Cognos 8 Stitch Queries over Teradata

[lawrensi]

Does anyone have any experience of multi-fact table stitch queries over Teradata? Specifically I am seeing coalesce/full outer joins stitches in the Cognos SQL but only with 'limited local' processing switched on.Ã,  I want to force these queries down to the database but setting the data source connection to 'database only' in Framework Manager results in an error message telling me that the sum function is not enabled. The Cognos knowledge base has a case for the exact error I am getting and tells me to amend the cogdmor_td.ini file. I have done this both locally and on the server but to no affect. I have also checked the Teradata ODBC driver to disable parsing.

Has anybody come across this?

[sir_jeroen]

Have you modelled your Framework dimensionally?
Are you using Multi-fact and/or multi-grain tables?

[tdyoda]

Yes.

ReportNet Addict needs another addiction (crack, perhaps?) since this has nothing to do with dimensional modelling.Ã,  If the question is "what to have for breakfast?", the answer MODEL DIMENSIONALLY isn't a great response.Ã,  LOL ;D

This has nothing to do with Teradata or the Teradata ODBC driver.Ã,  As a rule, we do not code the sum function.Ã,  ReportStudio has done a marvelous job of providing three ways to do the same thing, but provided nothing to tell you why they did (there may be an excellent reason just waiting to be made known).Ã, 

Since there so many ways to skin the cat  I can't guess which way you are trying to do the sum.Ã,  Ã, We pay close attention to the query subject and data layer definition of attribute, identifer, fact in the framework model.Ã,  This controls Automatic or Sum of the attribute.Ã,  This works much better than coding a sum function in our experience.

You can generate excellent queries with ReportStudio that align with Teradata better than any other BI product I've seen to date.Ã,  Start with a report and framework model and do like it is a lab problem.Ã,  Ã, You'll figure out how to accomplish your goal.Ã,  It is possible and hang in there, you're on the right track.

[sir_jeroen]

So.. and what's your problem? Before starting shouting at someone please think things over and give a solution to the problem.

Apparently you know so much about ReportNet and/or Teradata that it's better for me to stop .  I'm very interested in your solution so I'll be checking this topic out regularly...

Please let me learn from my mistake and give us the solution....


(Btw...Do you have ANY idea why I bothered asking if his model was dimensionally modelled?? )

[tdyoda]

Re-read my response.  As soon as we get more information we will proceed.

Do you realize that not all problems can be solved by dimensional modeling?  There is a problem set that solved by the outer join, which is why it is in the SQL language, and also why RS supports it with excellent control.  Meanwhile eat your own dog food and provide the reason it must be modeled dimensionally.

ReportStudio does not require local processing when all of the data is from one database platform.  However, there are certain functions that RS will do local processing without your consent.  Rank is an SQL 2003 specification that is fully supported by Teradata, but Cognos won't generate the complete syntax for and will switch to local processing is one example.  The partition by desc is the trigger.

[sir_jeroen]

Your response was that you were making fun of me... And I don't appreciate this kind of behaviour because what everybody is doing here is done in their own free time and effort. If everybody is starting to behave like you then soon no one will come here...

And furthermore... I can tell from experience that a lot of problems can be caused by wrong modelling so this is something you should ALWAYS have look at... But who am I to tell you that... You should probably already know all of that...
But as you seem to know so much about Cognos 8 (more than other users that are trying to help) I am eager to learn from you.... Apparently i'm just a rookie in your eyes... So please enlight me..

Btw: @lawrensi sorry that your question caused this "cat-fight"...