SQL Server对子查询使用不同的分组依据和排序依据列

Question

I'm trying to only display certain columns from a group by query using a grouping set. I also want to order by only certain other columns as well.

To do this I'm using a sub-query.

I have two questions:

1. Is there a better way to group by and order by diff columns?

2. Are there performance problems with using sub-query this way?

Query:

SELECT Region, State, Sales FROM (
    SELECT RegionID, Region, StateID, State, SUM(Sales) AS Sales 
    FROM Sales 
        GROUP BY 
            GROUPING SETS(
                (RegionId, Region, StateID, State), 
                (StateId, State), 
                ()
            )
    ) As T1 
ORDER BY RegionID, StateID

Reason why I want to do this is because I want to sort by different columns than I'm grouping by. But since I'm using a group by query I have to use all columns.

Answer 1

Your query is fine. However, I don't think you need the subquery:

SELECT Region, State, SUM(Sales) AS Sales
FROM Sales
GROUP BY GROUPING SETS( (Region, State), (State), () ) 
ORDER BY MAX(RegionID), MAX(StateID);

You can just use an aggregation function on the ids for ordering purposes.

Note that the subquery incurs very little performance penalty. The only issue I can think of is that the aggregation keys are slightly longer and there might be a very small performance penalty for that.