StringTokenizer和Split性能比较

朋友问我的问题，网上搜了下，都说StringTokenizer性能要好些，不过也有反对意见，还是自己试验下。

一百万以空格区分的字符串，拆分后重新组成不带空格的串。

import java.util.Random;
import java.util.StringTokenizer;
/**
 *
 * @author Jason Li 2014-6-3
 * StringTokenizer和Spilt速度比较
 *
 */
public class StringTest {

    public static void main(String[] args) {
        String str = buildString(1_000_000); //1.7新特性, 1000000
        long start;
        long end;

        System.out.println("-----------StringTokenizer start-----------");
        start = System.currentTimeMillis();
        StringTokenizer st = new StringTokenizer(str);
        StringBuilder sb = new StringBuilder();
        while(st.hasMoreTokens()){
            sb.append(st.nextToken());
        }
        end = System.currentTimeMillis();
        System.out.println("StringTokenizer time use:" + (end-start));

        System.out.println("-----------StringSpilt start-----------");
        start = System.currentTimeMillis();
        StringBuilder sb2 = new StringBuilder();
        String[] strs = str.split("\\s");
        for(String s: strs){
            sb2.append(s);
        }
        end = System.currentTimeMillis();
        System.out.println("StringSpilt time use:" + (end-start));

    }

    //建立一个长字符串，
    //其中有空格，以便拆分成length长度的n个字符串
    private static String buildString(int length) {
        StringBuilder sb = new StringBuilder();
        Random r =new Random();
        for (int i = 0; i <length;i++ ){
            for (int j = r.nextInt(10); j>0 ;j--){
                sb.append((char)('a' + r.nextInt(26)));
            }
            sb.append(" ");
        }
        return sb.toString();
    }
}

输出：

-----------StringTokenizer start-----------
StringTokenizer time use:131
-----------StringSpilt start-----------
StringSpilt time use:509

结论：

StringTokenizer确实更快些，至于为什么jdk里不推荐使用了，还要再研究