Re: GIN Index for low cardinality

Here is a real-world example from one of my databases where each value is about 17 characters long, and is present about 20 times:

gin: 411 MB
btree: 2167 MB
hash: 2159 MB

For what it is worth, that was 9.6 with freshly rebuilt indexes.

On 11.0 with freshly rebuilt indices, the hash index does get a bit smaller:

gin: 411 MB
btree: 2167 MB
hash: 1864 MB

Cheers,

Jeff