6 Tips for Managing Property Files with Spring

What could be simpler than property files? In an enterprise application, it turns out, many things! In this post I’ll take a look at a few subtle complexities to managing properties in an enterprise Java Spring application, and hopefully demonstrate how a little fore-thought in design can yield big savings in terms of time, confusion, bugs, and gray hair down the road.

Types of Properties

Before designing an approach to managing property files, it’s critical first to understand the types of properties are to be managed. Here are four characteristics to consider:

• Are some properties environment specific – in other words, do they have a different value in one environment than another? For example, a database connection string would differ in the TEST environment than it would the PROD environment.
• Do any properties contain sensitive information, like passwords or credentials, that can’t sit in plain text in the deployment unit (e.g. WAR, EAR)?
• Do any properties need to be located external from the deployment unit (e.g. on the file system rather than in the WAR, perhaps to be managed by a systems administrator)?
• Should properties be dynamic, in the sense that they can be updated at runtime (i.e. without requiring a restart of the application)?

Each of these special characteristics adds a degree of complexity to applications – necessitating additional infrastructure beyond the simple Java Properties class.

Spring’s PropertyPlaceholderConfigurer

Fortunately for us, the Spring framework, with the PropertyPlaceholderConfigurer and PropertyOverrideConfigurer, provides the features and hooks we need to manage these cases (with the exception “dynamic” properties). And true to its philosophy, Spring makes simple things simple, and complicated things possible.

The PropertyPlaceholderConfigurer allows properties to be pulled in to the application context file. For example, in the simplest case, a property “db.user” defined in database.properties could be pulled into the ${db.user} placeholder:

1. <beanid="propertyPlaceholderConfigurer"
2. class="org.springframework.beans.factory.config.PropertyPlaceholderConfigurer">
3. <propertyname="locations">
4. <list>
5. <value>classpath:database.properties</value>
6. </list>
7. </property>
8. </bean>
9. <beanid="dataSource"class="com.mchange.v2.c3p0.ComboPooledDataSource">
10. <propertyname="user"value="${db.user}"/>
11. <propertyname="password"value="${db.password}"/>
12. ...
13. </bean>

<bean id="propertyPlaceholderConfigurer"
class="org.springframework.beans.factory.config.PropertyPlaceholderConfigurer">
<property name="locations">
<list>
<value>classpath:database.properties</value>
</list>
</property>
</bean>
<bean id="dataSource" class="com.mchange.v2.c3p0.ComboPooledDataSource">
<property name="user" value="${db.user}"/>
<property name="password" value="${db.password}"/>
...
</bean>

The PropertyOverrideConfigurer works in the opposite way, pushing properties from property files into properties in the context file (without having to specifically define a place holder – e.g. ${db.user}). These are both well documented in the Spring API and documentation.

Tips

Given the different possible types of properties, and Spring’s property framework, here are a few tips for managing them:

1) Consider using a JVM System property that sets the environment of the machine (e.g. “my.env=TEST”), telling the configurer which property file to use. For example:

  <context:property-placeholder
location="classpath:db-${my.env}.properties"/>

If the “my.env” property was set to “TEST”, then obviously the PropertyPlacementConfigurer would look for a file called “db-TEST.properties”. For Tomcat, this property can be set in the admin console or defined in a startup script (e.g. “-Dmy.env=TEST”) – neither of which is very elegant. Alternatively, it is possible to use JNDI with Tomcat, defining “my.env” in the server.xml and the context.xml of the web app. (Note, there are of course many other ways to solve this environment-specific problem, but this is an easy and relatively straight-forward one.)

2) It may be necessary to set the ignoreUnresolvablePlaceholders to true for any PropertyPlaceholderConfigurer, which will ensure that a configurer won’t fail if it can’t find a property. Why would this be a good thing? Oftentimes, one context file will import other context files, and each may have their own configurer. If ignoreUnresolvablePlaceholders is set to false (the default), then one configure would fail if it couldn’t find the property, even if another configurer down-stream could find it (see good explanation here). Beware, however, since this will suppress warnings for legitimate missing properties, making for some tough-to-debug configuration problems.

3) To encrypt properties, subclass the PropertyPlacementConfigurer and override the convertPropertyValue() method. For example:

protected String convertPropertyValue(String strVal) {
//if strVal starts with “!!” then return EncryptUtil.decrypt(strVal)
// else return strVal
}

4) Consider using the systemPropertiesMode property of the configurer to override properties defined in property files with System properties. For one-off environment specific properties this can be a helpful solution, however, for defining many properties, this configuration can be cumbersone.

5) For properties that need to be managed outside of the WAR, consider using a System property to define where the file is located. For example, the property ${ext.prop.dir} could define some default directory on the file system where external property files are kept:

<context:property-placeholder
location="file:///${ext.prop.dir}db.properties"/>

This entails, however, that this property is set for any process that leverages the Spring container (e.g. add the param to the Run Configuration for integration/unit tests, etc.), otherwise the file would not be found. This can be a pain. To circumvent, the configurer can be overridden – changing the behavior such that it looks to the external directory only if the System property is set, otherwise it pulls from the classpath.

6) Beware of redundancy of environment-specific properties. For example, if the solution is to have one property file for each environment (e.g. “db-test.properties”, “db-dev.properties”, etc.), then maintaining these properties can be a bit of a nightmare – if a property “foo” is added, then it would have to be added to the property file for each environment (e.g. DEV, TEST, PROD, etc.). The PropertyOverrideConfigurer is appropriate to eliminate this redundancy, setting the default value in the application context itself, but then the overriding value in a separate file. It’s important, however, to document this well, since it can look a bit “magical” to an unsuspecting maintenance developer who sees one value specified in the context file, but another used at runtime.

Conclusion

Managing properties for an enterprise application is a little trickier than one might expect. With Spring’s property configurers, however, the toughest part is just knowing what you need – the rest comes out of the box, or with some minor extensions.

Hopefully a few of these tips will be useful for you. Please let me know some of your own!