Re: best way to "parameterize" a tablename?
From: Robert Klemme <shortcutter_at_googlemail.com>
Date: Fri, 22 Aug 2008 19:36:57 +0200
Message-ID: <6h8bpqFk8o5eU1@mid.individual.net>
>
> There's a reason for it, we basically don't want to intermingle
> one set of data with another. Think of the case where there's
> a million records of FOO and 100 records of BAR. An index
> won't be very useful, so you end up doing a lot of full
> table scans to extract the data, and we don't want to pay
> the big cost of FOO for the tiny data from BAR. The real
> case is a bit more complicated, but that's what it boils
> down to.
Date: Fri, 22 Aug 2008 19:36:57 +0200
Message-ID: <6h8bpqFk8o5eU1@mid.individual.net>
On 22.08.2008 02:13, mh_at_pixar.com wrote:
>> You have an odd schema is each table has the same structure but a different >> name.
>
> There's a reason for it, we basically don't want to intermingle
> one set of data with another. Think of the case where there's
> a million records of FOO and 100 records of BAR. An index
> won't be very useful, so you end up doing a lot of full
> table scans to extract the data, and we don't want to pay
> the big cost of FOO for the tiny data from BAR. The real
> case is a bit more complicated, but that's what it boils
> down to.
Did you ever consider partitioning that table? If you partition by FOO/BAR field then you do not need an index for Oracle to just access one partition.
Cheers
robert Received on Fri Aug 22 2008 - 12:36:57 CDT