On 2/6/18 17:15, Michael Paquier wrote: > On Tue, Feb 06, 2018 at 03:16:59PM -0500, Robert Haas wrote: >> What possible point can there be to such an expression? It's always 0. >> >> rhaas=# select distinct tgtype::smallint & 1 & 66 from >> generate_series(-32768,32767) tgtype; >> ?column? >> ---------- >> 0 >> (1 row) > > Of course you are right here. I just had a look at the patch again > after waking up and the current patch builds action_order based only on > action_timing and action_orientation, but it forgets event_manipulation. > For example row and statement triggers are correctly filtered, but the > ordering number includes both insert, update and delete types. So what > you need to use instead is (t.tgtype & 1), (t.tgtype & 66) as filter.
I think what I had meant to write was something like (t.tgtype & (1 | 66)) but maybe it's clearer to write it all out as you did. -- Peter Eisentraut http://www.2ndQuadrant.com/ PostgreSQL Development, 24x7 Support, Remote DBA, Training & Services