[postgis-users] GeomUnion Bug?

[Martin Davis]

Charlie,

Two observations:
- This does look like a bug, or at least a porting problem in GEOS.  In 
JTS, the envelope of a Point is itself a Point.  This means that the 
return value is always a valid geometry, and thus can be safely used in 
further operations.

- Are you sure you want to compute the union of the bounding boxes, 
rather than the bounding box that contains them both?  If the latter is 
the case, perhaps the ST_extent function is what you want.  (In raw 
GEOS, there should be an Envelope.expandToInclude function which does this)

Martin

Charlie Savage wrote:
> Hi Carl,
>
>> Charlie ,
>>
>> neither of those polygons are valid
>>
>> repeating points, self intersecting and the second one is not a 
>> closed linestring
>
> Interesting - that is what PostGis/Geos produces when you do this:
>
> select astext(envelope(GeomFromText('POINT(-104 40)')))
>
> "POLYGON((-104 40,-104 40,-104 40,-104 40,-104 40))"
>
> But you are right, PostGis doesn't think they are valid:
>
> select isvalid(envelope(GeomFromText('POINT(-104 40)')))
>
> False
>
> And this did work in older versions of GEOS - maybe it shouldn't have.
>
> So here is my problem then.  I have a map, a user enters a point.  I 
> want to store the bounding box.  So I use the envelope function as 
> above.  User enters another point.  Find its bounds.  Then union them 
> together.  I don't want to recalculate the bounds based on all the 
> points each time because its too expensive finding all the constituent 
> geometries.
>
> Ideas?
>
> Charlie
>
>
> ------------------------------------------------------------------------
>
> _______________________________________________
> postgis-users mailing list
> postgis-users at postgis.refractions.net
> http://postgis.refractions.net/mailman/listinfo/postgis-users
>   

-- 
Martin Davis
Senior Technical Architect
Refractions Research, Inc.
(250) 383-3022