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

 

Update analysis because it refers to items that are either no longer present

Started by Michael75, 07 Mar 2014 10:14:18 AM

Previous topic - Next topic

Michael75

We have a Sales cube which has been running pretty reliably for two years. It has 7 dimensional data sources, and 4 fact data sources. The cube is rebuilt nightly off the .pyj file, and the only Command Line Option used is -n2.

It's been brought to my attention that a large number of saved views are now bringing up the error which is the topic of this thread. There's a thread of the same name in the Cognos 8 Analysis Studio http://www.cognoise.com/index.php/topic,12479.0.html

which I'll refer to, but I'm on 10.1.1 and consider this to be a Transformer issue, so chose to start a new thread here. I know about MUNs and all that. There have been absolutely no changes made to this cube for 9 months now, so all that can have changed is the data.

I'm now busy investigating/fixing this, and since there are some areas of Transformer which aren't clear to me, I'll be asking for clarification on some topics.

I started by looking in the Categories map to see what had happened with the affected items, only to realise that quite a few of them do not exist here. After further study, I came to the conclusion that Transformer does not used the saved Categories when rebuilding the cube, but takes them from the data sources. Is this correct? And if so, what is the purpose of storing an out of date set of categories in the model?

It seems to me that if I added the -s Command Line Option, the Categories map would always be up to date. Is this correct, and if so, are there any associated pros and cons? The thread I referred to talks about saving categories in the model, but this option is not mentioned.

Next question. I've noticed that there is a small overlap between product and customer codes used as Category codes at the lowest levels of their respective dimensions. By that I mean that we have a customer 105261, which I can see in the Categories map, and a product 105261, which I can't see (presumably for the reasons mentioned above), but it's present in the product data source. Is this something that Transformer takes in its stride, or should I find a way to make these unique (and maybe risk breaking more views)?

Final question (for now). Is it necessarily a bad thing to have tildes in Category codes. I notice that there are many such cases in this cube, at levels where I know the data to be static, but as mentioned it has been running reliably up until now.

TIA
Michael

Michael75

QuoteI came to the conclusion that Transformer does not used the saved Categories when rebuilding the cube

I should add that I'm not referring to Calculated categories etc. in this sweeping statement  :)