问题
I have a case here where the switch statement contains about 40 cases of each returning a different configured object based on input. This method is shown as having too high cyclomatic complexity in the metrics and usually I would change this into a map of handler objects. But this switch sits in a piece of code where performance is all-important so I came up with the question of how a HashMap lookup and handler call compares to a switch block execution performance-wise. Anyone compared that yet? Is it worth considering? Or is there any faster lookup object for a finite number of int keys?
Cheers, Kai
回答1:
It seems that searching the values is slightly faster than switch approach. So if your code changes often focus on the hash approach, if it's polished and not going to change often you can use the switch approach without remorsals. If your application is in development by several programmers you should prefer the hash approach to prevent future bugs.
Switch version: 27.32191395 ns. Hash version: 26.98444367 ns.
Time gained with hash: 1.23%.
I tested it with the following code:
import java.util.Random;
/**
* @author ruslan.lopez
*/
public class CyclomaticVsHash {
private static final Random RNG = new Random();
static int[] myHash = { 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11,
12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28,
29, 30, 31, 32, 33, 34, 35, 36, 37, 38, 39, 40 };
/**
* @param args
* the command line arguments
*/
public static void main(String[] args) {
long iterations = 100000000;
warmUp(iterations);
System.out.println("Cycle1");
double individualTime = getAverageTimePerIterationc1(iterations);
iterations = 10000;
double totalTime = getTotalTimec1(iterations);
System.out.println("ns/iteration: " + individualTime);
System.out.println("Total time for " + iterations + " runs: "
+ totalTime);
System.out.println("Cycle2");
iterations = 100000000;
warmUp(iterations);
double individualTime1 = getAverageTimePerIterationc2(iterations);
iterations = 10000;
double totalTime1 = getTotalTimec2(iterations);
System.out.println("ns/iteration: " + individualTime1);
System.out.println("Total time for " + iterations + " runs: "
+ totalTime1);
}
public static void warmUp(long iterations) {
System.out.println("Starting warmup");
for (int i = 0; i < iterations; i++) {
runCycles();
runCycles1();
}
}
public static double getAverageTimePerIterationc1(long iterations) {
// test
System.out.println("Starting individual time test");
long timeTaken = 0;
for (int i = 0; i < iterations; i++) {
long startTime = System.nanoTime();
runCycles();
timeTaken += System.nanoTime() - startTime;
}
return (double) timeTaken / iterations;
}
public static long getTotalTimec1(long iterations) {
// test
System.out.println("Starting total time test");
long timeTaken = 0;
for (int i = 0; i < iterations; i++) {
long startTime = System.nanoTime();
runCycles();
timeTaken += System.nanoTime() - startTime;
}
return timeTaken;
}
public static double getAverageTimePerIterationc2(long iterations) {
// test
System.out.println("Starting individual time test");
long timeTaken = 0;
for (int i = 0; i < iterations; i++) {
long startTime = System.nanoTime();
runCycles1();
timeTaken += System.nanoTime() - startTime;
}
return (double) timeTaken / iterations;
}
public static long getTotalTimec2(long iterations) {
// test
System.out.println("Starting total time test");
long timeTaken = 0;
for (int i = 0; i < iterations; i++) {
long startTime = System.nanoTime();
runCycles1();
timeTaken += System.nanoTime() - startTime;
}
return timeTaken;
}
private static void runCycles() {
Integer num = RNG.nextInt();
int newnum;
switch (num) {
case 1:
newnum = num * 1;
break;
case 2:
newnum = num * 2;
break;
case 3:
newnum = num * 3;
break;
case 4:
newnum = num * 4;
break;
case 5:
newnum = num * 5;
break;
case 6:
newnum = num * 6;
break;
case 7:
newnum = num * 7;
break;
case 8:
newnum = num * 8;
break;
case 9:
newnum = num * 9;
break;
case 10:
newnum = num * 10;
break;
case 11:
newnum = num * 11;
break;
case 12:
newnum = num * 12;
break;
case 13:
newnum = num * 13;
break;
case 14:
newnum = num * 14;
break;
case 15:
newnum = num * 15;
break;
case 16:
newnum = num * 16;
break;
case 17:
newnum = num * 17;
break;
case 18:
newnum = num * 18;
break;
case 19:
newnum = num * 19;
break;
case 20:
newnum = num * 20;
break;
case 21:
newnum = num * 21;
break;
case 22:
newnum = num * 22;
break;
case 23:
newnum = num * 23;
break;
case 24:
newnum = num * 24;
break;
case 25:
newnum = num * 25;
break;
case 26:
newnum = num * 26;
break;
case 27:
newnum = num * 7;
break;
case 28:
newnum = num * 28;
break;
case 29:
newnum = num * 29;
break;
case 30:
newnum = num * 30;
break;
case 31:
newnum = num * 31;
break;
case 32:
newnum = num * 32;
break;
case 33:
newnum = num * 33;
break;
case 34:
newnum = num * 34;
break;
case 35:
newnum = num * 35;
break;
case 36:
newnum = num * 36;
break;
case 37:
newnum = num * 37;
break;
case 38:
newnum = num * 38;
break;
case 39:
newnum = num * 39;
break;
default:
newnum = num * 40;
break;
}
}
private static void runCycles1() {
Integer num = RNG.nextInt();
int nwenum = num > 0 && num < 39
? myHash[num - 1]
: myHash[39];
}
}
回答2:
I'm pretty sure that the internal representation of a switch uses some kind of lookup table that would have similar structure to a HashMap, so my guess is that any differences are negligable.
I would say performance is not worth considering here, just go with whichever solution gives you the cleanest code.
来源：https://stackoverflow.com/questions/16177842/java-switch-cyclomatic-complexity-vs-performance