Shortcut for adding to List in a HashMap

Question

I often have a need to take a list of objects and group them into a Map based on a value contained in the object. Eg. take a list of Users and group by Country.

My code for this usually looks like:

Map<String, List<User>> usersByCountry = new HashMap<String, List<User>>();
for(User user : listOfUsers) {
    if(usersByCountry.containsKey(user.getCountry())) {
        //Add to existing list
        usersByCountry.get(user.getCountry()).add(user);

    } else {
        //Create new list
        List<User> users = new ArrayList<User>(1);
        users.add(user);
        usersByCountry.put(user.getCountry(), users);
    }
}

However I can't help thinking that this is awkward and some guru has a better approach. The closest I can see so far is the MultiMap from Google Collections.

Are there any standard approaches?

Thanks!

Answer 1

No one comes to mind as far. I'd just optimize it as follows:

Map<String, List<User>> usersByCountry = new HashMap<String, List<User>>();
for(User user : listOfUsers) {
    List<User> users = usersByCountry.get(user.getCountry());
    if (users == null) {
        users = new ArrayList<User>();
        usersByCountry.put(user.getCountry(), users);
    }
    users.add(user);
}

Commons Collections has a LazyMap, but it's not parameterized. Google Collections doesn't have sort of a LazyMap or LazyList, but I recall that you could probably use a Function for this. Update: polygenelubricants has given a good example for that.

Answer 2

When I have to deal with a collection-valued map, I just about always wind up writing a little putIntoListMap() static utility method in the class. If I find myself needing it in multiple classes, I throw that method into a utility class. Static method calls like that are a bit ugly, but they're much cleaner than typing the code out every time. Unless multi-maps play a pretty central role in your app, IMHO it's probably not worth it to pull in another dependency.

Answer 3

It looks like your exact needs are met by LinkedHashMultimap in the GC library. If you can live with the dependencies, all your code becomes:

SetMultimap<String,User> countryToUserMap = LinkedHashMultimap.create();
// .. other stuff, then whenever you need it:
countryToUserMap.put(user.getCountry(), user);

insertion order is maintained (about all it looks like you were doing with your list) and duplicates are precluded; you can of course switch to a plain hash-based set or a tree set as needs dictate (or a list, though that doesn't seem to be what you need). Empty collections are returned if you ask for a country with no users, everyone gets ponies, etc - what I mean is, check out the API. It'll do a lot for you, so the dependency might be worth it.

Answer 4

A clean and readable way to add an element is the following:

String country = user.getCountry();
Set<User> users
if (users.containsKey(country))
{
    users = usersByCountry.get(user.getCountry());
}
else
{
    users = new HashSet<User>();
    usersByCountry.put(country, users);
}
users.add(user);

Note that calling containsKey and get is not slower than just calling get and testing the result for null.

Answer 5

Guava's Multimap really is the most appropriate data structure for this, and in fact, there is Multimaps.index(Iterable<V>, Function<? super V,K>) utility method that does exactly what you want: take an Iterable<V> (which a List<V> is), and apply the Function<? super V, K> to get the keys for the Multimap<K,V>.

Here's an example from the documentation:

For example,

  List<String> badGuys
      = Arrays.asList("Inky", "Blinky", "Pinky", "Pinky", "Clyde");
  Function<String, Integer> stringLengthFunction = ...;
  Multimap<Integer, String> index
      = Multimaps.index(badGuys, stringLengthFunction);
  System.out.println(index);

prints

 {4=[Inky], 5=[Pinky, Pinky, Clyde], 6=[Blinky]}

In your case you'd write a Function<User,String> userCountryFunction = ....

Answer 6

By using lambdaj you can obtain that result with just one line of code as it follows:

Group<User> usersByCountry = group(listOfUsers, by(on(User.class).getCountry()));

Lambdaj also offers lots of other features to manipulate collections with a very readable domain specific language.