citus/src/test/regress/expected/query_single_shard_table.out

3343 lines
173 KiB
Plaintext

CREATE SCHEMA query_single_shard_table;
SET search_path TO query_single_shard_table;
SET citus.next_shard_id TO 1620000;
SET citus.shard_count TO 32;
SET citus.shard_replication_factor TO 1;
SET client_min_messages TO NOTICE;
CREATE TABLE nullkey_c1_t1(a int, b int);
CREATE TABLE nullkey_c1_t2(a int, b int);
SELECT create_distributed_table('nullkey_c1_t1', null, colocate_with=>'none');
create_distributed_table
---------------------------------------------------------------------
(1 row)
SELECT create_distributed_table('nullkey_c1_t2', null, colocate_with=>'nullkey_c1_t1');
create_distributed_table
---------------------------------------------------------------------
(1 row)
INSERT INTO nullkey_c1_t1 SELECT i, i FROM generate_series(1, 8) i;
INSERT INTO nullkey_c1_t2 SELECT i, i FROM generate_series(2, 7) i;
CREATE TABLE nullkey_c2_t1(a int, b int);
CREATE TABLE nullkey_c2_t2(a int, b int);
SELECT create_distributed_table('nullkey_c2_t1', null, colocate_with=>'none');
create_distributed_table
---------------------------------------------------------------------
(1 row)
SELECT create_distributed_table('nullkey_c2_t2', null, colocate_with=>'nullkey_c2_t1', distribution_type=>null);
create_distributed_table
---------------------------------------------------------------------
(1 row)
INSERT INTO nullkey_c2_t1 SELECT i, i FROM generate_series(2, 7) i;
INSERT INTO nullkey_c2_t2 SELECT i, i FROM generate_series(1, 8) i;
CREATE TABLE nullkey_c3_t1(a int, b int);
SELECT create_distributed_table('nullkey_c3_t1', null, colocate_with=>'none');
create_distributed_table
---------------------------------------------------------------------
(1 row)
INSERT INTO nullkey_c3_t1 SELECT i, i FROM generate_series(1, 8) i;
RESET citus.shard_replication_factor;
CREATE TABLE reference_table(a int, b int);
SELECT create_reference_table('reference_table');
create_reference_table
---------------------------------------------------------------------
(1 row)
INSERT INTO reference_table SELECT i, i FROM generate_series(0, 5) i;
CREATE TABLE distributed_table(a int, b int);
SELECT create_distributed_table('distributed_table', 'a');
create_distributed_table
---------------------------------------------------------------------
(1 row)
INSERT INTO distributed_table SELECT i, i FROM generate_series(3, 8) i;
CREATE TABLE citus_local_table(a int, b int);
SELECT citus_add_local_table_to_metadata('citus_local_table');
citus_add_local_table_to_metadata
---------------------------------------------------------------------
(1 row)
INSERT INTO citus_local_table SELECT i, i FROM generate_series(0, 10) i;
CREATE TABLE postgres_local_table(a int, b int);
INSERT INTO postgres_local_table SELECT i, i FROM generate_series(5, 10) i;
CREATE TABLE articles_hash (
id bigint NOT NULL,
author_id bigint NOT NULL,
title varchar(20) NOT NULL,
word_count integer
);
INSERT INTO articles_hash VALUES ( 4, 4, 'altdorfer', 14551),( 5, 5, 'aruru', 11389),
(13, 3, 'aseyev', 2255),(15, 5, 'adversa', 3164),
(18, 8, 'assembly', 911),(19, 9, 'aubergiste', 4981),
(28, 8, 'aerophyte', 5454),(29, 9, 'amateur', 9524),
(42, 2, 'ausable', 15885),(43, 3, 'affixal', 12723),
(49, 9, 'anyone', 2681),(50, 10, 'anjanette', 19519);
SET citus.shard_replication_factor TO 1;
SELECT create_distributed_table('articles_hash', null, colocate_with=>'none');
NOTICE: Copying data from local table...
NOTICE: copying the data has completed
DETAIL: The local data in the table is no longer visible, but is still on disk.
