Tom Lane wrote:
>regression=# explain analyze select * from pg_proc WHERE NOT proisagg AND (pronamespace != (select oid from
pg_namespacewhere nspname = 'pg_catalog')
>regression(# OR oid in (select lanplcallfoid from pg_language) OR oid in (select lanvalidator from pg_language));
>
>
>
Yeah. I think we'd need to add "where langlcallfoid != 0" so we don't
pick up the internal/C/sql handlers. However, on closer inspection it
appears that doind all this in pg_dump would be lots more invasive than
I first thought. An alternative would be to adopt AndrewSN's suggestion
of an extra schema (or possibly two) created by initdb where we put
extra stuff. That would mean a very simple addition to initdb.c and no
changes to pg_dump, except ...
>Next question is whether there are any other places that would be
>affected besides createlang/droplang. I can't think of any offhand,
>but ...
>
>
>
>
me either, but I wonder if we should provide an option on pg_dump to
restore function handlers found in public to whatever we decide about
the above.
thoughts?
cheers
andrew