Optimize memory usage of a collection of Strings in Java

Question

I have a large number of name - value pairs (approx 100k) that I need to store in some sort of cache (say a hash map) where the value is a string with an average of about 30

Answer 1

Agreed with others on not using String.intern(): once you've put a string there, it will never go away. Look to early revisions of Xerces for why this is a bad idea.

A better solution is to use a WeakHashMap, wrapping the value in a WeakReference:

private Map> _map 
    = new WeakHashMap>();

public synchronized String intern(String str)
{
    WeakReference ref = _map.get(str);
    String s2 = (ref != null) ? ref.get() : null;
    if (s2 != null)
        return s2;
    str = new String(str);
    _map.put(str, new WeakReference(str));
    return str;
}

This code is from an article that I wrote on the Java reference objects. You'll find the explanation there.

EDIT: need to create a new string here (and I'll update the article) because the original might be a substring of a far larger character array. I thought that was fixed around JDK 1.3, but apparently not (at least not in 1.5).