我們都知道在Java 8 API添加了一個(gè)新的抽象稱為流Stream，可以讓你以一種聲明的方式處理數(shù)據(jù)。Stream 使用一種類似用 SQL 語(yǔ)句從數(shù)據(jù)庫(kù)查詢數(shù)據(jù)的直觀方式來(lái)提供一種對(duì) Java 集合運(yùn)算和表達(dá)的高階抽象。Stream API可以極大提高Java程序員的生產(chǎn)力，讓程序員寫出高效率、干凈、簡(jiǎn)潔的代碼。這種風(fēng)格將要處理的元素集合看作一種流， 流在管道中傳輸， 并且可以在管道的節(jié)點(diǎn)上進(jìn)行處理， 比如篩選， 排序，聚合等。元素流在管道中經(jīng)過(guò)中間操作（intermediate operation）的處理，最后由最終操作(terminal operation)得到前面處理的結(jié)果。
通過(guò)查看API能夠看到Java8 API為我們提供了Stream和parallelStream兩個(gè)不同的方法，那么同樣是流處理，這兩個(gè)方法又有什么區(qū)別呢？首先我們來(lái)看看以下的代碼：

public static void main(String[] args) {
        List<Integer> numberList = Arrays.asList(1,2,3,4,5,6,7,8,9);
        System.out.println("運(yùn)行結(jié)果：");
        // stream method
        numberList.stream().forEach(number -> {
            System.out.print(String.format("%d ",number));
        });
        System.out.println("\r");
        // parallelStream method
        numberList.parallelStream().forEach(number -> {
            System.out.print(String.format("%d ",number));
        });
        System.out.println("\r");
        // parallelStream method
        numberList.parallelStream().forEachOrdered(number -> {
            System.out.print(String.format("%d ",number));
        });
        System.out.println("\r");
}

通過(guò)多次運(yùn)行上述代碼，我們可以發(fā)現(xiàn)，通過(guò)parallelStream方法迭代集合，每次輸出的結(jié)果都不一樣，而通過(guò)steam方法或parallelStream方法并以forEachOrdered方式，每次執(zhí)行輸出的結(jié)果都是一樣的，并且順序符合集合元素的存放順序。
那么，為什么會(huì)造成這樣的結(jié)果差異呢，難道parallelStram是采用多線程并行的方式運(yùn)行？于是，我們進(jìn)一步修改下我們的代碼來(lái)驗(yàn)證一下猜測(cè)。

public static void main(String[] args) {
        System.out.println("運(yùn)行結(jié)果：");
        List<Integer> numberList = Arrays.asList(1,2,3,4,5,6,7,8,9);
        // stream method
        numberList.stream().forEach(number -> {
            System.out.println(String.format("Stream The Current Thread's ID is %d and output number %d ",Thread.currentThread().getId(),number));
        });
        System.out.println("\r");
        // parallelStream method
        numberList.parallelStream().forEach(number -> {
            System.out.println(String.format("ParallelStream The Current Thread's ID is %d and output number %d ",Thread.currentThread().getId(),number));
        });
        System.out.println("\r");
        // parallelStream method
        numberList.parallelStream().forEachOrdered(number -> {
            System.out.println(String.format("ParallelStream forEach Ordered The Current Thread's ID is %d and output number %d ",Thread.currentThread().getId(),number));
        });
        System.out.println("\r");
}

修改后代碼運(yùn)行結(jié)果如下：

運(yùn)行結(jié)果：
Stream The Current Thread's ID is 1 and output number 1 
Stream The Current Thread's ID is 1 and output number 2 
Stream The Current Thread's ID is 1 and output number 3 
Stream The Current Thread's ID is 1 and output number 4 
Stream The Current Thread's ID is 1 and output number 5 
Stream The Current Thread's ID is 1 and output number 6 
Stream The Current Thread's ID is 1 and output number 7 
Stream The Current Thread's ID is 1 and output number 8 
Stream The Current Thread's ID is 1 and output number 9 

ParallelStream The Current Thread's ID is 1 and output number 6 
ParallelStream The Current Thread's ID is 19 and output number 9 
ParallelStream The Current Thread's ID is 18 and output number 1 
ParallelStream The Current Thread's ID is 15 and output number 2 
ParallelStream The Current Thread's ID is 17 and output number 4 
ParallelStream The Current Thread's ID is 14 and output number 8 
ParallelStream The Current Thread's ID is 13 and output number 3 
ParallelStream The Current Thread's ID is 16 and output number 7 
ParallelStream The Current Thread's ID is 1 and output number 5 

ParallelStream forEach Ordered The Current Thread's ID is 15 and output number 1 
ParallelStream forEach Ordered The Current Thread's ID is 14 and output number 2 
ParallelStream forEach Ordered The Current Thread's ID is 14 and output number 3 
ParallelStream forEach Ordered The Current Thread's ID is 14 and output number 4 
ParallelStream forEach Ordered The Current Thread's ID is 14 and output number 5 
ParallelStream forEach Ordered The Current Thread's ID is 14 and output number 6 
ParallelStream forEach Ordered The Current Thread's ID is 14 and output number 7 
ParallelStream forEach Ordered The Current Thread's ID is 14 and output number 8 
ParallelStream forEach Ordered The Current Thread's ID is 14 and output number 9 

