[postgis-users] Problem : Within() incorrectly returns 'false'

[Arnaud Lesauvage]

Brent Wood a écrit :
> Mark Leslie a écrit :
>> Arnaud,
>> I've managed to track down the bug.  It seems that the algorithm
>> determines any point to be within a zero length segment, which your
>> polygon had a pair of.
>> I've made the fix and committed it.
> 
> A zero length segment implies a duplicate vertex. 
> 
> I understood such a situation to be an invalid topology, so should get picked
> up by isvalid(). Where a feature is invalid() a wrong answer from a query is
> always possible, and is not necessarily a bug (although it is nice to get these
> fixed :-)  
> 
> Are zero length segments one of the conditions picked up by isvalid()?

I haven't checked my polygon for 0-length segments but IsValid returns true, 
that's the first thing that my script checks.

--
Arnaud