Tom Lane wrote: > Richard Huxton <dev@xxxxxxxxxxxx> writes: > > Carl R. Brune wrote: > >> I should have added that I want to make further use of the > >> temporary table after the COMMIT -- the rollback approach you > >> propose makes it go away. > > > > In which case the transaction isn't READONLY. > > It does seem a bit inconsistent that we allow you to write into a > temp table during a "READONLY" transaction, but not to create/drop > one. I'm not excited about changing it though, as the tests to see if > the command is allowed would become vastly more complex. Temporary tables in the SQL standard are permanent objects, which is why creating or dropping them is a durable operation and not allowed in read-only transactions. It would probably make sense to allow creating or dropping PostgreSQL-style temporary tables, though. -- Peter Eisentraut http://developer.postgresql.org/~petere/