Best way to test performance?

Question

As there is no pure java cycle test method, I constructed a simple class to test the basic performance boost of using a StringBuffer over a standard String object.
Is the following code sample the best solution to detecting performance? Are there any obvious errors within my code? Am i forgetting anything?
Thanks
Antony Ribot
// Antony Ribot - 22nd May 2002 - String / StringBuffer Performance
// Under tests of (Integer.MAX_VALUE/2) I got the following results:
// String: 724903
// StringBuffer: 145409 (4.98 times faster)
public class StringPerf {
private long sStartTime, sEndTime, sbStartTime, sbEndTime;
final static int ITERATIONS = Integer.MAX_VALUE/2;
public static void main (String[] args) {
StringPerf sp = new StringPerf();
private StringPerf() {
stringTest();
stringBufferTest();
showResults(); // output final results of test
private void stringTest() {
String s = new String();
sStartTime = System.currentTimeMillis();
for (int i = 0; i < ITERATIONS; i++) {
s = s + "";
sEndTime = System.currentTimeMillis();
showResults("string", sStartTime, sEndTime); // output quickview result
private void stringBufferTest() {
StringBuffer sb = new StringBuffer();
sbStartTime = System.currentTimeMillis();
for (int i = 0; i < ITERATIONS; i++) {
sb.append("");
sbEndTime = System.currentTimeMillis();
showResults("String b", sbStartTime, sbEndTime); / output quickview result
private void showResults() {
System.out.println("\n--------------------------------------------------");
System.out.println("String test: " + (sEndTime - sStartTime) );
System.out.println("String b test: " + (sbEndTime - sbStartTime) );
private void showResults(String s, long start, long end) {
System.out.println(s + ": " + (end - start));
}

Answer 1

I have now created the following extra code:
...But I'm sure I've gone the wrong way around to solve the problem - am I really testing the speed of the substring() and deleteAtChar() or are they made unreliable by the way I am reassigning the text string?
e.g: tmpString = TEXTBLOCK; tmpBufString.append(TEXTBLOCK);     
Is there a better way to test this (I'm sure there is). :) One solution could be to create an very long String / StringBuffer initially and then remove the outer loop all together?
Currently, results have shown the String.subString() to be 1.6 times faster.
Thanks
Antony Ribot
private String TEXTBLOCK = "Strings are constant; their values cannot be changed after they are created.";
private String tmpString = "";
private StringBuffer tmpBufString = new StringBuffer();
private void stringRemTest() {
sRemStartTime = System.currentTimeMillis();
for (int j = 0; j < ITERATIONS/100; j++) {
tmpString = TEXTBLOCK;          // <--- is this line getting in the way of my results??
for (int i = 0; i < TEXTBLOCK.length(); i++) {
tmpString = tmpString.substring(1);
sRemEndTime = System.currentTimeMillis();
showResults("StringRem", sRemStartTime, sRemEndTime);
private void stringBufferRemTest() {
sbRemStartTime = System.currentTimeMillis();
for (int j = 0; j < ITERATIONS/100; j++) {
tmpBufString.append(TEXTBLOCK);          // <--- is this line getting in the way of my results??
for (int i = 0; i < TEXTBLOCK.length(); i++) {
tmpBufString = tmpBufString.deleteCharAt(0);
sbRemEndTime = System.currentTimeMillis();
showResults("String b rem", sbRemStartTime, sbRemEndTime);
}