Why does the SAX parser lose some character data or why is the data split
into several chunks?
If you read the SAX
documentation, you will find that SAX may deliver contiguous text as multiple calls to
characters, for reasons having to do with parser efficiency and input
buffering. It is the programmer's responsibility to deal with that
appropriately, e.g. by accumulating text until the next non-characters
Xerces will split calls to characters at the end of an internal buffer,
at a new line and also at a few other boundaries. You can never rely on contiguous
text to be passed in a single characters callback.
Why doesn't the SAX parser report ignorable whitespace for XML Schemas?
SAX is very clear that ignorableWhitespace is only called for
element content whitespace, which is defined in the context of a DTD.
The result of schema validation is the Post-Schema-Validation Infoset (PSVI).
Schema processors augment the base Infoset by adding new properties to
element and attribute information items, but not character information items.
Schemas do not change whether a character is element content whitespace.
Why is the Attributes parameter passed to startElement always a
reference to the same object?
Outside the scope of startElement, the value of the
Attributes parameter is undefined. For each instance of Xerces'
SAX parser, there exists only one Attributes instance which
is reused for every new set of attributes. Before each
startElement callback, the previous values in this object
will be overwritten. This is done for performance reasons in order
to reduce object creation. To persist a set of attributes
beyond startElement the object should be cloned, for
instance by using org.xml.sax.helpers.AttributesImpl.
Why does the SAX parser report that xmlns attributes have no namespace?
An erratum for the Namespaces in XML Recommendation put namespace declaration
attributes in the namespace "http://www.w3.org/2000/xmlns/". By default,
SAX2 (SAX 2.0.2) follows the original Namespaces in XML Recommendation, so
conforming parsers must report that these attributes have no namespace. To
configure the parser to report a namespace for such attributes, turn on
the xmlns-uris feature.
When using Xerces 2.6.2 (or prior) or other parser implementations
that do not support this feature, your code must handle this discrepancy
when interacting with APIs such as DOM and applications which expect a namespace
for xmlns attributes.
Is there any way I can determine what encoding an entity was
written in, or what XML version the document conformed to, if I'm
Yes. As of SAX 2.0.2 encoding and version information is made
available through the org.xml.sax.ext.Locator2
interface. In Xerces, instances of the SAX Locator interface
passed to a setDocumentLocator call will also implement
the Locator2 interface. You can determine the encoding
and XML version of the entity currently being parsed by calling the
getEncoding() and getXMLVersion() methods.