Wednesday, November 30, 2005

Best practices for testing J2EE database applications

Best practices for testing J2EE database applications

It wasn't too long ago that quality assurance (QA) teams played the leading (if not the only) role when it came to software testing. These days, however, developers are making an enormous contribution to software quality early in the development process, using automated unit-testing techniques. Most developers take for granted the need to use tools such as JUnit for comprehensive, automated testing, but there is little consensus on how to apply unit testing to the database operations associated with an application.

Let us have a look at the best practices that can help you get the most out of your database testing environment and can ensure that your applications are robust and resilient. The primary focus here is on J2EE applications interfacing with Oracle Database through JDBC, although the concepts also apply to applications written for other application environments, such as .NET.

Database Operations Tests:

Testing J2EE applications is often difficult and time-consuming at best, but testing the database operations portion of a J2EE application is especially challenging. Database operations tests must be able to catch logic errors—when a query returns the wrong data, for example, or when an update changes the database state incorrectly or in unexpected ways.

For example, say you have a USER class that represents a single USER table and that database operations on the USER table are encapsulated by a Data Access Object (DAO), UserDAO, as follows:

public interface UserDAO  {
  /**   * Returns the list of users      
              * with the given name
          * or at least the minimum age
  public List listUsers(String name, Integer minimumAge);
   * Returns all the users
   * in the database
  public List listAllUsers();

In this simple UserDAO interface, the listUsers() method should return all rows (from the USER table) that have the specified name or the specified value for minimum age. A test to determine whether you've correctly implemented this method in your own classes must take into consideration several questions:

  • Does the method call the correct SQL (for JDBC applications) or the correct query-filtering expression (for object role modeling [ORM]-based applications)?
  • Is the SQL- or query- filtering expression correctly written, and does it return the correct number of rows?
  • What happens if you supply invalid parameters? Does the method behave as expected? Does it handle all boundary conditions appropriately?
  • Does the method correctly populate the users list from the result set returned from the database?

Thus, even a simple DAO method has a host of possible outcomes and error conditions, each of which should be tested to ensure that an application works correctly. And in most cases, you'll want the tests to interact with the database and use real data—tests that operate purely at the individual class level or use mock objects to simulate database dependencies will not suffice. Database testing is equally important for read/write operations, particularly those that apply many changes to the database, as is often the case with PL/SQL stored procedures.

The bottom line: Only through a solid regime of database tests can you verify that these operations behave correctly.

The best practices in this article pertain specifically to designing tests that focus on these types of data access challenges. The tests must be able to raise nonobvious errors in data retrieval and modification that can occur in the data access abstraction layer. The article's focus is on database operations tests—tests that apply to the layer of the J2EE application responsible for persistent data access and manipulation. This layer is usually encapsulated in a DAO that hides the persistence mechanism from the rest of the application.

Best Practices

The following are some of the testing best practices:

Practice 1: Start with a "testable" application architecture.
Practice 2: Use precise assertions.
Practice 3: Externalize assertion data.
Practice 4: Write comprehensive tests.
Practice 5: Create a stable, meaningful test data set.
Practice 6: Create a dedicated test library.
Practice 7: Isolate tests effectively.
Practice 8: Partition your test suite.
Practice 9: Use an appropriate framework, such as DbUnit, to facilitate the process.

No comments: