I would strongly discourage regexes in this case!! It's a lot trickier than the split/hash solution (and a lot trickier than you think :). Take as an example the solution $data=~s!((.*)/.*/)\2!$1!g; from arhuman which doesn't work correctly even for your first example (NY/Ch/NY), as it leaves a '/' at the end of the string. And much more important it breaks in nearly all other cases. The problems include

• greedyness of .*
• correct removal of '/'
• consider sth like 'abc/def/abcd' (not a real life example but a similar thing might come up) and 'abc' will be deleted
• speed: the regex solution makes a comparison for pairs, the first word with all others, then the second word with all others. This results in an O(n^2) algorithm (i.e. double the number of elements in the string, quadruple the execution time) whereas the split/hash solution is O(n)

-- Hofmator