Creating dimension with several LTS problem

Question

Good morning fellow experts,
I've hit the following issue - I have two dim tables that I have no problem using separately, however, I want to merge them into 1 dimension for better drill-down capability. Here's the structure:
Table1 "Geo Dim" fields: location_id (Primary key), Country, State, City, Zip, Address, Name (lowest level)
Table2 "Buildings Dim" fields: building_id (Primary Key), location_id (could be FK to Table 1), Building Name
Fact table has building_id as FK
The goal is to be able to drill from Country to Individual Building. I also know that ideal variant is to create a view and merge the 2 dim tables, however, I'm doing a POC incorporating snowflake. I used this wonderful link (courtesy of Adrian Ward) http://majendi.blogspot.com/2009/01/obiee-snowflakes-and-stars-part-3.html as a guidance, however, I hit some problems. My question is - what's the proper way to model it in physical layer. Here's how I've currently designed it:
Physical:
Fact >-- (building_id key) Buildings >-- (location_id key) Geo
Logical Star:
Fact (with fact LTS) >--complex join Building_Geo
Building_Geo has 1 LTS with 2 mapped tables. A custom hierarchy is built. Logical key is building_id. Should I use 2 LTS instead of 1 and set hierarchy levels to a) building b) geo ?
Can someone point me as to what's wrong in this schema? It's resembling Header-Detail a lot - and I'm sensing that I'm just missing a small piece somewhere. Thanks

Answer 1

Hi mma,
My first solution started to work once I created/defined all hierarchical keys (simply dragging them wasn't enough and I missed that) . BI had no problem with single LTS for 2 mapped tables and generating correct sql. So, just a follow-up question - how does dimensional hierarchy behave in BMM if you don't define keys (I'd guess it'd throw a RPD consistency error). Also, Mma, do you think such schema would work when more dims would get thrown in? Also, except for circular joins - what other pitfalls might be there?
Physical:
Fact >-- (building_id key) Buildings >-- (location_id key) Geo
Logical Star:
Fact (with fact LTS) >--complex join Building_Geo
I marked as answered - but would appreciate if you could answer 1 more time.