Re: [boost] Re: boost/operators.hpp remarks

From: Sylvain Pion (Sylvain.Pion_at_[hidden])
Date: 2005-03-08 06:04:16

On Tue, Mar 08, 2005 at 09:29:31AM +0100, Daniel Frey wrote:
> Sylvain Pion wrote:
> >I have been using boost/operators.hpp recently, and found it
> >quite good at removing lines of code from my maintainance area :)
> >I have some remarks about it :
> >
> >1/
> >The mixed operators are only provided if you know the type in
> >advance and it is fixed.  For example, it is not possible for
> >example to mix with a template family of types.  E.g. let's
> >say that My_type has to provide mixed operators with all types
> >X<U>.  This is not possible to do with the current library.
> 
> Sounds reasonable, but I suspect some implementation problems. One might 
> try to provide new versions (with new names) of the 
> two-argument-template form with a template template parameter, but what 
> if X<U> is really a X<U,V=dummy<U> >?
> 
> FWIW, it seems to me that this is an unnecessary restriction of template 
> template parameters. Maybe someone of the standard gurus can explain to 
> me why I have to provide the arguments in the template declaration? IOW, 
> why not simply:
> 
> template< template X > class Y
> {
>    typedef X<int> Z; // Here(!), check if the expression is valid
>    typedef X<int,std::allocator<int> > Z2; // Again
> };
> 
> Of course, I'd like to hear about any other useful 
> technique/implementation...

The usual restriction of the template-template interface...
It is usually at this point that the rebind technique raises its ugly head :)

I am not familiar enough with MPL/enable_if/... to say if this can
be done nicely with those tools.  I just vaguely suspect it is.

> >Another related thing is how to express "I want My_type to mix
> >with all integral types".  Maybe there could be a version of
> >the operators that uses enable_if and tr1::is_integral somehow,
> >for this purpose ?
> 
> I'd rather go for a more independent version, like
> 
> template< typename T > class addable_integral<T>
>   : addable<T,int>,
>     addable<T,unsigned int>,
>     ...
> {};
> 
> [base class chaining omitted]
> 
> To sum it up, both ideas seem to be good. Do you have a patch? Something 
> that works, even if it's not yet boostified?

I do not have any patch.  I was just thinking loud about potential future
needs of mine or others.  If I get to it, I'll send a patch of course.

> >2/
> >I think it is "Euclidean Ring", not "Euclidian Ring" ("e" instead of
> >"i"), hence the names for euclidian_ring_operators<T> and similar,
> >are unfortunate.
> 
> From what I can see, both spellings seem to be used, where Google 
> offers "Do you mean Euclidean" for "Euclidian", but the latter has 
> 3.500.000 hits against 1.500.000 for Euclidean. Strange. Also, look at 
> this: <http://www.dictionary.net/euclidian>. So, my question is: What is 
> really correct? Can you (or someone else) give a reference?

Indeed both orthographs seem to be correct.  I had based my thought only on
the uses in Mathworld (http://mathworld.wolfram.com/EuclideanRing.html)
and CGAL.  Notice that the ratio is reversed when looking for
"euclidean geometry" (357,000 hits) versus "euclidian geometry" (60,000 hits).

-- 
Sylvain