Take n random elements from a List?

问题:

How can I take n random elements from an ArrayList<E>? Ideally, I'd like to be able to make successive calls to the take() method to get another x elements, without replacement.

回答1:

Two main ways.

1. Use Random#nextInt(int):

   List<Foo> list = createItSomehow();
   Random random = new Random();
   Foo foo = list.get(random.nextInt(list.size()));
   

   It's however not guaranteed that successive n calls returns unique elements.

2. Use Collections#shuffle():

   List<Foo> list = createItSomehow();
   Collections.shuffle(list);
   Foo foo = list.get(0);
   

   It enables you to get n unique elements by an incremented index (assuming that the list itself contains unique elements).

---

In case you're wondering if there's a Java 8 Stream approach; no, there isn't a built-in one. There's no such thing as Comparator#randomOrder() in standard API (yet?). You could try something like below while still satisfying the strict Comparator contract (although the distribution is pretty terrible):

List<Foo> list = createItSomehow();
int random = new Random().nextInt();
Foo foo = list.stream().sorted(Comparator.comparingInt(o -> System.identityHashCode(o) ^ random)).findFirst().get();

Better use Collections#shuffle() instead.

回答2:

Most of the proposed solutions till now suggest either a full list shuffle or successive random picking by checking uniqueness and retry if required.

But, we can take advantage of the Durstenfeld's algorithm (the most popular Fisher-Yates variant in our days).

Durstenfeld's solution is to move the "struck" numbers to the end of the list by swapping them with the last unstruck number at each iteration.

Due to the above, we don't need to shuffle the whole list, but run the loop for as many steps as the number of elements required to return. The algorithm ensures that the last N elements at the end of the list are 100% random if we used a perfect random function.

Among the many real-world scenarios where we need to pick a predetermined (max) amount of random elements from arrays/lists, this optimized method is very useful for various card games, such as Texas Poker, where you a-priori know the number of cards to be used per game; only a limited number of cards is usually required from the deck.

public static <E> List<E> pickNRandomElements(List<E> list, int n, Random r) {
    int length = list.size();

    if (length < n) return null;

    //We don't need to shuffle the whole list
    for (int i = length - 1; i >= length - n; --i)
    {
        Collections.swap(list, i , r.nextInt(i + 1));
    }
    return list.subList(length - n, length);
}

public static <E> List<E> pickNRandomElements(List<E> list, int n) {
    return pickNRandomElements(list, n, ThreadLocalRandom.current());
}

回答3:

If you want to successively pick n elements from the list and be able to do so without replacement over and over and over again, you are probably best of randomly permuting the elements, then taking chunks off in blocks of n. If you randomly permute the list you guarantee statistical randomness for each block you pick out. Perhaps the easiest way to do this would be to use Collections.shuffle.

回答4:

Simple and clear

   // define ArrayList to hold Integer objects
    ArrayList<Integer> arrayList = new ArrayList<>();

    for (int i = 0; i < maxRange; i++) {
        arrayList.add(i + 1);
    }

    // shuffle list
    Collections.shuffle(arrayList);

    // adding defined amount of numbers to target list
    ArrayList<Integer> targetList = new ArrayList<>();
    for (int j = 0; j < amount; j++) {
        targetList.add(arrayList.get(j)); 
    }

    return targetList;

回答5:

A fair way to do this is to go through the list, on the nth iteration calculating the probability of whether or not to pick the nth element, which is essentially the fraction of the number of items you still need to pick over the number of elements available in the rest of the list. For example:

public static <T> T[] pickSample(T[] population, int nSamplesNeeded, Random r) {
  T[] ret = (T[]) Array.newInstance(population.getClass().getComponentType(),
                                    nSamplesNeeded);
  int nPicked = 0, i = 0, nLeft = population.length;
  while (nSamplesNeeded > 0) {
    int rand = r.nextInt(nLeft);
    if (rand < nSamplesNeeded) {
      ret[nPicked++] = population[i];
      nSamplesNeeded--;
    }
    nLeft--;
    i++;
  }
  return ret;
}

(This code copied from a page I wrote a while ago on picking a random sample from a list.)

