Re: [boost] uBLAS formal review

From: Joerg Walter (jhr.walter_at_[hidden])
Date: 2002-06-28 00:26:22

----- Original Message -----
From: "Beman Dawes" <bdawes_at_[hidden]>
To: <boost_at_[hidden]>; <boost_at_[hidden]>
Sent: Friday, June 28, 2002 1:22 AM
Subject: RE: [boost] uBLAS formal review

> At 05:45 PM 6/27/2002, Powell, Gary wrote:
>
>  >  Should the include files be "filename.hpp" ? to fit the rest of boost.
>
> I'd really like to see us stick to the .hpp convention. Short term is can
> be hard on a developer used to some other scheme, but long term it is a
big
> consistency win if all boost libraries follow the same form.

Ok.

>  >Also shouldn't the include path be
>  >"boost/numerics/filename.hpp" ? doesn't that prevent more name
conflicts?
>
>  >It's been so long since I've had this issue come up.
>  >
>  >  Namespace, I'm torn whether it should be boost::numeric:ublas.... or
>  >boost::numeric::... as there may be more "vector" and matrix
>  >implementations coming along and we should be able to accomodate them.
>
> It seems that Boost continues to attract high-quality numerics
> libraries.  And that in turn will attract still more...
>
> That, coupled with the tendency for each major library to have
> "exception.hpp", etc., increases the chance of name clashes.  Thus give
> serious consideration to "boost/numerics/ublas/filename.hpp" and namespace
> boost::numeric::ublas.
>
> Then, to ease use:
>
>    namespace ub = boost::numeric::ublas;
>
> or whatever.  It helps to always use the same alias.

Ok, we'll synchronize file path and namespace. Only open question: should we
really prefix our preprocessor macros with BOOST_NUMERIC_UBLAS?

Regards

Joerg