Re: Review results: Special functions, Quaternions, Octonions

From: Hubert HOLIN (Hubert.Holin_at_[hidden])
Date: 2001-05-30 17:08:36

Paris (U.E.), le 30/05/2001

        Hi

--- In boost_at_y..., Jens Maurer <Jens.Maurer_at_g...> wrote:
> 
> The review period for Hubert Holin's Special functions, Quaternions,
> Octonions library is over.
> 
> The library has been accepted into boost, subject to the review

                Wonderful! My thanks to all, this has been a very 
enlightening experience!

> comments.  Here are the most prominent ones:
> 
>  - rewrite operator>> (Gary Powell)
> (It's templated on charT and traits, so it's unlikely you'll run
> into ODR trouble when separating it into subroutines.)

                This is clearly necessary. However, I would prefer, if 
possible, defer this to a later time. If the parser-building libraries 
being floated here do make their way into Boost, they will provide, I 
think, the best possible way to solve the problem. So I believe that by 
waiting a short spell we would win much. Otherwise, I'll do as 
suggested here.

>  - Avoid check for self-assignment in operator=() (Gary Powell)

                OK.

>  - We should postpone the namespace issue until there has been more
> discussion and clear guidelines.   For the moment, boost::octonion<>
> etc. appears to be acceptable, but maybe not perfect.

                OK. Can be easily changed at any time.

>  - Use <> for #include paths for internal boost consistency, even
> if proven wrong later.

                No problem.

>  - Make converting functions "explicit" (Gary Powell). Making them
> non-explicit later when more experience is available will not cause
> much harm with then-existing code, but the other way round will
> turn legal code illegal.  Take note of the decision in the docs.

                OK.

>  - (Here's an additional piece of information for the template-template
> overloads:  Template-template parameters are only matched by
> templates with the exact number of parameters.  Unlike defaulted
> function parameters, defaulted template parameters do count.
> Therefore, a type of form   X<T, Traits = default_traits> will not
> match the template-template overload for  TT<typename>.  Also, the
> EDG front-end started to support template-template parameters only
> very recently, so I expect regression failures on Alpha and IRIX.)
> 
>  - I'll have a look at the <iosfwd> issue during the final
> integration.
> 
>  - We must solve the subdirectory and sub-namespace issue sooner
> or later, but not necessarily during this review.  (I'm happy
> with boost/math/. "special_functions" has a well-defined meaning
> therein, I believe.)  While moving around things later is quite a
> pain, I feel that these issues should not prevent the integration
> of the current library.

                OK. Will move the directory special_functions, and the files 
quaternion.hpp and octonion.hpp into a directory math.

> However, as usual, the final decision is up to the author.

                I am also working on Peter Schmitteckert's request for 
avoidance of overflow for this version (some aspects tied to NaN and 
inf may have to be defered, though).

> Thanks to all reviewers.
> 
> Hubert, I'll be glad to integrate the source into the CVS when
> you send me the final version by private e-mail.

                My pleasure! Given my current schedule (and the fact that 
modifying the doc always takes longer than I expect), it should be done 
for the end of next week at le latest.

> I would like to remind everyone that replacing uploaded files
> under active review is discouraged, because it adds
> difficulties to the review process in that it requires
> tracking a moving target.
> 
> Jens Maurer
> Review Manager Special functions, Quaternions, Octonions

        Avec mes remerciements!

                        Hubert Holin
                        Hubert.Holin_at_[hidden]