Question

I have a table with INT primary key column. I do not want to use GUID or IDENTITY as primary key.

I wanted to know what is the best possible way to get the next ID without using an IDENTITY or GUID as primary key column in the table. I do not mind using GUID or IDENTITY in the table unless it’s the primary key column.

I have to find the next available ID (i.e get the MAX ID and increment it by 1), and use that:

SELECT @id=ISNULL(max(AlbumId)+1,1) FROM Albums

However, I want to prevent other applications from inserting into the table when I’m doing this so that we don’t have any problems.

NOTE: Kindly do not mark this question as DUPLICATE . I have gone through this answer and could not understand it properly. Would anyone be kind enough to explain me what is it they are doing? And can I use the same technique?

Answer 1

For SQL Server 2000 – 2008 R2, using INT IDENTITY is by far the easiest and most reliable way to do it.

Trying to do this yourself is like reinventing the wheel and carries a lot of ways in which it can go wrong – so why bother?? You need to (a) make sure this mechanism is concurrency safe (and just doing SELECT MAX(ID) + 1 is NOT safe!), and you (b) need to make sure your mechanism doesn’t become a major bottleneck in system performance, either…

What’s your problem with using INT IDENTITY as your primary key?? Doesn’t seem very rational…

For SQL Server 2012 and newer, you could also consider using a SEQUENCE (which is basically an IDENTITY that’s not specifically coupled to a single table)

Update: that other SO question that you mention in your post is doing basically a “do-it-yourself” simulation of a SEQUENCE:

• a table holds pairs of (sequencename, currentvalue)
• when you ask for a new value for a given sequence, the new value is determined by using just an UPDATE statement coupled with the OUTPUT clause

Why so complicated? This is the easiest and most efficient way (other than using an IDENTITY) to safely handle concurrent requests. The UPDATE statement will always exclusively lock that one row in the “sequence table” for the sequence you want to get a new ID from – thus preventing any other transactions to grab the same ID and get duplicated values – and it returns the new ID using the OUTPUT clause back to the caller.