Mutable objects and hashCode

问题

Have the following class:

public class Member {
private int x;
private long y;
private double d;

public Member(int x, long y, double d) {
    this.x = x;
    this.y = y;
    this.d = d;
}

@Override
public int hashCode() {
    final int prime = 31;
    int result = 1;
    result = prime * result + x;
    result = (int) (prime * result + y);
    result = (int) (prime * result + Double.doubleToLongBits(d));
    return result;
}

@Override
public boolean equals(Object obj) {
    if (this == obj) {
        return true;
    }
    if (obj instanceof Member) {
        Member other = (Member) obj;
        return other.x == x && other.y == y
                && Double.compare(d, other.d) == 0;
    }
    return false;
}

public static void main(String[] args) {
    Set<Member> test = new HashSet<Member>();
    Member b = new Member(1, 2, 3);
    test.add(b);
    System.out.println(b.hashCode());
    b.x = 0;
    System.out.println(b.hashCode());
    Member first = test.iterator().next();
    System.out.println(test.contains(first));
    System.out.println(b.equals(first));
           System.out.println(test.add(first));

}

}

It produces the following results:
30814 29853 false true true

Because the hashCode depends of the state of the object it can no longer by retrieved properly, so the check for containment fails. The HashSet in no longer working properly. A solution would be to make Member immutable, but is that the only solution? Should all classes added to HashSets be immutable? Is there any other way to handle the situation?

Regards.

回答1:

Objects in hashsets should either be immutable, or you need to exercise discipline in not changing them after they've been used in a hashset (or hashmap).

In practice I've rarely found this to be a problem - I rarely find myself needing to use complex objects as keys or set elements, and when I do it's usually not a problem just not to mutate them. Of course if you've exposed the references to other code by this time, it can become harder.

回答2:

Yes. While maintaining your class mutable, you can compute the hashCode and the equals methods based on immutable values of the class ( perhaps a generated id ) to adhere to the hashCode contract defined in Object class:

Whenever it is invoked on the same object more than once during an execution of a Java application, the hashCode method must consistently return the same integer, provided no information used in equals comparisons on the object is modified. This integer need not remain consistent from one execution of an application to another execution of the same application.

If two objects are equal according to the equals(Object) method, then calling the hashCode method on each of the two objects must produce the same integer result.

It is not required that if two objects are unequal according to the equals(java.lang.Object) method, then calling the hashCode method on each of the two objects must produce distinct integer results. However, the programmer should be aware that producing distinct integer results for unequal objects may improve the performance of hashtables.

Depending on your situation this may be easier or not.

class Member { 
    private static long id = 0;

    private long id = Member.id++;
    // other members here... 


    public int hashCode() { return this.id; }
    public boolean equals( Object o ) { 
        if( this == o ) { return true; }
        if( o instanceOf Member ) { return this.id == ((Member)o).id; }
        return false;
     }
     ...
 }

If you need a thread safe attribute, you may consider use: AtomicLong instead, but again, it depends on how are you going to use your object.

回答3:

Jon Skeet has listed all alternatives. As for why the keys in a Map or Set must not change:

The contract of a Set implies that at any time, there are no two objects o1 and o2 such that

o1 != o2 && set.contains(o1) && set.contains(o2) && o1.equals(o2)

Why that is required is especially clear for a Map. From the contract of Map.get():

More formally, if this map contains a mapping from a key k to a value v such that (key==null ? k==null : key.equals(k)), then this method returns v, otherwise it returns null. (There can be at most one such mapping.)

Now, if you modify a key inserted into a map, you might make it equal to some other key already inserted. Moreover, the map can not know that you have done so. So what should the map do if you then do map.get(key), where key is equal to several keys in the map? There is no intuitive way to define what that would mean - chiefly because our intuition for these datatypes is the mathematical ideal of sets and mappings, which don't have to deal with changing keys, since their keys are mathematical objects and hence immutable.

回答4:

As already mentioned, one can accept the following three solutions:

Use immutable objects; even when your class is mutable, you may use immutable identities on your hashcode implementation and equals checking, eg an ID-like value.
Similarly to the above, implement add/remove to get a clone of the inserted object, not the actual reference. HashSet does not offer a get function (eg to allow you alter the object later on); thus, you are safe there won't exist duplicates.
Exercise discipline in not changing them after they've been used, as @Jon Skeet suggests

But, if for some reason you really need to modify objects after being inserted to a HashSet, you need to find a way of "informing" your Collection with the new changes. To achieve this functionality:

You can use the Observer design pattern, and extend HashSet to implement the Observer interface. Your Member objects must be Observable and update the HashSet on any setter or other method that affects hashcode and/or equals.

Note 1: Extending 3, using 4: we may accept alterations, but those that do not create an already existing object (eg I updated a user's ID, by assigning a new ID, not setting it to an existing one). Otherwise, you have to consider the scenario where an object is transformed in such a way that is now equal to another object already existing in the Set. If you accept this limitation, 4th suggestion will work fine, else you must be proactive and define a policy for such cases.

Note 2: You have to provide both previous and current states of the altered object on your update implementation, because you have to initially remove the older element (eg use getClone() before setting new values), then add the object with the new state. The following snippet is just an example implementation, it needs changes based on your policy of adding a duplicate.

@Override
public void update(Observable newItem, Object oldItem) {
    remove(oldItem);
    if (add(newItem))
        newItem.addObserver(this);
}

I've used similar techniques on projects, where I require multiple indices on a class, so I can look up with O(1) for Sets of objects that share a common identity; imagine it as a MultiKeymap of HashSets (this is really useful, as you can then intersect/union indices and work similarly to SQL-like searching). In such cases I annotate methods (usually setters) that must fireChange-update each of the indices when a significant change occurs, so indices are always updated with the latest states.

回答5:

Theoretically (and more often than not practically too) your class either:

has a natural immutable identity that can be inferred from a subset of its fields, in which case you can use those fields to generate the hashCode from.
has no natural identity, in which case using a Set to store them is unnecessary, you could just as well use a List.

回答6:

Never change 'hashable field" after putting in hash based container.

As if you (Member) registered your phone number (Member.x) in yellow page(hash based container), but you changed your number, then no one can find you in the yellow page any more.

来源：https://stackoverflow.com/questions/4718009/mutable-objects-and-hashcode

标签

java

immutability

hashcode

hashset