Java String到短哈希码

Question

I want to "implement" a hash function from Strings to shorts, using the java standard hashCode() function of String object. I came up with the following simple implementation:

static short shortHashCode(String str)
{
   int strHashCode = str.hashCode();
   short shorterHashCode = (short) (strHashCode % Short.MAX_VALUE);
   return shorterHashCode;
}

1. Is my shortHashCode function a good hash function? Meaning is the chance of collisions small (chance that two different Strings will have the same hash code close to 1/Short.MAX_VALUE) ?
2. Is there a better way to implement hash function from Strings to shorts?

Answer 1

(short) (strHashCode % Short.MAX_VALUE);

is losing information unnecessarily.

 (short) (strHashCode % ((Short.MAX_VALUE + 1) << 1));

would not, but would be equivalent anyway to

 (short) strHashCode

since casting an integral type to a smaller integral type just truncates the most significant bits.

---

It also assumes that all bits have the same entropy, which may not be true. You could try and spread the entropy around:

 (short) (strHashCode ^ (strHashCode >>> 16))

which XORs the high 16 bits with the low 16 bits.

---

Meaning is the chance of collisions small (chance that two different Strings will have the same hash code close to 1/Short.MAX_VALUE) ?

java.lang.String.hashCode is not a cryptographically strong hash function , so it only has that property if an attacker can't control one or both inputs to force a collision.

If you expose it to strings from an untrusted source, you might see a much higher rate of hash collisions, possibly allowing an attacker to deny service.

Also, it is designed to tradeoff a small increase in collision rate for better performance, and cross-version stability. There are better string hashing functions out there.