resin tutorial basic Using Dependency-Injection with JDBC Databases

This tutorial describes the standard pattern for using a database in Resin, following the Dependency-Injection/Inversion-of-Control pattern.

Using a JDBC database is a three step process:

JDBC database access is based around the Factory pattern. With JDBC, javax.sql.DataSource is the Factory object. The <database> configures the DataSource and stores it in the JNDI resource map. The servlet will retrieve the DataSource and use it as a factory to obtain Connection objects, the main workhorse for using databases.

resin-web.xml configuration The JDBC query servlet. The JDBC initialization servlet.
CREATE TABLE jdbc_basic_brooms ( id INTEGER PRIMARY KEY auto_increment, name VARCHAR(128), cost INTEGER ); INSERT INTO jdbc_basic_brooms (name, cost) VALUES ('firebolt', 4000) INSERT INTO jdbc_basic_brooms (name, cost) VALUES ('nimbus 2001', 500) INSERT INTO jdbc_basic_brooms (name, cost) VALUES ('nimbus 2000', 300) INSERT INTO jdbc_basic_brooms (name, cost) VALUES ('cleansweep 7', 150) INSERT INTO jdbc_basic_brooms (name, cost) VALUES ('cleansweep 5', 100) INSERT INTO jdbc_basic_brooms (name, cost) VALUES ('shooting star', 50)

In Resin 3.0, the <database> tag configures the database pool and driver and saves the connection factory (DataSource) in JNDI. JNDI is just a global lookup tree available to all classes, making it straightforward to separate resource configuration from the application code.

The <driver> tag configures the database driver. The database vendor will make the driver classes available and describe the configuration variables. The thirdparty database page describes several important database configurations.

The <type> tag is the most important driver configuration item. It specifies the main Java driver class. For many drivers, you will have a choice of different drivers following different internal JDBC APIs. If you have a choice, you should try the drivers in the following order, after checking your database vendor's recommendations:

  1. JCA - Java Connection Architecture - this is a common driver interface for more than just JDBC. If possible, it's generally the best to choose.
  2. ConnectionPoolDataSource - JDBC driver which has extra hooks to help Resin pool the connections.
  3. Driver - old-style JDBC driver. Its main benefit is that it's generally always available as a fallback.
<web-app xmlns="http://caucho.com/ns/resin"> <database jndi-name="jdbc/basic"> <driver type="com.caucho.db.jca.ConnectionFactory"> <url>resin:WEB-INF/db</url> </driver> </database> </web-app>

The <url> specifies the location of the database. Each database driver will have a unique URL formal. In this case, the <url> specifies a directory for the database files. Other databases may specify a host and port.

The specific driver for this example, com.caucho.db.jca.ConnectionFactory is a simple database intended for examples and testing.

The servlet is configured with a DataSource to access JDBC. Resin allows two styles of configuration: Dependency Injection using bean-style setters and standard servlet <init-param> configuration. The Dependency Injection style is simpler, while the <init-param> style will work on all servlet engines. By creating a separate assemble() method, a servlet can take advantage of Resin's Dependency Injection and still be fully compatible with other servlet engines.

public class BasicServlet extends HttpServlet { private DataSource _ds; public void setDataSource(DataSource ds) { _ds = ds; } public void init() throws IOException, ServletException { if (_ds == null) { // the servlet could also create an assemble() method as // in the jdbc-basic tutorial throw new ServletException("data-source must be assigned"); } } ... }

Using dependency injection to configure servlets has some advantages over the init-param method:

  1. The servlet initialization code is simpler. The servlet doesn't need JNDI code.
  2. The configured values can be more complicated than the string-limitation of <init-param>.
  3. The DataSource itself isn't tied to JNDI, although JNDI will certainly remain the primary registry.

Enabling the Dependency Injection pattern is trivial: just add a setDataSource method as in the example above.

<servlet servlet-name="my-servlet" servlet-class="example.BasicServlet"> <init> <data-source>\${jndi("jdbc/basic")}</data-source> </init> </servlet>
... <servlet servlet-name="my-servlet" servlet-class="example.BasicServlet"> <init-param data-source="java:comp/env/jdbc/basic"/> </servlet> ...

The most important pattern when using JDBC is the following try/finally block. All database access should follow this pattern. Because connections are pooled, it's vital to close the connection no matter what kind of exceptions may be thrown So the conn.close() must be in a finally block.

Connection conn = _ds.getConnection(); try { ... } finally { conn.close(); }

The full example splits the database access into two methods to clarify the roles. The service retrieves the output writer from the servlet response and wraps any checked exceptions in a ServletException. Splitting the servlet method simplifies the doQuery method, so it can concentrate on the database access.

public void service(HttpServletRequest req, HttpServletResponse res) throws java.io.IOException, ServletException { PrintWriter out = res.getWriter(); try { doQuery(out); } catch (SQLException e) { throw new ServletException(e); } } private void doQuery(PrintWriter out) throws IOException, SQLException { Connection conn = _ds.getConnection(); try { String sql = "SELECT name, cost FROM jdbc_basic_brooms ORDER BY cost DESC"; Statement stmt = conn.createStatement(); ResultSet rs = stmt.executeQuery(sql); out.println("<table border='3'>"); while (rs.next()) { out.println("<tr><td>" + rs.getString(1)); out.println(" <td>" + rs.getString(2)); } out.println("</table>"); rs.close(); stmt.close(); } finally { conn.close(); } }