"Heikki Linnakangas" <heikki@enterprisedb.com> writes:
> Gregory Stark wrote:
>> For i/o-bound databases with very large indexes there should be an opportunity
>> where btree lookups are O(logn) and hash lookups can in theory be O(1).
>
> Ignoring the big-O complexity, if a hash index only stores a 32-bit hash code
> instead of the whole key, it could be a big win in storage size, and therefore
> in cache-efficiency and performance, when the keys are very long.
I think it has to show an improvement over an expression index over
(hashany(col)) and not just an improvement over an index over "col" due to col
being large.
-- Gregory Stark EnterpriseDB http://www.enterprisedb.com Ask me about EnterpriseDB's 24x7 Postgres support!