On Fri, 2015-11-20 at 16:39 +0000, Dimitre NovatchevB 
> I am aware of at least two very efficient algorithms:
> 
> B  1. Using *suffix-trees*B B -
> B  2. Using the hash of the search-string and scanning the given
> string

For substrings an n-gram approach can also be fast, but again it's a
sufficiently specialized optimization that I wouldn't expect most XPath
engines to do it unless they supported the Full Text extension.

Liam