SQL Full join on two columns - sql

Question

I'm trying to join two tables but the columns emp_id and scheme_id may be null or not populated in either table but if it is populated in either then I need to return the total pen_ee for that employee for each scheme (further table descriptions below). I can't edit the table structure and have to work with what I have.
I've been trying to use a full join to do this but don't understand if you can do a full join on two fields emp_id & scheme_id to get the required result.
Table PAYAUDPEN
This is the first two months of the year.
- Employee A has given 44.06 to scheme BMAL.
- Employee B has given 98.06 to scheme BMAL.
- Employee B has given 98.06 to scheme CLFL.
emp_id, period_id, scheme_id, pen_ee
A, 201601, BMAL, 22.03
A, 201602, BMAL, 22.03
B, 201601, BMAL, 98.06
B, 201602, CLFL, 98.06
Table PAYISPEN
This is the third & current month of the year. The system always puts the current month into this table)
- Employee A gave 22.03.
- Employee B gave 98.06.
(Note employee B did not contribute to the BMAL scheme again in month 3 which is part of the issue).
emp_id, scheme_id, pen_ee
A, BMAL, 22.03
B, CLFL, 98.06
Required Result
The SQL statement needs to return the 3 periods added together, for each employee for each scheme that they contributed to.
- Employee A would be 44.06 + 22.03=66.09 for scheme BMAL.
- Employee B would be 98.06 + NULL =98.06 for scheme BMAL.
- Employee B would be 98.06 + 98.06=196.12 for scheme CLFL.
A, BMAL, 66.09
B, BMAL, 98.06
B, CLFL, 196.12
To create the basics of the two tables and populate with the example data above run the following queries.
CREATE TABLE [dbo].[payaudpen](
[emp_id] [char](10) NOT NULL,
[period_id] [char](6) NOT NULL,
[scheme_id] [char](10) NOT NULL,
[pen_ee] [numeric](15, 2) NULL)
CREATE TABLE [dbo].[payispen](
[emp_id] [char](10) NOT NULL,
[scheme_id] [char](10) NOT NULL,
[pen_ee] [numeric](15, 2) NULL )
INSERT INTO payaudpen VALUES ('A','201601','BMAL','22.03'), ('A','201602','BMAL','22.03'), ('B','201601','BMAL','98.06'), ('B','201602','CLFL','98.06')
INSERT INTO payispen VALUES ('A','BMAL','22.03'), ('B','CLFL','98.06')
Current statement that I'm using:
SELECT a.emp_id,
a.scheme_id,
SUM(a.pen_ee)+AVG(b.pen_ee)
FROM payaudpen a
FULL JOIN payispen b
ON a.emp_id=b.emp_id
GROUP BY a.scheme_id, a.emp_id
Incorrect result
Doesn't return the correct value for employee B for each scheme.
A, BMAL, 66.09
B, BMAL, 196.12
B, CLFL, 196.12

Answer 1

You are trying to sum across two tables, use union all to make the tables into one relation with more rows, instead of join to make the tables into a relation with more columns:
WITH all_records AS (SELECT emp_id
, scheme_id
, pen_ee
FROM payispen
UNION ALL
SELECT emp_id
, scheme_id
, pen_ee FROM payaudpen)
SELECT emp_id, scheme_id, SUM(pen_ee)
FROM all_records
GROUP BY emp_id, scheme_id
Results:
emp_id scheme_id (No column name)
A BMAL 66.09
B BMAL 98.06
B CLFL 196.12

Answer 2

Apparently you want to join only rows that have both the same emp_id and the same scheme_id. This is possible in outer joins, just as it is in inner joins. I infer that you also want to merge the emp_id and scheme_id columns from the two tables so that when a does not provide them, they come from b, instead. This will do it:
SELECT
COALESCE(a.emp_id, b.emp_id) AS emp_id,
COALESCE(a.scheme_id, b.scheme_id) AS scheme_id,
SUM(a.pen_ee)+AVG(b.pen_ee) AS pen_ee
FROM
payaudpen a
FULL JOIN payispen b
ON a.emp_id = b.emp_id AND a.scheme_id = b.scheme_id
WHERE
COALESCE(a.emp_id, b.emp_id) in ('A','B')
AND (a.period_id IS NULL OR a.period_id in ('201601','201602'))
GROUP BY COALESCE(a.scheme_id, b.scheme_id), COALESCE(a.emp_id, b.emp_id)
Note the use of COALESCE() to handle the cases where table a does not provide emp_id or scheme_id; with SQL Server you could also use ISNULL() in its place. Note also the allowance for a.period_id IS NULL in the WHERE condition -- this is necessary (in conjunction with the COALESCE()ing) to include data from b rows that do not have corresponding a rows.

Answer 3

The key thing you need to change in your solution is to use ISNULL() to make sure the key appears when table b has data but not table a. Otherwise, you'll get rows which look like:
NULL | NULL | 98.06
I recommend:
SELECT ISNULL(a.emp_id,b.emp_id) AS emp_id
ISNULL(a.scheme_id, b.scheme_id) AS scheme_id
SUM(a.pen_ee)+AVG(b.pen_ee) AS pen_ee
FROM payaudpen a
FULL JOIN payispen b
ON a.emp_id=b.emp_id
AND a.scheme_id=b.scheme_id
WHERE a.emp_id in ('A','B')
and period_id in ('201601','201602')
GROUP BY ISNULL(a.emp_id,b.emp_id), ISNULL(a.scheme_id, b.scheme_id)