RE: [boost] Re: Formal Review: Template Metaprogramming Library ( MPL)

From: Aleksey Gurtovoy (agurtovoy_at_[hidden])
Date: 2002-07-26 22:07:26

Mat Marcus wrote:
> Aleksey> It shouldn't. Just remember that it's a class, not a 
> class template - that's
> Aleksey> the most important point (it should be easy, since 
> "class" is in the name
> Aelksey> here :).
> 
> One more try...
> 
> RIght, I am aware that it is a class. I have no problem 
> remebering what it does. I am trying to discuss pissible 
> improvements in what the name suggests to the user. 
> One of  the MPL's strenghts is that it provides a language for 
> metaprogramming. My intent in suggesting alternate names 
> (quoted metafunction, metafunctor,...) is to try to help 
> ensure that the language is in the domain of the user as 
> opposed to the implementor (a DSL as opposed to an ICCL).
> The phrase metafunction class suggests an 
> implementation detail to me. 

It suggests _the form_. The form is the only difference between metafunction

    template< typename T > struct f
    {
        typedef T type;
    };

and metafunction class

    struct f
    {
        template< typename T > struct apply
        {
            typedef T type;
        };
    };

> Or even worse, it can be misread 
> as "the class of the metafunction".

As I am not a native speaker, I cannot say how much it is a problem :). We
can always spell it with a hyphen, like many other terms in the standard: 

   metafunction-class

> 
> Why is the fact that it is a class and not a class template 
> the most important point? 

Because it immediately suggests you the metafunction's form.

> To me the most important point is the the fact that this form 
> of metafunction wrapping is what enables higher order 
> metafunctions, i.e. this is how one metafunction can be 
> returnedfrom another. 

That's somewhat a secondary knowledge - after you know the form of a
metafunction, you know what you can do with it.

> You could argue that this same principal is captured at 
> run time in the distinction between the word "function" 
> and the word "function-object" or "functor". I was looking 
> for a name which stressed this role, rather than 
> the (important) implementation detail that this is a class.
> 
> By the way, for the record, as evidenced by my comments and 
> MPL tutorials in the past, I think MPL is a fine piece of 
> work and I vote for formal acceptance into boost.

Mat, I appreciate your thoughts and suggestions. I am just not convinced
(yet) that the current vocabulary is such a poor fit.

Aleksey