Normally, you build a JAR file after your code source has been compiled, collecting the Java code (which has been segregated by package) into a single collection via either the jar command-line utility, or more commonly, the Ant jar task. The process is straightforward enough that I won't demonstrate it here, though later in the article we'll return to the topic of how JARs are constructed. For now, we just need to archive Hello, a stand-alone console utility that does the incredibly useful task of printing a message to the console, shown in Listing 1:
Listing 1. Archiving the console utility
package com.tedneward.jars;
public class Hello
{
public static void main(String[] args)
{
System.out.println("Howdy!");
}
}
The Hello utility isn't much, but it's a useful scaffold for exploring JAR files, starting with executing the code.
1. JARs are executable
Languages like .NET and C++ have historically had the advantage of being OS-friendly, in that simply referencing their name at the command-line (helloWorld.exe) or double-clicking their icon in the GUI shell would launch the application. In Java programming, however, a launcher application — java — bootstraps the JVM into the process, and we have to pass a command-line argument (com.tedneward.Hello) indicating the class whose main() method we want to launch.
These additional steps make it harder to create user-friendly applications in Java. Not only does the end user have to type all of these elements at the command-line, which many end users would rather avoid, but chances are good that he or she will somehow fat-finger it and get an obscure error back.
The solution is to make the JAR file "executable" so that the Java launcher will automatically know which class to launch when executing the JAR file. All we have to do is introduce an entry into the JAR file's manifest (MANIFEST.MF in the JAR's META-INF subdirectory), like so:
Listing 2. Show me the entrypoint!
Main-Class: com.tedneward.jars.Hello
The manifest is just a set of name-value pairs. Because the manifest can sometimes be touchy about carriage returns and whitespace, it's easiest to use Ant to generate it when building the JAR. In Listing 3, I've used the manifest element of the Ant jar task to specify the manifest:
Listing 3. Build me the entrypoint!
All a user has to do to execute the JAR file now is specify its filename on the command-line, via java -jar outapp.jar. In the case of some GUI shells, double-clicking the JAR file works just as well.
2. JARs can include dependency information
It seems that word of the Hello utility has spread, and so the need to vary the implementation has emerged. Dependency injection (DI) containers like Spring or Guice handle many of the details for us, but there's still a hitch: modifying code to include the use of a DI container can lead to results like what you see in Listing 4:
Listing 4. Hello, Spring world!
package com.tedneward.jars;
import org.springframework.context.*;
import org.springframework.context.support.*;
public class Hello
{
public static void main(String[] args)
{
ApplicationContext appContext =
new FileSystemXmlApplicationContext("./app.xml");
ISpeak speaker = (ISpeak) appContext.getBean("speaker");
System.out.println(speaker.sayHello());
}
}
Because the -jar option to the launcher overwrites whatever happens to be in the -classpath command-line option, Spring needs to be on the CLASSPATH and in the environment variable when you run this code. Fortunately, JARs permit a declaration of other JAR dependencies to appear in the manifest, which implicitly creates the CLASSPATH without you having to declare it, shown in Listing 5:
Listing 5. Hello, Spring CLASSPATH!
./lib/org.springframework.core-3.0.1.RELEASE-A.jar
./lib/org.springframework.asm-3.0.1.RELEASE-A.jar
./lib/org.springframework.beans-3.0.1.RELEASE-A.jar
./lib/org.springframework.expression-3.0.1.RELEASE-A.jar
./lib/commons-logging-1.0.4.jar" />
Notice that the Class-Path attribute contains a relative reference to the JAR files that the application depends on. You could also write this as an absolute reference or without a prefix entirely, in which case it would be assumed that the JAR files were in the same directory as the application JAR.
Unfortunately, the value attribute to the Ant Class-Path attribute has to appear in one line, because the JAR manifest can't cope with the idea of multiple Class-Path attributes. So, all of those dependencies have to appear on one line in the manifest file. Sure, it's ugly, but being able to java -jar outapp.jar is worth it!
