Description Marcel Stör 2010-06-08 08:17:52 EDT

Created attachment 171391 [details]
stack trace while parsing XML

One has to pay a serious performance penalty when the XML data source is used with an input stream -> allow to pass a XML document to BIRT.

[Scenario]
Our application creates XML content "on the fly" i.e. at runtime. We have a org.jdom.Document instance in memory. However, in order to use it as a XML data source for BIRT we need convert it into a byte[] which can then be wrapped by a ByteArrayInputStream and passed to BIRT:

task.getAppContext().put("org.eclipse.datatools.enablement.oda.xml.inputStream", xmlByteStream);

BIRT/ODA then does the exact opposite and parses the content with SAX (org.eclipse.datatools.enablement.oda.xml.util.SaxParser.run()). Needless to say that this is not very efficient.

[Proposal]
Allow to pass a XML document (org.w3c.dom.Document) using the app context like so:

task.getAppContext().put("org.eclipse.datatools.enablement.oda.xml.document", xmlDocument)

[Side note]
While debugging I noticed that for any given .rptdesign file several SaxParser instances were created (and run) - up to 8. It appears that for every table the whole XML document is parsed again! I suppose that even without the improvement described above this wouldn't be necessary.

[Side note 2]
Fix the naming typo in SuppressDuplciateReportExecutor.