How to compute shortest unique prefixes of a set of strings?

Question

It's a pretty common algorithm in command line parsing . Given a set of predefined long option names -- compute the shortest prefix that uniquely identifies one of these options. So for instance, for the following options:

-help
-hostname
-portnumber
-name
-polymorphic

This would be the output:

-he
-ho
-por
-n
-pol

I'm thinking two possible ways of doing this -- either as a tree:

               *
             / | \
            /  |  \
           H   N   P
          / \      |
         E   O     O
                  / \
                 R   L

Or by searching for substrings:

for (String s : strings) {
   for (int i = 1; i < s.length(); s++) {
      if (search(strings,s.substring(0,i)) == 1) {
          result.add(s.substring(0,i);
          break;
      }
   }
}

So, question is:

1. Which would you go for?
2. Am I missing an obvious third way?

Answer 1

The "tree" solution is a special case (well, actually it's pretty general) of a Patricia trie .

The first will generally be faster to lookup. The memory considerations probably aren't relevant to your context, since it's not permanently used and you're only performing the "lookup" once.

Answer 2

I'd do the tree, looks fine.

You could build up a hash of every possible distinct substring.

Hashmap<String, String> validSubs = new Hashmap<String, String>();
HashSet<String> usedSubs = new HashSet<String>();

for (String option : options) {
  for(int i = 0; i <= option.length; i++) {
    String sub = option.substring(0, i);
    if(usedSubs.contains(sub)) {
      validSubs.remove(sub);
    } else {
      validSubs.add(sub, option);
      usedSubs.add(sub);
    }
  }
}

Answer 3

Oh, yeah, the most obvious missing answer is to use a library that already does this. How to parse command line arguments in Java?