On 7 December 2013 21:34, Pavel Stehule <pavel.stehule@gmail.com> wrote: >> Well I was basically proposing that does_not_exist_skipping() be >> enhanced to report on non-existent types that form part of the object >> specification. I think this would affect the CAST, FUNCTION, AGGREGATE >> and OPERATOR cases, but should be a fairly trivial extension to the >> code that you've already added. > > > ok, updated patch is in attachment >
Cool. This looks good to me, except I found a corner case --- the type name for an operator may be "NONE", in which case the typeName in the list will be NULL, so that needs to be guarded against. Updated patch attached.
I think this is a good patch. It makes all the DROP...IF EXISTS commands consistently fault-tolerant, instead of the current 50/50 mix, and all the resulting NOTICEs give useful information about why objects don't exist and are being skipped.