#15 XPath white-space error

[Mirko Budszuhn]

Hello,

We found that VTD-XML behaved differently when we converted a node-set to a number.
This might be fixed by the recent patch, we do not use the git for testing.

We used the following document for testing.

<?xml version="1.0"?>
<bookstore>
    <book-id>123</book-id>
    <book-id>456</book-id>
    <book-id>789</book-id>
    <book-id>234</book-id>
</bookstore>

We executed the XPath expression number(/bookstore) on the libraries leading to the results shown below:

Testing library: nokogiri , xpath: number(/bookstore)
NaN
Testing library: rexml , xpath: number(/bookstore)
123.0
Testing library: xalan-j , xpath: number(/bookstore)
NaN
Testing library: jaxen , xpath: number(/bookstore)
NaN
Testing library: lxml , xpath: number(/bookstore)
nan
Testing library: VTD-XML , xpath: number(/bookstore)
1.23456789234E11

The results of VTD-XML (and REXML) differ significantly from the other outputs.

The relevant portion of the XPath 1.0 standard on how a node-set is converted to a number is the following:

a node-set is first converted to a string as if by a call to the string function and then converted in the same way as a string argument. (Chapter 4.4 Number Functions)

So we could rewrite the XPath expression to number(string(/bookstore)). This leads to the same results as before, which is right.

Important to note here is, that white-space in an XML Document is turned into a text-node children element.
The string value of an element node is defined like this in XPath 1.0:

The string-value of an element node is the concatenation of the string-values of all text node descendants of the element node in document order.[Chapter 5.2 Element Nodes]

So when we execute the XPath expression string(/bookstore) the result should contain with white-space as well. VTD-XML strips the white-space, which is non-standard behavior.

Executing just the string(node-set) function leads to the following results:

lxml, nokogiri, rexml, xalan-j, jaxen, xpath: string(/bookstore)

    123
    456
    789
    234

library: VTD-XML , xpath: string(/bookstore)
123456789234

When VTD-XML now calls the number function with the wrong string value like this number('123456789234'), it results in the wrong final result of 123456789234.

Thank you for your time and we hope this is helpful.