You may have noticed that Jostraca can be a bit slow. This tutorial will help you to speed up the code generation process, by showing you how to configure and use Jostraca more effectively. The main reason for performance problems comes from the cross-language support built into Jostraca. While this allows you to write templates using many different programming languages, it also means that Jostraca has to rely on external programs to actually do the code generation.

Jostraca generates code by first creating a CodeWriter program, and then by executing the CodeWriter program. After the very first time that you generate from a template using Jostraca, the CodeWriter exists. So later, if the template has not changed, you don't need to recreate and recompile the CodeWriter program. Instead, you can just run it directly.

To do this, use the -C command line option to Jostraca. This option turns template compilation off. Now, Jostraca does not try to create the CodeWriter, instead it just executes the existing one. Avoiding the compilation phase can save a lot time, especially for compiled languages like Java and C, where compilation can be time consuming.

This option is most useful when you are merely changing your meta data and the template remains unchanged.

If you are using a compiled language such as Java or C for the template script (the code inside the <% and %> markers), then the CodeWriter program first has to be compiled before it can be used. Compilation is usually done by calling an external program. Launching external programs is always going to cause a delay.

Given this problem, consider using an interpreted language for your template script. Using Perl, Python, Ruby, REBOL, Jython or similar can greatly increase code generation speed. These languages also have robust text processing abilities that make them more effective as scripting language. They also have large libraries of support code to help you access XML resources (if you store your meta data using XML). Of course, sometimes they are not appropriate, but particularly for smaller generation tasks they can be very effective.

Jostraca versions 0.4 and later always attempt to compile Java templates internally by creating an instance of com.sun.tools.javac.Main. This class is provided with the standard Java JDK from Sun and is located in the rt.jar jar file. Include rt.jar in your CLASSPATH and Jostraca will use it automatically. Because a separate javac process is not created, compilation is much faster.

If you have to use a compiled language, use the fastest compiler you can find. For Java, this means you should use the jikes compiler from IBM. To do this, rename the file local.conf.sample in the conf folder of the Jostraca distribution, to local.conf and uncomment the lines:

#main.ExternalCompiler = jikes #main.process.Compiler = org.jostraca.process.ExternalJavaCompiler

by removing the# symbol. You'll need to make sure that the path to the jikes executable is in your PATH environment variable (or just put in the full path to jikes). For other languages, you can use the same technique, but you'll have to determine the fastest compiler yourself.