[postgis-users] Problem with Contains in Postgis 1.2.1

Michael Fuhr mike at fuhr.org
Sat May 26 20:56:15 PDT 2007

On Tue, May 15, 2007 at 04:45:34AM -0600, Michael Fuhr wrote:
> Regarding the Contains() problem discussed in:
> http://postgis.refractions.net/pipermail/postgis-users/2007-April/015317.html
> I get true (as expected) for the test case in 1.2.2SVN.  However,
> the second of the following examples appears to be wrong.  The only
> difference between the polygons in each example is in the third and
> fourth points.  Relate() returns the same relationship in all three
> examples but Contains() incorrectly reports false in Example 2,
> where X in the third and fourth points of the polygon matches X in
> the point.
> select Contains(GeomFromText('POLYGON((0 0,0 20,10 20,10 30,30 30,30 0,0 0))'),
>                 GeomFromText('POINT(10 10)'));
>  contains 
> ----------
>  f
> (1 row)

Here's an excerpt of this query's output with PGIS_DEBUG_CALLS and
PGIS_DEBUG defined:

NOTICE:  point_in_polygon called.
NOTICE:  point_in_ring called.
NOTICE:  side result: -200.000000
NOTICE:  side result: -100.000000
NOTICE:  side result: -0.000000
NOTICE:  point on ring boundary between points 2, 3
NOTICE:  point_in_polygon: outside exterior ring.

The problem seems to be that the point (10 10) is incorrectly
considered to be on a ring boundary because it's neither left nor
right of the segment (10 20,10 30).  Isn't the following exit
condition in point_in_ring() premature?

   /* a point on the boundary of a ring is not contained. */
    if(fabs(side) < 1e-12) 
            lwnotice("point on ring boundary between points %d, %d", i, i+1);
            return 0;

Michael Fuhr

More information about the postgis-users mailing list