On Wed, 2003-10-29 at 09:03, Robert Treat wrote:
> I just noticed on one of my tables I have the following two indexes:
>
> 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 (more
data to pull off disk due to second item), so there may be a few
borderline cases that could switch to a sequential scan rather than an
index scan.