RecTree

[Paul Ramsey]

> On Mar 14, 2024, at 12:47 PM, Sandro Santilli <strk at kbt.io> wrote:
> 
> Using the liblwgeom one we're saving conversion which is good too!
> It looks like a clear win to switch over.
> 
> Can you see any reason to postpone such switch or should I push it ?

Yeah, I didn’t push originally years ago because I was uncertain about performance over a whole swath of use cases, and certainly could find cases (smaller objects) where the benefit was harder to measure or even seemed to be coming up slightly worse.
We might, at a minimum, need a “sizing gate” on the tree logic, to only deploy it when the inputs are of a certain size or larger.

P
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.osgeo.org/pipermail/postgis-devel/attachments/20240314/be3c2584/attachment.htm>