Fix NaN in distance calculation

[andreynovikov]

Existing implementation produces NaN for some points. Reference URL gives 404 and can not be checked anymore. Provided implementation can produce NaN too (by design) but works for all my tested cases (about million points). Compared by execution time methods are nearly equal - several milliseconds for 30000 points.

[devemux86]

We have various distance calculations in Mapsforge LatLongUtils class.
Is the Vincenty code from there?

[andreynovikov]

Don't think so. I've took it from my old application that I wrote long before I've got to know Mapsforge. I do not remember exactly but looks like I've adopted some JavaScript code that is widely distributed over Internet: http://stackoverflow.com/questions/12800045/distance-using-wgs84-ellipsoid/20615459#20615459

[devemux86]

Can you test LatLongUtils code?
Methods content seems similar and I'd prefer to borrow the existing from Mapsforge, so we can have a common base or fix it if there is a problem. 😄

[andreynovikov]

Ok, I'll test.

[devemux86]

I integrated Mapsforge API in #201.

So closing here, please test new methods.

[andreynovikov]

Everything works, thanks.

[devemux86]

Great, thanks for the feedback!