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

ev_cubename tables

Started by adityashah27, 19 Dec 2008 10:30:57 PM

Previous topic - Next topic

adityashah27

Hi All,

I would like to know how & why Planning Table Only Layoyt Publish process creates ev_cubename views.
I wasnt aware that it creates ev_ views for every associated et_ tables.
we are getting error message for table only publish i.e. unable to drop ev_cubename
and another error message is, unable to insert null into applicationcolumn table for ev_cubename object.

Appreciate your inputs on this.

Thanks,
Aditya


blackadder

The EV_ views were introduced to the Table publish when Incremental publishes came along in 8.2.
It simply applies a VISIBLE = 1 filter to the et table, so that if you run a report while an incremental publish is taking place, you will get consistently reported data .

Unfortunately its not as sophisticated as the EV_ view in the old View publish, which joins to all the dimension tables for each fact and returns all the dimension names. 

adityashah27

thats news to me. i did not realise this change at all ( i started working on 82/83 this year jan-feb)

do you have any docs or write-ups on this? this feature seems buggy.

blackadder

I've not been able to find any docs on it, but from my experimentation it works like this:
- As the Incremental publish runs, the newly created rows have Visible set to 0.
- The previously published rows co-exist and have Visible set to 1
- Right at the end of the publish job (as a single database transaction, to ensure data consistency), the old rows are deleted and the new ones have Visible set to 1

If your incremental publish is quick you won't notice it happening because it all happens very quickly. You would only notice it in action if you are running reports against the database at the time the publish was running, and happened to catch the point where two rows exist.

Its designed this way to ensure that data is neither lost nor double-counted during the course of the publish. In my experience it is not buggy.


I'm pretty sure that the Generate FM Model extension uses this filter too.

Hope that helps

adityashah27

Appreciate you shared your experience on this.

the reason i said buggy is while running manual full table only publish, the publish process creates et_ tables but due to some reasons its not creating all ev_ tables associated with et_ tables. i have noticed it happening for certain model design. for eg. dlist item names either blanck or special characters like '|', '-'.  these things replaced by 'n' for et_ tables but it has NULL as column name for associated ev_ tables.
Also when i uncheck certain cubes and run full publish, it drops et_ table but unable to drop associated ev_ tables (few of them)

thru documentation i wanted to figure out the structure of ev_ tables (associated with et_) and is there anything specific to these table due to which it creates NULL column names for certain dlist items.