COGNOiSe.com - The IBM Cognos Community

Legacy Business Intelligence => COGNOS ReportNet => Topic started by: bogdaboco on 12 Jan 2006 04:57:00 AM

Title: [Solved] Design: a very bad relationship to make a star shema...
Post by: bogdaboco on 12 Jan 2006 04:57:00 AM
Hello,

Sorry for beginning this topic in a wrong place, but I was lost.
And I am still lost:

How can I break a relationship 1.1 ---- 1.n with facts on each side (in each table) to make a star shema?

(cause if I keep the two tables with their facts, when I make a list with a fact of each table, one of the two facts (on the 1.1. side ) is duplicated n times.... )

Thx for your answers...

Title: Re: Design: a very bad relationship to make a star shema...
Post by: BIsrik on 12 Jan 2006 05:29:16 AM
try setting 1:1 ---1:1 on the both sides of relationship.

Srik
Title: Re: Design: a very bad relationship to make a star shema...
Post by: bogdaboco on 12 Jan 2006 06:27:58 AM
Already done, and it doesn't change anything... :(
Title: Re: Design: a very bad relationship to make a star shema...
Post by: Boris-A on 12 Jan 2006 09:46:02 AM
You should NEVER have facts linked directly to other facts.  The links should go through dimentions (with dimensional information defined).  I believe this is explained in the FM documentation.

Often the Time or products dimentions should be used to make those links.

Hope this was of some help
Title: Re: Design: a very bad relationship to make a star shema...
Post by: bogdaboco on 13 Jan 2006 09:06:33 AM
Effectively, I had to redefine my star shema. But this time,  I won't be easy...

Thx for your responses...





Title: Re: Design: a very bad relationship to make a star shema...
Post by: BIsrik on 16 Jan 2006 04:14:13 AM
plz update the topic...

Srik
Title: Re: [Solved] Design: a very bad relationship to make a star shema...
Post by: MFGF on 23 Jan 2006 06:07:56 AM
Hi,

This is an example of the need to create a "degenerate dimension".  In Framework Manager, create a new query subject (degenerate dimension) containing the common key(s) of the two fact tables.  Delete the direct relationship between the two fact query subjects, and replace with relationships from the degenerate dimension to each fact (1..1 --- 1..n in each case).  This should then result in a stitch query if you bring in data from both facts in a single query, and remove the double-counting you see currently.

Best regards,

MF.