Thread: Seq Scan cost shown to be (cost=10000000000.00..10000000001.10)

Hello,

I set enable_seqscan=off and also accidentally dropped the only index
on a table (actually, drop extension pg_bigm cascade) and observe
following:

postgres=# explain select * from testdata where name like '%gi%';
                                 QUERY PLAN
-----------------------------------------------------------------------------
 Seq Scan on testdata  (cost=10000000000.00..10000000001.10 rows=2 width=71)
   Filter: (name ~~ '%gi%'::text)
(2 rows)

Although, I suspect the (dropped index + enable_seqscan) causes this,
is the cost shown in explain output some kind of default max or
something like that for such abnormal cases?



--
Amit Langote

On Mon, May 27, 2013 at 12:42 AM, Amit Langote <amitlangote09@gmail.com> wrote:
> I set enable_seqscan=off and also accidentally dropped the only index
[...]
>  Seq Scan on testdata  (cost=10000000000.00..10000000001.10 rows=2 width=71)
[...]
> Although, I suspect the (dropped index + enable_seqscan) causes this,
> is the cost shown in explain output some kind of default max or
> something like that for such abnormal cases?

When you set enable_xxx=off, it not actually disables the xxx
operation, it sets the start cost to the high value (10000000000).

--
Kind regards,
Sergey Konoplev
PostgreSQL Consultant and DBA

Profile: http://www.linkedin.com/in/grayhemp
Phone: USA +1 (415) 867-9984, Russia +7 (901) 903-0499, +7 (988) 888-1979
Skype: gray-hemp
Jabber: gray.ru@gmail.com

>> Although, I suspect the (dropped index + enable_seqscan) causes this,
>> is the cost shown in explain output some kind of default max or
>> something like that for such abnormal cases?
>
> When you set enable_xxx=off, it not actually disables the xxx
> operation, it sets the start cost to the high value (10000000000).
>

Oh, okay, thanks!

--
Amit Langote

> When one uses “enable_” settings to adjust planner behavior, PostgreSQL
> just sets really high costs for the operations affected (like the one you
> see).
>
> As SeqScan is the only possible way to execute your query, it is still
> choosen.
>

I get it. Thanks!

--
Amit Langote