Learning and Using Jakarta Digester

by Philipp K. Janert, Ph.D.

10/23/2002

Turning an XML document into a corresponding hierarchy of Java bean objects is a fairly common task. In a previous article, I

described how to accomplish this using the standard SAX and DOM APIs.

Although powerful and flexible, both APIs are, in effect, too low-level for the specific task at hand. Furthermore, the

unmarshalling procedure itself requires a fair amount of coding: a parse-stack must be maintained when using SAX, and the

DOM-tree must be navigated when using DOM.

This is where the Apache Jakarta Commons Digester framework comes in.

The Jakarta Digester Framework

The Jakarta Digester framework grew out of the Jakarta Struts Web toolkit. Originally developed to process the central

struts-config.xml configuration file, it was soon recognized that the framework was more generally useful, and moved to the

Jakarta Commons project, the stated goal of which is to provide a "repository of reusable Java components." The most recent

version, Digester 1.3, was released on August 13, 2002.

The Digester class lets the application programmer specify a set of actions to be performed whenever the parser encounters

certain simple patterns in the XML document. The Digester framework comes with 10 prepackaged "rules," which cover most of

the required tasks when unmarshalling XML (such as creating a bean or setting a bean property), but each user is free to

define and implement his or her own rules, as necessary.

The Example Document and Beans

In this example, we will unmarshall the same XML document that we used in the previous article:

<?xml version="1.0"?>

<book>

<author>Author 1</author>

<title>Title 1</title>

</book>

<book>

<author>Author 2</author>

</book>

<name>Mag Title 1</name>

<headline>Some Headline</headline>

</article>

<headline>Another Headline</headline>

</article>

</magazine>

<book>

<author>Author 2</author>

<title>His Other Book</title>

</book>

<name>Mag Title 2</name>

<headline>Second Headline</headline>

</article>

</magazine>

</catalog>

The bean classes are also the same, except for one important change: In the previous article, I had declared these classes to

have package scope -- primarily so that I could define all of them in the same source file! Using the Digester framework,

this is no longer possible; the classes need to be declared as public (as is required for classes conforming to the JavaBeans

specification):

import java.util.Vector;

public class Catalog {

private Vector books;

private Vector magazines;

public Catalog() {

books = new Vector();

magazines = new Vector();

}

public void addBook( Book rhs ) {

books.addElement( rhs );

}

public void addMagazine( Magazine rhs ) {

magazines.addElement( rhs );

}

public String toString() {

String newline = System.getProperty( "line.separator" );

StringBuffer buf = new StringBuffer();

buf.append( "--- Books ---" ).append( newline );

for( int i=0; i<books.size(); i++ ){

buf.append( books.elementAt(i) ).append( newline );

}

buf.append( "--- Magazines ---" ).append( newline );

for( int i=0; i<magazines.size(); i++ ){

buf.append( magazines.elementAt(i) ).append( newline );

}

return buf.toString();

}

--------------------------------------------------------------------------------

public class Book {

private String author;

private String title;

public Book() {}

public void setAuthor( String rhs ) { author = rhs; }

public void setTitle( String rhs ) { title = rhs; }

public String toString() {

return "Book: Author='" + author + "' Title='" + title + "'";

}

--------------------------------------------------------------------------------

import java.util.Vector;

public class Magazine {

private String name;

private Vector articles;

public Magazine() {

articles = new Vector();

}

public void setName( String rhs ) { name = rhs; }

public void addArticle( Article a ) {

articles.addElement( a );

}

public String toString() {

StringBuffer buf = new StringBuffer( "Magazine: Name='" + name + "' ");

for( int i=0; i<articles.size(); i++ ){

buf.append( articles.elementAt(i).toString() );

}

return buf.toString();

}

--------------------------------------------------------------------------------

public class Article {

private String headline;

private String page;

public Article() {}

public void setHeadline( String rhs ) { headline = rhs; }

public void setPage( String rhs ) { page = rhs; }

public String toString() {

return "Article: Headline='" + headline + "' on page='" + page + "' ";

}

Specifying Patterns and Rules

The Digester class processes the input XML document based on patterns and rules. The patterns must match XML elements, based

on their name and location in the document tree. The syntax used to describe the matching patterns resembles the XPath match

patterns, a little: the pattern catalog matches the top-level <catalog> element, the pattern catalog/book matches a <book>

element nested directly inside a <catalog> element, but nowhere else in the document, etc.

All patterns are absolute: the entire path from the root element on down has to be specified. The only exception are patterns

containing the wildcard operator *: the pattern */name will match a <name> element anywhere in the document. Also note that

there is no need for a special designation for the root element, since all paths are absolute.

Whenever the Digester encounters one of the specified patterns, it performs the actions that have been associated with it. In

this, the Digester framework is of course related to a SAX parser (and in fact, the Digester class implements

org.xml.sax.ContentHandler and maintains the parse stack). All rules to be used with the Digester must extend

org.apache.commons.digester.Rule -- which in itself exposes methods similar to the SAX ContentHandler callbacks: begin() and

end() are called when the opening and closing tags of the matched element are encountered.

