Wrong distance calculated from pointToLineDistance

[mathildor]

Turf version:

6.5.0

Description of issue:

There seems to be a bug in the calculation of the minimum distance from a point to a line in turf. When calculating the length of a line showing the minimum distance between a point and a line I get a different value than when using the pointToLineDistance function.

Example illustrating the bug:

https://jsfiddle.net/h8c9pvfa/36/

[rowanwins]

Hi @mathildor

So I think this is a quirk of visualising data in EPSG:3857 (Web Mercator) and the data being in EPSG:4326 (WGS84).

Pulling your data into QGIS and visualising it in WGS84 shows their different trajectory. I also used QGIS to calculate the nearest point and then checked their respective lengths and they are pretty similar to what Turf is generating.

Hopefully that helps, if so feel free to close this issue :)

[mathildor]

Hi again,

I looked into this a bit further and I'm still confused.

If I use turf. nearestPointOnLine this function also gives 3,4m while turf. pointToLineDistance says 4,3m. As far as I can see in the documentation, these functions should give me the same results?

I updated the jsfiddle to show the two calculations:

https://jsfiddle.net/n5tsq0uy/4/

[mathildor]

Also, won't looking at the data in QGIS in epsg:4326 be misleading when finding the nearest point, since the coordinate-system has different scaling in the x- and y-axis? I think the point you're showing as the nearest point in QGIS is only the point that looks like the closes point in that projection, due to the scaling.

[mathildor]

@rowanwins have you had time to look further into this?

[zo0m]

Faced with the same issue:

const distanceToLine = turf.pointToLineDistance(point, line, distanceOptions);
const distanceToLinePoint = turf.distance(point, turf.nearestPointOnLine(line, point), distanceOptions);

I expected that these distances should be the same. turf.pointToLineDistance - returns the wrong value

[wchargin]

This is still broken in v7.0.0. I tested by creating a line at a given heading, projecting a point a small distance away from one of the endpoints perpendicular to the line, and comparing "distance to endpoint" with "distance to line". The error is effectively zero sometimes, but approaches 2% in my testing, varying continuously with the heading of the line:

Script: https://gist.github.com/wchargin/d673b48b8accf3027d47e644c4f7dddf

Here is a simple single example, for which PostGIS 3.4.2 (GEOS 3.12.1) says that the distance from point to line is 1000.26m but Turf says that it is 1017.6m:

$ cat test.sql
SELECT ST_Distance(
    ST_GeomFromGeoJSON('{"type":"Point","coordinates":[11.991689565382663,34.00578044047174]}')::geography,
    ST_GeomFromGeoJSON('{"type":"LineString","coordinates":[[12,34],[11.993027757380355,33.99311060965808]]}')::geography
);
$ psql -tf test.sql
 1000.26221222

Data: https://gist.github.com/wchargin/94c4be94b525fdb6ea022fcca4d6894d

(This datum was generated by asking Turf to give me a point that's 1000m off the line, so PostGIS seems more likely correct here.)

I do also find that this patch reduces the max error in my tests from 1.7e−2 to 1.5e−9, an improvement of roughly ten million times:

-  const lineDistance = turf.pointToLineDistance(target, line);
+  const lineDistance = turf.nearestPointOnLine(line, target).properties.dist;

Could pointToLineDistance be made to just call nearestPointOnLine, since that seems more accurate?