Shortest Common Superstring: find shortest string that contains all given string fragments

Given some string fragments, I would like to find the shortest possible single string ("output string") that contains all the fragments. Fragments can overlap each other in the output string.

Example:

For the string fragments:

BCDA
AGF
ABC

The following output string contains all fragments, and was made by naive appending:

BCDAAGFABC

However this output string is better (shorter), as it employs overlaps:

ABCDAGF
^
ABC
^
BCDA
^
AGF

I'm looking for algorithms for this problem. It's not absolutely important to find the strictly shortest output string, but the shorter the better. I'm looking for an algorithm better than the obvious naive one that would try appending all permutations of the input fragments and removing overlaps (which would appear to be NP-Complete).

I've started work on a solution and it's proving quite interesting; I'd like to see what other people might come up with. I'll add my work-in-progress to this question in a while.

• The problem seems to be NP-complete. If so, you won't be able to find a polynomial algorithm for determining the shortest string at all, but there might be polynomial algorithms which give approximate (not the shortest possible) solutions. – superM Sep 25 '12 at 10:59
• This blog post regarding NP-Complete is nice: codinghorror.com/blog/2008/11/… – occulus Sep 25 '12 at 11:04
• The blog's really nice, I read it all the time ))) – superM Sep 25 '12 at 11:06
• @superM this is similar enough to traveling salesman (each string a city and cost between cities = some number-overlap) – ratchet freak Sep 25 '12 at 11:34
• @ratchet freak, it is _ you could give small cost between cities if they have more common letters, and the biggest cost when they don't have any common letter at all – superM Sep 25 '12 at 11:41