[boost] Re: function_traits question/patch

From: Daniel Frey (daniel.frey_at_[hidden])
Date: 2003-05-15 11:17:38

Thomas Witt wrote:
> 
> Daniel,
> 
> Sorry I missed this posting.
> 
> Daniel Frey wrote:
> 
>> Thomas Witt wrote:
>>
>>>
>>> struct my_functor :
>>>   public boost::function_traits<int(double)>
>>> {
>>>  // ...
>>> };
>>
>>
>>
>> I like it, too. The only thing that I don't like is the name. Used 
>> like this, it no longer has traits semantics IMHO. Thus I'd prefer:
> 
> 
> I disagree. I think it has traits semantics. That what it does it 
> provides traits.

Maybe I should explain what I think a traits class is: It's a container 
for types and/or constants. It provides types/constants that depend on 
the "input parameters" at compile time. It is not a base class. Thus, I 
would have expected:

struct my_functor : boost::function_traits<int(double)>::base
{
   // ...
};

Or probably some other name for 'base' like 'functor'... (I know, some 
people will stone me for calling a 'function object' a 'functor' :).

Also, my suggestion to call it 'function<int(double)>' IMHO matches 
unary_function and binary_function better, so it might be easier for 
users that are used to the current style. But I don't really have strong 
feelings about it, so keeping the _traits is fine with me, too. Probably 
it's the least invasive approach from boost's perspective. :)

Regards, Daniel

-- 
Daniel Frey
aixigo AG - financial training, research and technology
Schloß-Rahe-Straße 15, 52072 Aachen, Germany
fon: [phone], fax: [phone]
eMail: daniel.frey_at_[hidden], web: http://www.aixigo.de