Re: redundent index?

On Wed, 29 Oct 2003 10:17:24 -0500, Rod Taylor <rbt@rbt.ca> wrote:
>On Wed, 2003-10-29 at 09:03, Robert Treat wrote:
>> Indexes: entity_watch_map_pkey primary key btree (entity_id, watch_id),
>>          ewm_entity_id btree (entity_id),
>>
>> I can't think of why the second index is there, as ISTM there is no
>> instance where the first index wouldn't be used in place of the second
>
>The cost in evaluating the first index will be a little higher

Yes, the actual cost may be a little higher.  But the cost estimation
might be significantly higher, so there can be border cases where the
planner chooses a sequential scan over a multi-column index scan while
a single-column index would correctly be recognized as being faster
...

Servus
 Manfred