[Proj] UNPP optimizations - was Re: Natural Earth projection

[support.mn at elisanet.fi]

Hello,

I started to be interested about the number of iteration steps
in the inverse (iterative) conversion.. we all need maximum
speed.. so:

you could call it (Usually Near the Previous Point)
UNPP optimization - what does it mean? Simply the code could
assume that the next point is very close the previous point converted
and since this is mostly the case speed up the inverse iteration..
so it could assume that the inverse is the previous inverse
calculated (what ever that was) and so lessen the number of
iteration steps.

I do not know if this optimization is already used but I am sure it
might be worth testing - at least in our systems the converted
points are usually very close each other? This could be behind
a switch for cases where the normal method would be faster (if
ever)?

regards: Janne.

--------------------------------------------------------------------------




OvV_HN [ovv at hetnet.nl] kirjoitti: 
> [Proj] Natural Earth projection
> Bernhard Jenny jennyb at geo.oregonstate.edu
> Thu Mar 8 18:46:20 EST 2012
> 
> 
> Hi Berhard,
> 
> Regarding a possible error in the inverse projection function Natural Earth:
> 
> Thanks for looking into this problem!
> Luckily it was my fault, excuses for that.
> 
> The problem was not quite standard (JavaScript) code in the original, which 
> I cleaned up. In this process a small typographic error was made, with large 
> consequences.
> Your code will undoubtedly pass many JavaScript interpreters, but not all.
> Especially the following two constructs are likely to give somewhere a 
> problem:
>     for(; ; )
> and
>     yc -= tol = f / fder;
> 
> My corrected code now goes all the way up to the poles, for instance:
> 
> R = 6371008.7714;
> lat = 90.0d, lon = 120.0d
> gives x = 6394608.652918302, y = 9062062.394736718
> back to lat, lon: 90.00000000000014, 120.00000000000055 d
> in 7 Newton iteration steps.
> 
> If someone could check the provided PROJ C-code, the Natural Earth 
> projection code could be added to the PROJ package, if so desired!
> 
> Oscar van Vlijmen
> 
> _______________________________________________
> Proj mailing list
> Proj at lists.maptools.org
> http://lists.maptools.org/mailman/listinfo/proj
>