Java Basics

So here you are, the eager Java developer, about to embrace JDBC (Java Database Connectivity), the next item on your Java technology checklist. If you followed my last article (JDJ, Vol. 5, issue 9), you've selected a database system and a JDBC driver to help you master this technology. Now you want to jump in and start writing code. Perhaps you've bought a JDBC book, read your JDBC driver documentation or collected various JDBC articles from magazines (such as this one). Unfortunately, many of these resources skim the introductory topics or, worse, offer seemingly conflicting code examples. This article discusses the details of connecting a Java application to a database using JDBC, including how the process has changed with the evolution of the Java programming language.

In theory, the basics of connecting a running Java application to a database are quite simple. Fundamentally, an application developer merely has to code to the JDBC API (see http://java.sun.com/j2se/1.3/docs/api/java/sql/package-summary.html for specific information on the J2SE 1.3 JDBC API). This API has seven classes. Only one, the DriverManager class, is commonly used by beginners. The bulk of the API is dominated by interfaces that must be implemented by the JDBC driver vendors. As previously discussed, this allows for a wide range in performance and flexibility due to specific implementation details. As is often the case, however, the devil is in the details, and there are a lot of details whenever databases are involved.

Before delving into them, however, a mental picture (see Figure 1 for an actual picture) of the process involved in connecting to a database can be useful in understanding why things behave the way they do. First, while seemingly obvious, the fundamental point to start with is that all Java code runs inside a specific Java Virtual Machine (JVM). Meanwhile, the database system of interest has its own interfaces and protocols, which are generally controlled by a server or daemon process. In order for these two server processes to communicate, a bridge must be established and that's accomplished via the JDBC driver. In this article we'll use a fictitious JDBC driver, aptly from the Acme Corporation, whose fully resolved class name is com.acme.jdbc.AcmeDriver. While database servers are in general explicitly designed to allow interprocess communication, the JVM, for obvious security reasons, is not. Therefore, the driver of another Java application also runs inside the same JVM as the Java database application and must provide this extra functionality.

Finding a Driver
Before it can be used, the JVM must "know" about the JDBC driver, which consists of two separate steps: loading and instantiation. The Java object that manages all of the JDBC drivers for a JVM is the DriverManager class. This single class is responsible for shouldering the burden of managing the pool of JDBC drivers that are available within a given JVM. The loading step can be done in two different fashions: static or dynamic loading. The first technique, which is rarely discussed in the popular literature, is static loading during the JVM initialization. This is accomplished by specifying the fully resolved driver (or drivers if multiple JDBC drivers are required) class to the JVM via the jdbc.drivers property:

java –Djdbc.drivers=com.acme.jdbc.AcmeDriver Test.java

One of the benefits of this approach is that the code doesn't need to be tied explicitly to a particular JDBC driver, allowing for changes in the actual driver used to be made by the system administrator (who ostensibly would be starting the JVM as a server process) without recompiling (and redistributing class files). Multiple-driver classes can be loaded in this fashion by separating them with colons, which could be important if the Java applications running inside the same JVM need to communicate to different databases.

The alternative approach is to dynamically load the JDBC driver within the actual Java application, which is done using the Java Reflection mechanism:

Class.forName(com.acme.jdbc.AcmeDriver) ;

This approach allows an application to dynamically load the requested driver (which can be specified at runtime). Of course, this requires that the driver class is actually in the CLASSPATH of the running JVM. This last point is one of the leading stumbling blocks for JDBC novices, as they will receive a ClassNotFoundException if the JVM can't find the requested class. If multiple drivers are loaded, any drivers listed in the jdbc.drivers property are registered first, followed by any dynamically loaded drivers.

Once a JDBC driver has been loaded into the JVM, it must be instantiated. The recommended method of designing drivers is to force them to be automatically instantiated during the loading process: that means that an application developer is able to load and instantiate the JDBC driver in one line of code. This is accomplished via a static initializer that creates a new instance of the driver class and registers the new driver object with the JDBC DriverManager object. As a result, the single line of code above is translated into the following steps:

1. JVM locates the requested driver class, which must be in the current CLASSPATH.
2. The JVM loads the driver class into memory.
3. The JVM executes the static initializer section of the driver class, which will create a new instance of the driver class and register this new instance with the DriverManager class.

