Question

A lot has been written about Nokogiri in terms of reading XML using XPath. However, what’s about using Nokogiri with XML containing XPath references.

In the example, the xml contains a XPath reference:

<elements>
 <element>
  <location>
   <longitude>...
   <latitude>...
  </location>
 </element>
 <element>
  <location reference="../../element/location"/>
 </element>
</elements>

Since both location elements are equal, only the first element is described in detail. The second just references the first.

Using Nokogiri, xml.xpath(‘//location’) returns two node instances as expected. The first node contains all child nodes. The second only a reference as an attribute of the second node instance.

Ok, assuming I want to request all longitude values, I would do xml.xpath(‘//location/longitude’). This returns only one node instance. However, since there are actually two elements of type “longitude”, I expected to receive two node instances, expecting Nokogiri to resolve XPath references…

How do I achieve this with Nokogiri?

Answer 1

You can collect location nodes that have actual values (non-reference nodes) and then collect all references separately, as the following snippet demonstrates:

require 'nokogiri'

xml = <<End
<elements>
 <element>
  <location>
   <longitude>45</longitude>
   <latitude>-70</latitude>
  </location>
 </element>
 <element>
  <location reference="../../element/location"/>
 </element>
</element>
End

doc = Nokogiri::XML(xml)

#Collect all the explicit longitudes
longitudes = doc.search('//location[not(@reference)]/longitude').map(&:text)

#Follow references to longitudes
doc.search('//location[@reference]').each do |location|
  reference = location.attribute('reference')
  longitudes << location.xpath("#{reference}/longitude").text
end

puts longitudes #=> ["45", "45"]

You can build on this technique to extract whatever information you want.