How can I implement multiple nameColumns that target the same table? - ssas

Question

I have two tables. The first one references the second one multiple times.
Some demo data:
id fk1 fk2 fk3
-------------------------------
4009 9419 2282 9005
4057 9419 2112 NULL
5480 NULL 4279 NULL
5989 NULL 1677 NULL
The second table contains the names for the foreign keys in table 1
id name
--------------------------------
1677 Bank Account No1
2112 Loyalty Account
2282 Sales Account
4279 Employee Account
9005 Warehouse No1
9419 Sales
I have to create a dimension for Table1. The fields fk1, fk2 and fk3 should use the nameColumn attribute, and show the values from table 2.
I tried creating this named query in the Data SourcE View, but if I try to put it into the cube, the deployment breaks without proper error.
SELECT [Table1].[id]
,fk1
,fk2
,fk3
,T2_fk1.name as fk1name
,T2_fk2.name as fk1name
,T2_fk3.name as fk1name
FROM [dbo].[Table1]
left join Table2 as T2_fk1
on Table2.id = [Table1].fk1
left join Table2 as T2_fk2
on Table2.id = [Table1].fk2
left join Table2 as T2_fk3
on Table2.id = [Table1].fk3
How can I implement multiple nameColumns that target the same table?

Answer 1

Even if NULL values can technically be used in Dimensions in Analysis Services, it is normally a bad idea to do so, as Analysis Services and the relational database have different ideas how they treat them: Analysis Services treats them as empty strings or numerical zeroes, while for the relational database, normally - with the exception of e. g. old versions of Oracle, which treat empty strings as null - these are different. Hence when Analysis Services issues a SQL statement containing DISTINCT or GROUP BY, it can happen that there is more than one row where Analysis Services only expects one, etc.
It is best practice to avoid nulls in attribute columns, as well as foreign key columns in the star schema. In measure columns, nulls are fine. Thus, you should change your statement to
SELECT [Table1].[id]
,coalesce(fk1, -1) as fk1
,coalesce(fk2, -1) as fk2
,coalesce(fk3, -1) as fk3
,coalesce(T2_fk1.name, '<unknown>') as fk1name
,coalesce(T2_fk2.name, '<unknown>') as fk1name
,coalesce(T2_fk3.name, '<unknown>') as fk1name
FROM [dbo].[Table1]
left join Table2 as T2_fk1
on Table2.id = [Table1].fk1
left join Table2 as T2_fk2
on Table2.id = [Table1].fk2
left join Table2 as T2_fk3
on Table2.id = [Table1].fk3
or whatever you chose to replace nulls with instead of -1 or <unknown>.