Actually if you look on CPAN there is a fork of XML::XPath done by James G. Smith as developer version. I experimented with it and it really worked for initial examples.

Unfortunately data structures differ from XML documents in various aspects: XML elements can be repeated but not in hashes, hash keys can be as strange as you like in contrast to element names. Arrays starting with index 1 in XPath is confusing to read on data structures. And a lot of code complexity for XML that's useless on data structures: for namespaces, attributes, etc., and missing features the other way around: eg. testing for the blessed types.

That's why I restarted it, with kind permission of James G. Smith to use his namespace.

Not that I already have all the features, of course ...