DeferredEntityReferenceImpl.javaAPI DocJava SE 5 API7383Fri Aug 26 14:55:44 BST


public class DeferredEntityReferenceImpl extends EntityReferenceImpl implements DeferredNode
EntityReference models the XML &entityname; syntax, when used for entities defined by the DOM. Entities hardcoded into XML, such as character entities, should instead have been translated into text by the code which generated the DOM tree.

An XML processor has the alternative of fully expanding Entities into the normal document tree. If it does so, no EntityReference nodes will appear.

Similarly, non-validating XML processors are not required to read or process entity declarations made in the external subset or declared in external parameter entities. Hence, some applications may not make the replacement value available for Parsed Entities of these types.

EntityReference behaves as a read-only node, and the children of the EntityReference (which reflect those of the Entity, and should also be read-only) give its replacement value, if any. They are supposed to automagically stay in synch if the DocumentType is updated with new values for the Entity.

The defined behavior makes efficient storage difficult for the DOM implementor. We can't just look aside to the Entity's definition in the DocumentType since those nodes have the wrong parent (unless we can come up with a clever "imaginary parent" mechanism). We must at least appear to clone those children... which raises the issue of keeping the reference synchronized with its parent. This leads me back to the "cached image of centrally defined data" solution, much as I dislike it.

For now I have decided, since REC-DOM-Level-1-19980818 doesn't cover this in much detail, that synchronization doesn't have to be considered while the user is deep in the tree. That is, if you're looking within one of the EntityReferennce's children and the Entity changes, you won't be informed; instead, you will continue to access the same object -- which may or may not still be part of the tree. This is the same behavior that obtains elsewhere in the DOM if the subtree you're looking at is deleted from its parent, so it's acceptable here. (If it really bothers folks, we could set things up so deleted subtrees are walked and marked invalid, but that's not part of the DOM's defined behavior.)

As a result, only the EntityReference itself has to be aware of changes in the Entity. And it can take advantage of the same structure-change-monitoring code I implemented to support DeepNodeList.

$Id:,v 1.19 2003/05/08 19:52:40 elena Exp $

Fields Summary
static final long
Serialization version.
protected transient int
Node index.
Constructors Summary
DeferredEntityReferenceImpl(DeferredDocumentImpl ownerDocument, int nodeIndex)
This is the deferred constructor. Only the fNodeIndex is given here. All other data, can be requested from the ownerDocument via the index.

    // Constructors

        super(ownerDocument, null);

        fNodeIndex = nodeIndex;

Methods Summary
public intgetNodeIndex()
Returns the node index.

        return fNodeIndex;
protected voidsynchronizeChildren()
Synchronize the children.

        // no need to synchronize again

        // get children
        DeferredDocumentImpl ownerDocument =
            (DeferredDocumentImpl) ownerDocument();
        ownerDocument.synchronizeChildren(this, fNodeIndex);
        setReadOnly(true, true);

protected voidsynchronizeData()
Synchronize the entity data. This is special because of the way that the "fast" version stores the information.

        // no need to sychronize again

        // get the node data
        DeferredDocumentImpl ownerDocument =
        name = ownerDocument.getNodeName(fNodeIndex);
        baseURI = ownerDocument.getNodeValue(fNodeIndex);