Update multiple columns based on different rows in another column sql server - sql

Question

I have a table
staff Type col1 col2
1 a 0 200
1 b 50 2000
1 c 2 200
2 a 0 100
2 b 25 1000
2 c 2 200
3 a 0 150
3 b 35 1500
3 c 2 200
I am just wondering if I could update both col1 and col2 based on the values in the Type column for each staff. My logic to update would be for example for staff 1
update col1 where Type =a is 200/(2000/50) =5
update col1 where Type =b is col1 - (200/(2000/50)) =50-5 = 45
update col2 where Type =b is 2000-200 = 1800
update col2 where Type =a to 0
Please see the resultant which I need to achieve is below and updated values(*)
staff Type col1 col2
1 a 5* 0*
1 b 45* 1800*
1 c 2 200
2 a 2.5* 0*
2 b 22.5* 900*
2 c 2 200
3 a 3.5* 0*
3 b 31.5* 1350*
3 c 2 200
Is it possible to get this done without using cursors/loops/functions/ctes?

Answer 1

Try below code
DECLARE #myTable TABLE (
staff INTEGER NOT NULL
,Type VARCHAR(1) NOT NULL
,col1 INTEGER NOT NULL
,col2 INTEGER NOT NULL
)
INSERT INTO #myTable VALUES(1,'a',0,200),
(1,'b',50,2000),
(1,'c',2,200),
(2,'a',0,100),
(2,'b',25,1000),
(2,'c',2,200),
(3,'a',0,150),
(3,'b',35,1500),
(3,'c',2,200)
UPDATE t1
SET Col1 = CASE WHEN Type ='a'
THEN col2/(Select Col2/Col1 FROM #myTable T2 WHERE T2.Type = 'b' AND T2.staff = T1.Staff)
WHEN Type ='b'
THEN col1 - ((Select Col2 FROM #myTable T2 WHERE T2.Type = 'a' AND T2.staff = T1.Staff)/(Col2/Col1))
ELSE COl1
END ,
Col2 = CASE WHEN Type = 'b'
THEN Col2 - (Select Col2 FROM #myTable T2 WHERE T2.Type = 'a' AND T2.staff = T1.Staff)
WHEN Type = 'a'
THEN 0
Else Col2
END
FROM #MyTable t1
SELECT * FROM #myTable

Answer 2

You can do it in one massive query, but you will likely need a lot of fail-safes to make it foolproof. For instance, you would've to account for a value of 0 in b:col1 or b:col2 (although assuming that you only want the cross multiplication, you can do a:col2*b:col1/b:col2 and it's much simpler for the computer (divisions are a mess) on top of having 1 less exception. The drawback is when you start working with massive numbers, but I'm gonna assume that you won't go over 1000000^2 so that's kinda irrelevant. Similarly, my code is gonna assume that there is always exactly one a row and a b row.
declare #test as table(
staff int,
Type char(1),
col1 decimal(6,2),
col2 decimal(6,2)
)
insert into #test values(1,'a',0,200);
insert into #test values(1,'b',50,2000);
insert into #test values(1,'c',2,200);
insert into #test values(2,'a',0,100);
insert into #test values(2,'b',25,1000);
insert into #test values(2,'c',2,200);
insert into #test values(3,'a',0,150);
insert into #test values(3,'b',35,1500);
insert into #test values(3,'c',2,200);
UPDATE
FirstTable
SET
FirstTable.col1 = CASE FirstTable.[type] WHEN 'a' THEN JoinedTable.col1 * FirstTable.col2 / JoinedTable.col2 ELSE FirstTable.col1 - FirstTable.col1 * JoinedTable.col2 / FirstTable.col2 END,
FirstTable.col2 = CASE FirstTable.[type] WHEN 'a' THEN 0 ELSE FirstTable.col2 - JoinedTable.col2 END
FROM
#test FirstTable
INNER JOIN
#test JoinedTable
ON
FirstTable.staff = JoinedTable.staff AND JoinedTable.[Type] = CASE FirstTable.[Type] WHEN 'a' THEN 'b' ELSE 'a' END
WHERE
FirstTable.[Type] in('a','b')
The behavior is as follow : if there is a 0 in a col2 for the Type b, it crashes. If there is more than one row for one staff with Type a or b, I believe it takes one of the 2 in a non deterministic fashion, which will lead to some non deterministic and potentially faulty results. If there is no Type a or b, it skips the staff entirely.
Otherwise (aka when it works), it simply updates rows with Type a and b based on their Type. Updating both in 1 query allow you to not use temporary tables and the likes since the new data doesn't override the old until the whole query is doen.
Also, try not to use keywords like Type for your columns or you may end up with some mistakes down the line.