how do you limit category generation in Transformer?

[drseuss]

i'm in the process of building a big cube. the model has 2 iqd's as sources. i haven't yet generated the cube, i just want to generate the categories locally (not on server) first, but transformer is hanging up. is there a way to limit or control category generation, say for a subset of the total rows of data? i tried generating categories just from one source, but transformer's hanging up here as well. any suggestions? is it possible?

thanks in advance.

 

[CosmicNerd]

What version of Powerplay have you got? In later versions you can generate a small test cube in the Run/Test build menu. This might help.

If you don't have this option, you could try using the Clean House option to get rid of some categories, if you can see them - may or may not be appropriate.

Or go into the category diagram and try suppressing or excluding categories for testing purposes.

What about creating a subset of the data and using that locally for testing? Can cause problems reconnecting to the real data though.

Hope one of these ideas helps.

 

[Shek]

1. You can copy the model and iqd in different folder and put filters in the sqls of the iqds
Say limit te geography to only one country by condition
select.....
from...
where country_name = 'INDIA' and so on for all iqds.

This filter needs to be present in the Measures iqds which refer to country table

 

[MiRRo]

You can also make an test iqd where you fill this iqd with halve of the records of your original iqd.
Best of luck!!

MiRRo

 

[BruceReed]

One more method to add to everyone else's suggestions, you can use Impromptu to filter & save these reports as iqd.

You have lots of choices!

Best luck
Bruce

 

[CanadianTechie]

Have you checked the Transformer log files for any clue as to what is happening?? Does this cube build fine on the server and just hang when building on your workstation? If not, have you tried ensuring that the DataWork directory is set to a different physical drive compared to the ModelWork directory (might not be possible depending on your disk configuration).

Another suggestion, when it comes to optimizing large cube builds, it is recommended that you have a different data source/iqd per dimension, including a separate data source/iqd for the measures. Further, the "ORDER" of the iqd/data sources is relevant, as the order of the sources should follow the left->right ordering of your dimensions, with the measures data source being last...you should see some significant improvement in performance....(i.e. it shouldn't hang on you!)

Let me know how you make out!

Pedro Cardoso
A.K.A CanadianTechie