Re: Wrong rows estimations with joins of CTEs slows queries by more than factor 500

Agreed.  I think the patch is still valuable on its own, although it
does not go down into MATERIALIZED case for further optimization.  Maybe
we can take another query as regression test to prove its value, in
which the CTE is not inlined without MATERIALIZED, such as

explain (costs off)
with x as (select unique1, unique2 from tenk1 b)
select count(*) from tenk1 a
where unique1 in (select unique1 from x x1) and
      unique1 in (select unique2 from x x2);
                            QUERY PLAN
------------------------------------------------------------------
 Aggregate
   CTE x
     ->  Seq Scan on tenk1 b
   ->  Hash Join
         Hash Cond: (a.unique1 = x2.unique2)
         ->  Nested Loop
               ->  HashAggregate
                     Group Key: x1.unique1
                     ->  CTE Scan on x x1
               ->  Index Only Scan using tenk1_unique1 on tenk1 a
                     Index Cond: (unique1 = x1.unique1)
         ->  Hash
               ->  HashAggregate
                     Group Key: x2.unique2
                     ->  CTE Scan on x x2
(15 rows)

vs

explain (costs off)
with x as (select unique1, unique2 from tenk1 b)
select count(*) from tenk1 a
where unique1 in (select unique1 from x x1) and
      unique1 in (select unique2 from x x2);
                            QUERY PLAN
------------------------------------------------------------------
 Aggregate
   CTE x
     ->  Seq Scan on tenk1 b
   ->  Hash Semi Join
         Hash Cond: (a.unique1 = x2.unique2)
         ->  Hash Semi Join
               Hash Cond: (a.unique1 = x1.unique1)
               ->  Index Only Scan using tenk1_unique1 on tenk1 a
               ->  Hash
                     ->  CTE Scan on x x1
         ->  Hash
               ->  CTE Scan on x x2
(12 rows)

I believe the second plan is faster in reality too.

Thanks
Richard