Source Organization

The description below uses the variable name $CATALINA_HOME to refer to the directory into which you have installed Tomcat 4, and is the base directory against which most relative paths are resolved. However, if you have configured Tomcat 4 for multiple instances by setting a CATALINA_BASE directory, you should use $CATALINA_BASE instead of $CATALINA_HOME for each of these references.

A key recommendation of this manual is to separate the directory hierarchy containing your source code (described in this section) from the directory hierarchy containing your deployable application (described in the preceding section). Maintaining this separation has the following advantages:

• The contents of the source directories can be more easily administered, moved, and backed up if the "executable" version of the application is not intermixed.
  
  
• Source code control is easier to manage on directories that contain only source files.
  
  
• The files that make up an installable distribution of your application are much easier to select when the deployment hierarchy is separate.

As we will see, the ant development tool makes the creation and processing of such directory hierarchies nearly painless.

The actual directory and file hierarchy used to contain the source code of an application can be pretty much anything you like. However, the following organization has proven to be quite generally applicable, and is expected by the example build.xml configuration file that is discussed below. All of these components exist under a top level project source directory for your application:

• docs/ - Documentation for your application, in whatever format your development team is using.
  
  
• src/ - Java source files that generate the servlets, beans, and other Java classes that are unique to your application. If your source code is organized in packages (highly recommended), the package hierarchy should be reflected as a directory structure underneath this directory.
  
  
• web/ - The static content of your web site (HTML pages, JSP pages, JavaScript files, CSS stylesheet files, and images) that will be accessible to application clients. This directory will be the document root of your web application, and any subdirectory structure found here will be reflected in the request URIs required to access those files.
  
  
• web/WEB-INF/ - The special configuration files required for your application, including the web application deployment descriptor (web.xml), tag library descriptors for custom tag libraries you have created, and other resource files you wish to include within your web application. Even though this directory appears to be a subdirectory of your document root, the Servlet Specification prohibits serving the contents of this directory (or any file it contains) directly to a client request. Therefore, this is a good place to store configuration information that is sensitive (such as database connection usernames and passwords), but is required for your application to operate successfully.

During the development process, two additional directories will be created on a temporary basis:

• build/ - When you execute a default build (ant), this directory will contain an exact image of the files in the web application archive for this application. For containers like Tomcat that allow you to deploy unpacked directory structures, you need only copy this directory structure into a subdirectory of $CATALINA_HOME/webapps in order to deploy and test it.
  
  
• dist/ - When you execute the ant dist target, this directory will be created. It will create an exact image of the binary distribution for your web application, including an license information, documentation, and README files that you have prepared.

Note that these two directories should NOT be archived in your source code control system, because they are deleted and recreated (from scratch) as needed during development. For that reason, you should not edit any source files in these directories if you want to maintain a permanent record of the changes, because the changes will be lost the next time that a build is performed.

External Dependencies

What do you do if your application requires JAR files (or other resources) from external projects or packages? A common example is that you need to include a JDBC driver in your web application, in order to operate. Different developers take different approaches to this problem. Some will encourage checking a copy of the JAR files you depend on into the source code control archives for every application that requires those JAR files. However, this can cause significant management issues when you use the same JAR in many applications - particular when faced with a need to upgrade to a different version of that JAR file. Therefore, this manual recommends that you NOT store a copy of the packages you depend on inside the source control archives of your applications. Instead, the external dependencies should be integrated as part of the process of deploying your application. In that way, you can always pick up the appropriate version of the JAR files from wherever your development system administrator has installed them, without having to worry about updating your application every time the version of the dependent JAR file is changed. In the example Ant build.xml file, we will demonstrate how to define build properties that let you configure the locations of the files to be copied, without having to modify build.xml when these files change. The build properties used by a particular developer can be customized on a per-application basis, or defaulted to "standard" build properties stored in the developer's home directory.

As mentioned earlier, it is highly recommended that you place all of the source files that comprise your application under the management of a source code control system like the Concurrent Version System (CVS). If you elect to do this, every directory and file in the source hierarchy should be registered and saved -- but none of the generated files. If you register binary format files (such as images or JAR libraries), be sure to indicate this to your source code control system.

