C++ 装饰器模式

装饰器模式的特点

首先，装饰器模式是结构型模式，准确来说是对象结构型模式。在不改变该类结构的情况下，给现有对象添加新的功能。总的来说，装饰器模式就是为了给现有对象添加新的功能，而设计模式的原则中的开闭原则是不尽量不改变类的现有结构的。一般为一个现有对象添加新的功能，我们往往能想到的方式有两种：其一，通过继承的方式；其二，通过组合的方式。但这两种方式都更改了该类的结构，实际上，装饰器可以看成这两种方式的结合体。下面分别举例说明这两种方式，最后再介绍装饰器模式。 例子如下：在二维平面有各种形状，比如有矩形、圆，进行画图，代码如下所示：

struct Point { double __x; double __y; Point(double x, double y):__x(x), __y(y){} }; class Shape { public: virtual void show() const = 0; }; class Circle : public Shape { private: Point center; double radius; public: Circle(Point pnt, double rds):center(pnt), radius(rds){} virtual void show() const { std::cout<<"I draw a Circle: "<<std::endl; std::cout<<"center: "<<"("<<center.__x<<","<<center.__y<<")"<<"\t"; std::cout<<"radius: "<<radius<<std::endl; } }; int main() { Shape* circle = new Circle(Point(3, 4), 1.0); circle->show(); return 0; }

现在我想添加新的功能，用红色画圆。

1.继承的方式添加新功能

先上代码：

struct Point { double __x; double __y; Point(double x, double y):__x(x), __y(y){} }; struct Color { unsigned short __r :8; unsigned short __g :8; unsigned short __b :8; Color(unsigned char r, unsigned char g, unsigned char b):__r(r),__g(g),__b(b){} }; class Shape { public: virtual void show() const = 0; }; class Circle : public Shape { private: Point center; double radius; public: Circle(Point pnt, double rds):center(pnt), radius(rds){} virtual void show() const { std::cout<<"I draw a Circle: "<<std::endl; std::cout<<"center: "<<"("<<center.__x<<","<<center.__y<<")"<<"\t"; std::cout<<"radius: "<<radius<<std::endl; } }; class ColorCircle : public Circle { private: Color color; public: ColorCircle(Point pnt, double rds, Color clr): Circle(pnt, rds), color(clr){} virtual void show() const { std::cout<<"Use RGB: "<<"("<<color.__r<<","<<color.__g<<","<<color.__b<<")"<<std::endl; Circle::show(); } }; int main() { Shape* circle = new ColorCircle(Point(3, 4), 1.0, Color(255, 0, 0)); circle->show(); return 0; }

这里用了一个ColorCircle 继承至Circle，这样也解决了新功能，看上去也没有修改原类的结构。而实际上，如果想使用红色画长方形呢？是不是又要用矩形的类继承一个新的类呢？也就是为每一个类添加新的功能，就要多一个类进行实现。这样显然不符合设计模式的原则。

2.使用组合的方式进行实现

代码如下：

struct Point { double __x; double __y; Point(double x, double y):__x(x), __y(y){} }; struct Color { unsigned short __r :8; unsigned short __g :8; unsigned short __b :8; Color(unsigned char r, unsigned char g, unsigned char b):__r(r),__g(g),__b(b){} }; class Shape { public: virtual void show() const = 0; }; class Circle : public Shape { private: Point center; double radius; public: Circle(Point pnt, double rds):center(pnt), radius(rds){} virtual void show() const { std::cout<<"I draw a Circle: "<<std::endl; std::cout<<"center: "<<"("<<center.__x<<","<<center.__y<<")"<<"\t"; std::cout<<"radius: "<<radius<<std::endl; } }; class ColorCircle { private: Circle circle; Color color; public: ColorCircle(Point pnt, double rds, Color clr):circle(pnt, rds),color(clr){} virtual void show() const { std::cout<<"Use RGB: "<<"("<<color.__r<<","<<color.__g<<","<<color.__b<<")"<<std::endl; circle.show(); } }; int main() { ColorCircle* circle = new ColorCircle(Point(3, 4), 1.0, Color(255, 0, 0)); circle->show(); return 0; }

上面通过ColorCircle类组合circle类，实现了用红笔画圆的新功能，但问题依旧是这样，用红笔画矩形呢？是不是有要用新的类组合矩形类？当然，有人会说：我用一个类组合各种形状，对各种形状一一实现新功能，原则上是可行的，但代码没有复用，而且，如果再新增一种形状，是不是又要修改组合的类呢？

装饰器模式实现添加新功能

首先通过上面的例子画出UML图，如图所示： 形状类Circle、Rectangle等继承抽象类Shape，而装饰器类ColorShapeDecorator等继承抽象类ShapeDecorator；而ShapeDecorator类既继承抽象类Shape也将Shape类组合，这样就形成了对象结构型装饰器模式。 通过C++代码实现如下：

struct Point { double __x; double __y; Point(double x, double y):__x(x), __y(y){} }; struct Color { unsigned short __r :8; unsigned short __g :8; unsigned short __b :8; Color(unsigned char r, unsigned char g, unsigned char b):__r(r),__g(g),__b(b){} }; class Shape { public: virtual void show() const = 0; virtual ~Shape() = default; }; class Circle : public Shape { private: Point center; double radius; public: Circle(Point pnt, double rds):center(pnt), radius(rds){} virtual void show() const { std::cout<<"I draw a Circle: "<<std::endl; std::cout<<"center: "<<"("<<center.__x<<","<<center.__y<<")"<<"\t"; std::cout<<"radius: "<<radius<<std::endl; } }; class ShapeDecorator : public Shape { protected: Shape* shape; public: virtual void show() const = 0; virtual ~ShapeDecorator() = default; }; class ColorShapeDecorator : public ShapeDecorator { private: Color color; public: ColorShapeDecorator(Shape* shp, Color clr):color(clr) { shape = shp; } virtual void show() const { std::cout<<"Use RGB: "<<"("<<color.__r<<","<<color.__g<<","<<color.__b<<")"<<std::endl; shape->show(); } virtual ~ColorShapeDecorator() { delete shape; } }; int main() { Shape* circle = new ColorShapeDecorator(new Circle(Point(3, 4), 1.0), Color(255, 0, 0)); circle->show(); return 0; }

装饰器模式的相关扩展设计

其实装饰器模式的扩展设计，一般有两种情况：其一，就是将上图UML类图的Shape类是具体类即不是抽象类；其二，就是上图的UML类图的ShapeDecorator抽象类变成具体类。设计方式和装饰器模式类似，这里不进行赘述，可以参考该博客。

装饰器的优点有二：

装饰类和被装饰类可以独立发展，不会相互耦合 动态的将责任附加到对象身上。

缺点是：多层装饰比较复杂。