A comparison of implementations of rotate.

Rotator.java

import java.util.List;


public interface Rotator {
    <T> void ror(int distance, List<T> list);


    <T> void rol(int distance, List<T> list);
}

Rotators.java

import static java.util.Collections.reverse;

import java.util.Collections;
import java.util.List;

public class Rotators {
    public static Rotator collections() {
        return CollectionsRotator.INSTANCE;
    }

    private static enum CollectionsRotator implements Rotator {
        INSTANCE;
        @Override public <T> void rol(int distance, List<T> list) {
            ror(list.size() - (distance % list.size()), list);
        }


        @Override public <T> void ror(int distance, List<T> list) {
            Collections.rotate(list, distance);
        }


        @Override public String toString() {
            return "java.util.Collections.rotate(List<?>, int)";
        }
    }


    public static Rotator reverseThrice() {
        return ReverseRotator.INSTANCE;
    }

    private static enum ReverseRotator implements Rotator {
        INSTANCE;
        public <T> void ror(int distance, List<T> list) {
            rol(list.size() - (distance % list.size()), list);
        }


        public <T> void rol(int distance, List<T> list) {
            final int l = list.size();
            if (l <= 1) { return; }
            final int n = modular(distance, l);
            if (n == 0) { return; }
            reverse(list);
            reverse(list.subList(0, l - n));
            reverse(list.subList(l - n, l));
        }


        private int modular(int n, int base) {
            if (n < 0) {
                return (base + (n % base)) % base;
            } else {
                return (n % base);
            }
        }


        @Override public String toString() {
            return "java.util.Collections.reverse(List<?>) x3";
        }
    }


    private Rotators() {}
}

RotatorTest.java

import static org.junit.Assert.assertEquals;

import java.util.List;
import java.util.Random;

import org.junit.Test;

import com.google.common.base.Charsets;
import com.google.common.primitives.Chars;


public class RotatorTest {
    @Test public void performance() {
        final int length = 10000;
        final long[] jdk = new long[length];
        final long[] lib = new long[length];
        final String s;
        {
            final Random r = new Random();
            final byte[] bytes = new byte[length];
            r.nextBytes(bytes);
            s = new String(bytes, Charsets.ISO_8859_1);
        }
        for (int i = 0; i < s.length(); i++) {
            final List<Character> l1 = Chars.asList(s.toCharArray());
            final List<Character> l2 = Chars.asList(s.toCharArray());
            final Rotator collections = Rotators.collections();
            final Rotator reverse = Rotators.reverseThrice();
            {
                final long before = System.nanoTime();
                collections.ror(i, l1);
                final long after = System.nanoTime();
                jdk[i] = after - before;
            }
            {
                final long before = System.nanoTime();
                reverse.ror(i, l2);
                final long after = System.nanoTime();
                lib[i] = after - before;
            }
            assertEquals(l1, l2);
        }
        System.out.println("jdk:" + sum(jdk) + "\r\nlib:" + sum(lib));
    }


    private long sum(long[] ls) {
        long sum = 0;
        for (long l : ls) {
            sum += l;
        }
        return sum;
    }
}

The reverse thrice implementation is based on Jon Bentley's Programming Pearls, section 2.3:
http://books.google.com.br/books?id=kse_7qbWbjsC&lpg=PA13&ots=DgsVzSzN3z&hl=en&pg=PA14#v=onepage&q&f=true

Collections.rotate() seems to be 2.5 times faster.

Hmm, changing everything to be based on arrays, takes off most of the overhead, now the implementation based on reverse is 27% faster than the custom implementation of the jdk.