Why do we need smart pointer and how to implement it.
Here are two simple questions.
Problem A
#include <string>
include <iostream>
using namespace std; class vehicle
{
public:
vehicle(const string& name);
virtual ~vehicle(){} void PrintOwnerInfo(); private:
string driver_name_;
}; vehicle::vehicle(const string& name):
driver_name_(name){
} void vehicle::PrintOwnerInfo(){
cout<<driver_name_<<endl;
} int main()
{
string driver_name = "Zhou You";
vehicle* pbicycle = new vehicle(driver_name);
pbicycle->PrintOwnerInfo();
return ;
}
Any problems in the first piece of code?In main function,I produce an vehicle object which is holden by "pbicycle",then I call PrintOwnerInfo to output the name of the driver.The point is I fail to delete the pointer pbicycle which holds an vehicle object,and this is the most common mistake new programmers always make(it calls memory leak).
Look at another piece of code.Problem B
class A
{
public:
A();
A(int num);
~A(){} void PrintNum();
private:
int num_;
}; A::A():num_(){
} A::A(int num):
num_(num){
} void A::PrintNum(){
cout<<num_<<endl;
} int main()
{
A *pa1 = new A();
pa1->PrintNum(); A *pa2 = pa1;//pa2 point to the same object with pa1
pa2->PrintNum(); delete pa1
pa2->PrintNum();//error happens return ;
}
In main function,I produce an object of class A,using pa1 to point to it.With a pointer of this object,I call PrintNum function to print the integer.Now that I have pa1,I want to create another pointer pa2 to point to the same object with pa1.The next thing I want to do is to delete pa1.Can I do that?Nooooo.With this inappropriate statement,I transform pa2 to a dangling pointer.The code above is pretty simple and easy,so we can understand it very quickly and find the drawbacks.But in real projects,such mistakes can be easily made and ignored.Is there any good solutions to avoid it?I mean we don't even need to care about these things.Actually this is part of memory management in C++,we can use smart pointer to help us to handle managing memory issue.
In the vehicle instance,We can write main function with the smart pointer like this.
int main()
{
string driver_name = "John Zhou";
SPvehicle spbicycle = new vehicle(driver_name);
spbicycle->PrintOwnerInfo(); return ;
}
"SPvehicle" is a smart pointer class for vehicle.I do not need to delete it When I finish using this pointer,because this pointer itself has the feature to release its object when it is not useful any more.So how to implement it? Basically,there are two points to take care.
1."spbicycle" should behave like an vehicle pointer does.It should support both dereferencing(*) operation and indirection(->) operation.
2.It should be able to release the object automatically.
Check out the code below.
class SPvehicle
{
public:
SPvehicle(){}
SPvehicle(vehicle *pvehicle);
~SPvehicle(); vehicle& operator*(){
return *pobjvehicle_;
} vehicle* operator->(){
return pobjvehicle_;
}
private:
vehicle *pobjvehicle_;
}; SPvehicle::SPvehicle(vehicle *pvehicle):
pobjvehicle_(pvehicle){
} SPvehicle::~SPvehicle(){
delete pobjvehicle_;
}
In the SPvehicle class,I overload 2 operators(*,->),then I can use it as a normal vehicle pointer.In the destructor,I delete pobjvehicle_ to release the memory space.
Class SPvehicle gives us a solution to manage the object automatically,but still remains problem B unsolved.So how to handle the dangling pointer?The object it points to is released by the other pointer,but it is not informed,and this is the direct reason it becomes dangling.For an object,can we remember the amount of the pointer which point to it?With the amount,we are able to know the right time to release the object(the amount decreasing to 0).Let's try.
