[postgis-devel] Making SFCGAL mandatory

[Hugo Mercier]

On 14/11/2018 19:51, Paul Ramsey wrote:
> 
> 
>> On Nov 14, 2018, at 10:36 AM, Raúl Marín Rodríguez
>> <rmrodriguez at carto.com <mailto:rmrodriguez at carto.com>> wrote:
>>
>> > But at some level it should be possible - so if someone wants to drop in boost-geometry they could, and won't have to be a part of PostGIS codebase. 
>>
>> And I guess this means we could have a GEOS-less build (would
>> it provide something like 'postgis-backend-geos'?).
>>
>> Would this kind of runtime selection/dependency make it harder to
>> debug reported issues? If so (my guess), I'd tread carefully.
> 
> More to the point, I think, is that adding levels of indirection
> increases the complexity of the whole thing, for relatively niche use
> cases. We had a lot of debugging issues arise out of the backend
> selection GUC, and I bet approximately nobody uses it. It would have
> been far more effective to just have the 6 functions in question added a
> ST_IntersectsCGAL() etc, etc, in terms of the real usage of PostGIS. The
> backend switch was too smart by 1/2.
> 

Probably. I don't remember exactly why we chose to use GUC. Functions in
their own schema like sfcgal.ST_Intersects() could have been a solution
as well (?)