EMLParser is slow to process large EML documents

[csjx]

The org.ecoinformatics.eml.EMLParser does not perform well when processing large EML documents (for instance, a document with 250 to 1000 attribute fully fleshed out elements defined). It can take 10, 30, 45 or more minutes to validate a document -- the duration scales with document size.

To try to alleviate this, change the parser to use a SAX-based model rather than a DOM.

org.ecoinformatics.eml.EMLParser uses two methods to validate a document: parseKeys() and parseKeyrefs(), both of which call getPathContent() and pass in an XPath selector. getPathContent() creates a DOM and passes back an org.w3.dom.NodeList.

See the attached file as an example.

eml250.xml.txt

[mbjones]

Need to test the EMLValidator against a large document. Its way faster than the old EMLParser. See #328.

[mbjones]

Tested the validator against the 4.3MB eml250.xml document above, and it is much faster -- now less than a second, down from many (tens?) of minutes with the old parser. I think this performance bug can be closed. Comments appreciated. Timing information is below:

$ time java -cp $CP $pkg/EMLValidator src/test/resources/eml250.xml
isValid: true

real	0m1.013s
user	0m2.627s
sys	0m0.149s

$ time java -cp $CP $pkg/EMLValidator src/test/resources/invalidEML/eml-error-annot-missing-id.xml 
isValid: false

real	0m0.312s
user	0m0.405s
sys	0m0.046s

$ time java -cp $CP $pkg/EMLValidator src/test/resources/eml-sample.xml 
isValid: true

real	0m0.337s
user	0m0.508s
sys	0m0.046s