[Xmlstar-devel] Should the first node be text node for xpath: /descendant::node()/rec/child::node()[1]

[Ming C. <cim...@ya...>]

Hi Noam,
 
Recently I get different outputs with those tools: XmlStar, XPathBuilder, XmlLint (win32 version, with libxml2 underneath) and XPathTester (online tool), for below XML file:
 
<?xml version="1.0" encoding="UTF-8"?>
<xml>
   <table>
      <rec id="1">
         <para type="error" position="1"/>
         <para type="error" position="2"/>
         <para type="error" position="3"/>
      </rec>
      <rec id="2">
         <para type="warning" position="1"/>
         <para type="warning" position="2"/>
         <para type="warning" position="3"/>
      </rec>
      <rec id="3">
         <para type="info" position="1"/>
         <para type="info" position="2"/>
         <para type="info" position="3"/>
      </rec>
   </table>
</xml>
 
While my original expected outputs (which really is XmlStar and XPathBuilder's output) are: "<para type="error" position="1"/><para type="warning" position="1"/><para type="info" position="1"/>", the XmlLint and XPathTester give some empty lines actually.
 
I asked this question to the libxml2 mailing list, and got below answers:
 
From: Liam R E Quin <li...@ho...>
To: Ming Chen <cim...@ya...> 
Cc: "xm...@gn..." <xm...@gn...> 
Sent: Tuesday, January 10, 2012 11:22 AM
Subject: Re: [xml] Does not support the expression : /descendant::node()/rec/child:­:node()[1]?
 
On Mon, 2012-01-09 at 18:54 -0800, Ming Chen wrote:

> According to the XPath spec (V2.0 section 3.2.3 Unabbreviated 
> Syntax) : child::node() selects all the children of the context node. Note 
> that no attribute nodes are returned, because attributes are not 
> children.

Note, libxml2 actually only supports XPath 1, not XPath 2.

However,
/descendant::node()/rec/child::node()[1]
will match text nodes, and you're getting the blank (whitespace-only)
text node that's the first child of elements, since your input is
"indented".

> Shouldn’t it have the same output as 
> /descendant::node()/rec/child::*[1] and /descendant::node()/rec/para[1]?
No.

The first child node in
<rec id="1">
<para type="error" position="1"/>
is the newline and spaces between id="1"> and <para.


Liam

-- 
Liam Quin - XML Activity Lead, W3C, http://www.w3.org/People/Quin/
Pictures from old books: http://fromoldbooks.org/


I think that explanation is legal. What's your opinion on this? 
 
 
Thanks,
Ming