What does this print:
from lxml import etree
doc = etree.fromstring('<a><b><c/></b></a>')
newdoc = etree.ElementTree(doc.find('b'))
print newdoc.xpath('/b/c')[0].xpath('/a')
The answer is: [<Element a at 817548c>]. The first point to note is
that xpath() against an element is only relative to that element: any
absolute XPaths enumerate from the top of the containing tree. The
second point is that the shallow copying of etree means that
_Element::xpath, unlike _ElementTree::xpath, evaluates absolute
paths from the top of the original underlying tree! So even though
there’s no <a> in newdoc, an absolute XPath on a child element can
still reach it.
Yuck.