|
Description
|
Image I/O describes image metadata in XML syntax using the W3C DOM,
and implements DOM specified interfaces in the org.w3c.dom package
In JDK 1.5 the internal use only XML "crimson" parser was replaced with "xerces"
and the DOM specification was updated from L2 to L3.
There were two consequences of this.
1) the DOM L3 specification added new methods as follows :
******
Node:
short compareDocumentPosition(Node other)
String getBaseURI()
Object getFeature(String feature, String version)
String getTextContent()
Object getUserData(String key)
boolean isDefaultNamespace(String namespaceURI)
boolean isEqualNode(Node arg)
boolean isSameNode(Node other)
String lookupNamespaceURI(String prefix)
String lookupPrefix(String namespaceURI)
void setTextContent(String textContent)
Object setUserData(String key, Object data, UserDataHandler handler)
Element extends Node
TypeInfo getSchemaTypeInfo()
void setIdAttribute(String name, boolean isId)
void setIdAttributeNode(Attr idAttr, boolean isId)
void setIdAttributeNS(String namespaceURI, String localName, boolean isId)
Attr extends Node
boolean isId()
TypeInfo getSchemaTypeInfo()
******
Unfortunately these are java interfaces, and so the engineer who worked
on updating the org.w3c.dom package needed to implement these methods
in javax.imageio. The short cut taken was to have them throw a
DOMException indicating they are not implemented.
Since Image I/O doesn't need any of these new methods this is
arguably OK. But no javadoc was added so these methods inherit
their specification which gives no indication of this and also
has confusing comments to "see the table above" when there is
no table in the doc for IIOMetaDataNode.
At the very least we should add javadoc to all these methods
indicating their actual behaviour.
Note that "Attr" is a return type from some methods on IIOMetaDataNode,
so it also needs to be mentioned that these return values will also
throw DOMException.
2. The second consequence, is that in setAttribute(name, value)
to verify that the name of an attribute was legal, Image I/O
needed to call in to Crimson. When Crimson was replaced, the
engineer doing this chose to copy the two Crimson classes
used into Image I/O rather than examining what was a
more appropriate solution.
The 1.4 DOM L2 javadoc had stated :
* @exception DOMException
* INVALID_CHARACTER_ERR: Raised if the specified name contains an
* illegal character.
However for L3 ( JDK 1.5) this was refined to :
DOMException - INVALID_CHARACTER_ERR: Raised if the specified name is not an XML name according to the XML version in use specified in the Document.xmlVersion attribute.
But Document.getXmlVersion() states in part :
If there is no declaration and if this document supports the "XML" feature, the value is "1.0". If this document does not support the "XML" feature, the value is always null. Changing this attribute will affect methods that check for invalid characters in XML names.
But IOMetaDataNode.getDocumentOwner() states :
Returns null, since IIOMetadataNodes do not belong to any Document.
Therefore, so as long as a character is a legal unicode character,
it should be permissible as a name of IIOMetaDataNode attribute,
and we do not need the crimson code to determine this.
Posted Date : 2007-05-17 22:11:05.0
|
