Question

I am new to sql and I have come across a problem on joining 3 tables.

I have 3 tables

1) USER_MASTER (userid,fname,lname)
This is the call center agent table

2) CALL_MASTER (callid, customername,phone,userid,calltype)
this is where the customer calls go into , user id is used to link which agent
picked up the call , calltype is the outcome of the call

3) CALL_TYPE_MASTER as

calltype typename
1        transferred
2        routed
3        disconnected

I want to list it like this

Agent name           Transferred    routed  disconnected
Balang Hector           7          1          2
Bonus Donna             0          1          2
Dalino Marie            8          1          1
Dela Cruz               0          0          1

I tried this query:

SELECT
USER_MASTER.FirstName,COUNT(CALL_MASTER.CallType) as Transferred 
FROM
(USER_MASTER LEFT JOIN CALL_MASTER ON CALL_MASTER.UserID=USER_MASTER.USERID) 
WHERE
CALL_MASTER.CallType=1
GROUP BY
USER_MASTER.FirstName

But the problem is I have to list another row as “routed” and “transferred”, but that would mean changing the where clause, and it would obviously not output “0” if it is not found.

Answer 1

This is, of course, a fairly standard PIVOT query. Here’s a version more idiomatic for SQL Server:

SELECT User_Master.name, [1] AS Transferred, [2] AS Routed, [3] AS Disconnected
FROM (SELECT userId, callType
      FROM Call_Master) Call_Master
PIVOT(COUNT(callType) FOR callType IN ([1], [2], [3])) Pivoted
JOIN User_Master
  ON User_Master.userId = Pivoted.userId

SQL Fiddle Demo

The aggregate appears to be happening before the join to User_Master, so the query should be able to use indices to fulfill the COUNT(*). Unfortunately there’s no way to automatically populate column aliases – you’d need dynamic SQL for that.

EDIT:

Explanations –

(SELECT userId, callType
 FROM Call_Master) Call_Master

This subquery is getting a list of columns over which the aggregate will be grouped and run. You can do any conditions for range checks, and similar. The intent is that the query should be what you’d write for a GROUP BY… just without that clause, and an aggregate.

PIVOT(COUNT(callType) FOR callType IN ([1], [2], [3])) Pivoted

This clause is telling the system “for the previous table-reference, run the given aggregate over the listed column, and for every change in the other (potentially different) column, put the result in a new column”. There are some caveats:

• All other columns will be included in what essentially amounts to a GROUP BY clause
• You can only specify one aggregate function
• You can only specify one column in the aggregate, and only a column (no math here). You’d need to do anything like that in the subquery
• The list of columns in the IN clause must contain all values your subquery returns, but can also have columns without results. Leaving out [2] will get a runtime error, but adding [4] just gets you a column with 0s
• The brackets are required, and if the column given in the FOR clause is character, you don’t use quotes around the values.

• An alias for the result table (Pivoted here) is required, and is in effect for all columns involved. There’s no Call_Master.userId column at this point in the query.

  JOIN User_Master
  ON User_Master.userId = Pivoted.userId

… And finally, a join to User_Master to translate user ids to names. Note that, because the aggregate happens “inside” the reference generated as part of the Pivot query, you don’t have to worry about strange things happening to the rest of your data.