Subject Re: [firebird-support] Is UPDATE in read-committed transaction atomic?
Author Ann W. Harrison
jaratulpas wrote:
> I have an application which uses read-committed transactions for
> performance reasons.

Have you actually tested the performance of the two modes, or
are you assuming that read-committed must be faster because it's
faster in other databases?

> Is the statement
> update object set cnt = cnt + 1 where id = 'a';
> safe, i.e. is the row locked for the duration of the update statement?

No, the row isn't locked at all, but yes, the statement is safe,
though it will lead to deadlocks.