SQL Server - Using joins in Update statement - sql

Question

I have a HRUser and an Audit table, both are in production with large number of rows.
Now I have added one more column to my HRUser table called IsActivated.
I need to create a one-time script which will be executed in production and populate data into this IsActivated column. After execution of this one-time script onwards, whenever the users activate their account, the HRUser table's IsActivated column will automatically be updated.
For updating the IsActivated column in the HRUser table, I need to check the Audit table whether the user has logged in till now.
UPDATE [dbo].HRUser
SET IsActivated = 1
FROM dbo.[UserAudit] A
JOIN dbo.[HRUser] U ON A.UserId = U.UserId
WHERE A.AuditTypeId = 14
AuditTypeId=14 means the user has logged in and user can login any number of times and every time the user logs in it will get captured in the UserAudit table...
The logic is that if the user has logged in at least once means the user is activated.
This cannot be tested on lower environments and it need to be directly executed on production as in lower environments we don’t have any data in the UserAudit table.
I am not really sure if that works as I have never used joins in update statement, I am looking for suggestions for any better approach than this for accomplishing my task

Answer 1

You could use EXISTS and a correlated subquery to filter on rows whose UserId has at least one audit event of id 14:
UPDATE h
SET IsActivated = 1
FROM [dbo].HRUser h
WHERE EXISTS (
SELECT 1 FROM
FROM dbo.[UserAudit] a
WHERE a.UserId = h.UserId AND a.AuditTypeId = 14
)
Note that there is no point reopening the target table in the subquery; you just need to correlate it with the outer query.

Answer 2

Two methods below. Method 1 is NOT recommended for tables "in production with large number of rows". But it is much easier to code. Method 2 works in production with no downtime.
Whichever method you choose: TEST it outside production. Copy the data from production. If you cannot do that, then build your own. Build a toy system. Highly recommended that you test at some level before running either method in production.
METHOD 1:
Updating on a join is straight forward. Use an alias. Reminder, this is NOT RECOMMENDED "with large number of rows" AND production running. The SQL Server optimizer most likely will escalate locks on both tables and block the tables until the update completes. IF you are taking an outage and are not concerned with how long the update takes, this method works.
UPDATE U
SET IsActivated = 1
FROM dbo.[UserAudit] A
JOIN dbo.[HRUser] U ON A.UserId = U.UserId
WHERE A.AuditTypeId = 14
METHOD 2:
IF you cannot afford to stop your production systems for this update (and most of us cannot), then I recommend that you do 2 things:
Set up a loop with the transaction inside the loop. This means that the optimizer will use row locks and not block the entire table. This method may take longer, but it will not block production. If the update takes longer, not a concern as long as the devops team never calls because production is blocked.
Capture rows to be updated outside the transaction. THEN, update based on a primary key (fastest). The total transaction time is how long the rows updated will be blocked.
Here is a toy example for looping.
-- STEP 1: get data to be updated
CREATE TABLE #selected ( ndx INT IDENTITY(1,1), UserId INT )
INSERT INTO #selected (UserId)
SELECT UserId
FROM dbo.[UserAudit] A
JOIN dbo.[HRUser] U ON A.UserId = U.UserId
WHERE A.AuditTypeId = 14
-- STEP 2: update on primary key in steps of 1000
DECLARE #RowsToUpdate INT = 1000
, #LastId INT = 0
, #RowCnt INT = 0
DECLARE #v TABLE(ndx INT, UserId INT)
WHILE 1=1
BEGIN
DELETE #v
INSERT INTO #v
SELECT TOP(#RowsToUpdate) *
FROM #selected WHERE ndx > #LastId
ORDER BY ndx
SET #RowCnt = ##ROWCOUNT
IF #RowCnt = 0
BREAK;
BEGIN TRANSACTION
UPDATE a
SET IsActivated = 1
FROM #v v
JOIN dbo.HRUser a ON a.Id = v.UserId
COMMIT TRANSACTION
SELECT #LastId = MAX(ndx) FROM #v
END