zoukankan      html  css  js  c++  java
  • 改进版的合并排序

    下面是改进版的合并排序,下面是jdk1.7的源码部分

    执行步骤如下:

    (1)如果比较的长度小于INSERTIONSORT_THRESHOLD插入排序的阈值,直接调用传统的插入排序进行比较

    (2)当大于插入排序的阈值时,采用合并排序算法,这里有个改进的地方,红色加亮部分,如果已经排好序的,不再进行比较,而是直接复制过去,提高效率

     private static void mergeSort(Object[] src,
                                      Object[] dest,
                                      int low, int high, int off,
                                      Comparator c) {
            int length = high - low;

            // Insertion sort on smallest arrays
            if (length < INSERTIONSORT_THRESHOLD) {
                for (int i=low; i<high; i++)
                    for (int j=i; j>low && c.compare(dest[j-1], dest[j])>0; j--)
                        swap(dest, j, j-1);
                return;
            }

            // Recursively sort halves of dest into src
            int destLow  = low;
            int destHigh = high;
            low  += off;
            high += off;
            int mid = (low + high) >>> 1;
            mergeSort(dest, src, low, mid, -off, c);
            mergeSort(dest, src, mid, high, -off, c);
       //改进地方
            // If list is already sorted, just copy from src to dest.  This is an
            // optimization that results in faster sorts for nearly ordered lists.
            if (c.compare(src[mid-1], src[mid]) <= 0) {
               System.arraycopy(src, low, dest, destLow, length);
               return;
            }

            // Merge sorted halves (now in src) into dest
            for(int i = destLow, p = low, q = mid; i < destHigh; i++) {
                if (q >= high || p < mid && c.compare(src[p], src[q]) <= 0)
                    dest[i] = src[p++];
                else
                    dest[i] = src[q++];
            }
        }

  • 相关阅读:
    网线帘幕动画
    图片缩放/旋转/平移/设置分辨率
    贝塞尔样条
    线性梯度画刷
    画七彩五角星
    kafka安装
    在windows远程提交任务给Hadoop集群(Hadoop 2.6)
    把Spark SQL的metadata存储到mysql
    使用IDEA开发SPARK提交remote cluster执行
    Netty的Channel
  • 原文地址:https://www.cnblogs.com/wzyxidian/p/5215203.html
Copyright © 2011-2022 走看看