How to upsert (update or insert) in SQL Server 2005

Solution 1:

Try to check for existence:

IF NOT EXISTS (SELECT * FROM dbo.Employee WHERE ID = @SomeID)

    INSERT INTO dbo.Employee(Col1, ..., ColN)
    VALUES(Val1, .., ValN)

ELSE

    UPDATE dbo.Employee
    SET Col1 = Val1, Col2 = Val2, ...., ColN = ValN
    WHERE ID = @SomeID

You could easily wrap this into a stored procedure and just call that stored procedure from the outside (e.g. from a programming language like C# or whatever you're using).

Update: either you can just write this entire statement in one long string (doable - but not really very useful) - or you can wrap it into a stored procedure:

CREATE PROCEDURE dbo.InsertOrUpdateEmployee
       @ID INT,
       @Name VARCHAR(50),
       @ItemName VARCHAR(50),  
       @ItemCatName VARCHAR(50),
       @ItemQty DECIMAL(15,2)
AS BEGIN
    IF NOT EXISTS (SELECT * FROM dbo.Table1 WHERE ID = @ID)
       INSERT INTO dbo.Table1(ID, Name, ItemName, ItemCatName, ItemQty)
       VALUES(@ID, @Name, @ItemName, @ItemCatName, @ItemQty)
    ELSE
       UPDATE dbo.Table1
       SET Name = @Name,
           ItemName = @ItemName,
           ItemCatName = @ItemCatName,
           ItemQty = @ItemQty
       WHERE ID = @ID
END

and then just call that stored procedure from your ADO.NET code

Solution 2:

You can use @@ROWCOUNT to check whether row should be inserted or updated:

update table1 
set name = 'val2', itemname = 'val3', itemcatName = 'val4', itemQty = 'val5'
where id = 'val1'
if @@ROWCOUNT = 0
insert into table1(id, name, itemname, itemcatName, itemQty)
values('val1', 'val2', 'val3', 'val4', 'val5')

in this case if update fails, the new row will be inserted

Solution 3:

You can check if the row exists, and then INSERT or UPDATE, but this guarantees you will be performing two SQL operations instead of one:

1. check if row exists
2. insert or update row

A better solution is to always UPDATE first, and if no rows were updated, then do an INSERT, like so:

update table1 
set name = 'val2', itemname = 'val3', itemcatName = 'val4', itemQty = 'val5'
where id = 'val1'

if @@ROWCOUNT = 0
  insert into table1(id, name, itemname, itemcatName, itemQty)
  values('val1', 'val2', 'val3', 'val4', 'val5')

This will either take one SQL operations, or two SQL operations, depending on whether the row already exists.

But if performance is really an issue, then you need to figure out if the operations are more likely to be INSERT's or UPDATE's. If UPDATE's are more common, do the above. If INSERT's are more common, you can do that in reverse, but you have to add error handling.

BEGIN TRY
  insert into table1(id, name, itemname, itemcatName, itemQty)
  values('val1', 'val2', 'val3', 'val4', 'val5')
END TRY
BEGIN CATCH
  update table1 
  set name = 'val2', itemname = 'val3', itemcatName = 'val4', itemQty = 'val5'
  where id = 'val1'
END CATCH

To be really certain if you need to do an UPDATE or INSERT, you have to do two operations within a single TRANSACTION. Theoretically, right after the first UPDATE or INSERT (or even the EXISTS check), but before the next INSERT/UPDATE statement, the database could have changed, causing the second statement to fail anyway. This is exceedingly rare, and the overhead for transactions may not be worth it.

Alternately, you can use a single SQL operation called MERGE to perform either an INSERT or an UPDATE, but that's also probably overkill for this one-row operation.

Consider reading about SQL transaction statements, race conditions, SQL MERGE statement.