Is string interning really useful?

Question

I was having a conversation about strings and various languages a while back, and the topic of string interning came up. Apparently Java and the .NET framework do this auto

Answer 1

The a.equals(b) is very fast for random strings. Its is only slow for Strings which are long and the same (or almost the same)

Random rand = new Random(1);
String[] list = new String[2000];
for(int i=0;i<list.length;i++)
    list[i] = "1234567"+Long.toString(rand.nextInt(36*37), 36); // semi random
int count = 0;
long start = System.nanoTime();
for(int i=0;i<list.length;i++)
    for(int j=0;j<list.length;j++)
        if (list[i].equals(list[j]))
            count++;
long time = System.nanoTime() - start;
System.out.printf("The average time for equals() was %,d ns.%n", time/list.length/list.length);

on a 2.3 GHz laptop prints

The average time for equals() was 19 ns.

If you intern() the first value and have to intern() one value to do the comparison

       if (list[i] == list[j].intern())

prints

The average time for equals() was 258 ns.

This is a common case as you often have one value you know is interned and a second which is input and is not intern'ed.

if you only use intern'ed Strings and == it, and don't count the cost, prints

The average time for equals() was 4 ns.

Which is many times faster if you are doing millions of comparison. However for a small number of comparisons, you are saving 8 ns but could be costing 250 ns more.

It may just be simpler to avoid intern() and use equals().