Saxon

This XPath expression returns an element node. The string value of the element node is "29", which is the result I get when I do

            Processor p = new Processor(false);
            XPathCompiler c = p.newXPathCompiler();
            c.declareNamespace("age", "http://www.w3.org/2003/01/geo/wgs84_pos#");
            XPathExecutable e = c.compile(xpath);
            XPathSelector s = e.load();
            s.setContextItem(p.newDocumentBuilder().build(new StreamSource(new StringReader(xml))));
            XdmItem val = s.evaluateSingle();
            assertEquals("result", "29", val.getStringValue());

The result you are displaying (in both cases) includes the name of the element node as well as its string value. It's not clear how you are getting this string; we need to know this because that would appear to be where the changed behaviour lies. So, could you please supply a simple repro that shows the effect?

Hello, Thanks for your very fast answer.

This is the test case that passes with 9.8.0-14 and fails with 9.9.0-2:

@Test
public void spaceNewLineBugReproducer() throws SaxonApiException {
    Processor p = new Processor(false);
    XPathCompiler c = p.newXPathCompiler();
    c.declareNamespace("age", "http://www.w3.org/2003/01/geo/wgs84_pos#");
    String xPathQuery="//Employees/Employee[1]/age:ag";;
    XPathExecutable e = c.compile(xPathQuery);
    XPathSelector selector = e.load();
    selector.setContextItem(p.newDocumentBuilder().build(new StreamSource(new StringReader(xmlDoc))));
    XdmValue nodes = selector.evaluate();
    XdmItem item = nodes.itemAt(0);
    assertEquals("<age:ag xmlns:age=\"http://www.w3.org/2003/01/geo/wgs84_pos#\">29</age:ag>",item.toString());
}

Regards

The specification of XdmNode.toString() says

     In the case of an element node, the result will be a well-formed
     XML document serialized as defined in the W3C XSLT/XQuery serialization specification,
     using options method="xml", indent="yes", omit-xml-declaration="yes".

Because this specifies indent="yes", the processor is free to add whitespace (in this case a newline character) after an element end tag.

I would need to do some digging to find out exactly why it was changed (addition of a newline at the end of a serialized file is always a debatable topic...) but both versions with and without the newline are within the spec. We did find some cases where the Saxon indentation algorithm wasn't conformant to the W3C specs, so there were changes in 9.9, but I'm pretty sure the W3C specs make the newline here optional.

If you want predictable and repeatable results you shouldn't be relying on indented serialization, because it's not sufficiently specified.

More specifically, what has changed is that the from 9.9 the serializer implements the process that the W3C spec calls "sequence normalization"; previously this was the responsibility of the client of the serializer. The practical implication of this is that in 9.9 the element node being serialized by the toString() call is wrapped in a document node; and the XML indenter always (in both 9.8 and 9.9) adds a final newline after the endDocument() event. The purpose of this is so that when XSLT or XQuery output is sent to the console, the XML result is not followed on the same line by the next thing sent to the console.

We could, and perhaps should, change XdmNode.toString() so this newline is somehow suppressed; but the fact that it is there isn't a bug.

Thanks a lot for your great support ! We will upgrade in next JMeter 5.1 and you'll be thanked.

Regards Philippe