HINT: To remove the local data, run: SELECT truncate_local_data_after_distributing_table($$query_single_shard_table.articles_hash$$)
create_distributed_table
---------------------------------------------------------------------
(1 row)
CREATE TABLE raw_events_first (user_id int, time timestamp, value_1 int, value_2 int, value_3 float, value_4 bigint, UNIQUE(user_id, value_1));
SELECT create_distributed_table('raw_events_first', null, colocate_with=>'none', distribution_type=>null);
create_distributed_table
---------------------------------------------------------------------
(1 row)
CREATE TABLE raw_events_second (user_id int, time timestamp, value_1 int, value_2 int, value_3 float, value_4 bigint, UNIQUE(user_id, value_1));
SELECT create_distributed_table('raw_events_second', null, colocate_with=>'raw_events_first', distribution_type=>null);
create_distributed_table
---------------------------------------------------------------------
(1 row)
CREATE TABLE agg_events (user_id int, value_1_agg int, value_2_agg int, value_3_agg float, value_4_agg bigint, agg_time timestamp, UNIQUE(user_id, value_1_agg));
SELECT create_distributed_table('agg_events', null, colocate_with=>'raw_events_first', distribution_type=>null);
create_distributed_table
---------------------------------------------------------------------
(1 row)
CREATE TABLE users_ref_table (user_id int);
SELECT create_reference_table('users_ref_table');
create_reference_table
---------------------------------------------------------------------
(1 row)
INSERT INTO raw_events_first VALUES (1, '1970-01-01', 10, 100, 1000.1, 10000), (3, '1971-01-01', 30, 300, 3000.1, 30000),
(5, '1972-01-01', 50, 500, 5000.1, 50000), (2, '1973-01-01', 20, 200, 2000.1, 20000),
(4, '1974-01-01', 40, 400, 4000.1, 40000), (6, '1975-01-01', 60, 600, 6000.1, 60000);
CREATE TABLE modify_fast_path(key int, value_1 int, value_2 text);
SELECT create_distributed_table('modify_fast_path', null);
create_distributed_table
---------------------------------------------------------------------
(1 row)
CREATE TABLE modify_fast_path_reference(key int, value_1 int, value_2 text);
SELECT create_reference_table('modify_fast_path_reference');
create_reference_table
---------------------------------------------------------------------
(1 row)
CREATE TABLE bigserial_test (x int, y int, z bigserial);
SELECT create_distributed_table('bigserial_test', null);
create_distributed_table
---------------------------------------------------------------------
(1 row)
RESET citus.shard_replication_factor;
CREATE TABLE append_table (text_col text, a int);
SELECT create_distributed_table('append_table', 'a', 'append');
create_distributed_table
---------------------------------------------------------------------
(1 row)
SELECT master_create_empty_shard('append_table') AS shardid1 \gset
SELECT master_create_empty_shard('append_table') AS shardid2 \gset
SELECT master_create_empty_shard('append_table') AS shardid3 \gset
COPY append_table (text_col, a) FROM STDIN WITH (format 'csv', append_to_shard :shardid1);
COPY append_table (text_col, a) FROM STDIN WITH (format 'csv', append_to_shard :shardid2);
CREATE TABLE range_table(a int, b int);
SELECT create_distributed_table('range_table', 'a', 'range');
create_distributed_table
---------------------------------------------------------------------
(1 row)
CALL public.create_range_partitioned_shards('range_table', '{"0","25"}','{"24","49"}');
INSERT INTO range_table VALUES (0, 1), (1, 2), (2, 3), (3, 4), (4, 5), (5, 6), (6, 50);
\set users_table_data_file :abs_srcdir '/data/users_table.data'
\set events_table_data_file :abs_srcdir '/data/events_table.data'
SET citus.shard_replication_factor TO 1;
CREATE TABLE users_table (user_id int, time timestamp, value_1 int, value_2 int, value_3 float, value_4 bigint);
SELECT create_distributed_table('users_table', null, colocate_with=>'none');
create_distributed_table
---------------------------------------------------------------------
(1 row)
\set client_side_copy_command '\\copy users_table FROM ' :'users_table_data_file' ' WITH CSV;'
:client_side_copy_command
CREATE TABLE non_colocated_users_table (id int, value int);
SELECT create_distributed_table('non_colocated_users_table', null, colocate_with => 'none');
create_distributed_table
