[gdal-dev] Proposed changed to solve inconsistant spatial filtering for null or empty geometries

[Edzer Pebesma]

On 11/01/2017 12:15 PM, Ari Jolma wrote:
> Even Rouault kirjoitti 01.11.2017 klo 13:02:
>>
>> Hi,
>>
>>  
>>
>> Nyall Dawson pointed to me in
>>
>> https://github.com/qgis/QGIS/pull/5501#issuecomment-340934947 about a
>> weird behaviour of OGR spatial filtering regarding null and empty
>> geometries.
>>
> ...
>>
>>  
>>
>>  
>>
>> Any thought ?
>>
> 
> Yes. Deep ones. If something is 'empty', is it 'a thing'? Can 'nothing'
> exist? Weird things should be expected to behave weirdly.

I think of it as the "there is no such thing for which ..." answer to
the question "is there a thing for which ...", e.g. if you'd do a left
join with a spatial predicate and there is no match.

And I agree, "no such thing" should never come out of a spatial filter.

I never understood the value of having all these typed EMPTYs in SFA,
and whether/how they'd differ from NULL geometries in OGR.

> 
> I would say such things should disappear, i.e., the PostGIS behavior is
> logical.
> 
> Ari
> 
> 
> 
> _______________________________________________
> gdal-dev mailing list
> gdal-dev at lists.osgeo.org
> https://lists.osgeo.org/mailman/listinfo/gdal-dev
> 

-- 
Edzer Pebesma
Institute for Geoinformatics
Heisenbergstrasse 2, 48151 Muenster, Germany
Phone: +49 251 8333081