This whole process should seem rather straightforward, so what cause is there for any concern? As a popular Java colloquialism states, "Write once, debug everywhere." The problem is that your Java code will need to run in a multitude of JVM implementations, all of which can have subtle variations. For example, some Microsoft virtual machines include a performance trick that loads Java classes differently than many other virtual machines. Unfortunately, this trick prevents the static initializer section from being executed until the JVM feels the class is actually needed. As a result, the driver is neither instantiated nor registered with the DriverManager object during the reflection process. To circumvent this "feature," the developer is required to perform these steps manually:

Class.forName(com.acme.jdbc.AcmeDriver).newInstance() ;

This is not only a "Microsoft problem," as even the Sun JDK 1.1 reference implementation does not work properly due to a race condition (see the JDBC FAQ for more information) that prevented the static initializer section from being processed. The workaround for this bug is for the user to explicitly create and register the driver class:

java.sql.DriverManager.registerDriver(new com.acme.jdbc.AcmeDriver()) ;

Fortunately, with the maturation of both Java and the JDBC driver implementations, these subtle variations are becoming less common (particularly if JDBC is confined to the server). This implies that the following code snippet is all that is required to explicitly instantiate and register a JDBC driver: try {

Class.forName("com.acme.jdbc.AcmeDriver") ;
}catch(java.lang.ClassNotFoundException e) {
System.err.print("ClassNotFoundException: ") ;
System.err.println(e.getMessage()) ;
}

Making the Connection
Once the driver class has been initialized, a Java application can request a connection from the DriverManager class. The DriverManager object selects the appropriate JDBC driver from its pool of registered drivers based on the specific JDBC URL passed to the DriverManager getConnection method. The DriverManager object tests each registered driver, using the provided URLs in the order in which they were registered. The first one that recognizes (i.e., establishes a connection) the JDBC URL is used to provide the actual database connection. Interestingly enough, this process is actually layered on top of the original antiquated method for establishing a database connection, which is occasionally still prominently mentioned in the documentation of certain JDBC drivers.

Driver driver = new com.acme.JdbcDriver() ;
Connection con = driver.connect(url) ;

A JDBC URL follows the standard URL syntax (i.e., Web addresses), which includes the database name, the database server and optional additional parameters such as username and password. Instead of the more familiar protocols such as HTTP or FTP, the jdbc protocol is used. This is followed by a subprotocol which is designated by the individual driver vendors and registered with Sun to prevent confusing duplications. The last part of the URL is a subname that provides a mapping to the actual database of interest. The subname section of a URL can vary significantly from vendor to vendor; for example, the following are all valid JDBC URLs:

// An ODBC registered data source
String url = "jdbc:odbc:DataSourceName" ;

// An Oracle 8i database with thin client connection
String url = "jdbc:oracle:thin:@server.acme.com:1521:jdbc" ;
// A Microsoft SQL Server database
String url = "jdbc:JTurbo://server.acme.com:1433/jdbc" ;

// A mSQL database
String url = "jdbc:msql://localhost:1114/jdbc" ;

Once the URL is known, obtaining the connection object is straightforward:

try {
Connection con = DriverManager.getConnection(url);
// Utilize the connection to make a query
con.close();
}catch(SQLException e) {
System.err.println("SQLException: " + e.getMessage());
}

Of course, the getConnection method has two additional signatures that allow the developer to pass additional information, such as a username and password, to the database.

The Evolution to Data Sources
While it might seem rather odd at first glance, the JDBC 2.0 specification introduced a new technique for establishing connections to a database, the DataSource object. In fact, the DriverManager class may be deprecated in future versions of Java. The main reason for this abrupt change is the emergence of server-side Java applications within the Enterprise framework. While a DriverManager object is limited in the functionality it can provide (i.e., it essentially serves as a JDBC driver pool), a DataSource object can represent an arbitrary data source, work with JNDI (Java Naming and Directory Interface), provide connection pooling (simplifies tracking license and database cursor limitations) and provide support for distributed transactions (which is important for Enterprise JavaBeans). Each of these new features provides powerful new functionality that exceeds the scope of this article.

While the future of Java Database Connectivity clearly lies with DataSource objects, the traditional DriverManager connection techniques will not disappear. Hopefully this article has helped to illuminate some of the finer points involved in connecting your Java application to a database. Once connected, the rest is SQL.