3. JARs can be implicitly referenced
If you have several different command-line utilities (or other applications) that make use of the Spring framework, it might be easier to put the Spring JAR files into a common location that all of the utilities can reference. Doing so avoids having multiple copies of JARs popping up all over the filesystem. The Java runtime's common location for JARs, known as the "extension directory," is by default located in the lib/ext subdirectory, underneath the installed JRE location.
The JRE is a customizable location, but it is so rarely customized within a given Java environment that it's entirely safe to assume that lib/ext is a safe place to store JARs, and that they will be implicitly available on the Java environment's CLASSPATH.
4. Java 6 allows classpath wildcards
In an effort to avoid huge CLASSPATH environment variables (which Java developers should have left behind years ago) and/or command-line -classpath parameters, Java 6 introduced the notion of the classpath wildcard. Rather than having to launch with each and every JAR file explicitly listed on an argument, the classpath wildcard lets you specify lib/*, and all of the JAR files listed in that directory (not recursively), in the classpath.
Unfortunately, the classpath wildcard doesn't hold for the previously discussed Class-Path attribute manifest entry. But it does make it easier to launch Java applications (including servers) for developer tasks such as code-gen tools or analysis tools.
5. JARs hold more than code
Spring, like so many parts of the Java ecosystem, depends on a configuration file that describes how the environment should be established. As written, Spring depends on an app.xml file that lives in the same directory as the JAR file — but it's entirely common for developers to forget to copy the configuration file alongside the JAR file.
Some configuration files are editable by a sysadmin, but a significant number of them (such as Hibernate mappings) are well outside of the sysadmin's domain, which leads to deployment bugs. A sensible solution would be to package the config file together with the code — and it's doable because a JAR is basically a ZIP in disguise. Just include config files in the Ant task or the jar command-line when building a JAR.
JARs can also include other types of files, not just configuration files. For instance, if my SpeakEnglish component wanted to access a properties file, I could set that up like Listing 6:
Listing 6. Respond at random
package com.tedneward.jars;
import java.util.*;
public class SpeakEnglish
implements ISpeak
{
Properties responses = new Properties();
Random random = new Random();
public String sayHello()
{
// Pick a response at random
int which = random.nextInt(5);
return responses.getProperty("response." + which);
}
}
Putting responses.properties into the JAR file means that there's one less file to worry about deploying alongside the JAR file. Doing that just requires including the responses.properties file during the JAR step.
Once you've stored your properties in a JAR, though, you might wonder how to get them back. If the data desired is co-located inside the same JAR file, as it is in the previous example, then don't bother trying to figure out the file location of the JAR and crack it open with a JarFile object. Instead, let the class's ClassLoader find it as a "resource" within the JAR file, using the ClassLoader getResourceAsStream() method shown in Listing 7:
Listing 7. ClassLoader locates a Resource
package com.tedneward.jars;
import java.util.*;
public class SpeakEnglish
implements ISpeak
{
Properties responses = new Properties();
// ...
public SpeakEnglish()
{
try
{
ClassLoader myCL = SpeakEnglish.class.getClassLoader();
responses.load(
myCL.getResourceAsStream(
"com/tedneward/jars/responses.properties"));
}
catch (Exception x)
{
x.printStackTrace();
}
}
// ...
}
You can follow this procedure for any kind of resource: configuration file, audio file, graphics file, you name it. Virtually any file type can be bundled into a JAR, obtained as an InputStream (via the ClassLoader) and used in whatever fashion suits your fancy.
In conclusion
This article has covered the top five things most Java developers don't know about JARs — at least based on history and anecdotal evidence. Note that all of these JARs-related tips are equally true for WARs. Some tips (the Class-Path and Main-Class attributes in particular) are less exciting in the case of WARs, because the servlet environment picks up the entire contents of a directory and has a predefined entry point. Still, taken collectively, these tips move us beyond the paradigm of "Okay, start by copying everything in this directory ..." In so doing, they also make deploying Java applications much simpler.
Next up in this series: 5 things you didn't know about performance monitoring for Java applications.
Author: Ted Neward
Ref Site: http://www.ibm.com/developerworks/java/library/j-5things6.html
No comments:
Post a Comment