[boost] Re: as_sequence<R(T0,T1)> ?

From: Tobias Schwinger (tschwinger_at_[hidden])
Date: 2004-10-10 13:09:36

Alexander Nasonov wrote:
> I have one addition to '[type_traits] "member_function_pointer_traits" ?'
> thread started long time ago (29 Jul 2004) by Tobias Schwinger.
> Is it makes sense to modify as_sequence to convert function type into a
> sequence? Eg:
> as_sequence<R   (T0,T1)> ---> vector<R,T0,T1>
> as_sequence<R(*)(T0,T1)> ---> vector<R,T0,T1>
> as_sequence<R(&)(T0,T1)> ---> vector<R,T0,T1>
> 
> as_sequence<R(C::*)(T0,T1)> ---> vector<R,C,T0,T1>

as_sequence<R(T0,T1)> reads great, but this would add a lot to the 
complexity of 'as_sequence':

First we have to idenitify if T is some sort of function (ptr/ref) type.
Although there are ways to detect (not decompose, that is) functions and 
even member function pointers without using an unrolled cascades of 
specializations for different function arities, it requires a very 
standard compliant compiler.
As we need to decompose the type and wrap its components into a vector 
in case it turns out to be a function anyway, we will instantiate one of 
these (more than 100) specializations.

Note that in order to do all this a big chunk of code needs to be 
included and processed (the unrolled cascade of specializations to 
identify/decompose the function type and the cascade of 'mpl::vector' up 
to the maximum arity supported).

This can significantly slow down the compilation time of a project 
(especially when separated into a lot of small translation units or when 
'as_sequence' is used often).

It may not be a good idea to impose these disadvantages on a "mostly 
harmless" user, probably not wanting to use this feature at all ;+).

However there is a possiblility to give control to the user:

As template specializations "slip through" to the namespace using them, 
a "dumb stub" could be added to 'as_sequence', never reporting that a 
type is a function. It could be "filled with life" afterwards when the 
user includes another header to enable 'as_sequence' for function types.

Still I am not too sure this is a good idea.
Comments about interest in this would be very helpful !

While we're at it - and in case you want to contribute to it:

My work on this in the files section is up to date and has been updated 
just recently.

Currently
   function_pointer_signature<R(*)(T0,T1)>::type is vector<T0,T1>
   function_pointer_signature<R(C::*)(T0,T1)>::type is vector<C,T0,T1>
   function_pointer_result<FuncPtr>::type is R (in both cases)

Best regards,

Tobias