|
Boost : |
From: Beman Dawes (bdawes_at_[hidden])
Date: 2002-10-12 07:46:15
At 03:49 AM 10/12/2002, Guillaume Melquiond wrote:
>Hi,
>
>Now that Boost 1.29 has been released, the Interval Library will sooner
or
>later be included in the main CVS. So I have a little question related to
>that: where should the library be located?
>
>There were some discusions about it during the review. But I don't
>remember that any consensus had been reached. The library is for the
>moment located directly under 'boost/'; but it may or may not be the best
>location (I don't have a strong opinion on this matter).
>
>Before answering, just remember that the current situation isn't really
>homogeneous:
>
> - rational is located under 'boost/'
Rational was an early Boost library, and the location was chosen before we
understood how large Boost is likely to grow, and before we had a fuller
appreciation of directory and namespace issues.
That doesn't in any way mean choices made for Rational were wrong, just
that it probably isn't the best library to use as a role model.
> - quaternion and octonion are located under 'boost/math/'
> - uBlas is located under 'boost/numeric/'
>
>If the answer was to be 'boost/numeric/', I have a related question. Some
>times ago, there was a discussion about changing 'numeric/' to
>'numerics/'. I personally was against such a change since it would break
>programs already relying on uBlas; and only one Boost header was slighty
>concerned, and only with compilers lacking quality of implementation. But
>a lot of people were favourable to this change.
IIRC, that issue turned out to be a non-problem.
>However, uBlas has now been officially released in Boost 1.29 and it is
>still located under 'boost/numeric/'. So more and more progams will be
>broken by such a change. Can I expect 'boost/numeric/' to stay as it is
or
>not?
As far as I know, the decision to stick with "numeric" should hold for the
foreseeable future.
> I'm asking this question, because if the consensus is to put the
>Interval Library in 'boost/numeric/', I would like to avoid to move it
one
>more time later (since it's not a small library).
You might wait a few days (particularly since some Boosters take the
weekend off) to see if anyone comes up will a killer argument for a
different approach, but otherwise I'd say go ahead.
--Beman
Boost list run by bdawes at acm.org, gregod at cs.rpi.edu, cpdaniel at pacbell.net, john at johnmaddock.co.uk