---------------------------------------------------------------------
(1 row)
INSERT INTO non_colocated_users_table (id, value) VALUES(1, 2),(2, 3),(3,4);
CREATE TABLE colocated_events_table (user_id int, time timestamp, event_type int, value_2 int, value_3 float, value_4 bigint);
SELECT create_distributed_table('colocated_events_table', null, colocate_with=>'users_table');
create_distributed_table
---------------------------------------------------------------------
(1 row)
\set client_side_copy_command '\\copy colocated_events_table FROM ' :'events_table_data_file' ' WITH CSV;'
:client_side_copy_command
CREATE TABLE non_colocated_events_table (user_id int, time timestamp, event_type int, value_2 int, value_3 float, value_4 bigint);
SELECT create_distributed_table('non_colocated_events_table', null, colocate_with=>'non_colocated_users_table');
create_distributed_table
---------------------------------------------------------------------
(1 row)
\set client_side_copy_command '\\copy non_colocated_events_table FROM ' :'events_table_data_file' ' WITH CSV;'
:client_side_copy_command
CREATE TABLE users_table_local AS SELECT * FROM users_table;
CREATE TABLE colocated_users_table (id int, value int);
SELECT create_distributed_table('colocated_users_table', null, colocate_with => 'users_table');
create_distributed_table
---------------------------------------------------------------------
(1 row)
INSERT INTO colocated_users_table (id, value) VALUES(1, 2),(2, 3),(3,4);
CREATE TABLE users_reference_table (like users_table including all);
SELECT create_reference_table('users_reference_table');
create_reference_table
---------------------------------------------------------------------
(1 row)
CREATE TABLE events_reference_table (like colocated_events_table including all);
SELECT create_reference_table('events_reference_table');
create_reference_table
---------------------------------------------------------------------
(1 row)
CREATE FUNCTION func() RETURNS TABLE (id int, value int) AS $$
SELECT 1, 2
$$ LANGUAGE SQL;
SET client_min_messages to DEBUG2;
-- simple insert
INSERT INTO nullkey_c1_t1 VALUES (1,2), (2,2), (3,4);
DEBUG: Creating router plan
INSERT INTO nullkey_c1_t2 VALUES (1,3), (3,4), (5,1), (6,2);
DEBUG: Creating router plan
INSERT INTO nullkey_c2_t1 VALUES (1,0), (2,5), (4,3), (5,2);
DEBUG: Creating router plan
INSERT INTO nullkey_c2_t2 VALUES (2,4), (3,2), (5,2), (7,4);
DEBUG: Creating router plan
-- simple select
SELECT * FROM nullkey_c1_t1 ORDER BY 1,2;
DEBUG: Distributed planning for a fast-path router query
DEBUG: Creating router plan
a | b
---------------------------------------------------------------------
1 | 1
1 | 2
2 | 2
2 | 2
3 | 3
3 | 4
4 | 4
5 | 5
6 | 6
7 | 7
8 | 8
(11 rows)
-- for update / share
SELECT * FROM modify_fast_path WHERE key = 1 FOR UPDATE;
DEBUG: Distributed planning for a fast-path router query
DEBUG: Creating router plan
key | value_1 | value_2
---------------------------------------------------------------------
(0 rows)
SELECT * FROM modify_fast_path WHERE key = 1 FOR SHARE;
DEBUG: Distributed planning for a fast-path router query
DEBUG: Creating router plan
key | value_1 | value_2
---------------------------------------------------------------------
(0 rows)
SELECT * FROM modify_fast_path FOR UPDATE;
DEBUG: Distributed planning for a fast-path router query
DEBUG: Creating router plan
key | value_1 | value_2
---------------------------------------------------------------------
(0 rows)
SELECT * FROM modify_fast_path FOR SHARE;
DEBUG: Distributed planning for a fast-path router query
DEBUG: Creating router plan
key | value_1 | value_2
---------------------------------------------------------------------
(0 rows)
-- cartesian product with different table types
-- with other table types
SELECT COUNT(*) FROM distributed_table d1, nullkey_c1_t1;
DEBUG: Router planner cannot handle multi-shard select queries
ERROR: cannot perform distributed planning on this query
DETAIL: Cartesian products are currently unsupported
SELECT COUNT(*) FROM reference_table d1, nullkey_c1_t1;
DEBUG: Creating router plan
count
---------------------------------------------------------------------
66
(1 row)
SELECT COUNT(*) FROM citus_local_table d1, nullkey_c1_t1;
DEBUG: Local tables cannot be used in distributed queries.
