在Kotlin代码中要慎用Java 8 的 Stream API

2020-05-08 16:35:52 浏览数 (1)

因为会有性能损失。

测试代码:

1.Java 中直接使用 Java 8 的 Stream API:

代码语言:javascript复制
package i;

import java.util.ArrayList;
import java.util.List;
import java.util.Random;

/**
 * @author: Jack
 * 2020-05-06 16:20
 */
public class StreamParallelDemo {
    public static void main(String[] args) {
        System.out.println(String.format("本计算机的核数:%d", Runtime.getRuntime().availableProcessors()));

        Random random = new Random();
        List<Integer> list = new ArrayList<>(100000000);

        for (int i = 0; i < 100000000; i  ) {
            list.add(random.nextInt(100));
        }

        long s = getCurrentTime();
        Integer ans = list.stream().reduce((a, b) -> a   b).get();
        System.out.println(ans);
        System.out.println(String.format("单线程计算耗时:%d ms", getCurrentTime() - s));

        s = getCurrentTime();
        ans = list.stream().parallel().reduce((a, b) -> a   b).get();
        System.out.println(ans);
        System.out.println(String.format("多线程计算耗时:%d ms", getCurrentTime() - s));

    }

    private static long getCurrentTime() {
        return System.currentTimeMillis();
    }
}

运行结果:

本计算机的核数:12 655269092 单线程计算耗时:3350 ms 655269092 多线程计算耗时:425 ms

2.Kotlin 使用 Java 8 的 Stream API:

代码语言:javascript复制
package i

import java.util.*

/**
 * @author: Jack
 * 2020-05-06 16:03
 */

fun main() {
    val n = 100000000
    val random = Random()
    val numbers = ArrayList<Int>(n)

    for (i in 0..n - 1) {
        numbers.add(random.nextInt(100))
    }

    run {
        val s = System.currentTimeMillis()
        val ans = singleSum(numbers)
        val t = System.currentTimeMillis()
        println(ans)
        println("singleSum Time:${t - s}ms")
    }
}

fun singleSum(numbers: List<Int>): Int {
    return numbers.stream().reduce { t: Int, u: Int -> t   u }.get()
}

运行结果:

655113317 singleSum Time:3509ms

代码语言:javascript复制
package i

import java.util.*

/**
 * @author: Jack
 * 2020-05-06 16:03
 */

fun main() {
    val n = 100000000
    val random = Random()
    val numbers = ArrayList<Int>(n)

    for (i in 0..n - 1) {
        numbers.add(random.nextInt(100))
    }

    run {
        val s = System.currentTimeMillis()
        val ans = parallelSum(numbers)
        val t = System.currentTimeMillis()
        println(ans)
        println("parallelSum Time:${t - s}ms")
    }
}

fun parallelSum(numbers: List<Int>): Int {
    return numbers.stream().parallel().reduce { t: Int, u: Int -> t   u }.get()
}

运行结果: 655149583 parallelSum Time:2911ms

看起来,Kotlin使用Java 8 的 parallel Stream API,似乎没有明显的性能提升(猜测:中间有Java-Kotlin互转带来的性能损耗?)。

3.直接使用Kotlin的 List的API:

代码语言:javascript复制
package i

import java.util.*

/**
 * @author: Jack
 * 2020-05-06 16:03
 */

fun main() {
    val n = 100000000
    val random = Random()
    val numbers = ArrayList<Int>(n)

    for (i in 0..n - 1) {
        numbers.add(random.nextInt(100))
    }

    run {
        val s = System.currentTimeMillis()
        val ans = listSum(numbers)
        val t = System.currentTimeMillis()
        println(ans)
        println("listSum Time:${t - s}ms")
    }
}

fun listSum(numbers: List<Int>): Int {
    return numbers.stream().parallel().reduce { t: Int, u: Int -> t   u }.get()
}

运行结果: 655344706 listSum Time:3235ms

4.使用Kotlin的Sequence API

代码语言:javascript复制
package i

import java.util.*

/**
 * @author: Jack
 * 2020-05-06 16:03
 */

fun main() {
    val n = 100000000
    val random = Random()
    val numbers = MutableList(n) { random.nextInt(100) }

    run {
        val s = System.currentTimeMillis()
        val ans = sequenceSum(numbers)
        val t = System.currentTimeMillis()
        println(ans)
        println("sequenceSum Time:${t - s}ms")
    }
}

fun sequenceSum(numbers: List<Int>): Int {
    return numbers.asSequence().reduce { t: Int, u: Int -> t   u }
}

运行结果:

655104311 sequenceSum Time:3176ms

思考题

为什么会有性能损失呢?

0 人点赞