1.1.17.2 template function

• When you use template function with real argument, compiler will extract type information form real argument, then generate real function code for you. So it will make compiler looks longer.
  swap(i1, i2); 
  swap(f1, f2); 
  // don’t need to specify int, compiler 
  //can extract type information from i1 
   
  // compiler will produce two swap functions bodies.
• Why we need a overload template function, Because not all type support the same operator. For example, in your template function, you use = operator, but when you use array for this template function type, compiler will report error.
  template <typename T> 
  void swap(T a[], T b[], int n) 
   
  template <typename T> 
  void swap(T &a, T &b )
• Overload is different with Specializations, Overload means that you have different function signatures. Specialization have the same function signature.
• Given a function name, you can have regular, template and explicit specialization temple. When pick a function, regular> specialization> template.
• function picking ranking from best to worst is:
  1. Exact math, regular function outranking template,
  2. Conversion by promotion
  3. Conversion by standard conversion.
  4. user-defined conversion.s
  int i ; 
  f(i); 
  f<>(i) // force compiler use template. 
   
  1) f(int) or template<typename T> f(T); 
  2) f(double)  // promotion conversion 
  3) f(char) // stand conversoin. usually give warning
• decltype and Trailing Return type(C++11 new feature)
  template<typename T1, typename T2> 
  void f(T1 x, T2 y){ 
  decltype(x+y) xpy = x+y; 
  } 
   
  //with return value; 
  template<typename T1, typename T2> 
  auto f(T1 x, T2 y) ->decltype(x+y) 
  { 
    x+y; 
  }