DEBUG: Wrapping relation "citus_local_table" "d1" to a subquery
DEBUG: Distributed planning for a fast-path router query
DEBUG: Creating router plan
DEBUG: generating subplan XXX_1 for subquery SELECT NULL::integer AS "dummy-1" FROM query_single_shard_table.citus_local_table d1 WHERE true
DEBUG: Plan XXX query after replacing subqueries and CTEs: SELECT count(*) AS count FROM (SELECT NULL::integer AS a, NULL::integer AS b FROM (SELECT intermediate_result."dummy-1" FROM read_intermediate_result('XXX_1'::text, 'binary'::citus_copy_format) intermediate_result("dummy-1" integer)) d1_1) d1, query_single_shard_table.nullkey_c1_t1
DEBUG: Creating router plan
count
---------------------------------------------------------------------
121
(1 row)
SELECT COUNT(*) FROM postgres_local_table d1, nullkey_c1_t1;
DEBUG: Local tables cannot be used in distributed queries.
DEBUG: Wrapping relation "postgres_local_table" "d1" to a subquery
DEBUG: generating subplan XXX_1 for subquery SELECT NULL::integer AS "dummy-1" FROM query_single_shard_table.postgres_local_table d1 WHERE true
DEBUG: Plan XXX query after replacing subqueries and CTEs: SELECT count(*) AS count FROM (SELECT NULL::integer AS a, NULL::integer AS b FROM (SELECT intermediate_result."dummy-1" FROM read_intermediate_result('XXX_1'::text, 'binary'::citus_copy_format) intermediate_result("dummy-1" integer)) d1_1) d1, query_single_shard_table.nullkey_c1_t1
DEBUG: Creating router plan
count
---------------------------------------------------------------------
66
(1 row)
-- with a colocated single-shard table
SELECT COUNT(*) FROM nullkey_c1_t1 d1, nullkey_c1_t2;
DEBUG: Creating router plan
count
---------------------------------------------------------------------
110
(1 row)
-- with a non-colocated single-shard table
SELECT COUNT(*) FROM nullkey_c1_t1 d1, nullkey_c2_t1;
DEBUG: router planner does not support queries that reference non-colocated distributed tables
ERROR: cannot perform distributed planning on this query
DETAIL: Cartesian products are currently unsupported
-- First, show that nullkey_c1_t1 and nullkey_c3_t1 are not colocated.
SELECT
(SELECT colocationid FROM pg_dist_partition WHERE logicalrelid = 'query_single_shard_table.nullkey_c1_t1'::regclass) !=
(SELECT colocationid FROM pg_dist_partition WHERE logicalrelid = 'query_single_shard_table.nullkey_c3_t1'::regclass);
?column?
---------------------------------------------------------------------
t
(1 row)
-- Now verify that we can join them via router planner because it doesn't care
-- about whether two tables are colocated or not but physical location of shards
-- when citus.enable_non_colocated_router_query_pushdown is set to on.
SET citus.enable_non_colocated_router_query_pushdown TO ON;
SELECT COUNT(*) FROM nullkey_c1_t1 JOIN nullkey_c3_t1 USING(a);
DEBUG: Creating router plan
count
---------------------------------------------------------------------
11
(1 row)
SET citus.enable_non_colocated_router_query_pushdown TO OFF;
SET citus.enable_repartition_joins TO ON;
SET client_min_messages TO DEBUG1;
SELECT COUNT(*) FROM nullkey_c1_t1 JOIN nullkey_c3_t1 USING(a);
count
---------------------------------------------------------------------
11
(1 row)
SET client_min_messages TO DEBUG2;
SET citus.enable_repartition_joins TO OFF;
RESET citus.enable_non_colocated_router_query_pushdown;
-- colocated join between single-shard tables
SELECT COUNT(*) FROM nullkey_c1_t1 JOIN nullkey_c1_t2 USING(a);
DEBUG: Creating router plan
count
---------------------------------------------------------------------
14
(1 row)
SELECT COUNT(*) FROM nullkey_c1_t1 LEFT JOIN nullkey_c1_t2 USING(a);
