[postgis-users] MakeLine(geometry, geometry) bug with 3d points

[strk at refractions.net]

Graham, 
thanks for the report, I've fixed this.
MakeLine should work on any-dimensioned geometry pair.

Mixed-dimensions in input are also accepted, with output
holding the wider coordinate set (0-padding the vertex
w/out that information).

You can either get latest version from the pgis_1_0 branch
(reccommended) or apply the flollowing patch:

diff -U2 -r1.71 -r1.71.2.1
--- lwgeom_api.c        23 Mar 2005 16:23:45 -0000      1.71
+++ lwgeom_api.c        31 Aug 2005 16:49:21 -0000      1.71.2.1
@@ -623,10 +623,7 @@

        pa->dims = 0;
-       TYPE_SETZM(pa->dims, hasz, hasm);
+       TYPE_SETZM(pa->dims, hasz?1:0, hasm?1:0);
        pa->npoints = npoints;


--strk;


On Tue, Aug 30, 2005 at 03:49:19PM -0700, Graham Davis wrote:
> Hi,
> 
> I've been trying to write a plpgsql trigger that automatically takes two 
> 3d points and creates a linestring from them.   Within the script, I 
> have selected the 2 points to be stored in plpgsql variables using 
> SELECT INTO statements.  The variables are declared as the same type as 
> the point column (which stores 3d points).  When I insert the new 
> linestring using MakeLine(geometry, geometry), if I have setup my 
> geometry column as a 3d linestring, I get a check constraint 
> "enforce_dims_geom" violation.  The following was used to create the 
> geometry column:
> 
> SELECT AddGeometryColumn('asset_position_lines', 'geom', 4326, 
> 'LINESTRING', 3);
> 
> It seems the MakeLine() function makes the points into 2d, and the 
> resulting linestring will not be a legal entry into the table.  So I 
> tried changing the column to be 2d, and the script runs, but the 
> function obviously still results in a linestring with only 2d points.  
> Not only that, it takes the extra coordinate from the first point (z) 
> and starts off the second point with that coordinate.
> 
> For example, if point1 is (20 40 5) and point2 is (30 50 5), and I call 
> MakeLine(point1, point2), the resulting linestring will have the 
> following two points A and B:  pointA is (20 40) and pointB is (5 30).
> 
> So I changed the geometry column back to 3d, and tried to force the 
> points and resulting line into 3d using force_3d().  The result was that 
> it did the same thing as listed above (it took only the x and y 
> coordinates of the first point, then inserted 0 as the z coordinate, 
> then took the remaining coordinates of the first point and started the 
> second point with that).  So using the same points in the above example, 
> the resulting line had two points A and B as follows: pointA (20 40 0) 
> and pointB (5 30 0).
> 
> It seems like the MakeLine() function only works with 2d points.  Is 
> this something that can be fixed to work with more dimensions?
> 
> For an easy way to illustrate how this does not work, simply type this 
> query into your postgres:
> 
> SELECT AsEWKT(MakeLine(MAKEPOINT(20,40,5), MAKEPOINT(30,50,5)));
> 
> The result should be something like  LINESTRING(20 40 5,30 50 5) but it 
> actually results in this:  LINESTRING(20 40,5 30)
> 
> As you can see, it only uses the x and y coordinates, and the z 
> coordinate of the first point gets carried over into the next point.
> 
> A similar query when trying to force the points into 3d:
> 
> SELECT AsEWKT(force_3d(MakeLine(force_3d(MAKEPOINT(20,40,5)), 
> force_3d(MAKEPOINT(30,50,5)))));
> 
> results in:  LINESTRING(20 40 0,5 30 0)
> 
> Currently I've coded a work around in my script which essentially uses 7 
> functions instead of the one now, and it is much slower (almost 50% 
> slower than using MakeLine) but it works.  It basically looks like this 
> (I removed all the concatenation parts so it's easier to read for you): 
> GeomFromEWKT('SRID=4326;LINESTRING( X(point1) Y(point1) Z(point1), 
> X(point2) Y(point2) Z(point2) )');
> 
> Anyways, this work around is much too slow for what I am trying to do, 
> and a MakeLine() that works with 3d points would be much better.  Is 
> anyone able to help me with this problem?  Thanks!
> 
> -- 
> Graham Davis
> Refractions Research Inc.
> gdavis at refractions.net
> 
> _______________________________________________
> postgis-users mailing list
> postgis-users at postgis.refractions.net
> http://postgis.refractions.net/mailman/listinfo/postgis-users