On most nested loops that I do explain/explain analyze on, the row estimation for the nested-loop itself is a product
ofthe inner nodes of the nested loop.
However in this case, I am stumped!
explain
select era.child_entity from entity_rel era join user_entity ue on ue.entity_id = era.parent_entity and
ue.user_id=12345
Nested Loop (cost=0.00..2903.37 rows=29107 width=4)
-> Index Only Scan using entity_pk on user_entity ue (cost=0.00..62.68 rows=2 width=4)
Index Cond: (user_id = 10954)
-> Index Scan using rel_parent on entity_rel era (cost=0.00..1261.85 rows=317 width=8)
Index Cond: (parent_entity = ue.entity_id)
How can the estimated number of rows for the nested loop node EXCEED the product of the 2 row estimates of the tables
beingjoined?
Not only does it exceed it - but it is orders of magnitude greater.
Am I missing something obvious here? I an see the nested loop row estimate being LESS but certainly not more.
PostgreSQL 9.2.4 on x86_64-pc-solaris2.10, compiled by gcc (GCC) 3.4.3 (csl-sol210-3_4-branch+sol_rpath), 64-bit