DEBUG: Creating router plan
count
---------------------------------------------------------------------
15
(1 row)
SELECT COUNT(*) FROM nullkey_c1_t1 FULL JOIN nullkey_c1_t2 USING(a);
DEBUG: Creating router plan
count
---------------------------------------------------------------------
15
(1 row)
SELECT COUNT(*) FROM nullkey_c1_t1 t1
LEFT JOIN LATERAL (
SELECT * FROM nullkey_c1_t2 t2 WHERE t2.b > t1.a
) q USING(a);
DEBUG: Creating router plan
count
---------------------------------------------------------------------
11
(1 row)
SELECT COUNT(*) FROM nullkey_c1_t1 t1
JOIN LATERAL (
SELECT * FROM nullkey_c1_t2 t2 WHERE t2.b > t1.a
) q USING(a);
DEBUG: Creating router plan
count
---------------------------------------------------------------------
4
(1 row)
SELECT COUNT(*) FROM nullkey_c1_t1 t1
WHERE EXISTS (
SELECT * FROM nullkey_c1_t2 t2 WHERE t2.b > t1.a
);
DEBUG: Creating router plan
count
---------------------------------------------------------------------
9
(1 row)
SELECT COUNT(*) FROM nullkey_c1_t1 t1
WHERE t1.b IN (
SELECT b+1 FROM nullkey_c1_t2 t2 WHERE t2.b = t1.a
);
DEBUG: Creating router plan
count
---------------------------------------------------------------------
2
(1 row)
SELECT COUNT(*) FROM nullkey_c1_t1 t1
WHERE t1.b NOT IN (
SELECT a FROM nullkey_c1_t2 t2 WHERE t2.b > t1.a
);
DEBUG: Creating router plan
count
---------------------------------------------------------------------
7
(1 row)
-- non-colocated inner joins between single-shard tables
SET client_min_messages to DEBUG1;
SET citus.enable_repartition_joins TO ON;
SELECT * FROM nullkey_c1_t1 JOIN nullkey_c2_t1 USING(a) ORDER BY 1,2,3;
a | b | b
---------------------------------------------------------------------
1 | 1 | 0
1 | 2 | 0
2 | 2 | 2
2 | 2 | 2
2 | 2 | 5
2 | 2 | 5
3 | 3 | 3
3 | 4 | 3
4 | 4 | 3
4 | 4 | 4
5 | 5 | 2
5 | 5 | 5
6 | 6 | 6
7 | 7 | 7
(14 rows)
SELECT * FROM (SELECT * FROM nullkey_c1_t1) nullkey_c1_t1 JOIN nullkey_c2_t1 USING(a) ORDER BY 1,2,3;
a | b | b
---------------------------------------------------------------------
1 | 1 | 0
1 | 2 | 0
2 | 2 | 2
2 | 2 | 2
2 | 2 | 5
2 | 2 | 5
3 | 3 | 3
3 | 4 | 3
4 | 4 | 3
4 | 4 | 4
5 | 5 | 2
5 | 5 | 5
6 | 6 | 6
7 | 7 | 7
(14 rows)
SELECT * FROM nullkey_c2_t1 JOIN (SELECT * FROM nullkey_c1_t1) nullkey_c1_t1 USING(a) ORDER BY 1,2,3;
a | b | b
---------------------------------------------------------------------
1 | 0 | 1
1 | 0 | 2
2 | 2 | 2
2 | 2 | 2
2 | 5 | 2
2 | 5 | 2
3 | 3 | 3
3 | 3 | 4
4 | 3 | 4
4 | 4 | 4
5 | 2 | 5
5 | 5 | 5
6 | 6 | 6
7 | 7 | 7
(14 rows)
SELECT COUNT(*) FROM nullkey_c1_t1 t1
JOIN LATERAL (
SELECT * FROM nullkey_c2_t2 t2 WHERE t2.b > t1.a
) q USING(a);
count
---------------------------------------------------------------------
2
(1 row)
SET citus.enable_repartition_joins TO OFF;
SET client_min_messages to DEBUG2;
-- non-colocated outer joins between single-shard tables
SELECT * FROM nullkey_c1_t1 LEFT JOIN nullkey_c2_t2 USING(a) ORDER BY 1,2,3 LIMIT 4;
DEBUG: router planner does not support queries that reference non-colocated distributed tables
DEBUG: push down of limit count: 4
ERROR: cannot push down this subquery
DETAIL: nullkey_c1_t1 and nullkey_c2_t2 are not colocated
SELECT * FROM nullkey_c1_t1 FULL JOIN nullkey_c2_t2 USING(a) ORDER BY 1,2,3 LIMIT 4;
DEBUG: router planner does not support queries that reference non-colocated distributed tables
DEBUG: push down of limit count: 4
ERROR: cannot push down this subquery
DETAIL: nullkey_c1_t1 and nullkey_c2_t2 are not colocated
SELECT * FROM nullkey_c1_t1 t1
LEFT JOIN LATERAL (
SELECT * FROM nullkey_c2_t2 t2 WHERE t2.b > t1.a
