3.1 The Java Interpreter

A Java interpreter is software that implements the Java virtual machine and runs Java applications. It can be a separate piece of software like the one that comes with the JDK, or part of a larger application like the Netscape Navigator Web browser. It's likely that the interpreter itself is written in a native, compiled language for your particular platform. Other tools, like Java compilers and development environments, can (and one could argue, should) be written in Java.

The Java interpreter performs all of the activities of the Java run-time system. It loads Java class files and interprets the compiled byte-code. It verifies compiled classes that are loaded from untrusted sources by applying the rules discussed in Chapter 1, Yet Another Language?. In an implementation that supports dynamic, or "just in time," compilation, the interpreter also serves as a specialized compiler that turns Java byte-code into native machine instructions.

Throughout the rest of this book, we'll be building both standalone Java programs and applets. Both are kinds of Java applications run by a Java interpreter. The difference is that a standalone Java application has all of its parts; it's a complete program that runs independently. An applet, as I described in Chapter 1, Yet Another Language?, is more like an embeddable program module; it relies on an applet viewer for support. Although Java applets are, of course, compiled Java code, the Java interpreter can't directly run them because they are used as part of a larger application. An applet-viewer application could be a Web browser like Sun's HotJava or Netscape Navigator, or a separate applet viewer application like the one that comes with Sun's Java Development Kit. All of Sun's tools, including HotJava, are written entirely in Java. Both HotJava and the applet viewer are standalone Java applications run directly by the Java interpreter; these programs implement the additional structure needed to run Java applets.

Sun's Java interpreter is called java. To start a standalone application with it, you specify an initial class to be loaded. You can also specify options to the interpreter, as well as any command-line arguments that are needed for the application:

% java [interpreter options] class name [program arguments] 

The class should be specified as a fully qualified class name including the class package, if any. Note, however, that you don't include the .class file extension. Here are a few examples:

% java animals.birds.BigBird
% java test

java searches for the class in the current class path, which is a list of locations where packages of classes are stored. I'll discuss the class path in detail in the next section, but for now you should know that you can set the class path with the -classpath option.

There are a few other interpreter options you may find useful. The -cs or -checksource option tells java to check the modification times on the specified class file and its corresponding source file. If the class file is out of date, it's automatically recompiled from the source. The -verify, -noverify, and -verifyremote options control the byte-code verification process. By default, java runs the byte-code verifier only on classes loaded from an untrusted source; this is the -verifyremote option. If you specify -verify, the byte-code verifier is run on all classes; -noverify means that the verifier is never run.

Once the class is loaded, java follows a very C-like convention and looks to see if the class contains a method called main(). If it finds an appropriate main()method, the interpreter starts the application by executing that method. From there, the application can start additional threads, reference other classes, and create its user interface or other structures, as shown in Figure 3.1.

In order to run, main() must have the right method signature. A method signature is a collection of information about the method, as in a C prototype or a forward function declaration in other languages. It includes the method's name, type, and visibility, as well as its arguments and return type. In this case, main() must be a public, static method that takes an array of String objects as its argument and does not return any value (void):

public static void main ( String [] myArgs ) 

Because main() is a public method, it can be accessed directly from any other class using the name of the class that contains it. We'll discuss the implications of visibility modifiers such as public in Chapter 5, Objects in Java.

The main() method's single argument, the array of String objects, holds the command-line arguments passed to java. As in C, the name that we give the parameter doesn't matter, only the type is important. Unlike C, the content of myArgs is a true array. There's no need for an argument count parameter, because myArgs knows how many arguments it contains and can happily provide that information:

int argc = myArgs.length; 

Java also differs from C in another respect here: myArgs[0]is the first command-line argument, not the name of the application. If you're accustomed to parsing C command-line arguments, you'll need to be careful not to trip over this difference.

The Java virtual machine continues to run until the main()method of its initial class file has returned, and until any threads that it started are complete. Special threads designated as "daemon" threads are silently killed when the rest of the application has completed.