mirror of https://github.com/citusdata/citus.git
267 lines
13 KiB
Plaintext
267 lines
13 KiB
Plaintext
-- ===================================================================
|
|
-- test recursive planning functionality on local tables
|
|
-- ===================================================================
|
|
CREATE SCHEMA subquery_local_tables;
|
|
SET search_path TO subquery_local_tables, public;
|
|
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;
|
|
-- foo is only on the local tables, thus can be replaced
|
|
-- bar is on the distributed tables with LIMIT, should be replaced
|
|
SELECT
|
|
foo.user_id
|
|
FROM
|
|
(SELECT
|
|
DISTINCT users_table_local.user_id
|
|
FROM
|
|
users_table_local, events_table_local
|
|
WHERE
|
|
users_table_local.user_id = events_table_local.user_id AND
|
|
event_type IN (1,2,3,4)
|
|
ORDER BY 1 DESC LIMIT 5
|
|
) as foo,
|
|
(SELECT
|
|
DISTINCT 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 LIMIT 5
|
|
) as bar
|
|
WHERE bar.user_id = foo.user_id
|
|
ORDER BY 1 DESC;
|
|
DEBUG: generating subplan XXX_1 for subquery SELECT DISTINCT users_table_local.user_id FROM subquery_local_tables.users_table_local, subquery_local_tables.events_table_local WHERE ((users_table_local.user_id OPERATOR(pg_catalog.=) events_table_local.user_id) AND (events_table_local.event_type OPERATOR(pg_catalog.=) ANY (ARRAY[1, 2, 3, 4]))) ORDER BY users_table_local.user_id DESC LIMIT 5
|
|
DEBUG: push down of limit count: 5
|
|
DEBUG: generating subplan XXX_2 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[5, 6, 7, 8]))) ORDER BY users_table.user_id DESC LIMIT 5
|
|
DEBUG: Plan XXX query after replacing subqueries and CTEs: 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, (SELECT intermediate_result.user_id FROM read_intermediate_result('XXX_2'::text, 'binary'::citus_copy_format) intermediate_result(user_id integer)) bar WHERE (bar.user_id OPERATOR(pg_catalog.=) foo.user_id) ORDER BY foo.user_id DESC
|
|
user_id
|
|
---------------------------------------------------------------------
|
|
6
|
|
5
|
|
4
|
|
3
|
|
2
|
|
(5 rows)
|
|
|
|
-- foo is only on the local tables, thus can be replaced
|
|
SELECT
|
|
foo.user_id
|
|
FROM
|
|
(SELECT
|
|
DISTINCT users_table_local.user_id
|
|
FROM
|
|
users_table_local, events_table_local
|
|
WHERE
|
|
users_table_local.user_id = events_table_local.user_id AND
|
|
event_type IN (1,2,3,4)
|
|
ORDER BY 1 DESC LIMIT 5
|
|
) as foo,
|
|
(SELECT
|
|
DISTINCT 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)
|
|
) as bar
|
|
WHERE bar.user_id = foo.user_id
|
|
ORDER BY 1 DESC;
|
|
DEBUG: generating subplan XXX_1 for subquery SELECT DISTINCT users_table_local.user_id FROM subquery_local_tables.users_table_local, subquery_local_tables.events_table_local WHERE ((users_table_local.user_id OPERATOR(pg_catalog.=) events_table_local.user_id) AND (events_table_local.event_type OPERATOR(pg_catalog.=) ANY (ARRAY[1, 2, 3, 4]))) ORDER BY users_table_local.user_id DESC LIMIT 5
|
|
DEBUG: Plan XXX query after replacing subqueries and CTEs: 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, (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[5, 6, 7, 8])))) bar WHERE (bar.user_id OPERATOR(pg_catalog.=) foo.user_id) ORDER BY foo.user_id DESC
|
|
user_id
|
|
---------------------------------------------------------------------
|
|
6
|
|
5
|
|
4
|
|
3
|
|
2
|
|
(5 rows)
|
|
|
|
-- subqueries in WHERE could be replaced even if they are on the local tables
|
|
SELECT DISTINCT user_id
|
|
FROM users_table
|
|
WHERE
|
|
user_id IN (SELECT DISTINCT value_2 FROM users_table_local WHERE value_1 = 1)
|
|
ORDER BY 1 LIMIT 5;
|
|
DEBUG: generating subplan XXX_1 for subquery SELECT DISTINCT value_2 FROM subquery_local_tables.users_table_local WHERE (value_1 OPERATOR(pg_catalog.=) 1)
|
|
DEBUG: Plan XXX query after replacing subqueries and CTEs: SELECT DISTINCT user_id FROM public.users_table WHERE (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))) ORDER BY user_id LIMIT 5
|
|
DEBUG: push down of limit count: 5
|
|
user_id
|
|
---------------------------------------------------------------------
|
|
1
|
|
2
|
|
3
|
|
4
|
|
5
|
|
(5 rows)
|
|
|
|
-- subquery in FROM -> FROM -> FROM should be replaced if
|
|
-- it contains onle local tables
|
|
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
|
|
ORDER BY 1 DESC
|
|
LIMIT 3;
|
|
DEBUG: generating subplan XXX_1 for subquery SELECT user_id, event_type FROM subquery_local_tables.events_table_local WHERE (value_2 OPERATOR(pg_catalog.<) 3) OFFSET 3
|
|
DEBUG: Plan XXX query after replacing subqueries and CTEs: SELECT DISTINCT user_id FROM (SELECT users_table.user_id FROM public.users_table, (SELECT bar.event_type, bar.user_id FROM (SELECT foo.event_type, users_table_1.user_id FROM public.users_table users_table_1, (SELECT intermediate_result.user_id, intermediate_result.event_type FROM read_intermediate_result('XXX_1'::text, 'binary'::citus_copy_format) intermediate_result(user_id integer, event_type integer)) foo WHERE (foo.user_id OPERATOR(pg_catalog.=) users_table_1.user_id)) bar) baz WHERE (baz.user_id OPERATOR(pg_catalog.=) users_table.user_id)) sub1 ORDER BY user_id DESC LIMIT 3
|
|
DEBUG: push down of limit count: 3
|
|
user_id
|
|
---------------------------------------------------------------------
|
|
6
|
|
5
|
|
4
|
|
(3 rows)
|
|
|
|
-- subquery in FROM -> FROM -> WHERE -> WHERE should be replaced if
|
|
-- it contains onle local tables
|
|
-- Later the upper level query is also recursively planned due to LIMIT
|
|
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_local WHERE event_type > 1 AND event_type <= 3 AND value_3 > 1)
|
|
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
|
|
ORDER BY 2 DESC, 1;
|
|
DEBUG: generating subplan XXX_1 for subquery SELECT user_id FROM subquery_local_tables.events_table_local WHERE ((event_type OPERATOR(pg_catalog.>) 1) AND (event_type OPERATOR(pg_catalog.<=) 3) AND (value_3 OPERATOR(pg_catalog.>) (1)::double precision))
|
|
DEBUG: push down of limit count: 5
|
|
DEBUG: generating subplan XXX_2 for subquery SELECT user_id FROM public.users_table WHERE ((value_2 OPERATOR(pg_catalog.>=) 5) AND (EXISTS (SELECT intermediate_result.user_id FROM read_intermediate_result('XXX_1'::text, 'binary'::citus_copy_format) intermediate_result(user_id integer))) 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(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_2'::text, 'binary'::citus_copy_format) intermediate_result(user_id integer))))) t GROUP BY t.user_id) q ORDER BY (array_length(events_table, 1)) DESC, user_id
|
|
user_id | array_length
|
|
---------------------------------------------------------------------
|
|
5 | 364
|
|
(1 row)
|
|
|
|
-- subquery (i.e., subquery_2) in WHERE->FROM should be replaced due to local tables
|
|
SELECT
|
|
user_id
|
|
FROM
|
|
users_table
|
|
WHERE
|
|
user_id IN
|
|
(
|
|
SELECT
|
|
user_id
|
|
FROM (
|
|
SELECT
|
|
subquery_1.user_id, count_pay
|
|
FROM
|
|
(
|
|
(SELECT
|
|
users_table.user_id,
|
|
'action=>1' AS event,
|
|
events_table.time
|
|
FROM
|
|
users_table,
|
|
events_table
|
|
WHERE
|
|
users_table.user_id = events_table.user_id AND
|
|
users_table.user_id >= 1 AND
|
|
users_table.user_id <= 3 AND
|
|
events_table.event_type > 1 AND events_table.event_type < 3
|
|
)
|
|
UNION
|
|
(SELECT
|
|
users_table.user_id,
|
|
'action=>2' AS event,
|
|
events_table.time
|
|
FROM
|
|
users_table,
|
|
events_table
|
|
WHERE
|
|
users_table.user_id = events_table.user_id AND
|
|
users_table.user_id >= 1 AND
|
|
users_table.user_id <= 3 AND
|
|
events_table.event_type > 2 AND events_table.event_type < 4
|
|
)
|
|
) AS subquery_1
|
|
LEFT JOIN
|
|
(SELECT
|
|
user_id,
|
|
COUNT(*) AS count_pay
|
|
FROM
|
|
users_table_local
|
|
WHERE
|
|
user_id >= 1 AND
|
|
user_id <= 3 AND
|
|
users_table_local.value_1 > 3 AND users_table_local.value_1 < 5
|
|
GROUP BY
|
|
user_id
|
|
HAVING
|
|
COUNT(*) > 1
|
|
LIMIT 10
|
|
) AS subquery_2
|
|
ON
|
|
subquery_1.user_id = subquery_2.user_id
|
|
GROUP BY
|
|
subquery_1.user_id,
|
|
count_pay) AS subquery_top
|
|
GROUP BY
|
|
count_pay, user_id
|
|
)
|
|
GROUP BY user_id
|
|
HAVING count(*) > 1 AND sum(value_2) > 29
|
|
ORDER BY 1;
|
|
DEBUG: generating subplan XXX_1 for subquery SELECT user_id, count(*) AS count_pay FROM subquery_local_tables.users_table_local WHERE ((user_id OPERATOR(pg_catalog.>=) 1) AND (user_id OPERATOR(pg_catalog.<=) 3) AND (value_1 OPERATOR(pg_catalog.>) 3) AND (value_1 OPERATOR(pg_catalog.<) 5)) GROUP BY user_id HAVING (count(*) OPERATOR(pg_catalog.>) 1) LIMIT 10
|
|
DEBUG: Plan XXX query after replacing subqueries and CTEs: SELECT user_id FROM public.users_table WHERE (user_id OPERATOR(pg_catalog.=) ANY (SELECT subquery_top.user_id FROM (SELECT subquery_1.user_id, subquery_2.count_pay FROM ((SELECT users_table_1.user_id, 'action=>1'::text AS event, events_table."time" FROM public.users_table users_table_1, public.events_table WHERE ((users_table_1.user_id OPERATOR(pg_catalog.=) events_table.user_id) AND (users_table_1.user_id OPERATOR(pg_catalog.>=) 1) AND (users_table_1.user_id OPERATOR(pg_catalog.<=) 3) AND (events_table.event_type OPERATOR(pg_catalog.>) 1) AND (events_table.event_type OPERATOR(pg_catalog.<) 3)) UNION SELECT users_table_1.user_id, 'action=>2'::text AS event, events_table."time" FROM public.users_table users_table_1, public.events_table WHERE ((users_table_1.user_id OPERATOR(pg_catalog.=) events_table.user_id) AND (users_table_1.user_id OPERATOR(pg_catalog.>=) 1) AND (users_table_1.user_id OPERATOR(pg_catalog.<=) 3) AND (events_table.event_type OPERATOR(pg_catalog.>) 2) AND (events_table.event_type OPERATOR(pg_catalog.<) 4))) subquery_1 LEFT JOIN (SELECT intermediate_result.user_id, intermediate_result.count_pay FROM read_intermediate_result('XXX_1'::text, 'binary'::citus_copy_format) intermediate_result(user_id integer, count_pay bigint)) subquery_2 ON ((subquery_1.user_id OPERATOR(pg_catalog.=) subquery_2.user_id))) GROUP BY subquery_1.user_id, subquery_2.count_pay) subquery_top GROUP BY subquery_top.count_pay, subquery_top.user_id)) GROUP BY user_id HAVING ((count(*) OPERATOR(pg_catalog.>) 1) AND (sum(value_2) OPERATOR(pg_catalog.>) 29)) ORDER BY user_id
|
|
user_id
|
|
---------------------------------------------------------------------
|
|
2
|
|
3
|
|
(2 rows)
|
|
|
|
SET client_min_messages TO DEFAULT;
|
|
-- Test https://github.com/citusdata/citus/issues/2717
|
|
create table test_dist (id int, table_name text, column_name text);
|
|
select create_distributed_table('test_dist','id');
|
|
create_distributed_table
|
|
---------------------------------------------------------------------
|
|
|
|
(1 row)
|
|
|
|
insert into test_dist values (1, 'test_dist', 'table_name');
|
|
with q as (
|
|
select icl.* from test_dist td
|
|
join information_schema.columns icl on icl.table_name::text = lower(td.table_name) and icl.column_name::text = lower(td.column_name)
|
|
) select column_name from q;
|
|
column_name
|
|
---------------------------------------------------------------------
|
|
table_name
|
|
(1 row)
|
|
|
|
\set VERBOSITY terse
|
|
DROP SCHEMA subquery_local_tables CASCADE;
|
|
NOTICE: drop cascades to 3 other objects
|
|
SET search_path TO public;
|