Boost logo

Geometry :

Subject: Re: [geometry] nsphere roadmap
From: Barend Gehrels (barend_at_[hidden])
Date: 2012-11-27 17:15:49

Hi Adam,

On 27-11-2012 0:10, Adam Wulkiewicz wrote:
> Barend Gehrels wrote:
>> There are many things to be finished and nsphere is certainly one of
>> them. However, if you want to do intersections with them, the best thing
>> is to convert them to polygons and use that. That is probably what will
>> be done internally too - the intersection between two circles is (often)
>> a polygon anyway, and we are far from supporting polygons with curved
>> segments.
> I have only one concern. The implementation of intersects() (and other
> boolean operations) would be typically very simple and fast for two
> circles. But it won't be consistent with polygon version of
> intersection(). I guess the way how intersects() checks geometries
> will be defined by a strategy?

True, I meant the output is a polygon. The intersection points are (will
be) calculated using circle/segment intersections, which is much faster,
and for "intersects" it can stop at the first intersection (as it
currently does too for poly/poly).
For intersection circle/circle calculating the intersection points is
also different, but the output (if any) is again a normal polygon (as
long as we don't support curved segments).

Thanks, Barend

Geometry list run by mateusz at