[postgis-users] polygons crossing 0/360 longitude

[Carlos Ferrão]

To calculate the two polygons I am using a PERL script on data
ingestion to build the insert statement..
If I have a polygon which is (lon/lat)
170 20, -170 20, 175 5, 170 10, 170 20
it becomes
polygon 1:
170 20, (-170+360=190) 20, 175 5, 170 10, 170 20

polygon 2:
(170-360=-190) 20, -170 20, (175-360=-185) 5, (170-360=-190) 10,
(170-360=-190) 20

Just insert the two polygons into a single row using the MULTIPOLYGON object.

Carlos.

On 6/21/06, Shane Byrne <shane at quake.mit.edu> wrote:
> Thanks Carlos, this is a good idea.
>
> Finding the problem polygons is easy enough, but how did you split them
> into two polygons along this meridian? Can I define a line and with that
> use some postgis function to split the polygon?
> I'd also like to copy over all the information in the other fields to
> each new polygon.
>
>
> Shane
>
> Carlos Ferrão wrote:
> > Hi,
> > I recently had a similar problem with products crossing the -180/180
> > longitude. The problem is that postgis doesn't connect the -180 to the
> > 180 as, for instance, Oracle Spatial Module.
> > The way I found to solve it was to do an algorithm to find the
> > polygons that cross the date line. For each one, I calculate 2
> > polygons, one in the negative coordinates and another in the positive
> > coordinates (you need to add/subtract 360). I add a multipolygon with
> > the information of the two polygons in a single row in the database
> > and it works perfectly.
> >
> > Hope it helps,
> > Carlos Ferrao
> > EOP - ESRIN - European Space Agency
> > Critical Software - www.criticalsoftware.com
> >
> > On 6/21/06, Shane Byrne <shane at quake.mit.edu> wrote:
> >> Hi,
> >> I have a postgres/postgis database with many (~80k) polygons, the
> >> vertices of which are stored in longitude (0-360) and latitude space.
> >> These polygons are typically small, but a few hundred of them span the
> >> prime meridian. This creates a problem for me when I search for polygon
> >> intersections. i.e. a polygon with a longitude range -5 to +5 is
> >> actually stored as +355 to +5 and so postgis thinks it intersects a
> >> whole bunch of polygons that it really doesn't.  I could change to the
> >> -180 to 180 longitude system but that only moves the problem to a
> >> different longitude.
> >>
> >> Is there a graceful way around this?
> >> I'm thinking of just making two tables, the other table would have
> >> longitudes ranging from -180 to +180 and doing two searches with
> >> longitude ranges chosen to avoid the problem area on each table. Does
> >> postgis have a way to handle cyclical coordinates?
> >>
> >> Thanks for any help,
> >> Shane
> >>
> >>
> >> _______________________________________________
> >> 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
> >
>
> --
>
> ______________________________________________________________
> Shane Byrne  -  University of Arizona  - Lunar & Planetary Lab
> ______________________________________________________________
> Email: shane at lpl.arizona.edu              Phone: (928)556-7235
> Web  : http://www.gps.caltech.edu/~shane  Fax  : (928)556-7014
> ______________________________________________________________
> Mail : USGS - Astrogeology Division,
>         2255 North Gemini Drive, Flagstaff, AZ 86001, US.
> ______________________________________________________________
> _______________________________________________
> postgis-users mailing list
> postgis-users at postgis.refractions.net
> http://postgis.refractions.net/mailman/listinfo/postgis-users
>