2014-02-24 02:44, Andreas Karlsson <andreas@proxel.se>:
> Note: The patches do not apply anymore due to changes to
> src/backend/utils/adt/Makefile.
I will fix it on the next version.
> I see, thanks for the explanation. But I am still not very fond of how that
> code is written since I find it hard to verify the correctness of it, but
> have no better suggestions.
We can split the selectivity estimation patch from the GiST support, add
it to the next commit fest for more review. In the mean time, I can
improve it with join selectivity estimation. Testing with different datasets
would help the most to reveal how true the algorithm is.
> Do you think the new index is useful even if you use the basic geo_selfuncs?
> Or should we wait with committing the patches until all selfuncs are
> implemented?
My motivation was to be able to use the cidr datatype with exclusion
constraints. I think the index would also be useful without proper
selectivity estimation functions. Range types also use geo_selfuncs for
join selectivity estimation.