Re: [boost] Fwd: [gsoc17] Regarding Boost.Geometry

Subject: Re: [boost] Fwd: [gsoc17] Regarding Boost.Geometry
From: Adam Wulkiewicz (adam.wulkiewicz_at_[hidden])
Date: 2017-03-21 14:03:42

Hi Aditya,

Aditya Gupta Via Boost wrote:
> Hi I am Aditya Gupta, IInd Yr. B. Tech. Computer Science at IIT Ropar and I
> wish to work on Project 3 of Boost.Geometry (Nearly antipodal points
> distance accuracy improvement.)
>
> I have completed the competency test
> <https://github.com/adityagupta1089/earth-geometry>, but I think maybe it
> was too small or have I misinterpreted, anyways.

So you used a sphere as a model of the Earth, haversine formula to 
calculate distance and compared it with Boost.Geometry 
spherical_equatorial coordinate system.

Try comparing it with Boost.Geometry geographic coordinate system. Is 
the result the same?
Why nearly antipodal points may be problematic?

Furthermore there are 3 various ways of calculating geographic distance 
in Boost.Geometry but they are not documented yet. In Boost.Geometry 
it's possible to specify how the algorithm calculates 
cordinate-system-specific parts by passing a strategy into a function as 
the last argument. By default, if you don't pass a strategy the 
algorithm use the default one (corresponding to the coordinate system of 
the pased Geometries). So if you passed geographic points one of the 
geographic strategies would be used.

See also this line of a unit test as an example, here the distance is 
calculated with one of the strategies passed explicitly into distance() 
algorithm:
https://github.com/boostorg/geometry/blob/develop/test/strategies/vincenty.cpp#L185

Regards,
Adam