Fixed distance computation between two coordinates

[Themacprod]

Hi,

• index.js : added the earth radius in the distance computation.
• test.js : fixed unit test + adding new tests

[jchristin]

Dear Turf Team,

Could you give us some inputs for this pull request? I am also interested in the fix.

Thanks,
Julien

[morganherlocker]

Sorry for the delay. This PR got lost in the shuffle. I'll take a look at this tomorrow.

[morganherlocker]

Ok, looking things over, my only concern is the performance drop. I'm going to tinker with it a bit and see if it can be made faster.

old

turf-distance x 4,941,766 ops/sec ±1.03% (93 runs sampled)

new

turf-distance x 3,596,891 ops/sec ±1.08% (92 runs sampled)

[morganherlocker]

Ok, I'm a bit confused here. Can you lay out a specific test case that was failing, that is now fixed? This method may be better, but I'm having to dig through a lot to determine what the relevant changes are. The example you linked to has a lot going on (buffers, points, etc. which are not related).

Basically just a test case with a couple screenshots is all I need to move forward.

[Themacprod]

Hi,

I think you should do your tests at a good distance from Ecuador, it will show you the impact of the latitude in the distance computation.

I've done a small jsfiddle to explain the issue (http://jsfiddle.net/Themacprod/r2fp5t66/3/)

It is easier to explain by drawing a turf-buffer (red buffer) :

If you trace a non-eleptical circle (green buffer) , you should have a buffer like that (I put it above the turf-buffer circle)

I placed 3 points :

• doesn't expect a featurecollection of points #1 : On the same latitude than the center maker AND on the green buffer
• critical distance calculation fix #2 : On the same longitude than the center marker AND on the green buffer
• Add degrees and radians to units parameters #3 : On the same latitude than the center marker AND on the red buffer

Here are my results
Distance from #1 (turf) = 0.113 km | Me = 0.4 km
Distance from #2 (turf) = 0.572 km | Me = 0.4 km
Distance from #3 (turf) = 0.405 km | Me = 0.283 km

Here are the Google result for the distance :

[morganherlocker]

Ok, the issue I'm having understanding your suggestion here is that turf-buffer does not use turf-distance in any way. The fact that buffers are elliptical is completely unrelated; that is due to the fact that JSTS uses planar geometry calculations instead of geodesic (something I am working on fixing in turf-buffer). Comparing the two buffers here, does not show an issue with turf-distance. turf-destination also uses geodesic calculations. Run turf-destination around a point, and you will see a nice perfect circle.

turf-distance already employs geodesic calculations. Basically what I need is a specific case where the the distance between 2 points is correct in some other program, but incorrect in turf-distance.

[Themacprod]

Here is a more explicit sample:

Marker #1 = [45.535017 , -73.645906 ]
Marker #2 = [45.53835 , -73.586447 ]

According to Google the distance between those 2 points is 4.65 km :

var Lat1 = 45.535017,
Lng1 = -73.645906,
Lat2 = 45.53835,
Lng2 = -73.586447;

var turfpoint1 = turfpoint([Lat1, Lng1], {name: 'Location 1'});
var turfpoint2 = turfpoint([Lat2, Lng2], {name: 'Location 2'});

var distanceturf1 = turfdistance(turfpoint1, turfpoint2, "kilometers");

console.log('Distance = ' + Math.round(distanceturf1 * 1000) / 1000 + ' km');

Distance = 6.614 km

With my changes I get 4.635 km (Simplified jsfiddle :http://jsfiddle.net/rq0b0gfb/2/)

[Themacprod]

Ok sorry guy, I feel a little bit stupid, unfortunatly I swap latitude and logitude in the turf point....
Forget this pull-request

[morganherlocker]

@Themacprod no worries at all. This stuff is really difficult to wrap one's head around (for me too). Thank you so much for the effort and following up with the tests/screenshots/examples/explanations/etc.

Seriously, investigations like this are all a part of the process, and hugely beneficial even when a PR does not ultimately get merged.

[nevi-me]

If I'm not too late; @morganherlocker what would be the ideal case in @Themacprod's last example? Should we get the same distance as what Google (and few others) get? Because that would be a matter of changing R to the value in https://github.com/Turfjs/turf-distance/issues/4.

I saw this PR last night but couldn't understand it, so thought I'd look at it this morning. The red flag for me was @Themacprod not substituting the current R with a different value, but calculating it each time, but glad that's resolved 👍