On Tue, 27 Aug 2024 at 13:40, Tom Lane <tgl@sss.pgh.pa.us> wrote:
> Yeah, it looks like that condition on "table_name" is not getting
> pushed down to the scan level anymore. I'm not sure why not,
> but will look closer tomorrow.
I was looking for the offending commit as at first I thought it might
be related to Memoize. It does not seem to be.
I get the following up until 2489d76c, and from then on, it's a subquery filter.
-> Index Scan using pg_class_relname_nsp_index on pg_class r_2
(cost=0.27..8.30 rows=1 width=8) (actual time=0.004..0.004 rows=0
loops=1)
Index Cond: (relname = 't_c56ng1_repository'::name)
Filter: ((relkind = ANY ('{r,p}'::"char"[])) AND
pg_has_role(relowner, 'USAGE'::text))
So looks like it was the "Make Vars be outer-join-aware." commit that
changed this.
David