How to resolve illegal cross join error

Question

Can someone please tell me how can we avoid illegal cross join error in modeling? If someone has any reference document , please share it.
I have 5 tables - Dim (A, B , C,E) and Fact( D). C is lookup table which is used to resolve the lookup code column in table E.
The relationships between these are
A--< B--< E >--C
& A--<D
My requirement is i have to expose Attributes of A,B ,C & E within Dimension.
These attributes (from B,C,E) will be treated as property of A.
Someone may query these attributes without selecting any column from fact table.
Thanks & Regards,
Ashish

Answer 1

Hi Ashish,
it's not the first time you come up with this kind of question and I'm wondering if you understand the principle of granularity when I read your questions.
On physical level, a fact table must always join with the lowest level of detail of your dimension table. Let me explain, by giving an example which looks like your situation.
Assume I have three tables:
Table "E" contains products, which has attributes like: product_id and product_name
Table "B" contains order line items, which has attributes like: order_line_item_id, order_id and product_id.
Table "A" contains orders, which has attributes like: order_id and customer_name.
Table "D" contains facts ("metrical data") about my order: which has attributes like: order_id, order_revenue.
This will be the diagram on physical level: E--<B--<A--<D
Here are my joins: E.product_id on B.produc_id, B.order_id on A.order_id, A.order_id--<D.order_id
Now, my question to you is: will I be able to get the revenue of a certain product?
The answer is: No, because I don't know what part of my order is spent on a certain product. The problem is thus that the data in the fact table isn't stored on that level of detail, or in other words the fact and dimension table don't share the same level of granularity.
How it should be:
If you want to get the revenue per product, you will need to have a second fact table "G", which has data which is stored on order line item level. This table contains the following attributes: order_line_item_id, product_id and order_line_item_revenue.
This will become your diagram on physical level:
B--<G>--E
Joins: B.order_line_item_id on G.order_line_item_id and E.product_id on G.product_id
Table E has become a dimension of fact table G.
The first physical diagram should look like this:
A--<D
If you want you can model both physical diagrams into one logical diagram, assuming that A and D (order dimension and fact table) are aggregates of B and G (order line items dimension and fact table). In that case you should read this blog item: http://obiee101.blogspot.com/2008/11/obiee-making-it-aggregate-aware.html
Regards,
Stijn