Boost logo

Geometry :

Subject: Re: [geometry] bounding object calculation
From: Barend Gehrels (barend_at_[hidden])
Date: 2013-08-19 16:50:19

Hi Adam,

On 16-8-2013 2:12, Adam Wulkiewicz wrote:
> Mateusz Loskot wrote:
>> On 15 August 2013 23:07, Barend Gehrels <barend_at_[hidden]> wrote:
>>> On 15-8-2013 23:18, Adam Wulkiewicz wrote:
>>>> I propose to choose a general name and implement a function
>>>> calculating a
>>>> bounding Geometry of arbitrary type. I'm using bounds() in the
>>>> Index to
>>>> return spatial index's bounding object. For example:
> [...]
>>>> This function could calculate bounding boxes, nspheres and convex
>>>> polygons/rings and in the future Geometries currently not available
>>>> like
>>>> oriented boxes, ellipsoids, capsules, cylinders and other
>>>> frequently used in
>>>> e.g. physics engines.
>>> Agreed this name is beter and more generic. My objection was mainly
>>> that
>>> envelope should not return circles, and because it is defined as
>>> such by
>>> OGC.
>>> So yes, bounds sounds good to me as a generic term for bounding_box,
>>> bounding_circle.
>>> And yes, then envelope is just a special case of bounds.
>>> Bounds is plural, maybe just boundary is enough?
> I'll answer to Barend's email here if that's ok.
> When I think of a boundry of a geometry it's more like a perimeter or
> an outer ring of a polygon. A geometry which closely encloses the
> internals, like the boundry of a country.


> As for 'bounds'. I'm not an expert in english so feel free to prove me
> wrong, but according to:
> (2)
> (3)
> may be used in plural form e.g. to describe some approximation of
> limits or a boundry.
> Considering mathematical definition of bounds (probably this isn't
> fully correct since bounds describe sets, not geometries) we can
> probably say that 1d box represents 2 bounds (lower and upper), 2d Box
> 4 bounds (left, right, top, bottom), etc.
> But I'm open to suggestion.
>> There is also option like mbr for minimum bounding rectangle :)
> I assume that this name could be a replacement for envelope which
> returns a Box. Am I right?

Agreed but that does not need replacement ;-)

> More generic, similar name could be mbg for minimum bounding geometry.
> Or bounding_geometry.

mbg is new we should not invent new abbreviations. I'm OK with bounds.

After reflection, I don't want to block envelope as returning an
nsphere, if this is OK for everyone, OK for me too. It will not change
anything of the signature so actually it only extends is meaning to
"circular" envelopes (and spheres in 3d). So it is backwards compatible.

Regards, Barend

Geometry list run by mateusz at