4/28/2009

Use string constants in annotations

Generally speaking, string literals should be replaced with string constants in source code. This rule also holds when it comes to annotation-based resource injections.

For field or setter injections, any string constants can be declared in the enclosing class. The following example shows how to declare an env-entry in web.xml and inject it to a servlet class.

public static final String ADMIN_EMAIL = "adminEmail";

@Resource(name=ADMIN_EMAIL)
private String adminEmail;

<env-entry>
<description>admin email</description>
<env-entry-name>adminEmail</env-entry-name>
<env-entry-value>admin@example.x</env-entry-value>
</env-entry>
The name attribute is mandatory and used to match this injection to the appropriate env-entry in web.xml. This env-entry can be injected into several other classes, and therefore it is desirable to declare its name as a constant. This string constant may also reside in a separate class (e.g., a Constants class, utility class, or even some interface), depending on how it is used.

For type-level injections, these string constants must be declared in a separate class in order to be referenced in type-level annotations. This is probably because any field declarations have not be parsed when type-level annotations in the same class are processed. The follwoing example shows how to inject a stateful EJB reference at a servlet class level.
import static test.Constants.FOO_EJB_NAME;
import static test.Constants.FOO_EJB_REF_NAME;

@EJB(name=FOO_EJB_REF_NAME, beanName=FOO_EJB_NAME, beanInterface=FooBean.class)
public class Servlet2 extends HttpServlet {...}
In the above injection, beanName is optional only if there is only one EJB with business interface FooBean, and hence no ambiguity in mapping the EJB reference to target EJB. name and beanInterface attributes are always required when the @EJB injection is used at type level.