Re: Reminder: Review Special functions, Quaternions, Octonions

From: Hubert HOLIN (Hubert.Holin_at_[hidden])
Date: 2001-05-25 17:28:23

Paris (U.E.), le 26/05/2001

--- In boost_at_y..., Beman Dawes <bdawes_at_a...> wrote:
> At 01:53 PM 5/25/2001, Daryle Walker wrote:
> 
>  >I don't think "special_functions" is a good name.  Relative to other math
>  >stuff, atanh, sinc_pi, and sinhc_pi aren't that special.  There can be
>  >special functions in other areas besides math.  So the name is ambiguous 
> in
>  >two ways.  I've read suggestions here before to start dividing the Boost
>  >files into sub-domains, before the main header and library subdirectories
>  >get too large.  Maybe it can start here, with a structure like:
>  >
>  >BOOST_ROOT/boost/numeric
>  >    + quaternion.hpp
>  >    + octonion.hpp
>  >    + trigonmetric.hpp (sinc_pi, future stuff)
>  >    + hyperbolic.hpp (atanh, sinhc_pi, future stuff)
>  >    + rational.hpp (moved from parent directory)
>  >
>  >BOOST_ROOT/boost/semi_numeric
>  >    + random.hpp (moved from parent directory)
>  >    + random/*.hpp (moved from parent directory)
>  >    + crc.hpp (moved from parent directory)
>  >
>  >With mirrored changes in the BOOST_ROOT/libs subdirectory.
> 
> While the details are certainly open to discussion, I think Daryle is right 
> about introducing sub-directories for numerics stuff.  After working with 
> current libraries already in sub-directories (graph/test/random, etc.), I 
> find the scheme very intuitive and useful.  Benefits greatly outweigh the 
> costs.
> 
> As to the distinction between "numeric" and "seminumeric",  here is what 
> Knuth had to say of his volume 2:
> 
> "The algorithms discussed in this book deal directly with numbers; yet I 
> believe they are properly called seminumerical, because they lie on the 
> borderline between numeric and symbolic calculation."
> 
> In his book, seminumerical includes random numbers and arithmetic 
> (including rational arithmetic).
> 
> To save endless taxonomy arguments, we might want to combine numeric and 
> seminumeric.
> 
> --Beman

        As far as sub-directories are concerned, I propose (and have 
presented as such in the updated version of my submission) 
"special_functions" . Despite the fact that any function can rightly be 
call "special", those that should be placed within should be in the 
spirit of the usual math collections of such (refer A&S...).

        I also believe we need more namespaces (among which "math"), but 
this is more radical.

                Hubert Holin
                Hubert.Holin_at_[hidden]