How does XPath deal with XML namespaces?

Question

How does XPath deal with XML namespaces?

If I use

/IntuitResponse/QueryResponse/Bill/Id

to parse the XML document below I get 0 nodes back.

<?xml version="1.0" encoding="UTF-8" standalone="yes"?>
<IntuitResponse xmlns="http://schema.intuit.com/finance/v3" 
                time="2016-10-14T10:48:39.109-07:00">
    <QueryResponse startPosition="1" maxResults="79" totalCount="79">
        <Bill domain="QBO" sparse="false">
            <Id>=1</Id>
        </Bill>
    </QueryResponse>
</IntuitResponse>

However, I'm not specifying the namespace in the XPath (i.e. http://schema.intuit.com/finance/v3 is not a prefix of each token of the path). How can XPath know which Id I want if I don't tell it explicitly? I suppose in this case (since there is only one namespace) XPath could get away with ignoring the xmlns entirely. But if there are multiple namespaces, things could get ugly.

Answer 1

XPath 1.0/2.0

Defining namespaces in XPath _{^{(recommended)}}

XPath itself doesn't have a way to bind a namespace prefix with a namespace. Such facilities are provided by the hosting library.

It is recommended that you use those facilities and define namespace prefixes that can then be used to qualify XML element and attribute names as necessary.

---

Here are some of the various mechanisms which XPath hosts provide for specifying namespace prefix bindings to namespace URIs.

^{_{(OP's original XPath, /IntuitResponse/QueryResponse/Bill/Id, has been elided to /IntuitResponse/QueryResponse.)}}

C#:

XmlNamespaceManager nsmgr = new XmlNamespaceManager(doc.NameTable);
nsmgr.AddNamespace("i", "http://schema.intuit.com/finance/v3");
XmlNodeList nodes = el.SelectNodes(@"/i:IntuitResponse/i:QueryResponse", nsmgr);

Google Docs:

Unfortunately, IMPORTXML() does not provide a namespace prefix binding mechanism. See next section, Defeating namespaces in XPath, for how to use local-name() as a work-around.

Java (SAX):

NamespaceSupport support = new NamespaceSupport();
support.pushContext();
support.declarePrefix("i", "http://schema.intuit.com/finance/v3");

Java (XPath):

xpath.setNamespaceContext(new NamespaceContext() {
    public String getNamespaceURI(String prefix) {
      switch (prefix) {
        case "i": return "http://schema.intuit.com/finance/v3";
        // ...
       }
    });

• Remember to call DocumentBuilderFactory.setNamespaceAware(true).
• See also: Java XPath: Queries with default namespace xmlns

JavaScript:

See Implementing a User Defined Namespace Resolver:

function nsResolver(prefix) {
  var ns = {
    'i' : 'http://schema.intuit.com/finance/v3'
  };
  return ns[prefix] || null;
}
document.evaluate( '/i:IntuitResponse/i:QueryResponse', 
                   document, nsResolver, XPathResult.ANY_TYPE, 
                   null );

Note that if the default namespace has an associated namespace prefix defined, using the nsResolver() returned by Document.createNSResolver() can obviate the need for a customer nsResolver().

Perl (LibXML):

my $xc = XML::LibXML::XPathContext->new($doc);
$xc->registerNs('i', 'http://schema.intuit.com/finance/v3');
my @nodes = $xc->findnodes('/i:IntuitResponse/i:QueryResponse');

Python (lxml):

from lxml import etree
f = StringIO('<IntuitResponse>...</IntuitResponse>')
doc = etree.parse(f)
r = doc.xpath('/i:IntuitResponse/i:QueryResponse', 
              namespaces={'i':'http://schema.intuit.com/finance/v3'})

Python (ElementTree):

namespaces = {'i': 'http://schema.intuit.com/finance/v3'}
root.findall('/i:IntuitResponse/i:QueryResponse', namespaces)

Python (Scrapy):

response.selector.register_namespace('i', 'http://schema.intuit.com/finance/v3')
response.xpath('/i:IntuitResponse/i:QueryResponse').getall()

PhP:

Adapted from @Tomalak's answer using DOMDocument:

$result = new DOMDocument();
$result->loadXML($xml);

$xpath = new DOMXpath($result);
$xpath->registerNamespace("i", "http://schema.intuit.com/finance/v3");

$result = $xpath->query("/i:IntuitResponse/i:QueryResponse");

See also @IMSoP's canonical Q/A on PHP SimpleXML namespaces.

