Randall Lucas <rlucas@tercent.net> writes:
> Do you have any suggestions on how this might
> better be implemented? It seems to me that with a partial index, there
> is not too much overhead and it doesn't seem too offensively
> un-boycecoddean.
FWIW, I'd go with the unique partial index too. It's a very efficient
solution. But it's not portable :-(. I'd be interested to hear what
people would do to solve this problem in bog-standard SQL.
regards, tom lane