seamusabshere wrote >> At READ COMMITTED isolation level, you should always get an atomic insert >> or update [1] > > I just think there are a lot of non-concurrent bulk loading and > processing workflows that could benefit from the performance advantages > of upsert (one trip to database). Bulk load raw data into UNLOGGED staging table LOCK production table UPDATE matched records INSERT unmatched records UNLOCK production table TRUNCATE staging table This seems like it would be sufficient for "non-concurrent bulk loading"... David J. -- View this message in context: http://postgresql.1045698.n5.nabble.com/Why-is-unique-constraint-needed-for-upsert-tp5812552p5812628.html Sent from the PostgreSQL - general mailing list archive at Nabble.com.