class RfCount
{
public:
RfCount(){}
RfCount(unsigned rfc):
rfc_(rfc){
} ~RfCount(){} unsigned AddRef(){
return ++rfc_;
} unsigned CutRef(){
return --rfc_;
} private:
unsigned rfc_;
}; class SPA
{
public:
SPA();
SPA(A *pa);
SPA(const SPA &spa);
~SPA(); A& operator*(){
return *pa_;
} A* operator->(){
return pa_;
} SPA& operator=(const SPA &spa); private:
RfCount* reference_count_;
A *pa_;
}; SPA::SPA():pa_(NULL),
reference_count_(new RfCount()){
cout<<"default SPA constructor."<<endl;
} SPA::SPA(A* pointera):
pa_(pointera),
reference_count_(new RfCount()){
cout<<"SPA constructor1 adds referencecount by 1."<<endl;
reference_count_->AddRef();
} SPA::SPA(const SPA& spa):
pa_(spa.pa_),
reference_count_(spa.reference_count_){
cout<<"SPA constructor2 adds referencecount by 1."<<endl;
reference_count_->AddRef();
} SPA::~SPA(){
cout<<"SPA destructor cut referencecount by 1."<<endl;
if(reference_count_->CutRef() == ){
cout<<"release the object."<<endl;
delete pa_;
}
} SPA& SPA::operator=(const SPA& spa){
if(this != &spa){
cout<<__FUNCTION__<<" "<<"refcount added by 1."<<endl;
reference_count_ = spa.reference_count_;
reference_count_->AddRef();
pa_ = spa.pa_;
}
return *this;
} void copypointer(SPA& spa){
SPA spa1;
spa1 = spa;
spa1->PrintNum();
} int main()
{
SPA spa1 = new A();
spa1->PrintNum();
copypointer(spa1);
spa1->PrintNum();
return ;
}
In main function,I produce a smart pointer spa1 at first(reference count is initialized to 1),then use it to call member function of class A which can prove that I can use this smart pointer as a normal A pointer.The next step is to call copypointer() function and it takes spa1 as the argument.In copypointer(),I reproduce a local smart pointer spa2 which is assigned with spa1(reference count up to 2),during the process of assignment,the referencecount is increased to 2.After function copypointer() finishes its task,spa2 just goes out of scope,and SPA destructor is called.In SPA destructor, referencecount is reduced by 1.Back to main funtion,when main ends up,SPA destructor is called again,and reference count is reduced to 0,and that's exactly the time we need to release A pointer inside class SPA,because there are no pointers referencing this A object.
OK,for problem A and B I post at the start of this article,we just figure out how to solve or avoid them,but the code look not cool.For problem A and B,we have to write 2 smart pointer classes respectively.so how about a generic class to implement the smart pointer.
class RfCount
{
public:
RfCount(){}
RfCount(unsigned rfc):
rfc_(rfc){
} ~RfCount(){} unsigned AddRef(){
return ++rfc_;
} unsigned CutRef(){
return --rfc_;
} private:
unsigned rfc_;
}; template <typename T>
class SP
{
public:
SP();
SP(T *pa);
SP(const SP<T>& sp);
~SP(); T& operator*(){
return *pt_;
} T* operator->(){
return pt_;
} SP& operator=(const SP &sp); unsigned AddRef(){
return reference_count_->AddRef();
} unsigned CutRef(){
return reference_count_->CutRef();
} private:
RfCount* reference_count_;
T *pt_;
}; template <typename T>
SP<T>::SP():pt_(NULL),
reference_count_(new RfCount()){
cout<<"default SP constructor."<<endl;
} template <typename T>
SP<T>::SP(T *pt):
pt_(pt),
reference_count_(new RfCount()){
cout<<"SP constructor1 adds referencecount by 1."<<endl;
reference_count_->AddRef();
} template <typename T>
SP<T>::SP(const SP<T>& sp):pt_(sp.pt_),
reference_count_(sp.reference_count_){
cout<<"SP constructor2 adds referencecount by 1."<<endl;
reference_count_->AddRef();
} template <typename T>
SP<T>::~SP(){
cout<<"SP destructor cut referencecount by 1."<<endl;
if(reference_count_->CutRef() == ){
cout<<"release the object."<<endl;
delete pt_;
}
} template <typename T>
SP<T>& SP<T>::operator=(const SP &sp){
if(this != &sp){
reference_count_ = sp.reference_count_;
cout<<__FUNCTION__<<" "<<"refcount added by 1."<<endl;
reference_count_->AddRef();
pt_ = sp.pt_;
}
return *this;
} void copypointer(SP<A>& spa){
SP<A> spa1;
spa1 = spa;
spa1->PrintNum();
} int main()
{
SP<A> spa1 = new A();
spa1->PrintNum();
copypointer(spa1);
spa1->PrintNum();
return ;
}
Using template,we can write a generic smart pointer class.
