If you are unable to create a new account, please email support@bspsoftware.com

 

News:

MetaManager - Administrative Tools for IBM Cognos
Pricing starting at $2,100
Download Now    Learn More

Main Menu

[SOLVED] Measure values suddenly multiplied by 1,000,000

Started by SueC, 16 Nov 2005 02:57:35 PM

Previous topic - Next topic

SueC

I know, most of you are thinking scaling, right?Ã,  Not quite.Ã,  Here's the scoop.

The same Transformer model was running on produciton and development servers and working just fine.Ã,  Sales Qty InScale = 0 and OutScale =0.Ã,  Same for Sales $.Ã, 

Now when I run the model in development all of the measure values are multiplied by 1,000,000.Ã,  Queries of the database show nothing strange with the data, even previewing it in transformer shows normal looking numbers.Ã, 

One interesting thing that I did notice is when I created a new model using the same iqd the OutScale setting defaulted to 6.

I've gotten nowhere with support on this and the development cube build time is driving me nuts it takes so long.Ã,  Does anyone have any idea why this started happening suddenly or how I might fix it?Ã,  Ã, ???

Sue

Darek

Check if your source is not decimal by any chance, with 6 decimal points. In any case, multiply the measure inside the SQL by one. Not only should it fix your problem, but also improve performance.

SueC

Updated 12/01/05    :(

I take it back.  The problem is masked in development now as the numbers are normal BUT my concern is this: why did it happen in the first place?  Something, somewhere changed.  Where do I look?  I'm thinking some field definition changed but I don't know where to look or what to look for.  I'm going to open another call with support but I really don't have much hope for them.

If anyone has any ideas, please post.  This all started when our DBA added some historical data for our Miami site (from an old software system) into the data warehouse.  He, of course, sees nothing wrong with the data and even did a reload without the Miami history for me - which didn't help.

************************
Darek -

It worked!Ã,  Thanks a 'million'.....get it....million.....oh, never mind.

:D

Sue

Draoued

Darek , you are saying :

QuoteIn any case, multiply the measure inside the SQL by one. Not only should it fix your problem, but also improve performance.


Could you explain why multiplying the measure inside the sql ( I suppose you mean in the IQD itself)Ã,  by 1 would improve the performance ??
Thank you.

Darek

The way I understand it, Transformer does not have to perform any data type conversions from a double type, and multiplying a measure by 1, does not change its value, while automatically making it a double. So Transformer spends less time reading the source. I've found it in some 5.x or 4.x related tip, but it seems to still hold true ... surprisingly.

Darek

SueC

Darn - I came back looking for the resolution for this problem and found that I never posted it.Ã,  Shame on me.Ã,  Good thing I sent the correction to support!

The problem was that the data fields were defined as NUMBER(25,6) in development and FLOAT(126).Ã,  Production did not create the inflated numbers.Ã,  The ETL analyst said he just took the Oracle default type (ut why they would be different on the two systems beats me).

Anyway, it happened again when the table was dropped and re-added.Ã,  At least it didn't take me 3 months to convince him that it was a db issues and not a Transformer issue this time!!!!

Smooch your Valentine everyone!

SueÃ,  :-*