Summing unrelated columns in two tables, then joining based on an ID

Question

I have two tables set up basically like this (extremely lite version):

Table 1 :

ID       Amt 1     Amt 2
-------------------------    
112     $20        $30
112     $50        $60
125     $75        $05

Table 2 :

ID       Amt 3     Amt 4
-------------------------
112       $25       $30
125       $40       $60
125      $110      $120

All the amount columns needed to be summed separately, the output to be similar to this:

Query results:

ID       Amt 1     Amt 2     Amt 3     Amt 4
--------------------------------------------
112     $70        $90       $25       $30
125     $75        $05       $150      $180

The joins I've tried result in the records duplicating (and summing) by a factor of the times the ID repeats in the second table. I have no primary key to connect these tables.

Any help would be greatly appreciated.

Thanks!

Answer 1

To solve this problem correctly, you need to do the aggregations separately. This will work even when both tables have multiple rows for the same id:

select id, sum(amt1), sum(amt2), sum(amt3), sum(amt4)
from ((select id, sum(amt1) as amt1, sum(amt2) as amt2, NULL as amt3, NULL as amt4
       from tbl1
       group by id
      ) union all
      (select id, NULL, NULL, sum(amt3), sum(amt4)
       from tbl2
       group by id
      )
     ) t
group by id

The above query gives the idea with a group by . Some people prefer a full outer join for this purpose:

select coalesce(t1.id, t2.id) as id, amt1, amt2, amt3, amt4
from (select id, sum(amt1) as amt1, sum(amt2) as amt2
       from tbl1
       group by id
      ) t1 full outer join
      (select id, sum(amt3) as amt3, sum(amt4) as amt4
       from tbl2
       group by id
      ) t2
      on t1.id = t2.id

The key is that the aggregations have to be done before any joins, so you don't have a problem of multiple rows.

Answer 2

On SQL Server 2005 or newer, you could use two CTE's (Common Table Expression) to do the summing/grouping per table, and then join the two - something like this:

;WITH CTE1 AS
(
    SELECT 
        ID, Amount1 = SUM(Amt1), Amount2 = SUM(Amt2)
    FROM
        dbo.SumTbl1
    GROUP BY
        ID  
),
CTE2 AS
(
    SELECT 
        ID, Amount3 = SUM(Amt3), Amount4 = SUM(Amt4)
    FROM
        dbo.SumTbl2
    GROUP BY
        ID  
)
SELECT CTE1.ID, Amount1, Amount2, Amount3, Amount4
FROM CTE1
INNER JOIN CTE2 ON CTE1.ID = CTE2.ID

This gives me your desired output