topic Re: Bug in esri.geometry.geometricDensify in 2.2? in ArcGIS JavaScript Maps SDK Questions

Bug in esri.geometry.geometricDensify in 2.2?

PhillipReed — Wed, 27 Apr 2011 18:51:03 GMT

This function seems to fail sometimes for points that are very far apart (20,000 km or so -- I'm not sure exactly where the error happens). Instead of returning a Polyline that is broken up into lots of little segments, it returns a Polyline with two points, one of which is the original start point and the second of which is NaN,NaN. For example, try this:

var pline = new esri.geometry.Polyline();
pline.addPath([[-48.6, 39.4], [130.9, -39.3]]);
var result = esri.geometry.geodesicDensify(pline, 10000);
if( isNaN( result.paths[0][1][0] ) ) {
    alert("It's a bug!");
}

Changing the distance by even a short amount seems to work perfectly -- for example, replace the -48.6 above with a -48.5, and it runs just fine. I've tested it in Firefox 4.0 and Chromium 12.0.747.0 in Ubuntu Linux 10.10, and it behaves exactly the same way in both browsers.

Does anybody else see this? Anybody know what's going wrong?

Re: Bug in esri.geometry.geometricDensify in 2.2?

derekswingley1 — Thu, 28 Apr 2011 23:01:02 GMT

This is a strange one...I think it is a bug. It looks like only it happens when the difference between x coords is between 179.5 and 180.5. How'd you find this?

Re: Bug in esri.geometry.geometricDensify in 2.2?

PhillipReed — Mon, 02 May 2011 13:46:02 GMT

I think I pretty much just got unlucky; I've got some code that is drawing lines of bearing, and uses 20,000 km as the default length. Some of them just happened to have coordinates that caused the function to fail.

Re: Bug in esri.geometry.geometricDensify in 2.2?

derekswingley1 — Mon, 02 May 2011 15:53:44 GMT

Can you log this as a bug with Esri support?

Re: Bug in esri.geometry.geometricDensify in 2.2?

JianHuang — Thu, 02 Jun 2011 01:38:13 GMT

Phillip,

This is a known limit of the algorithm itself, instead of a bug. When two points are nearly antipodal to each other, the algorithm couldn't converge to get the geodetic distance and azimuth. We will "fix" the issue by implementing another algorithm when the case happens.
Thanks.