【java】由Zookeeper sourcecode联想到的

一、导火索

  在看Zookeeper sourcecode的时候，发现server端的NIO*Factory 在channel上获取完SelectionKey列表后，进行了一个shuffle的操作保证公平性，code as followed:

ArrayList<SelectionKey> selectedList = new ArrayList<SelectionKey>(selected);
Collections.shuffle(selectedList);
for (SelectionKey k : selectedList) {
  ....
}

 说实话，对于我这样的资深工程师来说以前从没用过这个方法，来看看它是干什么的

    /**
     * Randomly permutes the specified list using a default source of
     * randomness.  All permutations occur with approximately equal
     * likelihood.
     （大致就是说它可以随机的交换list的元素，所有交换的可能性大致是相等的）
     * If the specified list does not
     * implement the {@link RandomAccess} interface and is large, this
     * implementation dumps the specified list into an array before shuffling
     * it, and dumps the shuffled array back into the list.  This avoids the
     * quadratic behavior that would result from shuffling a "sequential
     * access" list in place.
     */

 Note that,如果list没有实现RandomAccess interface 或者list比较大，那么它会将list convert to 一个array然后进行shuffle，最后在dump回list里，为什么要这么搞呢，因为这样可以避免二次行为 （相信很多人实际操作中第一想法会直接loop list元素，然后随机交换元素，这就是base knowledge不足啊）

int size = list.size();
        if (size < SHUFFLE_THRESHOLD || list instanceof RandomAccess) {//list比较小或者实现了RandomAccess接口,则直接搞就行了.
            for (int i=size; i>1; i--)
                swap(list, i-1, rnd.nextInt(i));
        } else {
            Object arr[] = list.toArray();

            // Shuffle array
            for (int i=size; i>1; i--)
                swap(arr, i-1, rnd.nextInt(i));

            // Dump array back into list
            ListIterator it = list.listIterator();
            for (int i=0; i<arr.length; i++) {
                it.next();
                it.set(arr[i]);
            }
        }

 

why?

 

看看RandomAccess,我挑几句注释来说明

/**
 * Generic list algorithms are encouraged to check whether the given list is an <tt>instanceof</tt> this interface before applying an algorithm that would provide poor performance if it were applied to a sequential access list,
 * and to alter their behavior if necessary to guarantee acceptable
 * performance.
 * 
 * <p>It is recognized that the distinction between random and sequential
 * access is often fuzzy.  For example, some <tt>List</tt> implementations
 * provide asymptotically linear access times if they get huge, but constant
 * access times in practice.  Such a <tt>List</tt> implementation
 * should generally implement this interface.  As a rule of thumb, a
 * <tt>List</tt> implementation should implement this interface if,
 * for typical instances of the class, this loop:
 * <pre>
 *     for (int i=0, n=list.size(); i &lt; n; i++)
 *         list.get(i);
 * </pre>
 * runs faster than this loop:
 * <pre>
 *     for (Iterator i=list.iterator(); i.hasNext(); )
 *         i.next();
 * </pre>
 */

 JDK建议我们如果是random access list,则用for(int i=0;i<list.size();i++){} 如果是sequence access list则用Iterator。因为最佳的方式就是用list instanceof RandomAccess来决定具体的算法.

 

其实数据比较大的时候,二者的效率还是有一定差距的.

 

JDK的source code还是值得去细细品味的

评论

4 楼 teasp 2013-04-12  

楼主作为资深工程师，居然没用过这个方法？

3 楼 cectsky 2013-04-12  

huangdj 写道

Collections.<strong>shuffle</strong>(selectedList); 
================
Collections.<strong>shuffle</strong>这个语法结构，看不懂，请教

格式错误,已改

2 楼 huangdj 2013-04-12  

Collections.<strong>shuffle</strong>(selectedList); 
================
Collections.<strong>shuffle</strong>这个语法结构，看不懂，请教

1 楼 chinaagan 2013-04-12  

实现了RandomAccess，比如数组，数组访问当然比迭代快多了。。。