Ruby (Nokogiri):

puts doc.xpath('/i:IntuitResponse/i:QueryResponse',
                'i' => "http://schema.intuit.com/finance/v3")

Note that Nokogiri supports removal of namespaces,

doc.remove_namespaces!

but see the below warnings discouraging the defeating of XML namespaces.

VBA:

xmlNS = "xmlns:i='http://schema.intuit.com/finance/v3'"
doc.setProperty "SelectionNamespaces", xmlNS  
Set queryResponseElement =doc.SelectSingleNode("/i:IntuitResponse/i:QueryResponse")

VB.NET:

xmlDoc = New XmlDocument()
xmlDoc.Load("file.xml")
nsmgr = New XmlNamespaceManager(New XmlNameTable())
nsmgr.AddNamespace("i", "http://schema.intuit.com/finance/v3");
nodes = xmlDoc.DocumentElement.SelectNodes("/i:IntuitResponse/i:QueryResponse",
                                           nsmgr)

SoapUI (doc):

declare namespace i='http://schema.intuit.com/finance/v3';
/i:IntuitResponse/i:QueryResponse

xmlstarlet:

-N i="http://schema.intuit.com/finance/v3"

XSLT:

<xsl:stylesheet version="1.0"
                xmlns:xsl="http://www.w3.org/1999/XSL/Transform"
                xmlns:i="http://schema.intuit.com/finance/v3">
   ...

---

Once you've declared a namespace prefix, your XPath can be written to use it:

/i:IntuitResponse/i:QueryResponse

---

Defeating namespaces in XPath _{^{(not recommended)}}

An alternative is to write predicates that test against local-name():

/*[local-name()='IntuitResponse']/*[local-name()='QueryResponse']

Or, in XPath 2.0:

/*:IntuitResponse/*:QueryResponse

Skirting namespaces in this manner works but is not recommended because it

• Under-specifies the full element/attribute name.

• Fails to differentiate between element/attribute names in different namespaces (the very purpose of namespaces). Note that this concern could be addressed by adding an additional predicate to check the namespace URI explicitly:

   /*[    namespace-uri()='http://schema.intuit.com/finance/v3' 
      and local-name()='IntuitResponse']
   /*[    namespace-uri()='http://schema.intuit.com/finance/v3' 
      and local-name()='QueryResponse']
  

  _{Thanks to Daniel Haley for the namespace-uri() note.}

• Is excessively verbose.

XPath 3.0/3.1

Libraries and tools that support modern XPath 3.0/3.1 allow the specification of a namespace URI directly in an XPath expression:

/Q{http://schema.intuit.com/finance/v3}IntuitResponse/Q{http://schema.intuit.com/finance/v3}QueryResponse

While Q{http://schema.intuit.com/finance/v3} is much more verbose than using an XML namespace prefix, it has the advantage of being independent of the namespace prefix binding mechanism of the hosting library. The Q{} notation is known as Clark Notation after its originator, James Clark. The W3C XPath 3.1 EBNF grammar calls it a BracedURILiteral.

_{Thanks to Michael Kay for the suggestion to cover XPath 3.0/3.1's BracedURILiteral.}

Answer 2

I use /*[name()='...'] in a google sheet to fetch some counts from Wikidata. I have a table like this

 thes    WD prop links   items
 NOM     P7749   3925    3789
 AAT     P1014   21157   20224

and the formulas in cols links and items are

=IMPORTXML("https://query.wikidata.org/sparql?query=SELECT(COUNT(*)as?c){?item wdt:"&$B14&"[]}","//*[name()='literal']")
=IMPORTXML("https://query.wikidata.org/sparql?query=SELECT(COUNT(distinct?item)as?c){?item wdt:"&$B14&"[]}","//*[name()='literal']")

respectively. The SPARQL query happens not to have any spaces...

I saw name() used instead of local-name() in Xml Namespace breaking my xpath!, and for some reason //*:literal doesn't work.