Question

I’m unable to figure out a sql query (using MS Sql Server). I’m trying to retrieve a single row from a dataset in which an item with one id can have more than one row. The part that is throwing me off is that the correct row should be based on a “hierarchy”. I have trying to throw a case statement at the problem.

Some sample data:

Id           Class       Date
100          Red         2012-12-12
100          Blue        2012-12-31
200          Red         2012-10-31
300          Green       2012-04-04
300          Blue        2011-09-01

I want to return a single row based on the value of Class.

Case When Red Then
    Date 
Case When Blue Then 
    Date
Case When Green Then 
    Date 
Else
    ''

My final dataset should look like this:

Id           Class       Date
100          Red         2012-12-12
200          Red         2012-10-31
300          Blue        2011-09-01

So, if one of the duplicate rows has a value of Red, use the date from that row first. Then blue, then green.

Been working on this one all day, playing around with subqueries, group bys, havings, case statements, derived tables. I’m quite rusty on my sql skills, as it’s been a while.

Any hints on the direction I should take?

Answer 1

You can try this

;WITH cte AS
  (
    SELECT Id, Class, Date,
      row_number() OVER (PARTITION BY Id
                         ORDER BY CASE Class
                                  WHEN 'Red' THEN 1
                                  WHEN 'Blue' THEN 2
                                  WHEN 'Green' THEN 3
                                  ELSE 4 END) as rn
    FROM MyTable
   )

SELECT Id, Class, Date
FROM cte
WHERE rn = 1