produce a WAR file and an EAR file for deploying to Oracle Application Server or an external application server that has been configured with the ADF Runtime Installer for running ADF-based applications.

Figure 62: Settings for the ADFToyStore.deploy Profile

The ToyStoreModel project contains an ADF business components deployment profile named ToyStoreModel.bcdeploy. This was created by selecting the ToyStoreService application module in the Application Navigator, and selecting Business Components Deployment... on the right-mouse menu. In the wizard that appears, we selected the Simple Archive Files option. By expanding the node in the navigator, you can see that an ADF deployment profile contains two "children" profiles which are defined to package up the compiled artifacts like...

• Custom component interfaces and message bundles from the *.common subpackages of the project into the ToyStoreModelCSCommon.jar file
• Component implementation files (*Impl.java) and XML definition files, into the ToyStoreModelCSMT.jar file

For a client like our Struts web controller layer that is using the ADF model layer in the same web container as a set of simple JavaBeans, both of these JAR files must be in the web container's classpath. For a remote client like a Swing application -- or a web-tier accessing a remotely-deployed application module deployed as an EJB Session Bean -- only the ToyStoreModelCSCommon.jar is required in the client tier. In that scenario, the ToyStoreModelCSMT.jar would be required in the class path of the EJB Session Bean supporting the "server tier" of that distributed application.

For completeness, Table 4 lists all the deployment profiles in use in the ADF Toy Store Demo.

Table 5 shows the additional supporting JAR files that are required for Struts/ADF support.

Deploying the Demo to an External Application Server

Using the ADFToyStore.deploy deployment profile, deploying and redeploying are a single step after having first performed two prerequisite steps:

1. Defined an application server connection for your target application server

   You can do this in the JDeveloper 10g Connections Navigator by right-clicking on the Application Server folder and picking New... from the right-mouse menu.

2. Run the ADF Runtime Installer to install the latest version of the Oracle ADF framework to your target server.

In this section we'll cover the basics to deploy the application using the IDE.

If you are deploying an Oracle ADF application built with JDeveloper 10g to any of the following servers, we offer automatic framework installation facilities:

• Standalone OC4J Container
• Oracle Application Server
• Apache Tomcat
• JBoss
• BEA WebLogic

To install the latest version of the ADF framework on one of the above servers, select Tools > ADF Runtime Installer > YourServerType. The wizard will ask for a root directory for your server installation. Once you provide that, the wizard will configure your server for you after you confirm the operation in the next step. It will produce a log file detailing all of the file copy and configuration file modification that it performed (as well as reporting any errors it ran into), and open that log file for your viewing in an IDE editor window.

To deploy the ADF Toy Store demo to one of these predefined application server connections, just click on the ADFToyStore.deploy node in the Application Navigator and select Deploy To > YourServerName in the right-mouse menu.

If are running a server other than Oracle Application Server or Standalone OC4J, you will need to manually setup the JDBC datasource mappings for the two datasources used in the Toy Store demo. Appendix 2 provides an example of those instructions for Tomcat servers.

Building, Packaging, and Deploying the Demo Using Ant

Apache Ant is the defacto standard for Java build tools. The ADF Toy Store Demo comes with a standard Ant build.xml file that you can use to build the demo from the command line. Table 6 gives an overview of all the Ant build targets and what they do.

Since JDeveloper 10g also supports Ant, you can build any of your Ant build targets right from the Application Navigator by clicking on the right-mouse menu as shown in Figure 63.

Figure 63: Deployment Profiles and Ant Build File in ADFToyStore Workspace

To build the ADFToyStore.ear file, pick the ear build target.

To build the ADFToyStore.war file from, pick the war build target.

When you use Ant from within the JDeveloper 10g IDE, it sets the jdev.home parameter to the home directory where the currently-running version of JDeveloper 10g lives. This way you can make references in your Ant build script to files that are relative to the JDeveloper installation home. The ADF Toy Store's build.xml file does this. This means, that when trying to build from the command line, if you try to perform a build target like this:

$ ant ear

You'll get a handy warning message telling you:

Buildfile: build.xml
init:
BUILD FAILED
C:\adftoystore\build.xml:52: Need to set jdev.home property to JDev home dir!

So, you'll need to run the build like this instead, passing a value for jdev.home on the ant command line:

$ ant -Djdev.home=C:\jdev\9051 ear

Where here C:\jdev\9051 is an example of what you would replace with your JDeveloper installation home directory.

I have setup Ant build targets named deploy and deployclean which will try to perform the OC4J Standalone installation from within the build script. You'll likely need to edit the properties in the section of build.xml that looks like this to get the deploy-from-Ant to work, supplying values that make sense for your environment.

  <!--
   | Local OC4J Connection Information for "deploy" task
   +-->
  <property name="oc4j.admin.jar" value="${jdev.home}/j2ee/home/admin.jar"/>
  <property name="oc4j.url" value="ormi://localhost/"/>
  <property name="oc4j.username" value="admin"/>
  <property name="oc4j.password" value="welcome"/>
  <property name="j2ee.app.name" value="ADFToyStore"/>
  <property name="j2ee.web.app.name" value="ADFToyStore"/>
  <property name="oc4j.web.site" value="http-web-site"/>
  <property name="j2ee.context.root" value="ADFToyStore"/>

These parameters are used in the deploy task as arguments passed to the command-line deployment utility admin.jar.

