Vincenzo Romano <vincenzo.romano@xxxxxxxxx> writes: > On Sunday 22 July 2007 19:20:08 Tom Lane wrote: >> With what index, pray tell? > In the original setup, the "UNIQUE" constraint had been dropped > *before* doing the tests. So the "slow" case is without the > UNIQUE constraint but with an index on NOT NULL fields. You haven't said where you think this index is coming from. regards, tom lane