Trail: Java API for XML Processing (JAXP)
Lesson: Streaming API for XML


The StAX API exposes methods for iterative, event-based processing of XML documents. XML documents are treated as a filtered series of events, and infoset states can be stored in a procedural fashion. Moreover, unlike SAX, the StAX API is bidirectional, enabling both reading and writing of XML documents.

The StAX API is really two distinct API sets: a cursor API and an iterator API. These two API sets are explained in greater detail later in this lesson, but their main features are briefly described below.

Cursor API

As the name implies, the StAX cursor API represents a cursor with which you can walk an XML document from beginning to end. This cursor can point to one thing at a time, and always moves forward, never backward, usually one infoset element at a time.

The two main cursor interfaces are XMLStreamReader and XMLStreamWriter. XMLStreamReader includes accessor methods for all possible information retrievable from the XML Information model, including document encoding, element names, attributes, namespaces, text nodes, start tags, comments, processing instructions, document boundaries, and so forth; for example:

public interface XMLStreamReader {
    public int next() throws XMLStreamException;
    public boolean hasNext() throws XMLStreamException;

    public String getText();
    public String getLocalName();
    public String getNamespaceURI();
    // ... other methods not shown

You can call methods on XMLStreamReader, such as getText and getName, to get data at the current cursor location. XMLStreamWriter provides methods that correspond to StartElement and EndElement event types; for example:

public interface XMLStreamWriter {
    public void writeStartElement(String localName) throws XMLStreamException;
    public void writeEndElement() throws XMLStreamException;
    public void writeCharacters(String text) throws XMLStreamException;
    // ... other methods not shown

The cursor API mirrors SAX in many ways. For example, methods are available for directly accessing string and character information, and integer indexes can be used to access attribute and namespace information. As with SAX, the cursor API methods return XML information as strings, which minimizes object allocation requirements.

Iterator API

The StAX iterator API represents an XML document stream as a set of discrete event objects. These events are pulled by the application and provided by the parser in the order in which they are read in the source XML document.

The base iterator interface is called XMLEvent, and there are subinterfaces for each event type listed in XMLEvent table. The primary parser interface for reading iterator events is XMLEventReader, and the primary interface for writing iterator events is XMLEventWriter. The XMLEventReader interface contains five methods, the most important of which is nextEvent, which returns the next event in an XML stream. XMLEventReader implements java.util.Iterator, which means that returns from XMLEventReader can be cached or passed into routines that can work with the standard Java Iterator; for example:

public interface XMLEventReader extends Iterator {
    public XMLEvent nextEvent() throws XMLStreamException;
    public boolean hasNext();
    public XMLEvent peek() throws XMLStreamException;
    // ...

Similarly, on the output side of the iterator API, you have:

public interface XMLEventWriter {
    public void flush() throws XMLStreamException;
    public void close() throws XMLStreamException;
    public void add(XMLEvent e) throws XMLStreamException;
    public void add(Attribute attribute) throws XMLStreamException;
    // ...

Iterator Event Types

XMLEvent Types Defined in the Event Iterator API

Event Type



Reports the beginning of a set of XML events, including encoding, XML version, and standalone properties.


Reports the start of an element, including any attributes and namespace declarations; also provides access to the prefix, namespace URI, and local name of the start tag.


Reports the end tag of an element. Namespaces that have gone out of scope can be recalled here if they have been explicitly set on their corresponding StartElement.


Corresponds to XML CData sections and CharacterData entities. Note that ignorable white space and significant white space are also reported as Character events.


Character entities can be reported as discrete events, which an application developer can then choose to resolve or pass through unresolved. By default, entities are resolved. Alternatively, if you do not want to report the entity as an event, replacement text can be substituted and reported as Characters.


Reports the target and data for an underlying processing instruction.


Returns the text of a comment.


Reports the end of a set of XML events.


Reports as java.lang.String information about the DTD, if any, associated with the stream, and provides a method for returning custom objects found in the DTD.


Attributes are generally reported as part of a StartElement event. However, there are times when it is desirable to return an attribute as a standalone Attribute event; for example, when a namespace is returned as the result of an XQuery or XPath expression.


As with attributes, namespaces are usually reported as part of a StartElement, but there are times when it is desirable to report a namespace as a discrete Namespace event.

Note that the DTD, EntityDeclaration, EntityReference, NotationDeclaration, and ProcessingInstruction events are only created if the document being processed contains a DTD.

Example of Event Mapping

As an example of how the event iterator API maps an XML stream, consider the following XML document:

<?xml version="1.0"?>
<BookCatalogue xmlns="">
        <Title>Yogasana Vijnana: the Science of Yoga</Title>
        <Cost currency="INR">11.50</Cost>

This document would be parsed into eighteen primary and secondary events, as shown in the following table. Note that secondary events, shown in curly braces ({}), are typically accessed from a primary event rather than directly.

Example of Iterator API Event Mapping








isCData = false data = "\n" IsWhiteSpace = true



qname = BookCatalogue: attributes = null namespaces = {BookCatalogue" ->"}



qname = Book attributes = null namespaces = null



qname = Title attributes = null namespaces = null



isCData = false data = "Yogasana Vijnana: the Science of Yoga\n\t" IsWhiteSpace = false



qname = Title namespaces = null



qname = ISBN attributes = null namespaces = null



isCData = false data = "81-40-34319-4\n\t" IsWhiteSpace = false



qname = ISBN namespaces = null



qname = Cost attributes = {"currency" -> INR} namespaces = null



isCData = false data = "11.50\n\t" IsWhiteSpace = false



qname = Cost namespaces = null



isCData = false data = "\n" IsWhiteSpace = true



qname = Book namespaces = null



isCData = false data = "\n" IsWhiteSpace = true



qname = BookCatalogue: namespaces = {BookCatalogue" ->"}




There are several important things to note in this example:

Choosing between Cursor and Iterator APIs

It is reasonable to ask at this point, “What API should I choose? Should I create instances of XMLStreamReader or XMLEventReader? Why are there two kinds of APIs anyway?”

Development Goals

The authors of the StAX specification targeted three types of developers:

Given these wide-ranging development categories, the StAX authors felt it was more useful to define two small, efficient APIs rather than overloading one larger and necessarily more complex API.

Comparing Cursor and Iterator APIs

Before choosing between the cursor and iterator APIs, you should note a few things that you can do with the iterator API that you cannot do with the cursor API:

Similarly, keep some general recommendations in mind when making your choice:

Previous page: Why StAX?
Next page: Using StAX