Re: Design question

From: Noons <wizofoz2k_at_yahoo.com.au>
Date: Wed, 18 Feb 2009 00:18:19 +1100
Message-ID: <gned5f$q2v$1_at_reader.motzarella.org>

matthias.hoys_at_gmail.com wrote,on my timestamp of 17/02/2009 7:57 PM:

> - in a DWH, why are Time dimensions often split-up into different
> columns as day, week, year (number datatype) instead of just one Date
> column ?

Oh, I forgot:

In the original post you mention lots of queries based on "week across years", or "month across years".
That might be better handled with this type of date "soft-partitioning", rather than a simple DATE column.

It lets you do indexing tuned for the specific category of query and might also help with setting up MVs.

Given that you stated data doesn't change that often, I'd lean heavy to the MV approach as well: they are eminently suitable to situations where updates are not that frequent. Received on Tue Feb 17 2009 - 07:18:19 CST