回答6:

Use the following class:

import java.util.Enumeration;
import java.util.Random;

public class RandomPermuteIterator implements Enumeration<Long> {
    int c = 1013904223, a = 1664525;
    long seed, N, m, next;
    boolean hasNext = true;

    public RandomPermuteIterator(long N) throws Exception {
        if (N <= 0 || N > Math.pow(2, 62)) throw new Exception("Unsupported size: " + N);
        this.N = N;
        m = (long) Math.pow(2, Math.ceil(Math.log(N) / Math.log(2)));
        next = seed = new Random().nextInt((int) Math.min(N, Integer.MAX_VALUE));
    }

    public static void main(String[] args) throws Exception {
        RandomPermuteIterator r = new RandomPermuteIterator(100);
        while (r.hasMoreElements()) System.out.print(r.nextElement() + " ");
    }

    @Override
    public boolean hasMoreElements() {
        return hasNext;
    }

    @Override
    public Long nextElement() {
        next = (a * next + c) % m;
        while (next >= N) next = (a * next + c) % m;
        if (next == seed) hasNext = false;
        return  next;
    }
}

回答7:

Keep selecting a random element and make sure you do not choose the same element again:

public static <E> List<E> selectRandomElements(List<E> list, int amount)
{
    // Avoid a deadlock
    if (amount >= list.size())
    {
        return list;
    }

    List<E> selected = new ArrayList<>();
    Random random = new Random();
    int listSize = list.size();

    // Get a random item until we got the requested amount
    while (selected.size() < amount)
    {
        int randomIndex = random.nextInt(listSize);
        E element = list.get(randomIndex);

        if (!selected.contains(element))
        {
            selected.add(element);
        }
    }

    return selected;
}

In theory this could run endlessly but in practise it is fine. The closer you get the whole original list the slower the runtime of this gets obviously but that is not the point of selecting a random sublist, is it?

回答8:

The following class retrieve N items from a list of any type. If you provide a seed then on each run it will return the same list, otherwise, the items of the new list will change on each run. You can check its behaviour my running the main methods.

import java.util.ArrayList;
import java.util.Arrays;
import java.util.Collections;
import java.util.List;
import java.util.Random;

public class NRandomItem<T> {
    private final List<T> initialList;

    public NRandomItem(List<T> list) {
        this.initialList = list;
    }

    /**
     * Do not provide seed, if you want different items on each run.
     * 
     * @param numberOfItem
     * @return
     */
    public List<T> retrieve(int numberOfItem) {
        int seed = new Random().nextInt();
        return retrieve(seed, numberOfItem);
    }

    /**
     * The same seed will always return the same random list.
     * 
     * @param seed,
     *            the seed of random item generator.
     * @param numberOfItem,
     *            the number of items to be retrieved from the list
     * @return the list of random items
     */
    public List<T> retrieve(int seed, int numberOfItem) {
        Random rand = new Random(seed);

        Collections.shuffle(initialList, rand);
        // Create new list with the number of item size
        List<T> newList = new ArrayList<>();
        for (int i = 0; i < numberOfItem; i++) {
            newList.add(initialList.get(i));
        }
        return newList;
    }

    public static void main(String[] args) {
        List<String> l1 = Arrays.asList("Foo", "Bar", "Baz", "Qux");
        int seedValue = 10;
        NRandomItem<String> r1 = new NRandomItem<>(l1);

        System.out.println(String.format("%s", r1.retrieve(seedValue, 2)));
    }
}

回答9:

As noted in other answers, Collections.shuffle is not very efficient when the source list is big, because of the copying. Here is a Java 8 one-liner that:

• efficient enough on random access lists like ArrayList if you don't need many elements from the source
• doesn't modify the source
• DOES NOT guarantee uniqueness, if it's not super important for you. If you pick 5 from a hundred, there's a very good chance the elements will be unique.

Code:

private static <E> List<E> pickRandom(List<E> list, int n) {
  return new Random().ints(n, 0, list.size()).mapToObj(list::get).collect(Collectors.toList());
}

However, for a list with no quick random access (like LinkedList) the complexity would be n*O(list_size).