[postgis-users] Buffer() Arc Tolerance

[Paul Ramsey]

Richard_D_Levine at raytheon.com wrote:

> LLOL (Lunatic LOL)
> 
> My project is safety critical.  Our safety engineers want the polygon
> scribed outside of the circle instead of inside it to ensure that the
> radius is covered at a minimum.  There's just no talking to those folks.  I
> add the arc tolerance to the radius before calling Make_Buffer_Polygon() to
> solve the problem.
> 
> If the extra parameter is exposed, I can figure out the rest.  An
> icosidodecagon is overkill for my stuff most of the time.  Using a value
> smaller than 8 should enhance performance.

You'll have to pick this up off of the 0.9 CVS branch when it gets 
added. Should be a day or so (or you can submit the patch to us yourself :)

> Does the world really need a word for 32 sided polygon?  Icosidodecagon has
> the same number of characters as 32sidedpolygon.
> 
> Thanks,
> 
> Rick
> 
> 
> Richard_D_Levine at raytheon.com wrote:
> 
> 
>>Forgive me for using the O word, but my application is currently using
>>Oracle Spatial, and I want to try it out on PostGIS.  Oracle Spatial's
>>Buffer() function (Make_Buffer_Polygon()) takes arc tolerance as a
>>parameter.  For a point, it scribes a equal-sided polygon inside the
> 
> circle
> 
>>described by the point and buffer distance (radius).  The arc tolerance
> 
> is
> 
>>the farthest distance from any side of the polygon to the circumference
> 
> of
> 
>>the (imaginary) circle.  The number of points in the returned polygon
> 
> from
> 
>>Oracle is determined by the radius of the circle and the arc tolerance.
> 
> 
> I see. I don't know if I like the way Oracle has exposed that particular
> knob (seems needlessly obfuscating the underlying action, which is the
> creation of N edges to approximate the arc). However, I can see that the
> occasional lunatic user (*cough*) might actually need to twiddle it.
> 
> 
>>>From what you said below, it appears that the PostGIS Buffer() will
> 
> return
> 
>>an octagon.  Is that correct?
> 
> 
> No, as Martin noted, the parameter refers to the number of segments per
> 90 degrees, so applying it circularly, you will get an icosidodecagon (I
> love the internet). http://www.google.com/search?q=icosidodecagon
> 
> 
>>It appears I've hit a snag, because I need to construct arbitrarily
> 
> complex
> 
>>polygons by taking the union of buffer polygons around points and
>>(multi)lines.
>>
>>Any ideas?
> 
> 
> We can add a parameter to the Buffer() postgis exposes easily enough,
> since GEOS/JTS allows this parameter to be modified.
> 
> Paul
> 
> 
>>For nitty gritty the details, sometime you have to "use the source,
> 
> Luke".
> 
>> From PostGIS source:
>>
>>     Geometry *g3 = g1->buffer(width);
>>
>>So, PostGIS is not defining a stroke angle, or using anything other than
>>the defaults, it's the GEOS code which is controlling what is done.
>>
>>Hitting the GEOS code, source/operation/buffer is the place to go, but
>>the documentation of DEFAULT_QUADRANT_SEGMENTS seems a bit thin. I can
>>tell you that the default is 8, which is what PostGIS uses.
>>
>>What do you mean by the arc tolerance? The number of vertices used to
>>approximate the circular structure?
>>
>>Richard_D_Levine at Raytheon.com wrote:
>>
>>
>>>If I use Buffer() on a point to approximate a geodetic circle, how do I
>>>determine the arc tolerance of the returned polygon to the circle?
>>>
>>>Is there any documentation that contains answers to questions like this?
>>>I've read the PostGIS documentation cover to cover and the function
>>>explanations are at a very high level (mostly copied from the OGC spec.)
> 
> 
> _______________________________________________
> postgis-users mailing list
> postgis-users at postgis.refractions.net
> http://postgis.refractions.net/mailman/listinfo/postgis-users
> 
> 
> 
> 
> 
> _______________________________________________
> postgis-users mailing list
> postgis-users at postgis.refractions.net
> http://postgis.refractions.net/mailman/listinfo/postgis-users