On Fri, 2005-03-25 at 15:10, Terry Lee Tucker wrote:
> Hi,
>
> If I were to set the value of pg_class.indisunique on a unique index to False
> inside a transaction so I could juggle sequence numbers around on a table
> with a unique two element index, and then set it back again to its proper
> value, all in the same transaction, would that allow me to temorarily
> override the unique index behavior? Is it safe to temporarily change the
> value of that column?
You could, but, in order to ensure there are no duplicates, you'd have
to check after turning it back on to see if there were and delete them
by some logic that made sense for your transactional methodologies.
I wouldn't recommend it really, because I'm willing to bet there are
race conditions I'm not thinking of that could bite you in the behind.