First: KUDO !!! The release notes are extremely promising in regard to performance improvements :-)
I've made some (dirty) tests with BRIN and btree_gin. (on a smalll Windows laptop ...)
just a few remarks:
- btree_gin deserve a better description than that:
"However, they are useful for GIN testing and as a base for developing other GIN operator classes."
I came to similar times between btree and gin for indexes on "category" columns (ca 20 to 200 distinct values) For me, gin clearly wins here thanks to the index size difference.
I reached the same conclusion for things with higher distinct values, but still several copies of each distinct value. except I don't think we should change the description until the BETWEEN issue is resolved. That is a pretty serious limitation, I think.
Or at least, if you we invite people to use it for this purpose, we would have to warn them that it is not suitable for range queries. It wouldn't be so bad if it merely didn't support them well, but as things are now it positively pulls the planner away from better options, because it looks falsely attractive.
I've looked into doing myself, but I'm afraid it is beyond me.
