split logic when loading slowly changing dimensions
I know when loading slowly changing dimensions you load as follows:
source -> Table comparison -> History preseving -> Key generation -> target;
What if now when loading I want the following:
For a changing dimension, I need to do something;
For a historical dimension, I need to do something else;
For a brand new one, I need to do something total new.
How can I achieve this? I want to split my logic and cater for all 3 scenarios and run scripts for each splits.
Thanks and regards,
Edited by: gsecure on Apr 8, 2010 9:07 AM
I always question when people suggest they need to do this, especially with a type 2 SCD.
Whilst you can use the map operation method, you need to use it with caution as multiple instances of the same target table aren't generally a good idea as you can run into concurrency/sequencing issues, as controlling the order of inserts and updates isn't straight-forward.
Can you explain what needs to be done differently for each type of operation? The main reason I ask is that the transforms automatically handle the from/to dates logic and when to generate a surrogate ID.