-- -- MULTI_JOIN_PRUNING -- -- Check that join-pruning works for joins between two relations. For now -- we only check for join-pruning between locally partitioned relations. In the -- future we want to check for pruning between re-partitioned relations as well. SET citus.explain_distributed_queries TO off; SET client_min_messages TO DEBUG2; SELECT sum(l_linenumber), avg(l_linenumber) FROM lineitem, orders WHERE l_orderkey = o_orderkey; DEBUG: Router planner cannot handle multi-shard select queries DEBUG: join prunable for intervals [-2147483648,-1] and [0,2147483647] DEBUG: join prunable for intervals [0,2147483647] and [-2147483648,-1] sum | avg --------------------------------------------------------------------- 36089 | 3.0074166666666667 (1 row) SELECT sum(l_linenumber), avg(l_linenumber) FROM lineitem, orders WHERE l_orderkey = o_orderkey AND l_orderkey > 9030; DEBUG: Router planner cannot handle multi-shard select queries DEBUG: join prunable for intervals [-2147483648,-1] and [0,2147483647] DEBUG: join prunable for intervals [0,2147483647] and [-2147483648,-1] sum | avg --------------------------------------------------------------------- 17999 | 3.0189533713518953 (1 row) -- Shards for the lineitem table have been pruned away. Check that join pruning -- works as expected in this case. SELECT sum(l_linenumber), avg(l_linenumber) FROM lineitem, orders WHERE l_orderkey = o_orderkey AND l_orderkey > 20000; DEBUG: Router planner cannot handle multi-shard select queries DEBUG: join prunable for intervals [-2147483648,-1] and [0,2147483647] DEBUG: join prunable for intervals [0,2147483647] and [-2147483648,-1] sum | avg --------------------------------------------------------------------- | (1 row) -- Partition pruning left three shards for the lineitem and one shard for the -- orders table. These shard sets don't overlap, so join pruning should prune -- out all the shards, and leave us with an empty task list. select * from pg_dist_shard where logicalrelid='lineitem'::regclass or logicalrelid='orders'::regclass order by shardid; logicalrelid | shardid | shardstorage | shardminvalue | shardmaxvalue --------------------------------------------------------------------- lineitem | 360000 | t | -2147483648 | -1 lineitem | 360001 | t | 0 | 2147483647 orders | 360002 | t | -2147483648 | -1 orders | 360003 | t | 0 | 2147483647 (4 rows) set client_min_messages to debug3; SELECT sum(l_linenumber), avg(l_linenumber) FROM lineitem, orders WHERE l_orderkey = o_orderkey AND l_orderkey > 6000 AND o_orderkey < 6000; DEBUG: no shard pruning constraints on lineitem found DEBUG: shard count after pruning for lineitem: 2 DEBUG: no shard pruning constraints on orders found DEBUG: shard count after pruning for orders: 2 DEBUG: Router planner cannot handle multi-shard select queries DEBUG: no shard pruning constraints on lineitem found DEBUG: shard count after pruning for lineitem: 2 DEBUG: no shard pruning constraints on orders found DEBUG: shard count after pruning for orders: 2 DEBUG: join prunable for intervals [-2147483648,-1] and [0,2147483647] DEBUG: join prunable for intervals [0,2147483647] and [-2147483648,-1] DEBUG: assigned task to node localhost:xxxxx DEBUG: assigned task to node localhost:xxxxx sum | avg --------------------------------------------------------------------- | (1 row) set client_min_messages to debug2; -- Make sure that we can handle filters without a column SELECT sum(l_linenumber), avg(l_linenumber) FROM lineitem, orders WHERE l_orderkey = o_orderkey AND false; DEBUG: Creating router plan sum | avg --------------------------------------------------------------------- | (1 row) SELECT sum(l_linenumber), avg(l_linenumber) FROM lineitem INNER JOIN orders ON (l_orderkey = o_orderkey) WHERE false; DEBUG: Creating router plan sum | avg --------------------------------------------------------------------- | (1 row) -- These tests check that we can do join pruning for tables partitioned over -- different type of columns including varchar, array types, composite types -- etc. This is in response to a bug we had where we were not able to resolve -- correct operator types for some kind of column types. EXPLAIN (COSTS OFF) SELECT count(*) FROM array_partitioned_table table1, array_partitioned_table table2 WHERE table1.array_column = table2.array_column; DEBUG: Router planner cannot handle multi-shard select queries DEBUG: join prunable for intervals [{},{AZZXSP27F21T6,AZZXSP27F21T6}] and [{BA1000U2AMO4ZGX,BZZXSP27F21T6},{CA1000U2AMO4ZGX,CZZXSP27F21T6}] DEBUG: join prunable for intervals [{BA1000U2AMO4ZGX,BZZXSP27F21T6},{CA1000U2AMO4ZGX,CZZXSP27F21T6}] and [{},{AZZXSP27F21T6,AZZXSP27F21T6}] QUERY PLAN --------------------------------------------------------------------- Aggregate -> Custom Scan (Citus Adaptive) citus.explain_distributed_queries: false (3 rows) EXPLAIN (COSTS OFF) SELECT count(*) FROM composite_partitioned_table table1, composite_partitioned_table table2 WHERE table1.composite_column = table2.composite_column; DEBUG: Router planner cannot handle multi-shard select queries DEBUG: join prunable for intervals [(a,3,b),(b,4,c)] and [(c,5,d),(d,6,e)] DEBUG: join prunable for intervals [(c,5,d),(d,6,e)] and [(a,3,b),(b,4,c)] QUERY PLAN --------------------------------------------------------------------- Aggregate -> Custom Scan (Citus Adaptive) citus.explain_distributed_queries: false (3 rows) -- Test that large table joins on partition varchar columns work EXPLAIN (COSTS OFF) SELECT count(*) FROM varchar_partitioned_table table1, varchar_partitioned_table table2 WHERE table1.varchar_column = table2.varchar_column; DEBUG: Router planner cannot handle multi-shard select queries DEBUG: join prunable for intervals [AA1000U2AMO4ZGX,AZZXSP27F21T6] and [BA1000U2AMO4ZGX,BZZXSP27F21T6] DEBUG: join prunable for intervals [BA1000U2AMO4ZGX,BZZXSP27F21T6] and [AA1000U2AMO4ZGX,AZZXSP27F21T6] QUERY PLAN --------------------------------------------------------------------- Aggregate -> Custom Scan (Citus Adaptive) citus.explain_distributed_queries: false (3 rows) SET client_min_messages TO WARNING; DROP TABLE varchar_partitioned_table; DROP TABLE array_partitioned_table; DROP TABLE composite_partitioned_table;