The body() method is called for the content nested inside of the matched element, and finally, there is a finish() method,

which is called once processing of the closing tag is complete, to provide a hook to do possible final clean-up chores. Most

application developers will not have to concern themselves with these functions, however, since the standard rules that ship

with the framework are likely to provide all desired functionality.

To unmarshal a document, then, create an instance of the org.apache.commons.digester.Digester class, configure it if

necessary, specify the required patterns and rules, and finally, pass a reference to the XML file to the parse() method. This

is demonstrated in the DigesterDriver class below. (The filename of the input XML document must be specified on the command

line.)

import org.apache.commons.digester.*;

import java.io.*;

import java.util.*;

public class DigesterDriver {

public static void main( String[] args ) {

try {

Digester digester = new Digester();

digester.setValidating( false );

digester.addObjectCreate( "catalog", Catalog.class );

digester.addObjectCreate( "catalog/book", Book.class );

digester.addBeanPropertySetter( "catalog/book/author", "author" );

digester.addBeanPropertySetter( "catalog/book/title", "title" );

digester.addSetNext( "catalog/book", "addBook" );

digester.addObjectCreate( "catalog/magazine", Magazine.class );

digester.addBeanPropertySetter( "catalog/magazine/name", "name" );

digester.addObjectCreate( "catalog/magazine/article", Article.class );

digester.addSetProperties( "catalog/magazine/article", "page", "page" );

digester.addBeanPropertySetter( "catalog/magazine/article/headline" );

digester.addSetNext( "catalog/magazine/article", "addArticle" );

digester.addSetNext( "catalog/magazine", "addMagazine" );

File input = new File( args[0] );

Catalog c = (Catalog)digester.parse( input );

System.out.println( c.toString() );

} catch( Exception exc ) {

exc.printStackTrace();

}

After instantiating the Digester, we specify that it should not validate the XML document against a DTD -- because we did not

define one for our simple Catalog document. Then we specify the patterns and the associated rules: the ObjectCreateRule

creates an instance of the specified class and pushes it onto the parse stack. The SetPropertiesRule sets a bean property to

the value of an XML attribute of the current element -- the first argument to the rule is the name of the attribute, the

second, the name of the property.

Whereas SetPropertiesRule takes the value from an attribute, BeanPropertySetterRule takes the value from the raw character

data nested inside of the current element. It is not necessary to specify the name of the property to set when using

BeanPropertySetterRule: it defaults to the name of the current XML element. In the example above, this default is being used

in the rule definition matching the catalog/magazine/article/headline pattern. Finally, the SetNextRule pops the object on

top of the parse stack and passes it to the named method on the object below it -- it is commonly used to insert a finished

bean into its parent.

Note that it is possible to register several rules for the same pattern. If this occurs, the rules are executed in the order

in which they are added to the Digester -- for instance, to deal with the <article> element, found at

catalog/magazine/article, we first create the appropriate article bean, then set the page property, and finally pop the

completed article bean and insert it into its magazine parent.

Invoking Arbitrary Functions

It is not only possible to set bean properties, but to invoke arbitrary methods on objects in the stack. This is accomplished

using the CallMethodRule to specify the method name and, optionally, the number and type of arguments passed to it.

Subsequent specifications of the CallParamRule define the parameter values to be passed to the invoked functions. The values

can be taken either from named attributes of the current XML element, or from the raw character data contained by the current

element. For instance, rather than using the BeanPropertySetterRule in the DigesterDriver implementation above, we could have

achieved the same effect by calling the property setter explicitly, and passing the data as parameter:

digester.addCallMethod( "catalog/book/author", "setAuthor", 1 );

digester.addCallParam( "catalog/book/author", 0 );

The first line gives the name of the method to call (setAuthor()), and the expected number of parameters (1). The second line

says to take the value of the function parameter from the character data contained in the <author> element and pass it as

first element in the array of arguments (i.e., the array element with index 0). Had we also specified an attribute name

(e.g., digester.addCallParam( "catalog/book/author", 0, "author" );), the value would have been taken from the respective

attribute of the current element instead.

One important caveat: confusingly, digester.addCallMethod( "pattern", "methodName", 0 ); does not specify a call to a method

taking no arguments -- instead, it specifies a call to a method taking one argument, the value of which is taken from the

character data of the current XML element! We therefore have yet another way to implement a replacement for

BeanPropertySetterRule:

digester.addCallMethod( "catalog/book/author", "setAuthor", 0 );

To call a method that truly takes no parameters, use digester.addCallMethod( "pattern", "methodName" );.

Summary of Standard Rules

Below are brief descriptions of all of the standard rules.

Creational

ObjectCreateRule: Creates an object of the specified class using its default constructor and pushes it onto the stack; it is

popped when the element completes. The class to instantiate can be given through a class object or the fully-qualified class

name.

FactoryCreateRule: Creates an object using a specified factory class and pushes it onto the stack. This can be useful for

classes that do not provide a default constructor. The factory class must implement the

org.apache.commons.digester.ObjectCreationFactory interface.

Property Setters

SetPropertiesRule: Sets one or several named properties in the top-level bean using the values of named XML element

attributes. Attribute names and property names are passed to this rule in String[] arrays. (Typically used to handle XML

constructs like <article page="10">.)

