MERGE statement using OUTPUT with WHERE condition - sql

Question

I'm trying to insert data from one system into another so I keep an intermediate mapping table to keep the ids from old and new table.
I'm using a MERGE condition, is there any way to
DECLARE #TenantId INT = 1
MERGE dbo.[Account] AS t
USING (SELECT m.[AccountId],
m.[TenantId],
a.[ID_Account],
a.[Account_No],
a.[Account_Name],
FROM [Client1].dbo.[Account] a
LEFT JOIN migration.[Account] m ON m.[ID_Account] = a.[ID_Account] AND m.[TenantId] = #TenantId
) AS s
ON (t.[AccountId] = s.[AccountId] AND t.[TenantId] = s.[TenantId])
WHEN NOT MATCHED THEN
INSERT ([TenantId], [Number], [Name], [Active] )
VALUES (#TenantId, s.[Account_No], s.[Account_Name], 1)
OUTPUT #TenantId, inserted.[AccountId], s.[ID_Account] INTO migration.[Account];
This is fine, but if I try to run it a second time, the records are inserted again on my migration.[Account] table keeping repeated data. Is there any way to put a where condition on the output?

Answer 1

When you check to see if values from the source exist in the target, you check to see if the AccountId from migration.Account is equal to the AccountId from dbo.Account. You output the results to the migration table, but the original record still exists so it will be included in subsequent executions of the merge.
To fix your problem you need to either:
delete the original record
Add an isUpdated column in the migration table that you update after insertion and will prevent the source from being loaded.
Update the AccountId with the new AccountId so it will match the condition on subsequent executions.
Have the new record somehow cancel the original in the source record.
Here's a method with a new isUpdated and updates the accountId column:
UPDATE m SET
isUpdated = 1,
AccountId = o.NewAccountId
FROM migration.Account [m]
INNER JOIN (
MERGE dbo.[Account] AS t
USING (SELECT m.[AccountId],
m.[TenantId],
a.[ID_Account],
a.[Account_No],
a.[Account_Name],
FROM [Client1].dbo.[Account] a
LEFT JOIN migration.[Account] m ON m.[ID_Account] = a.[ID_Account]
AND m.[TenantId] = #TenantId
WHERE m.isUpdated IS NULL OR m.isUpdated = 0
) AS s
ON (t.[AccountId] = s.[AccountId] AND t.[TenantId] = s.[TenantId])
WHEN NOT MATCHED THEN
INSERT ([TenantId], [Number], [Name], [Active] )
VALUES (#TenantId, s.[Account_No], s.[Account_Name], 1)
OUTPUT #TenantId, s.AccountId [OriginalAccountId], inserted.[NewAccountId]
) [o] ON o.TenantId = m.TenantId AND o.OriginalAccountId = m.AccountId

Answer 2

What you want to do is do updates on when matched:
MERGE into dbo.[Account] AS destination
USING (SELECT m.[AccountId],
m.[TenantId],
a.[ID_Account],
a.[Account_No],
a.[Account_Name],
FROM [Client1].dbo.[Account] a
LEFT JOIN migration.[Account] m ON m.[ID_Account] = a.[ID_Account] AND m.[TenantId] = #TenantId
) AS holdinarea
ON (destination.[AccountId] = holdinarea.[AccountId] AND destination.[TenantId] = holdinarea.[TenantId])
WHEN MATCHED THEN
UPDATE set [AccountId]=holdinarea.accountid,
[TenantId]=holdinarea.tenantID
WHEN NOT MATCHED THEN
INSERT ([TenantId], [Number], [Name], [Active] )
VALUES (holdingarea.TenantId, holdingarea.[Account_No], holdingarea.[Account_Name], 1);