Subject: Re: [boost] GGL Review
From: Barend Gehrels (barend_at_[hidden])
Date: 2009-11-16 12:31:53
Hi Brandon, Jonathan,
> I believe that this is because for many use-cases, it isn't worth the
> run-time performance and development-time trade-off.
GGL provides the following approaches, let me summarize:
1) library users can use points with double, long double or float
2) for higher numerical robustness: users can call algorithms using
another calculation type (e.g. GMP or CLN, ...)
3) idem: users can use points with GMP or CLN coordinates
> There are many buggy floating point boolean op libraries out there.
> Why do we need another?
Can you explain this? I've done several benchmarks comparing 7 libraries
and I think they all reported the correct results. Do you refer to cgal,
geos, gpc, terralib, gtl (b.p.), wykobi or ggl? Or do you use them in
other domains than we've tested them?
> And are we talking about numerical floating point issues, or
> algorithmic bugs? I don't consider those to be the same thing.
I totally agree with this. Algorithmic bugs should not be there, not in
any library, of course. Numerical floating point issues can be addressed
using arbitrary arithmetic types, and besides that other measures can be
taken using the default floating point types. Only the very last thing
is not done in GGL.
Boost list run by bdawes at acm.org, gregod at cs.rpi.edu, cpdaniel at pacbell.net, john at johnmaddock.co.uk