Re: [boost] Re: [MPL] Type detection

From: Joel de Guzman (joel_at_[hidden])
Date: 2003-10-24 00:31:17

Brian McNamara <lorgon_at_[hidden]> wrote:
> On Fri, Oct 24, 2003 at 09:25:54AM +0800, Joel de Guzman wrote:
>> Well, I was kinda hoping that you have a trick up your sleeve :->
> 
> :)
> 
>> If you can think of something better, I'm all ears.
> 
> Perhaps I am stuck in a mental rut, but it's hard for me to imagine
> anything better than the enable_if solution for something like this.
> After all:
> 
>  - you want to write a template function which only works on arguments
>    that meet some certain criteria
> 
> and, lo,
> 
>  - enable_if ensures that the function only 'exists' for arguments that
>    meet criteria specified by a value/metafunction, and
> 
>  - writing the (is_XXX) metafunction to define the criteria is
>    straightforward
> 
> So, I'm not sure why this solution would seem unsatisfying, or why you'd
> be looking for something "better".  (Or maybe I'm putting words in your
> mouth by suggesting you find it dissatisfying.)

No, it's not dissatisfying at all. I like enable_if.

> What qualities would a "better" solution have?  More concise to write?
> Faster compile-time?  Something else?  Maybe I do have the right trick
> somewhere in my bag-o-tricks, but I'm not sure what I'm looking for. :)

Perhaps, "something better" in my book, would mean something that 
works on more compilers, even those that do not have SFINAE. FWIW, 
CRTP works on more compilers. Whether I like it or not, I still have to 
support VC6 and Borland.

Cheers,
-- 
Joel de Guzman
http://www.boost-consulting.com
http://spirit.sf.net