1 教科书里的单例模式
我们都很清楚一个简单的单例模式该怎样去实现:构造函数声明为private或protect防止被外部函数实例化,内部保存一个private static的类指针保存唯一的实例,实例的动作由一个public的类方法代劳,该方法也返回单例类唯一的实例。
上代码:
#pragma once
#include<iostream>
class Singleton_Test
{
private:
Singleton_Test()
{
}
~Singleton_Test()
{
}
static Singleton_Test *m_Obj;
public:
Singleton_Test* GetInstance();
};
Singleton_Test* Singleton_Test::m_Obj = NULL;
Singleton_Test* Singleton_Test::GetInstance()
{
if (m_Obj == NULL)
{
m_Obj = new Singleton_Test();
std::cout << "Creat the Singleton .." << std::endl;
}
std::cout << "Get the Singleton" << std::endl;
return m_Obj;
}
#include<iostream>
#include"scoped_ptr_test.hpp"
int main()
{
Singleton_Test *Mytest=NULL;
Mytest->GetInstance();
Singleton_Test *Test_02=NULL;
Test_02->GetInstance();
getchar();
return 0;
}
2 懒汉与饿汉这是一个很棒的实现,简单易懂。但这是一个完美的实现吗?不!该方法是线程不安全的,考虑两个线程同时首次调用instance方法且同时检测到p是NULL值,则两个线程会同时构造一个实例给p,这是严重的错误!同时,这也不是单例的唯一实现!
单例大约有两种实现方法:懒汉与饿汉。
懒汉:故名思义,不到万不得已就不会去实例化类,也就是说在第一次用到类实例的时候才会去实例化,所以上边的经典方法被归为懒汉实现;
饿汉:饿了肯定要饥不择食。所以在单例类定义的时候就进行实例化。
特点与选择:
由于要进行线程同步,所以在访问量比较大,或者可能访问的线程比较多时,采用饿汉实现,可以实现更好的性能。这是以空间换时间。
在访问量较小时,采用懒汉实现。这是以时间换空间。
3 线程安全的懒汉实现
线程不安全,怎么办呢?最直观的方法:加锁。
方法1:加锁的经典懒汉实现:
#pragma once
#include<iostream>
#include<boost/thread/mutex.hpp>
#include<boost/thread/thread.hpp>
/*
the point of ther
*/
class Singleton_Test
{
private:
Singleton_Test()
{
}
~Singleton_Test()
{
}
static Singleton_Test *m_Obj;
static boost::mutex m_mutex;
public:
Singleton_Test* GetInstance();
};
Singleton_Test* Singleton_Test::m_Obj = NULL;
boost::mutex Singleton_Test::m_mutex;
Singleton_Test* Singleton_Test::GetInstance()
{
boost::unique_lock<boost::mutex> lock(m_mutex);
if (m_Obj == NULL)
{
m_Obj = new Singleton_Test();
std::cout << "Creat the Singleton .." << std::endl;
}
std::cout << "Get the Singleton" << std::endl;
return m_Obj;
}
#include<iostream>
#include"scoped_ptr_test.hpp"
void FunctionCreatSingleton()
{
Singleton_Test *Test = NULL;
Test->GetInstance();
}
int main()
{
//Singleton_Test *Mytest=NULL;
//Mytest->GetInstance();
//Singleton_Test *Test_02=NULL;
//Test_02->GetInstance();
boost::thread *ThreadGroup[5];
for (int i = 0;i < 5;i++)
{
ThreadGroup[i] = new boost::thread(&FunctionCreatSingleton);
}
getchar();
return 0;
}
此方法也很容易实现,在instance函数里定义一个静态的实例,也可以保证拥有唯一实例,在返回时只需要返回其指针就可以了。推荐这种实现方法,真得非常简单。方法2:内部静态变量的懒汉实现
#pragma once
#include<iostream>
#include<boost/thread/mutex.hpp>
#include<boost/thread/thread.hpp>
/*
the point of ther
*/
class Singleton_Test
{
private:
Singleton_Test()
{
}
~Singleton_Test()
{
}
static Singleton_Test *m_Obj;
static boost::mutex m_mutex;
public:
Singleton_Test* GetInstance();
};
//Singleton_Test* Singleton_Test::m_Obj = new Singleton_Test();
Singleton_Test* Singleton_Test::m_Obj =NULL ;
boost::mutex Singleton_Test::m_mutex;
Singleton_Test* Singleton_Test::GetInstance()
{
boost::unique_lock<boost::mutex> lock(m_mutex);
if (m_Obj == NULL)
{
m_Obj = new Singleton_Test();
std::cout << "Creat the sinleton .." << std::endl;
}
std::cout << "Get the sinleton .." << std::endl;
return m_Obj;
}
为什么我不讲“线程安全的饿汉实现”?因为饿汉实现本来就是线程安全的,不用加锁。为啥?自己想!4 饿汉实现
#pragma once
#include<iostream>
#include<boost/thread/mutex.hpp>
#include<boost/thread/thread.hpp>
/*
the point of ther
*/
class Singleton_Test
{
private:
Singleton_Test()
{
}
~Singleton_Test()
{
}
static Singleton_Test *m_Obj;
static boost::mutex m_mutex;
public:
Singleton_Test* GetInstance();
};
Singleton_Test* Singleton_Test::m_Obj = new Singleton_Test();
boost::mutex Singleton_Test::m_mutex;
Singleton_Test* Singleton_Test::GetInstance()
{
if (m_Obj == NULL)
{
std::cout << "Creat the sinleton .." << std::endl;
}
std::cout << "Get the sinleton .." << std::endl;
return m_Obj;
}