Re: [boost] [Review Request] Inclusion of the Boost.Polygon Voronoi Library

Subject: Re: [boost] [Review Request] Inclusion of the Boost.Polygon Voronoi Library
From: Andrii Sydorchuk (sydorchuk.andriy_at_[hidden])
Date: 2012-04-14 19:35:20

> Binaries for command-line versions of the algorithm are freely available
>> for Mac OSX, Windows, and Linux.
>
>
> Seems that I'll be able to do comparison with S-Hull-Pro also!
>

Those guys apparently didn't test their commercial version evaluation
binaries:

Input:
3 2 points
1 1
-1 -1
1 -1

Result:
Segmentation fault

Input:
4 2 points
10000 10000
-10000 -10000
10000 -10000
-10000 10000

Result:
linear structure, aborting. // Don't know what that means, but it doesn't
generate any output. Looks like they don't support coordinates with
absolute value higher then 10 000.

I hope those evaluation binaries don't correspond to the actual code,
because it looks quite disappointing. I'll try to find if they have some
documentation describing such behavior or list of limitations input set of
points should satisfy. Also by default their binary generates triangulation
of 20 000 random points and it takes 0.071 seconds on my machine (it also
saves that data to the file, so some time should be subtracted).
Boost.Polygon Voronoi constructs Voronoi diagram of 20 000 random points
within 0.043 seconds and that with coordinates spread through the whole
signed integer range plus coordinates of the inscribed circles are at place
within 64 EPS precision.

Andrii