Problem solve Get help with specific problems with your technologies, process and projects.

Cross join effects

I have a real dilemma. I have two counts and totals I'd like to show on a single row, using the following three tables:

EmpNo LastName
 1    Jones   
 2    Harris  
 3    Smith   

idno empno amount
  1    1    1,200
  2    2    2,600
  3    2    1,400
  4    2    1,000
  5    3    3,200
  6    3    2,800

idno empno amount
  1    1     100
  2    1     200
  3    2     500
  4    3     200
  5    3     175
  6    3      25
  7    3     300

I would like to get this result:

EmpNo Count of  Sum of    Count of  Sum of
      Hardware  Hardware  Software  Software
  1      1       1,200       2        300
  2      3       5,000       1        500
  3      2       6,000       4        700

Instead, I get this result:

EmpNo Count of  Sum of    Count of  Sum of
      Hardware  Hardware  Software  Software
  1      2       2,400       2        300
  2      3       5,000       3      1,500
  3      8      24,000       8      1,400

This is my SQL statement:

SELECT DISTINCTROW SalesPerson.empno, 
Count(HardwareSales.amount) AS [Count Of Hardware], 
Sum(HardwareSales.amount) AS [Sum Of Hardware], 
Count(SoftwareSales.amount) AS [Count Of Software], 
Sum(SoftwareSales.amount) AS [Sum Of Software]
FROM (SalesPerson INNER JOIN SoftwareSales 
ON SalesPerson.empno = SoftwareSales.empno) 
INNER JOIN HardwareSales 
ON SalesPerson.empno = HardwareSales.empno
GROUP BY SalesPerson.empno, 
SalesPerson.lastname, SalesPerson.firstname, 
HardwareSales.empno, SoftwareSales.empno;

Please help as I hope to learn a lot from this query and others to come.

Thanks for the splendid example.

You have two unrelated one-to-many relationships, from employee to software, and from employee to hardware, being combined in a query. The important point is that while multiple software rows are related to an employee, and multiple hardware rows are related to an employee, the individual software and hardware rows are unrelated to each other.

The effects of joining them are easy to see in your example. The software and hardware counts for each employee are multiplied together. The sums are inflated by a factor equal to the number of unrelated rows. What has happened is that every single software row for an employee is matched with every possible hardware row for that employee. Thus, for each employee you get a "Cartesian product" of software and hardware rows, resulting in cross join effects.

Here's one way to solve the problem:

select SalesPerson.empno
     , Hcount as [Count Of Hardware]
     , Hsum   as [Sum Of Hardware]
     , Scount as [Count Of Software]
     , Ssum   as [Sum Of Software]
  from (
  join (
       select empno
            , count(amount) as Scount
            , sum(amount)   as Ssum
         from SoftwareSales 
       group by empno
       ) as S
    on SalesPerson.empno = S.empno        
  join (
       select empno
            , count(amount) as Hcount
            , sum(amount)   as Hsum
         from HardwareSales 
       group by empno
       ) as H
    on SalesPerson.empno = H.empno

This is Microsoft Access, right? When you paste this query into the SQL View window and save it, Access may replace the inner parentheses around the subqueries with square brackets and a period, but the query should still continue to work. Alternatively, you could save each of the subqueries as a query, then refer to them in the main query's FROM clause, in effect using the saved queries the way other databases use VIEWs.

The secret is to join each employee to only one row for software totals and only one row for hardware totals.

Dig Deeper on Oracle and SQL

Start the conversation

Send me notifications when other members comment.

Please create a username to comment.