Getting Started on Your Own ADF-Based Applications

In the sections above, we analyzed all the key aspects of the ADF Toy Store demo to better understand each of the "moving parts" in a typical J2EE web application adhering to an MVC architecture. Along the way, we noted numerous Oracle JDeveloper 10g features for simplifying the development of our model, view, and controller-layer components. For example, JDeveloper 10g offers integrated support for:

• Modeling for your Java classes and ADF components using UML, with three-way code, metadata, and model synchronization
• Designing your database schema and constraints visually
• Binding your user interfaces to any kind of business service data, using a consistent, visual, and declarative approach
• Building business services productively by extending the ADF Business Components.
• Creating, configuring, and evolving every aspect of your ADF-based J2EE application through synchronized design time tools like the Application Navigator, Structure Window, Property Inspector, Wizards, and Object Editors.
• Visually Editing your Struts Page Flow editing to more easily understand and modify all of the configuration info in your struts-config.xml
• Visually Designing your JSP pages for more easily designing web pages.
• Running and debugging your application on the embedded Oracle Containers for J2EE (OC4J) container without going through a lengthy deploy step each time.

To get more step-by-step assistance in beginning to use Oracle ADF in your own J2EE application development, the Oracle by Example Series for Oracle JDeveloper 10g (on our OTN web site) offers a number of step-by-step tutorials that can help you get started. After going through the Installing the Sample Schemas and Establishing a Database Connection setup steps, try any or all of the following for some additional "practice" with using ADF, JSP, and Struts:

• Developing an End-to-End Web Application Using the Default Technology Scope
• Creating a Business Model using ADF Business Components
• Creating Custom Business Methods Using ADF Business Components
• Adding a BI Beans Graph in an ADF Business Components/JSP/Struts Application

While this article focused exclusively on using ADF with JSP pages, the following two tutorials illustrates two alternative client technologies that we support for your application building:

• Creating an ADF JClient Application for Business Components
• Developing Applications with Oracle ADF UIX

Conclusion

By exploring the details of this ADF Toy Store application built using Oracle ADF and Apache Struts, we've seen how you can build J2EE-compliant applications by standing on the shoulders of existing J2EE frameworks. By doing so, we've seen that the code that we actually write in minimal, and when required, is code that is focused directly on the business application problem at hand. Both Struts and ADF make extensive use of XML-based configuration information for their framework components, which further simplifies development by driving a lot of framework behavior from this metadata instead of using a heavy code generation approach.

Along the way, we've seen that the Oracle JDeveloper 10g IDE provides excellent support for putting both Struts and ADF to use in real application scenarios. Over 800 developers inside Oracle Corporation alone are using the ADF framework daily for building their self-service web applications as part of the Oracle e-Business Suite. Hundreds of external customers are using JDeveloper and ADF as well, as evidenced by the enthusiastic questions we get every day on the Oracle Technology Network JDeveloper Discussion Forum -- a forum you can use as well as you experiment with Struts and ADF.

Hopefully after discovering the power of framework-based J2EE development, you will be asking yourself, "How could I think of building my next application any other way?"

Appendix 1: Known Issues

• If you are you are using DHCP to get an automatically-assigned IP address, sometimes when JDeveloper launches your default browser and starts the embedded OC4J, you may see an HTTP error saying the ADF Toy Store home page does not exist. Typically this can be corrected by experimenting with different options for how the Embedded OC4J server refers to the local machine in the URL. One option that always works for me is setting the preference to use localhost as your machine name in the URL as shown in Figure 64. The access this preference page, select Tools | Embedded OC4J Server Preferences... menu option, and visit the Startup category.

  
  Figure 64: Changing the IDE Preference for Embedded OC4J Host Name

• If you see an error like the following (additional details removed for clarity) when you click on one of the categories to browse items in the store, this means that the database connection has failed. Check that the database is running and if it is, double-check the connection details that you have provided in the two JDeveloper named connections (toystore and toystore_statemgmt) in the setup instructions.

  500 Internal Server Error

JBO-30003: The application pool (ToyStoreServiceLocal) failed to checkout an application
            module due to the following exception:

oracle.jbo.DMLException: JBO-26061: Error while opening JDBC connection.
  :
## Detail 0 ##
  :
java.sql.SQLException: Io exception: Connection refused

(DESCRIPTION=(TMP=)(VSNNUM=153092608)(ERR=12500)
  (ERROR_STACK=
    (ERROR=(CODE=12500)(EMFI=4))
    (ERROR=(CODE=12560)(EMFI=4))
    (ERROR=(CODE=530)(EMFI=4))
    (ERROR=(BUF='32-bit Windows Error: 2: No such file or directory'))
  )
)


Appendix 2: Configuring Toy Store Datasources on Apache Tomcat

In JDeveloper 10g, there is automatic support for installing the ADF Runtime on Tomcat, as well defining Application Server connections to a Tomcat server and deploying to it. In fact, with a Tomcat-type application server connection defined, you can just click right-mouse Deploy to option on the ADFToyStore.deploy deployment profile in the ToyStoreViewController project to carry out the deployment.

