Geometry Boost List: Re: [geometry] overlay_invalid_input_exception in 1.50

Subject: Re: [geometry] overlay_invalid_input_exception in 1.50
From: Barend Gehrels (barend_at_[hidden])
Date: 2012-08-31 05:16:53

Hi,

Welcome to the list!

On 31-8-2012 11:09, V wrote:
> Hello world
>
> I encounter a overlay_invalid_input_exception while using the
> boost::geometry::intersection(p1,p2,out)
>
> I don't get why it works with some polygons and not with other, any help
> would be very welcome
>
> this work :
> str1 : POLYGON((530 -6890, 3530 -6890, 3680 -8240, 530 -6890))
> str2 : POLYGON((6926 1523, 5926 -10976, 8426 -3476, -4073 -10976, 3426
> -5976, -4073 -976, 6926 1523))
>
> and not this :
> str1 : POLYGON((4280 160, 5930 1810, 6980 1510, 6980 -4490, 5930 -7640, 4280
> -7490, 4280 160))
> str2 : POLYGON((6926 1523, 5926 -10976, 8426 -3476, -4073 -10976, 3426
> -5976, -4073 -976, 6926 1523))
>
> The read_wkt(str1,p1) et (str2,p2) get the right polygon each time.
> The intersection(p1,p2,out) crashes only in the second case.
>

It does not crash but reports an exception.

Your second polygon POLYGON((6926 1523, 5926 -10976, 8426 -3476, -4073 
-10976, 3426 -5976, -4073 -976, 6926 1523)) is self-intersecting:

The first intersection with this invalid polygon succeeds, because it 
probably happens to intersect it in a non-intersecting piece of it. 
Boost.Geometry does not checks self-intersectons before the process (for 
performance reasons). However, the second intersectoin is problematic, 
and then the polygons are checked on validity.

Regards, Barend