BeanPropertySetterRule: Sets a named property on the top-level bean to the character data enclosed by the current XML

element. (Example: <page>10</page>.)

SetPropertyRule: Sets a property on the top-level bean. Both the property name, as well as the value to which this property

will be set, are given as attributes to the current XML element. (Example: <article key="page" value="10" />.)

Parent/Child Management

SetNextRule: Pops the object on top of the stack and passes it to a named method on the object immediately below. Typically

used to insert a completed bean into its parent.

SetTopRule: Passes the second-to-top object on the stack to the top-level object. This is useful if the child object exposes

a setParent method, rather than the other way around.

SetRootRule: Calls a method on the object at the bottom of the stack, passing the object on top of the stack as argument.

Arbitrary Method Calls

CallMethodRule: Calls an arbitrary named method on the top-level bean. The method may take an arbitrary set of parameters.

The values of the parameters are given by subsequent applications of the CallParamRule.

CallParamRule: Represents the value of a method parameter. The value of the parameter is either taken from a named XML

element attribute, or from the raw character data enclosed by the current element. This rule requires that its position on

the parameter list is specified by an integer index.

Specifying Rules in XML: Using the xmlrules Package

So far, we have specified the patterns and rules programmatically at compile time. While conceptually simple and

straightforward, this feels a bit odd: the entire framework is about recognizing and handling structure and data at run time,

but here we go fixing the behavior at compile time! Large numbers of fixed strings in source code typically indicate that

something is being configured (rather than programmed), which could be (and probably should be) done at run time instead.

The org.apache.commons.digester.xmlrules package addresses this issue. It provides the DigesterLoader class, which reads the

pattern/rule-pairs from an XML document and returns a digester already configured accordingly. The XML document configuring

the Digester must comply with the digester-rules.dtd, which is part of the xmlrules package.

Below is the contents of the configuration file (named rules.xml) for the example application. I want to point out several

things here.

Patterns can be specified in two different ways: either as attributes to each XML element representing a rule, or using the

<pattern> element. The pattern defined by the latter is valid for all contained rule elements. Both ways can be mixed, and

<pattern> elements can be nested -- in either case, the pattern defined by the child element is appended to the pattern

defined in the enclosing <pattern> element.

The <alias> element is used with the <set-properties-rule> to map an XML attribute to a bean property.

Finally, using the current release of the Digester package, it is not possible to specify the BeanPropertySetterRule in the

configuration file. Instead, we are using the CallMethodRule to achieve the same effect, as explained above.

<?xml version="1.0"?>

<digester-rules>

<set-properties-rule pattern="catalog" >

</set-properties-rule>

<call-method-rule pattern="author" methodname="setAuthor"

paramcount="0" />

<call-method-rule pattern="title" methodname="setTitle"

paramcount="0" />

<set-next-rule methodname="addBook" />

</pattern>

<call-method-rule pattern="name" methodname="setName" paramcount="0" />

<set-properties-rule>

</set-properties-rule>

<call-method-rule pattern="headline" methodname="setHeadline"

paramcount="0" />

<set-next-rule methodname="addArticle" />

</pattern>

<set-next-rule methodname="addMagazine" />

</pattern>

</digester-rules>

Since all the actual work has now been delegated to the Digester and DigesterLoader classes, the driver class itself becomes

trivially simple. To run it, specify the catalog document as the first command line argument, and the rules.xml file as the

second. (Confusingly, the DigesterLoader will not read the rules.xml file from a File or an org.xml.sax.InputSource, but

requires a URL -- the File reference in the code below is therefore transformed into an equivalent URL.)

import org.apache.commons.digester.*;

import org.apache.commons.digester.xmlrules.*;

import java.io.*;

import java.util.*;

public class XmlRulesDriver {

public static void main( String[] args ) {

try {

File input = new File( args[0] );

File rules = new File( args[1] );

Digester digester = DigesterLoader.createDigester( rules.toURL() );

Catalog catalog = (Catalog)digester.parse( input );

System.out.println( catalog.toString() );

} catch( Exception exc ) {

exc.printStackTrace();

}

Conclusion

This concludes our brief overview of the Jakarta Commons Digester package. Of course, there is more. One topic ignored in

this introduction are XML namespaces: Digester allows you to specify rules that only act on elements defined within a certain

namespace.

We mentioned briefly the possibility of developing custom rules, by extending the Rule class. The Digester class exposes the

customary push(), peek(), and pop() methods, giving the individual developer freedom to manipulate the parse stack directly.

Lastly, note that there is an additional package providing a Digester implementation which deals with RSS (Rich-Site-

Summary)-formatted newsfeeds. The Javadoc tells the full story.

References

Jakarta Commons Digester Homepage

"Simple XML Parsing with SAX and DOM"

Jakarta Struts Homepage

"Java & XML Data Binding" -- XML Data Binding addresses the general problem of making XML data available in applications.

Programming Jakarta Struts -- the upcoming book on Jakarta Struts at O'Reilly.

Philipp K. Janert, Ph.D. is a Software Project Consultant, server programmer, and architect.