c – 基于所有派生策略的基本策略工作获取模板类专业化

我的政策来自基本政策.某些类专门用于派生策略,而其他类专门用于基本策略,可以使用所有派生策略.

我遇到的问题是代码重复太多(主要是构造函数和类本身的一些样板代码).下面的代码可以更好地解释我的意思：

struct BasePolicy {};
struct DerivedPolicy1 : public BasePolicy {};
struct DerivedPolicy2 : public BasePolicy {};
//... more policies deriving from BasePolicy (or other BasePolicies)
struct AnotherPolicy {};

template <typename T>
struct Foo;

// This struct can work for all BasePolicy types which includes all derivations
// of it (but doesn't because it is specialized for BasePolicy only)
template<>
struct Foo<BasePolicy>
{
  //... many constructors along with code
};

template<>
struct Foo<AnotherPolicy>
{
  //... more code
};

/* Would like to avoid the following code as it duplicates the above when it 
   comes to constructors and other things such as typedefs */
//template<>
//struct Foo<DerivedPolicy1> : Foo<BasePolicy>
//{
//  //... same constructors as Foo<BasePolicy>
//};
//
//template<>
//struct Foo<DerivedPolicy2> : Foo<BasePolicy>
//{
//  //... same constructors as Foo<BasePolicy>
//};

int main()
{
  // would like this to compile without uncommenting the commented out code and
  // without having the client (i.e. the line below) to somehow get the base
  // type of the policy (although if it can be done transparently, that will
  // work)
  Foo<DerivedPolicy1> a; 
};

派生政策是否有任何方法可以被专门针对基本政策的类接受？我希望客户不要做任何额外的事情.

以下是无效的C代码,但我希望发生这样的事情(如果你记住上面的代码)：

template<>
struct Foo<BasePolicy | DerivedPolicy1 | DerivedPolicy2>
{
  //... many constructors along with code
};

解决方法:

这是SFINAE的一个案例.

template< typename Policy, ...some_condition... >
struct Foo<Policy>
{
 ...
};

你应该确定什么是some_condition.您可以指定Policy派生自BasePolicy：

template< typename Policy, enable_if< is_base<BasePolicy, Policy> > >

或者您可以明确列出允许的策略：

template< typename Policy,
           enable_if_c <is_same<Policy, BasePolicy>::value || 
                         is_same<Policy, DerivedPolicy1>::value> || 
                         ...whatever...
                        >
         >

标签：c,templates,code-duplication
来源： https://codeday.me/bug/20190902/1790483.html