We recommended (in the previous section) that you should not store the contents of the build/ and dist/ directories created by your development process in the source code control system. An easy way to tell CVS to ignore these directories is to create a file named .cvsignore (note the leading period) in your top-level source directory, with the following contents:

	build dist build.properties

The reason for mentioning build.properties here will be explained in the Processes section.

Detailed instructions for your source code control environment are beyond the scope of this manual. However, the following steps are followed when using a command-line CVS client:

• To refresh the state of your source code to that stored in the the source repository, go to your project source directory, and execute cvs update -dP.
  
  
• When you create a new subdirectory in the source code hierarchy, register it in CVS with a command like cvs add {subdirname}.
  
  
• When you first create a new source code file, navigate to the directory that contains it, and register the new file with a command like cvs add {filename}.
  
  
• If you no longer need a particular source code file, navigate to the containing directory and remove the file. Then, deregister it in CVS with a command like cvs remove {filename}.
  
  
• While you are creating, modifying, and deleting source files, changes are not yet reflected in the server repository. To save your changes in their current state, go to the project source directory and execute cvs commit. You will be asked to write a brief description of the changes you have just completed, which will be stored with the new version of any updated source file.

CVS, like other source code control systems, has many additional features (such as the ability to tag the files that made up a particular release, and support for multiple development branches that can later be merged). See the links and references in the Introduction for more information.

We will be using the ant tool to manage the compilation of our Java source code files, and creation of the deployment hierarchy. Ant operates under the control of a build file, normally called build.xml, that defines the processing steps required. This file is stored in the top-level directory of your source code hierarchy, and should be checked in to your source code control system.

Like a Makefile, the build.xml file provides several "targets" that support optional development activities (such as creating the associated Javadoc documentation, erasing the deployment home directory so you can build your project from scratch, or creating the web application archive file so you can distribute your application. A well-constructed build.xml file will contain internal documentation describing the targets that are designed for use by the developer, versus those targets used internally. To ask Ant to display the project documentation, change to the directory containing the build.xml flie and type:

	ant -projecthelp

To give you a head start, a basic build.xml file is provided that you can customize and install in the project source directory for your application. This file includes comments that describe the various targets that can be executed. Briefly, the following targets are generally provided:

• clean - This target deletes any existing build and dist directories, so that they can be reconstructed from scratch. This allows you to guarantee that you have not made source code modifications that will result in problems at runtime due to not recompiling all affected classes.
  
  
• prepare - This target "prepares" the build directory, creating subdirectories as required. A common use of this target is to copy static files (documentation, HTML pages, and JSP pages) from the source directory to the build directory. When executed, this target will only create directories if they do not exist, and only copy files if the destination file does not exist, or the source version of the file is newer. This target is generally invoked indirectly, by virtue of a depends attribute on some other task.
  
  
• compile - This target is used to compile any source code that has been changed since the last time compilation took place. The resulting class files are created in the build directory, so that they can be executed when the application is deployed. Because this command is executed so often during development, it is normally made the "default" target so that a simple ant command will execute it.
  
  
• all - This target is a short cut for running the clean target, followed by the compiile target. Thus, it guarantees that you will recompile the entire application, to ensure that you have not unknowingly introduced any incompatible changes.
  
  
• deploy - This target is used to install your application into a servlet container (we will continue to use Tomcat in our examples) so that it can be tested. If your application requires external JAR files, they will be copied to the /WEB-INF/lib directory at deployment time.
  
  
• javadoc - This target creates Javadoc API documentation for the Java classes in this web application. The example build.xml file assumes you want to include the API documentation with your app distribution, so it generates the docs in a subdirectory of the dist directory. Because you normally do not need to generate the Javadocs on every compilation, this target is usually a dependency of the dist target, but not of the compile target.
  
  
• dist - This target creates a distribution directory for your application, including any required documentation, the Javadocs for your Java classes, and a web application archive (WAR) file that will be delivered to system administrators who wish to install your application. Because this target also depends on the deploy target, the web application archive will have also picked up any external dependencies that were included at deployment time.