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

print join'/',grep{if(!$u{$_}){$u{$_}=1}}split'/',$data;
[download]

Well, I have two comments: a regex is a cool approach to take, and a regex is the wrong approach to take. You said yourself you could split and use a hash. That's what I'd do. I'd use a regex if I were trying to impress someone (which I'm not). But if you are, then here's my offering:

$cities = "Here/There/Everywhere/There/Again/Here/Too";

$cities =~ s{
  ([^/]+)      # non-slashes (the city): \1
  /            # a /
  (?=          # look ahead for:...
    (?:        # group (not capture)
      [^/]* /  # city and /
    )*?        # 0 or more times, non-greedily
    \1         # is the city here again?
    (?: /|$ )  # a / or end-of-string
  )
}
{}gx;
[download]

_____________________________________________________
Jeff japhy Pinyan: Perl, regex, and perl hacker.
s++=END;++y(;-P)}y js++=;shajsj<++y(p-q)}?print:??;

Right argument, japhy!! But it's an interesting problem nevertheless. Your regex breaks on: New York/York/Boston => New York/Boston My fix makes it sadly less elegant:

    1 while s{
      (/|^)   # add: capture slash or start of string: $1
      ([^/]+)      # non-slashes (the city): \2
      /            # a /
      (?=          # look ahead for:...
        (?:        # group (not capture)
          [^/]* /  # city and /
        )*?        # 0 or more times, non-greedily
        \2         # is the city here again?
        (?: /|$ )  # a / or end-of-string
      )
    }
    {$1}gx;  # add: replacement
[download]

Now this is close to my approach:  1 while s#(/|^)((?>[^/]*))/(?=(?:.*?/)?\2(?:/|$))#$1#g; The advantage of both solutions is that as little as possible is replaced. The disadvantage is that a regex isn't the right way to do it ;) - for all the reasons already given in this thread.

-- Hofmator