Interface XmlOutput

All Known Implementing Classes:
C14nXmlOutput, DOMOutput, FastInfosetStreamWriterOutput, ForkXmlOutput, IndentingUTF8XmlOutput, MTOMXmlOutput, SAXOutput, StAXExStreamWriterOutput, UTF8XmlOutput, XMLEventWriterOutput, XmlOutputAbstractImpl, XMLStreamWriterOutput

public interface XmlOutput
Well-formed XML writer.

Implementations of this interface is used to connect XMLSerializer to the actual target. This allows XMLSerializer to be API agnostic.

Notes

JAXBContext assigns indices to URIs and local names that are statically known by using NameList. XmlOutput implementation can use these indices to improve the performance. For example, those namespace URI indices can be turned into prefixes quickly.

XmlOutput still allows arbitrary namepsace URIs / local names to be written.

The NamespaceContextImpl object, which is shared between XmlOutput and XMLSerializer, keeps track of the in-scope namespace bindings. By the time the beginStartTag(com.sun.xml.bind.v2.runtime.Name) method is called, all the namespace bindings for the new element is already declared. Similarly, after the endTag(com.sun.xml.bind.v2.runtime.Name) method is called, in-scope bindings will be removed. This book keeping is all done outside XmlOutput.

XmlOutput and XMLSerializer uses indices to reference prefixes/URIs to be written. NamespaceContextImpl can convert prefix indices to URIs and the string representations of prefixes. Binding from indices to URIs and prefixes do not change while indices are "in scope", so XmlOutput is again expected to take advantage of this to improve the perofmrnace.

prefix index 0 is reserved for "xml", and this binding is assumed to be always there. NamespaceContextImpl can handle this index correctly, but this binding will never be reported to XmlOutput through beginStartTag(com.sun.xml.bind.v2.runtime.Name).

One pecurilar behavior of a NamespaceContextImpl object is that it tries to define redundant xmlns="" on the root element. Implementations of XmlOutput is encouraged to check for this and avoid generating redundant namespace declarations.

Call Sequence

XMLSerializer calls the writer methods in the following order:

 CALLSEQUENCE  :=  startDocument ELEMENT endDocument
               |   ELEMENT   // for fragment

 ELEMENT       :=  beginStartTag attribute* endStartTag CONTENTS endTag

 CONTENTS      :=  (ELEMENT | text)*
 
TODO: for FI, consider making attribute values from Strings to CharSequences.
Author:
Kohsuke Kawaguchi