CODEWITHSUNDEEP
sql-server
Alvin Das
Alvin Das

Reputation: 95

Update multiple rows with the select statement : SQL Server

An old problem I see, but after two days looking around, I just couldn't fix it. I have two tables: employee and random_values

employee has one column: emp_no with 3000 rows

random_values has one column: rand_val with 100 rows

I am trying to replace 100 rows in emp_no with 100 rows from rand_val.

I have the following:

UPDATE employee
SET employee.emp_no=random_values.rand_val
   FROM random_values
WHERE employee.emp_no IN (SELECT emp_no FROM employee
                GROUP BY emp_no HAVING COUNT(*)>1)

Results: After running the emp_no gets 100 rows OF THE SAME VALUE, namely the first value from rand_val

Goal: emp_no should get 100 DIFFERENT rows from rand_val

Any help is appreciated. Thanks.

Upvotes: 1

Views: 10894

Answers (1)

Robert McKee
Robert McKee

Reputation: 21477

Generate two queries, one that returns the emp_no's you want to replace with a second column that is numbered sequentially. The other that does the same for the random's. Then in the update, join them based on the numbers.

UPDATE employee
SET e.emp_no=r.rand_val
FROM (query1) e
JOIN (query2) r
  ON e.number=r.number

Likely query1 would be something like:

SELECT emp_no,row_number() OVER(ORDER BY emp_no) as number
FROM employee
WHERE employee.emp_no IN (SELECT emp_no FROM employee
                GROUP BY emp_no HAVING COUNT(*)>1)

Likely query2 would be something like:

SELECT rand_val,row_number() OVER(ORDER BY NEWID()) as number
FROM random_values

So the whole thing would be:

UPDATE employee
SET e.emp_no=r.rand_val
FROM (
    SELECT emp_no,row_number() OVER(ORDER BY emp_no) as number
    FROM employee
    WHERE employee.emp_no IN (SELECT emp_no FROM employee
                GROUP BY emp_no HAVING COUNT(*)>1)) e
JOIN (
    SELECT rand_val,row_number() OVER(ORDER BY NEWID()) as number
    FROM random_values) r
  ON e.number=r.number

It's been a while, you might have to join them back to the employee table as well like this:

UPDATE employee
SET e.emp_no=r.rand_val
FROM employee
JOIN
(
    SELECT emp_no,row_number() OVER(ORDER BY emp_no) as number
    FROM employee
    WHERE employee.emp_no IN (SELECT emp_no FROM employee
                GROUP BY emp_no HAVING COUNT(*)>1)) e
  ON employee.emp_no=e.emp_no
JOIN (
    SELECT rand_val,row_number() OVER(ORDER BY NEWID()) as number
    FROM random_values) r
  ON e.number=r.number

But you still might have difficulties with the duplicates both getting assigned a new number (or the same random number) unless you can differentiate the emp_no's by some other field. I assume your real employee table has more than just 1 column, and you would need to reference both the emp_no and the unique field in the join to make sure they get assigned different numbers, or eliminate one of the two duplicates in query1.

Upvotes: 2

Related Questions