Question

I have a table which looks something like this:

--------EXAMPLE---------------------------------------------------------
Cat1           Cat2          Value
a              NULL          131
b              e1            143
a              e2            135
c              NULL          136
a              e1            132
a              NULL          13
c              e4            1

I want to select the row from above which has the highest value for various types in Cat1. So, my answer would look like the following:

--------EXAMPLE---------------------------------------------------------
Cat1           Cat2          Value
b              e1            143
a              e2            135
c              NULL          136

I am using Microsoft SQL and I am looking for something efficient as I have about 40 million rows in my table that needs manipulation.

Thanks!!

Answer 1

You can use a CTE with DENSE_RANK:

WITH cte 
     AS (SELECT cat1, 
                cat2, 
                Value, 
                Rank = Dense_rank() 
                         OVER ( 
                           partition BY cat1 
                           ORDER BY Value DESC) 
         FROM   dbo.tablename) 
SELECT cat1, 
       cat2, 
       value 
FROM   cte 
WHERE  rank = 1 

Use Row_Number instead of Dense_Rank if you just want a single (arbitrary) record if there are multiple records with the same highest value.