Geometry Boost List: Re: [geometry] Linestring buffer

Subject: Re: [geometry] Linestring buffer
From: Barend Gehrels (barend_at_[hidden])
Date: 2012-07-22 11:13:58

Hi,

On 19-7-2012 17:43, aimes00 wrote:
> Hi Barend,
>
> First of all, I don't need rounded end-caps at all! Maybe is an important
> feature for some people but I think that it the least important thing!
>
> I've been testing the algorithm for linestrings and it seems to work ok.
> I've test around 12000 geometries and I've only got 197 errors. It could
> seem a bad result but it is less than the 2% of the tests.
>
> Errors occur when, I don't know why, the algorithm return an empty list.
>
> My test are based in geographic coordinates (over carthesian-based
> geometries) and I'm trying to make a buffer of 0.000018 in both sides
> (around 2 meters). The lines don't have any strange thing, so I don't have
> any clue about the failure. I will try to attach the document which contains
> the lines that make it fail. Maybe you'll find the issue quicker than me.
> http://boost-geometry.203548.n3.nabble.com/file/n4025063/BufferProblems.txt
> BufferProblems.txt

I can reproduce this. I tried the first 20 cases and 18 of them failed 
for me (the second and 11th were OK somehow). This needs more attention 
from my side. I cannot yet say why these cases are failing. I will come 
back to this.

Thanks again for your clear report - enough to do...

Regards, Barend