Boost logo

Boost :

Subject: Re: [boost] numeric directory/namespace
From: Stefan Seefeld (stefan_at_[hidden])
Date: 2018-08-14 22:58:47


On 08/14/18 18:42, Robert Ramey via Boost wrote:
> I'm preparing to make adjustments to safe_numerics directory and
> namespace in preparation to incorporating safe_numerics into the
> official boost distributions.  I have a couple of decisions to make
> that someone/everyone will want to weigh in on.
>
> There is already a directory boost_root/libs/numeric which seems to be
> an "umbrella" for libraries dealing with numerics.  It includes
> conversion, interval, odeint and ublas.  There is an index.html in
> there also.  But instead of referring to the above libraries, it
> points to the ublas documentation.
>
> So it's sort of tempting the place safe numerics in libs as a sub
> directory.  But I really have strong reservations about the whole
> concept of have this "umbrella" directory in the first place.
> Personally I'd just prefer to use boost_root/libs/safe_numerics.  I
> don't think the extra structure adds anything and makes it harder to
> implement tools which scan the directories.  It would also mean that I
> should ? use #include <boost/numeric/safe_numeric/ ....> and I'm not
> sure want to do that.
I fully agree, for all the reasons you provide. I've been getting into
the ublas project over recent months (mentored two GSoC students working
on Boost.uBLAS), and was thinking about proposing to move it out of the
'numerics/' subdirectory (with suitable proxy headers for backward
compatibility, of course).

>
> Also I expect I'll want to sync the namespace
> boost::numeric::safe_numeric:: ... with the file system.

Right.

>
> It just seems more complex and flexible than it has to be.  I see the
> future of boost as diminishing coupling between libraries and I see
> the current setup as not helping with this.

Yes !!

>
> Please don't wait to answer until I've made the changes.

Cheers,

Stefan

-- 
      ...ich hab' noch einen Koffer in Berlin...
    

Boost list run by bdawes at acm.org, gregod at cs.rpi.edu, cpdaniel at pacbell.net, john at johnmaddock.co.uk