? operator seems slower than regular gin about hstore_ops HOT 2 CLOSED

Actually, yes, it was expected. gin_hstore_hash_ops uses partial match for keys (iterates all key-value pairs for each key) while default opclass do plain entry scan in this case. gin_hstore_hash_ops could extract keys as separate entries, but I choose partial match for them in order to have smaller index which is faster for update. Also, note that you could use ?| operator in your query.

from hstore_ops.

Okay so not a bug. I'll give ?| a try. Another question I had not quite related to this one -- but I put a note on Oleg's blog about it though I think my note may have been flagged as spam.

I thought the hstore gin would get some query speed benefit in 9.4 with changes to gin, but didn't see any there in my quick tests when comparing with my 9.3 install. Though the test I had done still had my postgis geometries as part of the table. Haven't looked at size of index though. Is it certain operators and maybe I wasn't testing the right ones. Was just testing @> and ?

from hstore_ops.