Re: Multicolumn index for single-column queries? - Mailing list pgsql-general

From Harald Fuchs
Subject Re: Multicolumn index for single-column queries?
Date
Msg-id 87muknwhle.fsf@protecting.net
Whole thread Raw
In response to Multicolumn index for single-column queries?  (rihad <rihad@mail.ru>)
Responses Re: Multicolumn index for single-column queries?  (Gavin Flower <GavinFlower@archidevsys.co.nz>)
Re: Multicolumn index for single-column queries?  (Gavin Flower <GavinFlower@archidevsys.co.nz>)
List pgsql-general
Andreas Kretschmer <andreas@a-kretschmer.de> writes:

> Am 18.04.19 um 08:52 schrieb rihad:
>> Hi. Say there are 2 indexes:
>>
>>     "foo_index" btree (foo_id)
>>
>>     "multi_index" btree (foo_id, approved, expires_at)
>>
>>
>> foo_id is an integer. Some queries involve all three columns in
>> their WHERE clauses, some involve only foo_id.
>> Would it be ok from general performance standpoint to remove
>> foo_index and rely only on multi_index? I know that
>> PG would have to do less work updating just one index compared to
>> updating them both, but wouldn't searches
>> on foo_id alone become slower? 
>
> it depends .
>
> it depends on the queries you are using, on your workload. a
> multi-column-index will be large than an index over just one column,
> therefore you will have more disk-io when you read from such an index.

I think it also depends on the average number of rows having the same foo_id.




pgsql-general by date:

Previous
From: Tom Lane
Date:
Subject: Re: PostgreSQL ping/pong to client
Next
From: Ron
Date:
Subject: Re: Multicolumn index for single-column queries?