Re: [boost] [intrusive] Comments on interface changes and stateful value_traits (long)

From: Ion Gaztañaga (igaztanaga_at_[hidden])
Date: 2007-08-26 15:12:40

David Abrahams wrote:
> What is the advantage to that over
> 
> make_list
>   < base_hook<my_base_hook>
>   , constant_time_size<false>
>   >::type l;

Just that programmers might found it more familiar, like an STL 
container, but I must say that your approach gives me much more 
guarantee of a single type, something that is difficult to achieve with:

set<T, set_options<...>::type >;

because I can't define the same options structure as:

set<T, set_options<pred<std::less<T> >, ...>::type >;

... because I have no access to the T type of the set from set_options 
(I could pass it again, but that would be verbose. Whereas if I define:

make_set<T, pred<std::less<T>>, ...>::type l;

I have all the needed data to produce a single type. Although it's a bit 
different from standard container definitions, your suggestion might be 
the right thing to do to achieve both single type and less typing. So I 
guess that I can offer both Tobias' approach:

set<T, pred<std::greater<T>, >

and your approach for people that want to make sure that the same types 
are being created when specifying the same options.

Thanks,

Ion