//词汇拙计啊,乃们就将就看吧,有啥问题给俺指正出来呀。
Why do we need smart pointer and how to implement it.的更多相关文章
- [CareerCup] 13.8 Smart Pointer 智能指针
13.8 Write a smart pointer class. A smart pointer is a data type, usually implemented with templates ...
- 理解smart pointer之三:unique_ptr
unique_ptr最先在boost中被定义,后来被C++标准委员会选中为C++11的feature之一. std::unique_ptr is a smart pointer that retain ...
- c++ smart pointer
智能指针(smart pointer)是存储指向动态分配(堆)对象指针的类,用于生存期控制,能够确保自动正确的销毁动态分配的对象,防止内存泄露.它的一种通用实现技术是使用引用计数(reference ...
- c++(smart pointer)
(一)首先对智能指针有一些概念性的了解 **********本部分内容摘自开源中国社区http://my.oschina.net/u/158589/blog/28994******** 1.什么是智能 ...
- c/c++ 标准库 智能指针( smart pointer ) 是啥玩意儿
标准库 智能指针( smart pointer ) 是啥玩意儿 一,为什么有智能指针??? c++程序员需要自己善后自己动态开辟的内存,一旦忘了释放,内存就泄露. 智能指针可以帮助程序员"自 ...
- C++ smart pointer智能指针
在C++中,程序员可以直接操作内存,给编程增加了不少的灵活性.但是灵活性是有代价的,程序员必须负责自己负责释放自己申请的内存,否则就会出现内存泄露.智能指针就是为了解决这个问题而存在的.它和其他指 ...
- Effective C++ Item 17 Store newed objects in smart pointer in standalone statements
If you trying to do multiple things in one statement, you should think carefully abnormal behavior e ...
- Smart pointer 智能指针小总结
Smart pointer line 58之后smart pointer里的计数已经是0,所以会真正释放它引用的对象,调用被引用对象的析构函数.如果继续用指针访问,会出现如下图的内存访问异常.所以说如 ...
- smart pointer
smart pointer是一种abstract data type,它可以模仿指针的行为,而且额外提供了一系列诸如自己主动内存管理.边界检查等特性,这些特性是为了在保证效率的基础上降低因为对指针的不 ...
随机推荐
- (redis缓存更新策略)postgres 9.4.1 && redis 3.7.0 && redis_fdw_REL9_4_STABLE
首先下载redis_fdw,这里要注意下载的版本.(https://github.com/pg-redis-fdw/redis_fdw) 一开始,我下载了REL9_4_STABLE_pre2.8版本, ...
- python 自动化之路 logging日志模块
logging 日志模块 http://python.usyiyi.cn/python_278/library/logging.html 中文官方http://blog.csdn.net/zyz511 ...
- docker中搭建jenkins环境
想在docker中搭建一个jenkins环境.开始的时候我想在先pull centos,然后在里面自己搭建环境,搭建后的环境可以运行,但是不知道挂载哪的数据,这也是我不熟悉jenkins的原因. 同事 ...
- 高性能网络I/O框架-netmap源码分析
from:http://blog.chinaunix.net/uid-23629988-id-3594118.html 博主这篇文章写的很好 感觉很有借签意义 值得阅读 高性能网络I/O框架-netm ...
- ES6 语法简介
参考: http://es6.ruanyifeng.com/ 总结学习 JavaScript语言下一代标准,2015年6月正式发布. 1.let和const命令 let用作变量声明,只在代码块内有效 ...
- Linux下使用openvpn客户端
安装 root 权限用户下 :yum install openvpn fedora23的yum可能会切换到dnf源上安装不必在意,等待就ok. 稍等片刻将自动安装好openvpn需要的软件包.安装完成 ...
- css(html)背景图优化合并
图片本身的优化: 图像质量要求和图像文件大小决定你用什么格式的图片,用较小的图片文件呈现较好的图像质量. 当图片色彩过于丰富且无透明要求时,建议采用jpg格式并保存为较高质量. 当图片色彩过于丰富又有 ...
- 使用GetLogicalDriveStrings获取驱动器根路径
使用GetLogicalDriveStrings获取驱动器根路径,并使用自定义的GetDriveInfo函数获取驱动器的属性. VS2012 + win7 x64下调试通过. #include < ...
- linux的getcwd和readlink的区别
针对linux下的程序,有两个路径: 1>运行程序的路径; 2>可执行文件所在的路径 例如: 如果我在/home/yongchao下执行 $ ./temp/test 那么 运行程序的 ...
- sql server 与C#数据类型对应表