However, there is one manual step left to do regarding the configuration of the two JDBC datasources that the Toy Store application uses: jdbc/toystoreDS and jdbc/toystore_statemgmtDS. Similar manual steps would be required to map the logical JDBC datasource names to physical JDBC connections with any other J2EE server as well. For the record, this section explains how to configure these two datasources on the Tomcat server.

The ADF Toy Store demo's web.xml file contains the following two resource reference definitions for JDBC datasources named jdbc/toystoreDS and jdbc/toystore_statemgmtDS:

  <resource-ref>
    <res-ref-name>jdbc/toystoreDS</res-ref-name>
    <res-type>javax.sql.DataSource</res-type>
    <res-auth>Container</res-auth>
  </resource-ref>
  <resource-ref>
    <res-ref-name>jdbc/toystore_statemgmtDS</res-ref-name>
    <res-type>javax.sql.DataSource</res-type>
    <res-auth>Container</res-auth>
  </resource-ref>


This is the J2EE container-independent resource definition which is always paired with container-specific setup information for these resources. To provide the Tomcat-specific datasource configuration details for the "/ADFToyStore" web application context, edit the TOMCAT_HOME/conf/server.xml file, and find the place inside the appropriate <Host> element, where you can paste the additional ADF Toy Store <Context> definition. It will look like this:

<Server>
  <Service>
    <Engine>
      <Host>
         :
        <Context>
          <!-- Existing other <Context> entries -->
        </Context>
         :
        <!--
         | Paste in Entire <Context> element show below right here
         +-->
       </Host>
     </Engine>
   </Service>
</Server>

Paste in the following ADF Toy Store <Context> entry at the indicated location. You will need to edit the value of the JDBC url parameter to have the appropriate @machine:port:sid string to match the database where you have created the TOYSTORE and TOYSTORE_STATEMGMT users and installed the ADF Toy Store tables.

        <!--
         | BEGIN ADF Toy Store Tomcat DataSource Config Section
         | Change JDBC URL info to match your database!!
         +-->
        <Context path="ADFToyStore" docBase="ADFToyStore">

          <Logger className="org.apache.catalina.logger.FileLogger"
                  prefix="localhost_ADFToyStore." 
                  suffix=".log"
                  timestamp="true"/>
            <Resource name="jdbc/toystoreDS"
                      auth="Container"
                      type="oracle.jdbc.pool.OracleConnectionPoolDataSource"/> 
            <Resource name="jdbc/toystore_statemgmtDS"
                      auth="Container"
                      type="oracle.jdbc.pool.OracleConnectionPoolDataSource"/> 
            <ResourceParams name="jdbc/toystoreDS">
              <parameter>
                <name>factory</name>
                <value>oracle.jdbc.pool.OracleDataSourceFactory</value>
              </parameter>
              <parameter>
                <name>url</name>
                <value>jdbc:oracle:thin:toystore/toystore@127.0.0.1:1521:ORCL</value>
              </parameter>
            </ResourceParams>
            <ResourceParams name="jdbc/toystore_statemgmtDS">
              <parameter>
                <name>factory</name>
                <value>oracle.jdbc.pool.OracleDataSourceFactory</value>
              </parameter>
              <parameter>
                <name>url</name>
                <value>jdbc:oracle:thin:toystore_statemgmt/toystore@127.0.0.1:1521:ORCL</value>
              </parameter>
            </ResourceParams>
       </Context>
        <!--
         | END ADF Toy Store Tomcat DataSource Config Section
         +-->

The ADF Toy Store Demo comes with a testing servlet in the ToyStoreViewController.jpr project named toystore.controller.TestJDBCDatasources. The web.xml file is pre-configured to map this servlet to the path "/testjdbcdatasources". So, by pointing your browser at the URL:

http://yourtomcatmachine:yourport/ADFToyStore/testjdbcdatasources

You will see the following returned in the browser if the two required datasources are properly configured:

Successfully looked-up 'java:comp/env/jdbc/toystoreDS' DataSource
Successfully issued a ROLLBACK command using connection from this datasource
Successfully looked-up 'java:comp/env/jdbc/toystore_statemgmtDS' DataSource
Successfully issued a ROLLBACK command using connection from this datasource

After you have performed these steps, you should be able to point your browser at:

http://yourtomcatmachine:yourport/ADFToyStore

to run the ADF Toy Store demo on Tomcat.

References

1. Expert One-on-One: J2EE Design and Development (Johnson) from Wrox
2. EJB Design Patterns: Advanced Patterns, Processes, and Idioms (Marinescu) from John Wiley & Sons
3. Core J2EE Patterns: Best Practices and Design Strategies (Alur, Crupi, Malks) from Prentice Hall