-- =================================================================== -- test recursive planning functionality on views -- =================================================================== CREATE SCHEMA subquery_view; SET search_path TO subquery_view, public; SET citus.next_shard_id TO 1512000; CREATE TABLE users_table_local AS SELECT * FROM users_table; CREATE TABLE events_table_local AS SELECT * FROM events_table; SET client_min_messages TO DEBUG1; CREATE VIEW view_without_subquery AS SELECT DISTINCT users_table.value_1 FROM users_table, events_table WHERE users_table.user_id = events_table.user_id AND event_type IN (1,2,3,4) ORDER BY 1 DESC; SELECT * FROM view_without_subquery ORDER BY 1 DESC LIMIT 5; DEBUG: generating subplan XXX_1 for subquery SELECT DISTINCT users_table.value_1 FROM public.users_table, public.events_table WHERE ((users_table.user_id OPERATOR(pg_catalog.=) events_table.user_id) AND (events_table.event_type OPERATOR(pg_catalog.=) ANY (ARRAY[1, 2, 3, 4]))) ORDER BY users_table.value_1 DESC DEBUG: Plan XXX query after replacing subqueries and CTEs: SELECT value_1 FROM (SELECT intermediate_result.value_1 FROM read_intermediate_result('XXX_1'::text, 'binary'::citus_copy_format) intermediate_result(value_1 integer)) view_without_subquery ORDER BY value_1 DESC LIMIT 5 value_1 --------------------------------------------------------------------- 5 4 3 2 1 (5 rows) CREATE VIEW view_without_subquery_second AS SELECT DISTINCT users_table.user_id FROM users_table, events_table WHERE users_table.user_id = events_table.user_id AND event_type IN (1,2,3,4) ORDER BY 1 DESC LIMIT 5; SELECT * FROM view_without_subquery_second ORDER BY 1; DEBUG: push down of limit count: 5 DEBUG: generating subplan XXX_1 for subquery SELECT DISTINCT users_table.user_id FROM public.users_table, public.events_table WHERE ((users_table.user_id OPERATOR(pg_catalog.=) events_table.user_id) AND (events_table.event_type OPERATOR(pg_catalog.=) ANY (ARRAY[1, 2, 3, 4]))) ORDER BY users_table.user_id DESC LIMIT 5 DEBUG: Plan XXX query after replacing subqueries and CTEs: SELECT user_id FROM (SELECT intermediate_result.user_id FROM read_intermediate_result('XXX_1'::text, 'binary'::citus_copy_format) intermediate_result(user_id integer)) view_without_subquery_second ORDER BY user_id user_id --------------------------------------------------------------------- 2 3 4 5 6 (5 rows) -- subqueries in FROM clause with LIMIT should be recursively planned CREATE VIEW subquery_limit AS SELECT user_id FROM (SELECT DISTINCT users_table.user_id FROM users_table, events_table WHERE users_table.user_id = events_table.user_id AND event_type IN (1,2,3,4) ORDER BY 1 DESC LIMIT 5 ) as foo ORDER BY 1 DESC; SELECT * FROM subquery_limit ORDER BY 1 DESC; DEBUG: push down of limit count: 5 DEBUG: generating subplan XXX_1 for subquery SELECT DISTINCT users_table.user_id FROM public.users_table, public.events_table WHERE ((users_table.user_id OPERATOR(pg_catalog.=) events_table.user_id) AND (events_table.event_type OPERATOR(pg_catalog.=) ANY (ARRAY[1, 2, 3, 4]))) ORDER BY users_table.user_id DESC LIMIT 5 DEBUG: Plan XXX query after replacing subqueries and CTEs: SELECT user_id FROM (SELECT foo.user_id FROM (SELECT intermediate_result.user_id FROM read_intermediate_result('XXX_1'::text, 'binary'::citus_copy_format) intermediate_result(user_id integer)) foo ORDER BY foo.user_id DESC) subquery_limit ORDER BY user_id DESC user_id --------------------------------------------------------------------- 6 5 4 3 2 (5 rows) -- subqueries in FROM clause with GROUP BY non-distribution column should be recursively planned CREATE VIEW subquery_non_p_key_group_by AS SELECT * FROM (SELECT DISTINCT users_table.value_1 FROM users_table, events_table WHERE users_table.user_id = events_table.user_id AND event_type IN (1,2,3,4) ORDER BY 1 ) as foo ORDER BY 1 DESC; SELECT * FROM subquery_non_p_key_group_by ORDER BY 1 DESC; DEBUG: generating subplan XXX_1 for subquery SELECT DISTINCT users_table.value_1 FROM public.users_table, public.events_table WHERE ((users_table.user_id OPERATOR(pg_catalog.=) events_table.user_id) AND (events_table.event_type OPERATOR(pg_catalog.=) ANY (ARRAY[1, 2, 3, 4]))) ORDER BY users_table.value_1 DEBUG: Plan XXX query after replacing subqueries and CTEs: SELECT value_1 FROM (SELECT foo.value_1 FROM (SELECT intermediate_result.value_1 FROM read_intermediate_result('XXX_1'::text, 'binary'::citus_copy_format) intermediate_result(value_1 integer)) foo ORDER BY foo.value_1 DESC) subquery_non_p_key_group_by ORDER BY value_1 DESC value_1 --------------------------------------------------------------------- 5 4 3 2 1 0 (6 rows) CREATE VIEW final_query_router AS SELECT * FROM (SELECT users_table.value_2 FROM users_table, events_table WHERE users_table.user_id = events_table.user_id AND event_type IN (1,2,3,4) GROUP BY users_table.value_2 ORDER BY 1 DESC ) as foo, (SELECT users_table.value_3 FROM users_table, events_table WHERE users_table.user_id = events_table.user_id AND event_type IN (5,6,7,8) GROUP BY users_table.value_3 ORDER BY 1 DESC ) as bar WHERE foo.value_2 = bar.value_3 ORDER BY 2 DESC, 1; SELECT * FROM final_query_router ORDER BY 1; DEBUG: generating subplan XXX_1 for subquery SELECT users_table.value_2 FROM public.users_table, public.events_table WHERE ((users_table.user_id OPERATOR(pg_catalog.=) events_table.user_id) AND (events_table.event_type OPERATOR(pg_catalog.=) ANY (ARRAY[1, 2, 3, 4]))) GROUP BY users_table.value_2 ORDER BY users_table.value_2 DESC DEBUG: generating subplan XXX_2 for subquery SELECT users_table.value_3 FROM public.users_table, public.events_table WHERE ((users_table.user_id OPERATOR(pg_catalog.=) events_table.user_id) AND (events_table.event_type OPERATOR(pg_catalog.=) ANY (ARRAY[5, 6, 7, 8]))) GROUP BY users_table.value_3 ORDER BY users_table.value_3 DESC DEBUG: Plan XXX query after replacing subqueries and CTEs: SELECT value_2, value_3 FROM (SELECT foo.value_2, bar.value_3 FROM (SELECT intermediate_result.value_2 FROM read_intermediate_result('XXX_1'::text, 'binary'::citus_copy_format) intermediate_result(value_2 integer)) foo, (SELECT intermediate_result.value_3 FROM read_intermediate_result('XXX_2'::text, 'binary'::citus_copy_format) intermediate_result(value_3 double precision)) bar WHERE ((foo.value_2)::double precision OPERATOR(pg_catalog.=) bar.value_3) ORDER BY bar.value_3 DESC, foo.value_2) final_query_router ORDER BY value_2 value_2 | value_3 --------------------------------------------------------------------- 0 | 0 1 | 1 2 | 2 3 | 3 4 | 4 5 | 5 (6 rows) CREATE VIEW final_query_realtime AS SELECT * FROM (SELECT users_table.value_2 FROM users_table, events_table WHERE users_table.user_id = events_table.user_id AND event_type IN (1,2,3,4) GROUP BY users_table.value_2 ORDER BY 1 DESC ) as foo, (SELECT users_table.user_id FROM users_table, events_table WHERE users_table.user_id = events_table.user_id AND event_type IN (5,6,7,8) ORDER BY 1 DESC ) as bar WHERE foo.value_2 = bar.user_id ORDER BY 2 DESC, 1 DESC LIMIT 3; SELECT DISTINCT ON (users_table.value_2) users_table.value_2, time, value_3 FROM final_query_realtime, users_table WHERE users_table.user_id = final_query_realtime.user_id ORDER BY 1 DESC, 2 DESC, 3 DESC LIMIT 3; DEBUG: generating subplan XXX_1 for subquery SELECT users_table.value_2 FROM public.users_table, public.events_table WHERE ((users_table.user_id OPERATOR(pg_catalog.=) events_table.user_id) AND (events_table.event_type OPERATOR(pg_catalog.=) ANY (ARRAY[1, 2, 3, 4]))) GROUP BY users_table.value_2 ORDER BY users_table.value_2 DESC DEBUG: push down of limit count: 3 DEBUG: generating subplan XXX_2 for subquery SELECT foo.value_2, bar.user_id FROM (SELECT intermediate_result.value_2 FROM read_intermediate_result('XXX_1'::text, 'binary'::citus_copy_format) intermediate_result(value_2 integer)) foo, (SELECT users_table.user_id FROM public.users_table, public.events_table WHERE ((users_table.user_id OPERATOR(pg_catalog.=) events_table.user_id) AND (events_table.event_type OPERATOR(pg_catalog.=) ANY (ARRAY[5, 6, 7, 8]))) ORDER BY users_table.user_id DESC) bar WHERE (foo.value_2 OPERATOR(pg_catalog.=) bar.user_id) ORDER BY bar.user_id DESC, foo.value_2 DESC LIMIT 3 DEBUG: Plan XXX query after replacing subqueries and CTEs: SELECT DISTINCT ON (users_table.value_2) users_table.value_2, users_table."time", users_table.value_3 FROM (SELECT intermediate_result.value_2, intermediate_result.user_id FROM read_intermediate_result('XXX_2'::text, 'binary'::citus_copy_format) intermediate_result(value_2 integer, user_id integer)) final_query_realtime, public.users_table WHERE (users_table.user_id OPERATOR(pg_catalog.=) final_query_realtime.user_id) ORDER BY users_table.value_2 DESC, users_table."time" DESC, users_table.value_3 DESC LIMIT 3 DEBUG: push down of limit count: 3 value_2 | time | value_3 --------------------------------------------------------------------- 5 | Thu Nov 23 16:28:38.455322 2017 | 4 4 | Thu Nov 23 10:22:39.468816 2017 | 3 3 | Thu Nov 23 15:55:08.493462 2017 | 3 (3 rows) CREATE VIEW subquery_in_where AS SELECT DISTINCT user_id FROM users_table WHERE user_id IN (SELECT DISTINCT value_2 FROM users_table WHERE value_1 >= 1 AND value_1 <= 20 ORDER BY 1 LIMIT 5); SELECT * FROM subquery_in_where ORDER BY 1 DESC; DEBUG: push down of limit count: 5 DEBUG: generating subplan XXX_1 for subquery SELECT DISTINCT value_2 FROM public.users_table WHERE ((value_1 OPERATOR(pg_catalog.>=) 1) AND (value_1 OPERATOR(pg_catalog.<=) 20)) ORDER BY value_2 LIMIT 5 DEBUG: Plan XXX query after replacing subqueries and CTEs: SELECT user_id FROM (SELECT DISTINCT users_table.user_id FROM public.users_table WHERE (users_table.user_id OPERATOR(pg_catalog.=) ANY (SELECT intermediate_result.value_2 FROM read_intermediate_result('XXX_1'::text, 'binary'::citus_copy_format) intermediate_result(value_2 integer)))) subquery_in_where ORDER BY user_id DESC user_id --------------------------------------------------------------------- 4 3 2 1 (4 rows) -- subquery in FROM -> FROM -> WHERE should be replaced due to LIMIT CREATE VIEW subquery_from_from_where AS SELECT user_id, array_length(events_table, 1) FROM ( SELECT user_id, array_agg(event ORDER BY time) AS events_table FROM ( SELECT u.user_id, e.event_type::text AS event, e.time FROM users_table AS u, events_table AS e WHERE u.user_id = e.user_id AND u.user_id IN ( SELECT user_id FROM users_table WHERE value_2 >= 5 AND EXISTS (SELECT user_id FROM events_table WHERE event_type > 1 AND event_type <= 3 AND value_3 > 1 AND user_id = users_table.user_id) AND NOT EXISTS (SELECT user_id FROM events_table WHERE event_type > 3 AND event_type <= 4 AND value_3 > 1 AND user_id = users_table.user_id) LIMIT 5 ) ) t GROUP BY user_id ) q; SELECT * FROM subquery_from_from_where ORDER BY 2 DESC, 1; DEBUG: push down of limit count: 5 DEBUG: generating subplan XXX_1 for subquery SELECT user_id FROM public.users_table WHERE ((value_2 OPERATOR(pg_catalog.>=) 5) AND (EXISTS (SELECT events_table.user_id FROM public.events_table WHERE ((events_table.event_type OPERATOR(pg_catalog.>) 1) AND (events_table.event_type OPERATOR(pg_catalog.<=) 3) AND (events_table.value_3 OPERATOR(pg_catalog.>) (1)::double precision) AND (events_table.user_id OPERATOR(pg_catalog.=) users_table.user_id)))) AND (NOT (EXISTS (SELECT events_table.user_id FROM public.events_table WHERE ((events_table.event_type OPERATOR(pg_catalog.>) 3) AND (events_table.event_type OPERATOR(pg_catalog.<=) 4) AND (events_table.value_3 OPERATOR(pg_catalog.>) (1)::double precision) AND (events_table.user_id OPERATOR(pg_catalog.=) users_table.user_id)))))) LIMIT 5 DEBUG: Plan XXX query after replacing subqueries and CTEs: SELECT user_id, array_length FROM (SELECT q.user_id, array_length(q.events_table, 1) AS array_length FROM (SELECT t.user_id, array_agg(t.event ORDER BY t."time") AS events_table FROM (SELECT u.user_id, (e.event_type)::text AS event, e."time" FROM public.users_table u, public.events_table e WHERE ((u.user_id OPERATOR(pg_catalog.=) e.user_id) AND (u.user_id OPERATOR(pg_catalog.=) ANY (SELECT intermediate_result.user_id FROM read_intermediate_result('XXX_1'::text, 'binary'::citus_copy_format) intermediate_result(user_id integer))))) t GROUP BY t.user_id) q) subquery_from_from_where ORDER BY array_length DESC, user_id user_id | array_length --------------------------------------------------------------------- 5 | 364 (1 row) -- subquery in FROM -> FROM -> FROM should be replaced if -- it contains onle local tables CREATE VIEW subquery_from_from_where_local_table AS SELECT DISTINCT user_id FROM ( SELECT users_table.user_id FROM users_table, ( SELECT event_type, user_id FROM (SELECT event_type, users_table.user_id FROM users_table, (SELECT user_id, event_type FROM events_table_local WHERE value_2 < 3 OFFSET 3) as foo WHERE foo.user_id = users_table.user_id ) bar ) as baz WHERE baz.user_id = users_table.user_id ) as sub1; WARNING: "view subquery_from_from_where_local_table" has dependency to "table events_table_local" that is not in Citus' metadata DETAIL: "view subquery_from_from_where_local_table" will be created only locally HINT: Distribute "table events_table_local" first to distribute "view subquery_from_from_where_local_table" SELECT * FROM subquery_from_from_where ORDER BY 1 DESC LIMIT 3; DEBUG: push down of limit count: 5 DEBUG: generating subplan XXX_1 for subquery SELECT user_id FROM public.users_table WHERE ((value_2 OPERATOR(pg_catalog.>=) 5) AND (EXISTS (SELECT events_table.user_id FROM public.events_table WHERE ((events_table.event_type OPERATOR(pg_catalog.>) 1) AND (events_table.event_type OPERATOR(pg_catalog.<=) 3) AND (events_table.value_3 OPERATOR(pg_catalog.>) (1)::double precision) AND (events_table.user_id OPERATOR(pg_catalog.=) users_table.user_id)))) AND (NOT (EXISTS (SELECT events_table.user_id FROM public.events_table WHERE ((events_table.event_type OPERATOR(pg_catalog.>) 3) AND (events_table.event_type OPERATOR(pg_catalog.<=) 4) AND (events_table.value_3 OPERATOR(pg_catalog.>) (1)::double precision) AND (events_table.user_id OPERATOR(pg_catalog.=) users_table.user_id)))))) LIMIT 5 DEBUG: Plan XXX query after replacing subqueries and CTEs: SELECT user_id, array_length FROM (SELECT q.user_id, array_length(q.events_table, 1) AS array_length FROM (SELECT t.user_id, array_agg(t.event ORDER BY t."time") AS events_table FROM (SELECT u.user_id, (e.event_type)::text AS event, e."time" FROM public.users_table u, public.events_table e WHERE ((u.user_id OPERATOR(pg_catalog.=) e.user_id) AND (u.user_id OPERATOR(pg_catalog.=) ANY (SELECT intermediate_result.user_id FROM read_intermediate_result('XXX_1'::text, 'binary'::citus_copy_format) intermediate_result(user_id integer))))) t GROUP BY t.user_id) q) subquery_from_from_where ORDER BY user_id DESC LIMIT 3 DEBUG: push down of limit count: 3 user_id | array_length --------------------------------------------------------------------- 5 | 364 (1 row) SET citus.enable_repartition_joins to ON; CREATE VIEW repartition_view AS SELECT count(*) FROM ( SELECT DISTINCT users_table.value_2 FROM users_table, events_table WHERE users_table.user_id = events_table.value_2 AND users_table.user_id < 2 ) as foo, ( SELECT user_id FROM users_table ) as bar WHERE foo.value_2 = bar.user_id; SELECT * FROM repartition_view; DEBUG: generating subplan XXX_1 for subquery SELECT DISTINCT users_table.value_2 FROM public.users_table, public.events_table WHERE ((users_table.user_id OPERATOR(pg_catalog.=) events_table.value_2) AND (users_table.user_id OPERATOR(pg_catalog.<) 2)) DEBUG: generating subplan XXX_2 for subquery SELECT count(*) AS count FROM (SELECT intermediate_result.value_2 FROM read_intermediate_result('XXX_1'::text, 'binary'::citus_copy_format) intermediate_result(value_2 integer)) foo, (SELECT users_table.user_id FROM public.users_table) bar WHERE (foo.value_2 OPERATOR(pg_catalog.=) bar.user_id) DEBUG: Plan XXX query after replacing subqueries and CTEs: SELECT count FROM (SELECT intermediate_result.count FROM read_intermediate_result('XXX_2'::text, 'binary'::citus_copy_format) intermediate_result(count bigint)) repartition_view count --------------------------------------------------------------------- 58 (1 row) CREATE VIEW all_executors_view AS SELECT count(*) FROM ( SELECT value_2 FROM users_table WHERE user_id = 15 OFFSET 0 ) as foo, ( SELECT user_id FROM users_table OFFSET 0 ) as bar, ( SELECT DISTINCT users_table.value_2 FROM users_table, events_table WHERE users_table.user_id = events_table.value_2 AND users_table.user_id < 2 ) baz, ( SELECT user_id FROM users_table_local WHERE user_id = 2 ) baw WHERE foo.value_2 = bar.user_id AND baz.value_2 = bar.user_id AND bar.user_id = baw.user_id; WARNING: "view all_executors_view" has dependency to "table users_table_local" that is not in Citus' metadata DETAIL: "view all_executors_view" will be created only locally HINT: Distribute "table users_table_local" first to distribute "view all_executors_view" SELECT * FROM all_executors_view; DEBUG: generating subplan XXX_1 for subquery SELECT value_2 FROM public.users_table WHERE (user_id OPERATOR(pg_catalog.=) 15) OFFSET 0 DEBUG: generating subplan XXX_2 for subquery SELECT user_id FROM public.users_table OFFSET 0 DEBUG: generating subplan XXX_3 for subquery SELECT DISTINCT users_table.value_2 FROM public.users_table, public.events_table WHERE ((users_table.user_id OPERATOR(pg_catalog.=) events_table.value_2) AND (users_table.user_id OPERATOR(pg_catalog.<) 2)) DEBUG: generating subplan XXX_4 for subquery SELECT user_id FROM subquery_view.users_table_local WHERE (user_id OPERATOR(pg_catalog.=) 2) DEBUG: generating subplan XXX_5 for subquery SELECT count(*) AS count FROM (SELECT intermediate_result.value_2 FROM read_intermediate_result('XXX_1'::text, 'binary'::citus_copy_format) intermediate_result(value_2 integer)) foo, (SELECT intermediate_result.user_id FROM read_intermediate_result('XXX_2'::text, 'binary'::citus_copy_format) intermediate_result(user_id integer)) bar, (SELECT intermediate_result.value_2 FROM read_intermediate_result('XXX_3'::text, 'binary'::citus_copy_format) intermediate_result(value_2 integer)) baz, (SELECT intermediate_result.user_id FROM read_intermediate_result('XXX_4'::text, 'binary'::citus_copy_format) intermediate_result(user_id integer)) baw WHERE ((foo.value_2 OPERATOR(pg_catalog.=) bar.user_id) AND (baz.value_2 OPERATOR(pg_catalog.=) bar.user_id) AND (bar.user_id OPERATOR(pg_catalog.=) baw.user_id)) DEBUG: Plan XXX query after replacing subqueries and CTEs: SELECT count FROM (SELECT intermediate_result.count FROM read_intermediate_result('XXX_5'::text, 'binary'::citus_copy_format) intermediate_result(count bigint)) all_executors_view count --------------------------------------------------------------------- 0 (1 row) SET citus.enable_repartition_joins to OFF; -- the same query, but this time the CTEs also live inside a subquery CREATE VIEW subquery_and_ctes AS SELECT * FROM ( WITH cte AS ( WITH local_cte AS ( SELECT * FROM users_table_local ), dist_cte AS ( SELECT user_id FROM events_table, (SELECT DISTINCT value_2 FROM users_table OFFSET 0) as foo WHERE events_table.user_id = foo.value_2 AND events_table.user_id IN (SELECT DISTINCT value_1 FROM users_table ORDER BY 1 LIMIT 3) ) SELECT dist_cte.user_id FROM local_cte join dist_cte on dist_cte.user_id=local_cte.user_id ) SELECT count(*) as cnt FROM cte, (SELECT DISTINCT users_table.user_id FROM users_table, events_table WHERE users_table.user_id = events_table.user_id AND event_type IN (1,2,3,4) ORDER BY 1 DESC LIMIT 5 ) as foo WHERE foo.user_id = cte.user_id ) as foo, users_table WHERE foo.cnt > users_table.value_2; WARNING: "view subquery_and_ctes" has dependency to "table users_table_local" that is not in Citus' metadata DETAIL: "view subquery_and_ctes" will be created only locally HINT: Distribute "table users_table_local" first to distribute "view subquery_and_ctes" SELECT * FROM subquery_and_ctes ORDER BY 3 DESC, 1 DESC, 2 DESC, 4 DESC LIMIT 5; DEBUG: CTE cte is going to be inlined via distributed planning DEBUG: CTE local_cte is going to be inlined via distributed planning DEBUG: CTE dist_cte is going to be inlined via distributed planning DEBUG: generating subplan XXX_1 for subquery SELECT user_id, "time", value_1, value_2, value_3, value_4 FROM subquery_view.users_table_local DEBUG: push down of limit count: 3 DEBUG: generating subplan XXX_2 for subquery SELECT DISTINCT value_1 FROM public.users_table ORDER BY value_1 LIMIT 3 DEBUG: generating subplan XXX_3 for subquery SELECT DISTINCT value_2 FROM public.users_table OFFSET 0 DEBUG: push down of limit count: 5 DEBUG: generating subplan XXX_4 for subquery SELECT DISTINCT users_table.user_id FROM public.users_table, public.events_table WHERE ((users_table.user_id OPERATOR(pg_catalog.=) events_table.user_id) AND (events_table.event_type OPERATOR(pg_catalog.=) ANY (ARRAY[1, 2, 3, 4]))) ORDER BY users_table.user_id DESC LIMIT 5 DEBUG: generating subplan XXX_5 for subquery SELECT count(*) AS cnt FROM (SELECT dist_cte.user_id FROM ((SELECT intermediate_result.user_id, intermediate_result."time", intermediate_result.value_1, intermediate_result.value_2, intermediate_result.value_3, intermediate_result.value_4 FROM read_intermediate_result('XXX_1'::text, 'binary'::citus_copy_format) intermediate_result(user_id integer, "time" timestamp without time zone, value_1 integer, value_2 integer, value_3 double precision, value_4 bigint)) local_cte JOIN (SELECT events_table.user_id FROM public.events_table, (SELECT intermediate_result.value_2 FROM read_intermediate_result('XXX_3'::text, 'binary'::citus_copy_format) intermediate_result(value_2 integer)) foo_1 WHERE ((events_table.user_id OPERATOR(pg_catalog.=) foo_1.value_2) AND (events_table.user_id OPERATOR(pg_catalog.=) ANY (SELECT intermediate_result.value_1 FROM read_intermediate_result('XXX_2'::text, 'binary'::citus_copy_format) intermediate_result(value_1 integer))))) dist_cte ON ((dist_cte.user_id OPERATOR(pg_catalog.=) local_cte.user_id)))) cte, (SELECT intermediate_result.user_id FROM read_intermediate_result('XXX_4'::text, 'binary'::citus_copy_format) intermediate_result(user_id integer)) foo WHERE (foo.user_id OPERATOR(pg_catalog.=) cte.user_id) DEBUG: Plan XXX query after replacing subqueries and CTEs: SELECT cnt, user_id, "time", value_1, value_2, value_3, value_4 FROM (SELECT foo.cnt, users_table.user_id, users_table."time", users_table.value_1, users_table.value_2, users_table.value_3, users_table.value_4 FROM (SELECT intermediate_result.cnt FROM read_intermediate_result('XXX_5'::text, 'binary'::citus_copy_format) intermediate_result(cnt bigint)) foo, public.users_table WHERE (foo.cnt OPERATOR(pg_catalog.>) users_table.value_2)) subquery_and_ctes ORDER BY "time" DESC, cnt DESC, user_id DESC, value_1 DESC LIMIT 5 DEBUG: push down of limit count: 5 cnt | user_id | time | value_1 | value_2 | value_3 | value_4 --------------------------------------------------------------------- 432 | 1 | Thu Nov 23 17:30:34.635085 2017 | 3 | 4 | 4 | 432 | 1 | Thu Nov 23 17:23:03.441394 2017 | 5 | 4 | 3 | 432 | 3 | Thu Nov 23 17:18:51.048758 2017 | 1 | 5 | 5 | 432 | 3 | Thu Nov 23 17:10:35.959913 2017 | 4 | 3 | 1 | 432 | 5 | Thu Nov 23 16:48:32.08896 2017 | 5 | 2 | 1 | (5 rows) CREATE VIEW subquery_and_ctes_second AS SELECT time, event_type, value_2, value_3 FROM ( WITH cte AS MATERIALIZED ( WITH local_cte AS MATERIALIZED ( SELECT * FROM users_table_local ), dist_cte AS MATERIALIZED ( SELECT user_id FROM events_table ) SELECT dist_cte.user_id FROM local_cte join dist_cte on dist_cte.user_id=local_cte.user_id ) SELECT DISTINCT cte.user_id FROM users_table, cte WHERE users_table.user_id = cte.user_id AND users_table.user_id IN (WITH cte_in_where AS (SELECT DISTINCT value_2 FROM users_table WHERE value_1 >= 1 AND value_1 <= 20 ORDER BY 1 LIMIT 5) SELECT * FROM cte_in_where) ORDER BY 1 DESC ) as foo, events_table WHERE foo.user_id = events_table.value_2; WARNING: "view subquery_and_ctes_second" has dependency to "table users_table_local" that is not in Citus' metadata DETAIL: "view subquery_and_ctes_second" will be created only locally HINT: Distribute "table users_table_local" first to distribute "view subquery_and_ctes_second" SELECT * FROM subquery_and_ctes_second ORDER BY 3 DESC, 2 DESC, 1 DESC LIMIT 5; DEBUG: CTE cte_in_where is going to be inlined via distributed planning DEBUG: generating subplan XXX_1 for CTE cte: WITH local_cte AS MATERIALIZED (SELECT users_table_local.user_id, users_table_local."time", users_table_local.value_1, users_table_local.value_2, users_table_local.value_3, users_table_local.value_4 FROM subquery_view.users_table_local), dist_cte AS MATERIALIZED (SELECT events_table.user_id FROM public.events_table) SELECT dist_cte.user_id FROM (local_cte JOIN dist_cte ON ((dist_cte.user_id OPERATOR(pg_catalog.=) local_cte.user_id))) DEBUG: generating subplan XXX_1 for CTE local_cte: SELECT user_id, "time", value_1, value_2, value_3, value_4 FROM subquery_view.users_table_local DEBUG: generating subplan XXX_2 for CTE dist_cte: SELECT user_id FROM public.events_table DEBUG: Plan XXX query after replacing subqueries and CTEs: SELECT dist_cte.user_id FROM ((SELECT intermediate_result.user_id, intermediate_result."time", intermediate_result.value_1, intermediate_result.value_2, intermediate_result.value_3, intermediate_result.value_4 FROM read_intermediate_result('XXX_1'::text, 'binary'::citus_copy_format) intermediate_result(user_id integer, "time" timestamp without time zone, value_1 integer, value_2 integer, value_3 double precision, value_4 bigint)) local_cte JOIN (SELECT intermediate_result.user_id FROM read_intermediate_result('XXX_2'::text, 'binary'::citus_copy_format) intermediate_result(user_id integer)) dist_cte ON ((dist_cte.user_id OPERATOR(pg_catalog.=) local_cte.user_id))) DEBUG: push down of limit count: 5 DEBUG: generating subplan XXX_2 for subquery SELECT DISTINCT value_2 FROM public.users_table WHERE ((value_1 OPERATOR(pg_catalog.>=) 1) AND (value_1 OPERATOR(pg_catalog.<=) 20)) ORDER BY value_2 LIMIT 5 DEBUG: generating subplan XXX_3 for subquery SELECT DISTINCT cte.user_id FROM public.users_table, (SELECT intermediate_result.user_id FROM read_intermediate_result('XXX_1'::text, 'binary'::citus_copy_format) intermediate_result(user_id integer)) cte WHERE ((users_table.user_id OPERATOR(pg_catalog.=) cte.user_id) AND (users_table.user_id OPERATOR(pg_catalog.=) ANY (SELECT cte_in_where.value_2 FROM (SELECT intermediate_result.value_2 FROM read_intermediate_result('XXX_2'::text, 'binary'::citus_copy_format) intermediate_result(value_2 integer)) cte_in_where))) ORDER BY cte.user_id DESC DEBUG: Plan XXX query after replacing subqueries and CTEs: SELECT "time", event_type, value_2, value_3 FROM (SELECT events_table."time", events_table.event_type, events_table.value_2, events_table.value_3 FROM (SELECT intermediate_result.user_id FROM read_intermediate_result('XXX_3'::text, 'binary'::citus_copy_format) intermediate_result(user_id integer)) foo, public.events_table WHERE (foo.user_id OPERATOR(pg_catalog.=) events_table.value_2)) subquery_and_ctes_second ORDER BY value_2 DESC, event_type DESC, "time" DESC LIMIT 5 DEBUG: push down of limit count: 5 time | event_type | value_2 | value_3 --------------------------------------------------------------------- Thu Nov 23 21:54:46.924477 2017 | 6 | 4 | 5 Wed Nov 22 21:24:22.849224 2017 | 5 | 4 | 1 Wed Nov 22 21:05:25.194441 2017 | 5 | 4 | 1 Thu Nov 23 04:01:12.29256 2017 | 4 | 4 | 3 Thu Nov 23 09:33:16.992454 2017 | 3 | 4 | 1 (5 rows) CREATE VIEW deep_subquery AS SELECT count(*) FROM ( SELECT avg(min) FROM ( SELECT min(users_table.value_1) FROM ( SELECT avg(event_type) as avg_ev_type FROM ( SELECT max(value_1) as mx_val_1 FROM ( SELECT avg(event_type) as avg FROM ( SELECT cnt FROM (SELECT count(*) as cnt, value_2 FROM users_table GROUP BY value_2) as level_1, users_table WHERE users_table.user_id = level_1.cnt ) as level_2, events_table WHERE events_table.user_id = level_2.cnt GROUP BY level_2.cnt ) as level_3, users_table WHERE user_id = level_3.avg GROUP BY level_3.avg ) as level_4, events_table WHERE level_4.mx_val_1 = events_table.user_id GROUP BY level_4.mx_val_1 ) as level_5, users_table WHERE level_5.avg_ev_type = users_table.user_id GROUP BY level_5.avg_ev_type ) as level_6, users_table WHERE users_table.user_id = level_6.min GROUP BY users_table.value_1 ) as bar; SELECT * FROM deep_subquery; DEBUG: generating subplan XXX_1 for subquery SELECT count(*) AS cnt, value_2 FROM public.users_table GROUP BY value_2 DEBUG: generating subplan XXX_2 for subquery SELECT avg(events_table.event_type) AS avg FROM (SELECT level_1.cnt FROM (SELECT intermediate_result.cnt, intermediate_result.value_2 FROM read_intermediate_result('XXX_1'::text, 'binary'::citus_copy_format) intermediate_result(cnt bigint, value_2 integer)) level_1, public.users_table WHERE (users_table.user_id OPERATOR(pg_catalog.=) level_1.cnt)) level_2, public.events_table WHERE (events_table.user_id OPERATOR(pg_catalog.=) level_2.cnt) GROUP BY level_2.cnt DEBUG: generating subplan XXX_3 for subquery SELECT max(users_table.value_1) AS mx_val_1 FROM (SELECT intermediate_result.avg FROM read_intermediate_result('XXX_2'::text, 'binary'::citus_copy_format) intermediate_result(avg numeric)) level_3, public.users_table WHERE ((users_table.user_id)::numeric OPERATOR(pg_catalog.=) level_3.avg) GROUP BY level_3.avg DEBUG: generating subplan XXX_4 for subquery SELECT avg(events_table.event_type) AS avg_ev_type FROM (SELECT intermediate_result.mx_val_1 FROM read_intermediate_result('XXX_3'::text, 'binary'::citus_copy_format) intermediate_result(mx_val_1 integer)) level_4, public.events_table WHERE (level_4.mx_val_1 OPERATOR(pg_catalog.=) events_table.user_id) GROUP BY level_4.mx_val_1 DEBUG: generating subplan XXX_5 for subquery SELECT min(users_table.value_1) AS min FROM (SELECT intermediate_result.avg_ev_type FROM read_intermediate_result('XXX_4'::text, 'binary'::citus_copy_format) intermediate_result(avg_ev_type numeric)) level_5, public.users_table WHERE (level_5.avg_ev_type OPERATOR(pg_catalog.=) (users_table.user_id)::numeric) GROUP BY level_5.avg_ev_type DEBUG: generating subplan XXX_6 for subquery SELECT avg(level_6.min) AS avg FROM (SELECT intermediate_result.min FROM read_intermediate_result('XXX_5'::text, 'binary'::citus_copy_format) intermediate_result(min integer)) level_6, public.users_table WHERE (users_table.user_id OPERATOR(pg_catalog.=) level_6.min) GROUP BY users_table.value_1 DEBUG: generating subplan XXX_7 for subquery SELECT count(*) AS count FROM (SELECT intermediate_result.avg FROM read_intermediate_result('XXX_6'::text, 'binary'::citus_copy_format) intermediate_result(avg numeric)) bar DEBUG: Plan XXX query after replacing subqueries and CTEs: SELECT count FROM (SELECT intermediate_result.count FROM read_intermediate_result('XXX_7'::text, 'binary'::citus_copy_format) intermediate_result(count bigint)) deep_subquery count --------------------------------------------------------------------- 0 (1 row) CREATE VIEW result_of_view_is_also_recursively_planned AS SELECT user_id FROM (SELECT DISTINCT users_table.user_id FROM users_table, events_table WHERE users_table.user_id = events_table.user_id AND event_type IN (1,2,3,4) ORDER BY 1 DESC LIMIT 5 ) as foo ORDER BY 1 DESC; SELECT * FROM (SELECT * FROM result_of_view_is_also_recursively_planned, events_table WHERE events_table.value_2 = result_of_view_is_also_recursively_planned.user_id ORDER BY time DESC LIMIT 5 OFFSET 4 ) as foo ORDER BY time DESC LIMIT 5; DEBUG: push down of limit count: 5 DEBUG: generating subplan XXX_1 for subquery SELECT DISTINCT users_table.user_id FROM public.users_table, public.events_table WHERE ((users_table.user_id OPERATOR(pg_catalog.=) events_table.user_id) AND (events_table.event_type OPERATOR(pg_catalog.=) ANY (ARRAY[1, 2, 3, 4]))) ORDER BY users_table.user_id DESC LIMIT 5 DEBUG: push down of limit count: 9 DEBUG: generating subplan XXX_2 for subquery SELECT result_of_view_is_also_recursively_planned.user_id, events_table.user_id, events_table."time", events_table.event_type, events_table.value_2, events_table.value_3, events_table.value_4 FROM (SELECT foo.user_id FROM (SELECT intermediate_result.user_id FROM read_intermediate_result('XXX_1'::text, 'binary'::citus_copy_format) intermediate_result(user_id integer)) foo ORDER BY foo.user_id DESC) result_of_view_is_also_recursively_planned, public.events_table WHERE (events_table.value_2 OPERATOR(pg_catalog.=) result_of_view_is_also_recursively_planned.user_id) ORDER BY events_table."time" DESC OFFSET 4 LIMIT 5 DEBUG: Plan XXX query after replacing subqueries and CTEs: SELECT user_id, user_id_1 AS user_id, "time", event_type, value_2, value_3, value_4 FROM (SELECT intermediate_result.user_id, intermediate_result.user_id_1 AS user_id, intermediate_result."time", intermediate_result.event_type, intermediate_result.value_2, intermediate_result.value_3, intermediate_result.value_4 FROM read_intermediate_result('XXX_2'::text, 'binary'::citus_copy_format) intermediate_result(user_id integer, user_id_1 integer, "time" timestamp without time zone, event_type integer, value_2 integer, value_3 double precision, value_4 bigint)) foo(user_id, user_id_1, "time", event_type, value_2, value_3, value_4) ORDER BY "time" DESC LIMIT 5 user_id | user_id | time | event_type | value_2 | value_3 | value_4 --------------------------------------------------------------------- 2 | 3 | Thu Nov 23 16:44:41.903713 2017 | 4 | 2 | 2 | 2 | 5 | Thu Nov 23 16:11:02.929469 2017 | 4 | 2 | 0 | 4 | 5 | Thu Nov 23 14:40:40.467511 2017 | 1 | 4 | 1 | 3 | 2 | Thu Nov 23 14:02:47.738901 2017 | 1 | 3 | 2 | 3 | 6 | Thu Nov 23 14:00:13.20013 2017 | 3 | 3 | 3 | (5 rows) SET client_min_messages TO DEFAULT; CREATE TABLE reference_table (text_col text, int_col int); SELECT create_reference_table('reference_table'); create_reference_table --------------------------------------------------------------------- (1 row) SELECT public.coordinator_plan_with_subplans($Q$ EXPLAIN (COSTS OFF) WITH cte AS ( SELECT application_name AS text_col FROM pg_stat_activity ) SELECT * FROM reference_table JOIN cte USING (text_col); $Q$); coordinator_plan_with_subplans --------------------------------------------------------------------- Custom Scan (Citus Adaptive) Task Count: 1 (2 rows) CREATE TABLE dist_table(text_col text, int_col int); SELECT create_distributed_table('dist_table', 'text_col'); create_distributed_table --------------------------------------------------------------------- (1 row) SELECT public.coordinator_plan_with_subplans($Q$ EXPLAIN (COSTS OFF) WITH cte AS ( SELECT application_name AS text_col FROM pg_stat_activity ) SELECT * FROM dist_table JOIN cte USING (text_col); $Q$); coordinator_plan_with_subplans --------------------------------------------------------------------- Custom Scan (Citus Adaptive) -> Distributed Subplan XXX_1 -> Function Scan on pg_stat_get_activity s -> Distributed Subplan XXX_2 -> Custom Scan (Citus Adaptive) Task Count: 1 Task Count: 4 (7 rows) CREATE OR REPLACE VIEW view_on_views AS SELECT pg_stat_activity.application_name, pg_locks.pid FROM pg_stat_activity, pg_locks; SELECT public.coordinator_plan_with_subplans($Q$ EXPLAIN (COSTS OFF) WITH cte AS ( SELECT application_name AS text_col FROM view_on_views ) SELECT * FROM reference_table JOIN cte USING (text_col); $Q$); coordinator_plan_with_subplans --------------------------------------------------------------------- Custom Scan (Citus Adaptive) Task Count: 1 (2 rows) SELECT public.coordinator_plan_with_subplans($Q$ EXPLAIN (COSTS OFF) WITH cte AS ( SELECT application_name AS text_col FROM view_on_views ) SELECT * FROM dist_table JOIN cte USING (text_col); $Q$); coordinator_plan_with_subplans --------------------------------------------------------------------- Custom Scan (Citus Adaptive) -> Distributed Subplan XXX_1 -> Nested Loop -> Function Scan on pg_stat_get_activity s -> Function Scan on pg_lock_status l Task Count: 4 (6 rows) SET client_min_messages TO WARNING; DROP SCHEMA subquery_view CASCADE; SET search_path TO public;