刚才看到这篇博客。
http://www.cnblogs.com/WindBlog/archive/2011/07/21/2112452.html大家一般都认为用Hash的办法。不过其实还有更高效的算法。
计算机图形学中,有个八叉树量化法,是用来从24颜色中查找重复颜色,并且进行计数归并的算法。它的算法思想是八叉树一共8层,每层都有8个节点,每一条路径从根到页正好对应8个位.
而颜色RGB 三个数字正好可以拼成一个由0-7数字组成的8位数字,于是正好可以用八叉树算法进行插入。
比如:
八叉树比较复杂,不过对于我们的这个需求来说,其实比较简单。
我们这个算法是10叉树,每层都保存了0-9 10个数字。层数就是手机号码的长度。
手机号的第一位就是第一层,只需遍历到最后一层即可判断是否重复。
于是让我们来实现这个十叉树。效率都和回复中的Linq做比较。
View Code
1 using System;
2 using System.Collections.Generic;
3 using System.Linq;
4 using System.Text;
5
6 namespace ConsoleApplication1
7 {
8 class Program
9 {
10 static void Main(string[] args)
11 {
12 //示例数组,存放手机号
13 string[] mobileArray = new string[100000];// { "13900001234", "13900001235", "13900001236", "13900001237", "13900001234" };
14
15 for (int i = 0; i < 100000; i++)
16 {
17 mobileArray[i] = "1390000"
18 + (i.ToString().Length > 4 ? i.ToString().Substring(0, 4) : (i.ToString() + "0000").Substring(0, 4));
19 }
20
21 ////linq语句来实现【select mobile from tmpTable group by mobile having count(*)>1】的效果
22 var selMobile = from n in mobileArray group n by n into g where g.Count() > 1 select g.Distinct();// select g;
23
24
25
26 System.Diagnostics.Stopwatch sw = new System.Diagnostics.Stopwatch();
27 sw.Reset();
28 sw.Start();
29 int count1 = 0;
30 //通过两层循环输出重复的手机号
31 foreach (var mobile in selMobile)
32 {
33 foreach (string multiMobile in mobile)
34 {
35 count1++;
36 //Console.WriteLine(multiMobile);
37 }
38 }
39
40 sw.Stop();
41
42 Console.WriteLine("Linq共有重复号" + count1+"耗时"+ sw.ElapsedTicks );
43
44 TenNodeTree tree = new TenNodeTree();
45 TenNodeTree tree2 = new TenNodeTree();
46
47 sw.Reset();
48 sw.Start();
49 int count2 = 0;
50 //mobileArray = new string[] { "13900001234", "13900001235", "13900001236", "13900001237", "13900001234", "13900001236" };
51 foreach (var item in mobileArray)
52 {
53 if (!tree.Add(item))
54 {
55 if (tree2.Add(item))
56 {
57 count2++;
58 }
59 }
60
61 }
62
63 sw.Stop();
64
65 Console.WriteLine("十叉树共有重复号" + count1 + "耗时" + sw.ElapsedTicks);
66 Console.ReadLine();
67
68 }
69
70 class TenNodeTree
71 {
72 public TenNode Root = new TenNode();
73
74 public bool Add(string no)
75 {
76 TenNode cnode = Root;
77 bool isadd = false ;
78 for (int i = 0; i < no.Length ; i++)
79 {
80 char k = no[i];
81
82 if (cnode.Child[k] == null)
83 {
84 isadd = true;
85 cnode.Child[k] = new TenNode();
86 }
87 cnode = cnode.Child[k];
88 }
89
90 return isadd;
91
92 }
93
94 }
95
96 class TenNode
97 {
98 public TenNode[] Child = new TenNode[256];
99 }
100
101
102 }
103 }
不过,运行一下,你会发现效率完全不是 Linq的对手:
Linq共有重复号9000耗时143185 十叉树共有重复号9000耗时411221
但是,你可不要以为这个算法有问题,要知道Linq是经过高度优化的,我们的算法的实现还有优化空间。让我们来优化它!
怎么优化?指针!有了指针,C#的性能可以提高N倍,见指针版代码:
View Code
using System;
using System.Collections.Generic;
using System.Linq;
using System.Text;
namespace ConsoleApplication1
{
unsafe class Program
{
static void Main(string[] args)
{
//示例数组,存放手机号
string[] mobileArray = new string[100000];// { "13900001234", "13900001235", "13900001236", "13900001237", "13900001234" };
for (int i = 0; i < 100000; i++)
{
mobileArray[i] = "1390000"
+ (i.ToString().Length > 4 ? i.ToString().Substring(0, 4) : (i.ToString() + "0000").Substring(0, 4));
}
////linq语句来实现【select mobile from tmpTable group by mobile having count(*)>1】的效果
var selMobile = from n in mobileArray group n by n into g where g.Count() > 1 select g.Distinct();// select g;
System.Diagnostics.Stopwatch sw = new System.Diagnostics.Stopwatch();
sw.Reset();
sw.Start();
int count1 = 0;
//通过两层循环输出重复的手机号
foreach (var mobile in selMobile)
{
foreach (string multiMobile in mobile)
{
count1++;
//Console.WriteLine(multiMobile);
}
}
sw.Stop();
Console.WriteLine("Linq共有重复号" + count1+"耗时"+ sw.ElapsedTicks );
TenNodeTree tree = new TenNodeTree();
TenNodeTree tree2 = new TenNodeTree();
sw.Reset();
sw.Start();
int count2 = 0;
//mobileArray = new string[] { "13900001234", "13900001235", "13900001236", "13900001237", "13900001234", "13900001236" };
foreach (var item in mobileArray)
{
fixed (char* no = item)
{
if (!tree.Add(no , 11 ))
{
if (tree2.Add(no,11))
{
count2++;
}
}
}
}
sw.Stop();
Console.WriteLine("十叉树共有重复号" + count1 + "耗时" + sw.ElapsedTicks);
Console.ReadLine();
}
class TenNodeTree
{
public TenNode Root = new TenNode();
public bool Add(char* no,int len)
{
TenNode cnode = Root;
bool isadd = false ;
for (int i = 0; i < len; i++)
{
char k = *no;
if (cnode.Child[k-48] == null)
{
isadd = true;
cnode.Child[k-48] = new TenNode();
}
cnode = cnode.Child[k-48];
no++;
}
return isadd;
}
}
class TenNode
{
public TenNode[] Child = new TenNode[10];
}
}
}
Linq共有重复号9000耗时139310 十叉树共有重复号9000耗时69545
如何?效率已达到Linq的1倍!
这还不算完,我们还没有使用Release模式呢!
Linq共有重复号9000耗时141970 十叉树共有重复号9000耗时35843
Release后,性能又提升1倍!
大家不妨用其他语言来实现下,比比效率如何?
C#还是很强的,HOHO
==================================
今天又做了测试,发现我家的老笔记本上,是十叉树占优,但是公司的电脑上是HashSet比较快。
不过十叉树应该还没有达到最优化,应该是分配节点时的开销过大导致。暂时想不出更好的优化方法-_-
==================================
五分钟后再次测试,十叉树只需在初始化时预先分配一个节点池,即可完胜HashSet.不过,此法或有胜之不武的嫌疑,哈哈。
也就是说,不算实例化的时间,十叉树胜,算实例化时间,哈希胜,