Documentation - are SP-GiST indexes really better for overlapping bounding boxes?

[Laurenz Albe]

​The documentation claims that

The performance tests reveal that SP-GiST indexes are especially beneficial when there are many overlapping objects, that is, with so-called “spaghetti data”.

That is counter-intuitive, given that SP-GiST indexes use disjoint bounding boxes, and ​Paul Ramsey's recent blog post confirms that:

When there is less overlap, then SPGIST will out-perform GIST.

Perhaps the documentation should be fixed.

[robe]

I think the Spgist spatial index was added by MobilityDB group and presumably the description of it's use was probably added by them. I personally haven't noticed much of a difference between spgist and gist indexes for most PostGIS cases. Perhaps pramsey can update the docs accordingly since I have no such metric that they are better.

[Paul Ramsey <pramsey@…>]

In 3b1a12e/git:

Remove un-backed assertion that SPGIST indexes are best for highly overlapping data. Closes #4907