Timing analysis on HashSet and TreeSet Help

Question

Need help with this problem. I need to test two implementations of the java.util.Set interface for time efficiency. The implementations are java.util.HashSet and java.util.TreeSet. The code to construct each follows like this:
import java.util.*;
....main()
Set s = new HashSet();
// test
s = new TreeSet();
// test
Sample Output: (HashSet versus TreeSet: 1000000 insertions timed every 100 insertions. )
Test run for HashSet
     Total elapsed time (ms) for HashSet: 12137
     Average time (ms) to insert 100 elements: 1.2137
Test run for TreeSet
     Total elapsed time (ms) for TreeSet: 14607
     Average time (ms) to insert 100 elements: 1.4607
CODING DETAILS:
1. What to insert? Randomly generated integers. Here's a code slice to illustrate:
Set s = new HashSet();
     Random r = new Random();
     long start = new Date().getTime(),
     mark = start,
     stop = 0;
     for (int i = 0; i < tries; i++) {      // tries is 1,000,000
     s.add(new Integer(r.nextInt()));
     if ((i % segment) == 0) {          // segment is 100
     stop = new Date().getTime();
     long diff = stop - mark;
2. Need to save the elapsed time per 100 insertions. It can be generated by adding up the 100-insertions times. Like in the above code, the variable diff records the time per 100 insertions, and
it save the different values of diff.

Answer 1

Are you adding up the time intervals for every 100 insertions to get the total time it took? The current time provided by the runtime may not be accurate due to some caching done in the OS level. It can be as bad as 50 ms (in windows 98, I think). So adding up such short time intervals can give you inacurate results. Also, you should use System.currentTimeMillis() rathen than new Date().getTime()
Theoritically, Treeset gives you guranteed O(logN) while HashSet gives you O(1) as long as you have a decent hashing algorithm. Since, you are getting the int from a pseudo-random genetor (and Integer.hashcode() returns the int value it represents), HashSet should perorm better for such large number of data.
Here is a slightly changed version of your code,
import java.util.*;
public class TreeHashPlay {
public static void main(String[] args) {
double hs = measure(new HashSet());
double ts = measure(new TreeSet());
System.out.println("ration TreeSet/HashSet: "+ ts/hs);
static double measure(Set set) {
long end = 0;
long start;
final int UP = 1000000;
Random rand = new Random();
start = System.currentTimeMillis();
for(int i = 0; i < UP; i++) {
set.add(new Integer(rand.nextInt()));
end = System.currentTimeMillis();
long duration = end - start;
System.out.println("impl: "+set.getClass()+" total time: "+ duration+" total item: "+ UP
+" average time: "+(double)duration/UP);
return (double)duration/UP;
}//class ends
and Here is a sample output:
impl: class java.util.HashSet total time: 11530 total item: 1000000 average time: 0.01153
impl: class java.util.TreeSet total time: 18460 total item: 1000000 average time: 0.01846
ratio TreeSet/HashSet: 1.6010407632263661
Hope it helps.