Misleading javadoc of CacheBuilder

[vermahim]

Hi,

Javadoc of CacheBuilder says it uses LRU as eviction policy, but my test says otherwise..

`private static void testCacheEvictionPolicy() {
int maxSz = 3;
LoadingCache<Integer, Integer> cache = CacheBuilder.newBuilder().maximumSize(maxSz).build(new CacheLoader<Integer, Integer>() {
@OverRide
public Integer load(Integer integer) throws Exception {
return integer;
}
});
callCache(cache, 3);
printCache(cache);
for (int i = 0; i < 1000; i++) {
callCache(cache, 1);
}
callCache(cache, 4);
printCache(cache);
}

private static void callCache(LoadingCache<Integer, Integer> cache, int n) {
    for (int i = 1; i <= n; i++) {
        try {
            cache.get(i);
        } catch (ExecutionException e) {
            throw new RuntimeException(e);
        }
    }

}

private static void printCache(LoadingCache<Integer, Integer> cache) {
    System.out.println("cache begin");
    for (Map.Entry<Integer, Integer> entry : cache.asMap().entrySet()) {
        System.out.println(entry.getKey() + " : " + entry.getValue());
    }
    System.out.println("cache end");
}`

output:

cache begin
3 : 3
2 : 2
1 : 1
cache end
cache begin
3 : 3
2 : 2
4 : 4
cache end

[lowasser]

What the cache does is LRU, but on a per-segment basis: the cache is divided into concurrencyLevel segments, each of which does LRU internally with a size cap of maximumSize/concurrencyLevel.

[kluever]

It's probably worth adding this clarification to the javadocs.