Re: Strange (and good) side effect of partitioning ?

I wrote:
> There's no specific mechanism in Postgres that would cause "X between 20
> and 10" to be reduced to constant-false

Wait, I take that back. There is a mechanism that can conclude that
"X >= 20" and "X <= 10" are contradictory, but it's not applied by
default. Observe:

regression=# set constraint_exclusion = default;
SET
regression=# explain select * from tenk1 where unique1 between 20 and 10;
QUERY PLAN
-----------------------------------------------------------------------------
Index Scan using tenk1_unique1 on tenk1 (cost=0.29..8.30 rows=1 width=244)
Index Cond: ((unique1 >= 20) AND (unique1 <= 10))
(2 rows)

regression=# set constraint_exclusion = on;
SET
regression=# explain select * from tenk1 where unique1 between 20 and 10;
QUERY PLAN
------------------------------------------
Result (cost=0.00..0.00 rows=0 width=0)
One-Time Filter: false
(2 rows)

The default value of constraint_exclusion is "partition", which means
(you guessed it) that it's applied only to potential partitioning
constraints. This is a heuristic based on the typical payoff of
excluding whole partitions versus skipping an empty index scan.
But if you have a workload where it's really worth spending
planner cycles looking for self-contradictory queries, you can
turn it on.

regards, tom lane