Re: [boost] Re: [type_traits] "member_function_pointer_traits" ?

From: Tobias Schwinger (tschwinger_at_[hidden])
Date: 2004-07-29 15:09:51

> Please, no more trait blobs! IMO you should separate the traits into
> individual meta functions that works on all function pointer types. For
> instance:
>
>   function_arity<T>::type
>     Self explanatory.
>
I agree, because of:

- MPL-interoperability

- Currently only the arity-part works on compilers without support for 
partial template specialization. This clearly delimits compiler support.

- Including blobs of unused code unnecessarily increases compilation time.

>   function_signature<T>::type
>     MPL sequence containing all types of the function signature,
>     including return type and type of *this if T is a member function
>     pointer.
>
Is it a good idea to separate this even further ?

  function_return<T>::type
  function_input<T>::type

And only valid on member function pointers:

  function_this<T>::type
  is_const_function<T>::type/value
  is_volatile_function<T>::type/value

A 'function_signature' metafunction could then be implemented on top of 
these.

As I mentioned in the original post:  A function type can exist without 
a pointer while a member function type can not.
How to address this ? Handle all three cases or require pointer types ?

I would appreciate any suggestions.

> And while we're at it, why not deprecate function_traits?
>
It was just a -careful- request !

-
Tobias