Re: bitmap index and star transformation must be single colum?
- From: Greg Rahn <greg@xxxxxxxxxxxxxxxxxx>
- To: Adam Musch <ahmusch@xxxxxxxxx>
- Date: Thu, 7 Oct 2010 09:51:13 -0700
In dimensional modeling, keys are (should be) surrogate keys so there is
really no reason for a multi-column *dimension* PK, but that does not mean
that the logical PK of a *fact* row cant be a subset of the total
dimensional keys and thus > 1 column.
From "The Data Warehouse
Toolkit<http://www.amazon.com/Data-Warehouse-Toolkit-Complete-Dimensional/dp/0471200247>
":
"Every join between dimension and fact tables in the data warehouse should
be based on meaningless integer surrogate keys. You should avoid using the
natural operational production codes. None of the data warehouse keys should
be smart, where you can tell something about the row just by looking at the
key."
On Thu, Oct 7, 2010 at 8:47 AM, Adam Musch <ahmusch@xxxxxxxxx> wrote:
> I don't necessarily know that I agree that it's "good design" to
> always have single column primary keys - I can come up with a use case
> (VPD along a client/customer dimension) that's leading me down the
> path to multicolumn keys. Without the compound key, one can
> cross-link data from client A to client B very, very easily.
>
> --
Regards,
Greg Rahn
http://structureddata.org
Other related posts:
