Re: Would Oracle use index on count(*)?

From: Jim Morgan <jimmorgan_at_csi.com>
Date: Tue, 14 Apr 1998 09:45:39 -0400
Message-ID: <6gvp7k$1cos$1@rtpnews.raleigh.ibm.com>

I saw this thread late, but I ran into this once before and was wondering why count(*) took so long. I tried count(colname) but that performed just as poorly. It kind of makes sense that count(primarykey) would use the index and thus perform much better, but what if you had a compound primary key?

Is there such a thing as count(col1, col2, col3) where col1, col2 and col3 make up a composite primary key?

Also, I imagine the discussion of count(column) using an index holds true whether you're talking about a primary key or ANY index. True?

--
Regards,
Jim

Jonathan Lewis wrote in message <01bd67a7$5b5c31d0$294b989e_at_WORKSTATION>...
>
>
>
>Eugen Nyffeler <eugen.nyffeler_at_ubs.com> wrote in article
><35331BA9.86E49A6E_at_ubs.com>...
>>
>> If you use a count(*) the full table scan is coosen. If you use
>> count(primary key field)
>> an index scan shoud be used. I don't know the exact Oracle
>internals for
>> this, but
>> i think one reason is to make the codeing simple (e.g. what do you
>do
>> when there
>> are several indexes but none is unique, or even indexes with null
>value
>> ??)
>>
>
Received on Tue Apr 14 1998 - 08:45:39 CDT