出于好奇,我对时差进行了这个小测试。 事实证明,使用预编译模式比使用String.matches方法快5倍以上。
import java.util.regex.Pattern;
/**
* @author Rajind Ruparathna
*/
public class MatchesTest {
public static void main(String Args[]) {
String first = "@\\{message.headers\\.?([^\\}]*)\\}";
String second = "@\\{message.headers.wolla\\}";
long start, end, total;
float avg;
int NUM_OF_ITERATIONS = 100;
Pattern pattern = Pattern.compile(first);
total = 0;
start = 0;
end = 0;
avg = 0;
for (int i=0; i< NUM_OF_ITERATIONS; i++) {
start = System.nanoTime();
pattern.matcher(second).matches();
end = System.nanoTime();
total = total + (end - start);
}
avg = total/NUM_OF_ITERATIONS;
System.out.println("Duration pre compiled: " + avg);
total = 0;
start = 0;
end = 0;
avg = 0;
for (int i=0; i< NUM_OF_ITERATIONS; i++) {
start = System.nanoTime();
first.matches(second);
end = System.nanoTime();
total = total + (end - start);
}
avg = total/NUM_OF_ITERATIONS;
System.out.println("In place compiled: " + avg);
}
}
输出(纳秒):
Duration pre compiled: 4505.0
In place compiled: 44960.0
附言 此测试是快速而又肮脏的测试,可能不符合性能基准测试的惯例。 如果您想获得高度准确的结果,请使用微基准测试工具。