Streaming XML Reader fills up memory?

[Jonatan Samoocha]

After encountering a memory leak I stripped down my code to the following minimal fragment that reproduces the problem:

    
package [PROPRIETARY_PACKAGE_NAME];

import java.io.File;
import java.io.FileInputStream;

import javax.xml.stream.XMLEventReader;
import javax.xml.stream.XMLInputFactory;
import javax.xml.stream.events.XMLEvent;


public class Tesje {

	public static void main(String[] args) {
		System.out.println("Starting read");
		
		try {
			XMLInputFactory xmlIFac = XMLInputFactory.newInstance();
			FileInputStream iStream = new FileInputStream(new File(
					"C:\\temp\\bbPerfTest\\import2.xml"));
			
			XMLEventReader xmlReader = xmlIFac.createXMLEventReader(iStream);
			
			XMLEvent e = xmlReader.nextEvent();
			
			while (! e.isEndDocument()) {
				e = xmlReader.nextEvent();
			}
			
			iStream.close();
		} catch (Exception e) {
			e.printStackTrace();
		}
		
		System.out.println("Finished Read");
	}	
}

The file "import2.xml" has a size of approximately 15Mb. When running this code, the following output (and stack trace) is written:

Starting read
Exception in thread "main" java.lang.OutOfMemoryError: Java heap space
	at weblogic.xml.babel.scanner.ScannerState.addToBuffer(ScannerState.java:287)
	at weblogic.xml.babel.scanner.ScannerState.read(ScannerState.java:395)
	at weblogic.xml.babel.scanner.ScannerState.checkedRead(ScannerState.java:623)
	at weblogic.xml.babel.scanner.CharData.read(CharData.java:65)
	at weblogic.xml.babel.scanner.Scanner.startState(Scanner.java:296)
	at weblogic.xml.babel.scanner.Scanner.scan(Scanner.java:178)
	at weblogic.xml.babel.baseparser.BaseParser.accept(BaseParser.java:533)
	at weblogic.xml.babel.baseparser.StartElement.parse(StartElement.java:77)
	at weblogic.xml.babel.baseparser.BaseParser.parseElement(BaseParser.java:453)
	at weblogic.xml.babel.baseparser.BaseParser.parseSome(BaseParser.java:326)
	at weblogic.xml.stax.XMLStreamReaderBase.advance(XMLStreamReaderBase.java:199)
	at weblogic.xml.stax.XMLStreamReaderBase.next(XMLStreamReaderBase.java:243)
	at weblogic.xml.stax.XMLEventReaderBase.parseSome(XMLEventReaderBase.java:189)
	at weblogic.xml.stax.XMLEventReaderBase.nextEvent(XMLEventReaderBase.java:122)
	at [PROPRIETARY_PACKAGE_NAME].Test.main(Test.java:27)

Using a heap dump analyzer, I found out that some char[] internal to that weblogic parser was filled with the contents of the whole file. There are 2 strange things in this situation:

1. Why is the javax.xml.stream.XMLEventReader instance using the weblogic.xml.stax.* package internally? Is this configured somewhere? Even stranger, I couldn't find any of the weblogic.xml.stax.* classes on my classpath.
2. Why is the whole file loaded in memory while the code iterates through it event by event? Do I need to configure anything to reduce some internal buffer size, and hence the overall memory footprint?

Many thanks,

Jonatan

[William Brogden]

1. I suspect that somewhere there must be a system property that specifies the weblogic.xml ... package for the default parser.

A normal SAX parser would not be so wasteful.

Bill

[Jonatan Samoocha]

The problem seems to be classpath-related: when the above code is moved to an empty project with just the JAXB jars on the classpath, everything runs ok. I'll report about the offending classes if able to find them.