Disconnected from the target VM, address: '127.0.0.1:52976', transport: 'socket'

Process finished with exit code 0

通過(guò)上面的運(yùn)行結(jié)果，我們可以看到通過(guò)ParallelStream方法迭代的方法，是采用多線程的，可以看過(guò)每次輸出都是不同的線程ID，而ParallelStream(). forEach Ordered是在多線程的基礎(chǔ)上，保證了數(shù)據(jù)的順序輸出。到此，我們驗(yàn)證了我們的猜測(cè)ParallelStream方法是多線程的，而關(guān)于線程是否并行的驗(yàn)證，我們需進(jìn)一步修改下我們的代碼，于是有了下面的代碼：

 public static void main(String[] args) throws InterruptedException {
        System.out.println("運(yùn)行結(jié)果：");
        List<Integer> numberList = Arrays.asList(1,2,3,4,5,6,7,8,9);
        //for
        Long forBegin = System.currentTimeMillis();
        for(Integer number : numberList){
            //System.out.println(String.format("For The Current Thread's ID is %d and output number %d ",Thread.currentThread().getId(),number));
            Thread.sleep(1000);
        }
        System.out.println(String.format("For execute time cost %d ms",System.currentTimeMillis()-forBegin));
        System.out.println("\r");
        // stream method
        Long streamBegin = System.currentTimeMillis();
        numberList.stream().forEach(number -> {
            //System.out.println(String.format("Stream The Current Thread's ID is %d and output number %d ",Thread.currentThread().getId(),number));
            try {
                Thread.sleep(1000);
            } catch (InterruptedException e) {
                e.printStackTrace();
            }
        });
        System.out.println(String.format("Stream execute time cost %d ms",System.currentTimeMillis()-streamBegin));
        System.out.println("\r");
        // parallelStream method
        Long parallelStreamBegin = System.currentTimeMillis();
        numberList.parallelStream().forEach(number -> {
            //System.out.println(String.format("ParallelStream The Current Thread's ID is %d and output number %d ",Thread.currentThread().getId(),number));
            try {
                Thread.sleep(1000);
            } catch (InterruptedException e) {
                e.printStackTrace();
            }
        });
        System.out.println(String.format("ParallelStream execute time cost %d ms",System.currentTimeMillis()-parallelStreamBegin));
        System.out.println("\r");
        // parallelStream method
        Long parallelStreamForEachOrderBegin = System.currentTimeMillis();
        numberList.parallelStream().forEachOrdered(number -> {
            //System.out.println(String.format("ParallelStream forEachOrdered The Current Thread's ID is %d and output number %d ",Thread.currentThread().getId(),number));
            try {
                Thread.sleep(1000);
            } catch (InterruptedException e) {
                e.printStackTrace();
            }
        });
        System.out.println(String.format("ParallelStream forEachOrdered execute time cost %d ms",System.currentTimeMillis()-parallelStreamForEachOrderBegin));
        System.out.println("\r");
}

這里我們加入了傳統(tǒng)的for循環(huán)迭代方式，加入一起比較，由于要體現(xiàn)多線程并行的優(yōu)勢(shì)，這里我們將每次循環(huán)里加入線程休眠1秒鐘，運(yùn)行后的結(jié)果如下：

運(yùn)行結(jié)果：
For execute time cost 9032 ms

Stream execute time cost 9079 ms

ParallelStream execute time cost 2011 ms

ParallelStream forEachOrdered execute time cost 9037 ms

通過(guò)運(yùn)行結(jié)果，我們可以看到parallelStream().forEach方式耗時(shí)最短，而另外其他3種方式運(yùn)行的耗時(shí)都幾乎接近。因此，我們可以斷定我們的猜測(cè)是正確的，parallelStream().forEach是通過(guò)多線程并行的方式來(lái)執(zhí)行我們的代碼，而parallelStream(). forEachOrdered也是采用多線程，但由于加入了順序執(zhí)行約束，故程序是采用多線程同步的方式運(yùn)行的，最終耗時(shí)與for、stream兩種單線程執(zhí)行的耗時(shí)接近，但parallelStream(). forEachOrdered由于是多線程，與for、stream兩種單線程的方式相比，優(yōu)勢(shì)在于很好的利用了CPU多核的資源。感興趣的同學(xué)可以通過(guò)以下代碼查看CPU的核數(shù)，并通過(guò)jstack dump出堆棧來(lái)查看線程對(duì)CPU使用的情況。

System.out.println("系統(tǒng)一共有"+Runtime.getRuntime().availableProcessors()+"個(gè)cpu");