SEO

#1 (**permalink**) 04-07-2008, 08:38 AM

I found this thread about whether to update fact tables in data
warehouses, other than for errors in the data.

http://groups.google.com/group/micro... b42e0ce71039

The thread was from 2003, so I thought I would start this one.

It seems to me that the reason you wouldn't update data in a data
warehouse is that each row represents an event, and if the data has to
be updated later on to reflect something that was unknown when the
record was created, it was probably because a new event has occured to
create the piece of data.

Any thoughts?

#2 (**permalink**) 04-09-2008, 04:49 AM

"Joe" <darwinshamster@gmail.com> wrote in message
news:5a38ed1b-62b3-4190-b62c-95198877f594@h1g2000prh.googlegroups.com...
>I found this thread about whether to update fact tables in data
> warehouses, other than for errors in the data.
>
> http://groups.google.com/group/micro... b42e0ce71039
>
> The thread was from 2003, so I thought I would start this one.
>
> It seems to me that the reason you wouldn't update data in a data
> warehouse is that each row represents an event, and if the data has to
> be updated later on to reflect something that was unknown when the
> record was created, it was probably because a new event has occured to
> create the piece of data.
>
> Any thoughts?

If an update is needed to produce some correct and useful result then do the
update. Unless you have some specific data model or universe of discourse in
mind I don't see any other basis for saying whether an update is appropriate
or not.

--
David Portas

#3 (**permalink**) 04-09-2008, 05:06 PM

On Apr 8, 5:02*pm, "David Portas"
<REMOVE_BEFORE_REPLYING_dpor...@acm.org> wrote:
> "Joe" <darwinshams...@gmail.com> wrote in message
>
> news:5a38ed1b-62b3-4190-b62c-95198877f594@h1g2000prh.googlegroups.com...
>
> >I found this thread about whether to update fact tables in data
> > warehouses, other than for errors in the data.
>
> >http://groups.google.com/group/micro...er.datawarehou...
>
> > The thread was from 2003, so I thought I would start this one.
>
> > It seems to me that the reason you wouldn't update data in a data
> > warehouse is that each row represents an event, and if the data has to
> > be updated later on to reflect something that was unknown when the
> > record was created, it was probably because a new event has occured to
> > create the piece of data.
>
> > Any thoughts?
>
> If an update is needed to produce some correct and useful result then do the
> update. Unless you have some specific data model or universe of discourse in
> mind I don't see any other basis for saying whether an update is appropriate
> or not.
>
> --
> David Portas

This article doesn't have quite the approach I was thinking of, but
does show how to handle late arriving facts without updating the fact
table.

http://www.kimballgroup.com/html/des...ensionRows.pdf

Joe

Thread Tools
Show Printable Version Email this Page
Display Modes
Linear Mode Switch to Hybrid Mode Switch to Threaded Mode

SEO

Updating data in a data warehouse