Question

I have the following data in the table (I’m using SQL Server 2008R2):

Customer Code, Device Expiry Date, Device Name, Log ID

S001, NULL, Dvc A, 1
S002, NULL, Dvc B, 2
S002, NULL, Dvc A, 3
S003, 2010-01-01, Dvc A, 4
S004, 2011-06-25, Dvc B, 5
S004, 2012-02-10, Dvc A, 5

Log ID column is based on running number.

I want to select one record for each customer code based on the following criteria:
– If Device Expiry Date is not null then take the record with the latest date
– If all the expiry date is null then take the latest record (maximum log ID)

Hence, the selected records will be:

Customer Code, Device Expiry Date, Device Name, Log ID
S001, NULL, Dvc A, 1
S002, NULL, Dvc A, 3
S003, 2010-01-01, Dvc A, 4
S004, 2012-02-10, Dvc A, 5

The output of the query will be further joined with master customer code:

Customer Code, Member Type
S001, Silver
S002, Gold
S003, Silver
S004, Silver
S005, Gold

When joined, the end result will be:

Customer Code, Member Type, Device Expiry Date, Device Code
S001, Silver, NULL, Dvc A
S002, Gold, NULL, Dvc A
S003, Silver, 2010-01-01, Dvc A
S004, Silver, 2012-02-10, Dvc A
S005, Gold, NULL, NULL

What I had on mind is to use several sub queries just to get the Devices records. Please help with some idea on getting the above result. Thank you in advance.

Answer 1

You could try the following (untested)

;WITH q AS (
  SELECT  *, rn = ROW_NUMBER() OVER (PARTITION BY [Customer Code] ORDER BY [Log ID] DESC, CASE WHEN ISNULL([Device Expiry Date]) THEN 0 ELSE [Device Expiry Date]) END DESC)
  FROM    Devices
)
SELECT  *
FROM    q
WHERE   rn = 1

The idea is to add a rownumber to each row that will

• reset for each customer
• starts with the largest Log ID
• and uses a dummy date for those records that have a NULL date

All that’s required now is to select all records from the resultset where this rownumber = 1.