) q USING(a) ORDER BY 1,2,3 OFFSET 3 LIMIT 4;
DEBUG: router planner does not support queries that reference non-colocated distributed tables
DEBUG: push down of limit count: 7
ERROR: cannot push down this subquery
DETAIL: nullkey_c1_t1 and nullkey_c2_t2 are not colocated
SELECT COUNT(*) FROM nullkey_c1_t1 t1
LEFT JOIN LATERAL (
SELECT * FROM nullkey_c2_t2 t2 WHERE t2.b > t1.a
) q USING(a);
DEBUG: router planner does not support queries that reference non-colocated distributed tables
ERROR: cannot push down this subquery
DETAIL: nullkey_c1_t1 and nullkey_c2_t2 are not colocated
SELECT COUNT(*) FROM nullkey_c1_t1 t1
WHERE EXISTS (
SELECT * FROM nullkey_c2_t2 t2 WHERE t2.b > t1.a
);
DEBUG: router planner does not support queries that reference non-colocated distributed tables
ERROR: cannot push down this subquery
DETAIL: nullkey_c2_t2 and nullkey_c1_t1 are not colocated
SELECT COUNT(*) FROM nullkey_c1_t1 t1
WHERE t1.b IN (
SELECT b+1 FROM nullkey_c2_t2 t2 WHERE t2.b = t1.a
);
DEBUG: router planner does not support queries that reference non-colocated distributed tables
ERROR: cannot push down this subquery
DETAIL: nullkey_c2_t2 and nullkey_c1_t1 are not colocated
SELECT COUNT(*) FROM nullkey_c1_t1 t1
WHERE t1.b NOT IN (
SELECT a FROM nullkey_c2_t2 t2 WHERE t2.b > t1.a
);
DEBUG: router planner does not support queries that reference non-colocated distributed tables
ERROR: cannot push down this subquery
DETAIL: nullkey_c2_t2 and nullkey_c1_t1 are not colocated
-- join with a reference table
SELECT COUNT(*) FROM nullkey_c1_t1, reference_table WHERE nullkey_c1_t1.a = reference_table.a;
DEBUG: Creating router plan
count
---------------------------------------------------------------------
8
(1 row)
WITH cte_1 AS
(SELECT * FROM nullkey_c1_t1, reference_table WHERE nullkey_c1_t1.a = reference_table.a ORDER BY 1,2,3,4 FOR UPDATE)
SELECT COUNT(*) FROM cte_1;
DEBUG: Creating router plan
count
---------------------------------------------------------------------
8
(1 row)
-- join with postgres / citus local tables
SELECT * FROM nullkey_c1_t1 JOIN postgres_local_table USING(a) ORDER BY 1,2,3;
DEBUG: Local tables cannot be used in distributed queries.
DEBUG: Wrapping relation "postgres_local_table" to a subquery
DEBUG: generating subplan XXX_1 for subquery SELECT a, b FROM query_single_shard_table.postgres_local_table WHERE true
DEBUG: Plan XXX query after replacing subqueries and CTEs: SELECT nullkey_c1_t1.a, nullkey_c1_t1.b, postgres_local_table.b FROM (query_single_shard_table.nullkey_c1_t1 JOIN (SELECT postgres_local_table_1.a, postgres_local_table_1.b FROM (SELECT intermediate_result.a, intermediate_result.b FROM read_intermediate_result('XXX_1'::text, 'binary'::citus_copy_format) intermediate_result(a integer, b integer)) postgres_local_table_1) postgres_local_table USING (a)) ORDER BY nullkey_c1_t1.a, nullkey_c1_t1.b, postgres_local_table.b
DEBUG: Creating router plan
a | b | b
---------------------------------------------------------------------
5 | 5 | 5
6 | 6 | 6
7 | 7 | 7
8 | 8 | 8
(4 rows)
SELECT * FROM nullkey_c1_t1 JOIN citus_local_table USING(a) ORDER BY 1,2,3;
DEBUG: Local tables cannot be used in distributed queries.
DEBUG: Wrapping relation "citus_local_table" to a subquery
DEBUG: Distributed planning for a fast-path router query
DEBUG: Creating router plan
DEBUG: generating subplan XXX_1 for subquery SELECT a, b FROM query_single_shard_table.citus_local_table WHERE true
DEBUG: Plan XXX query after replacing subqueries and CTEs: SELECT nullkey_c1_t1.a, nullkey_c1_t1.b, citus_local_table.b FROM (query_single_shard_table.nullkey_c1_t1 JOIN (SELECT citus_local_table_1.a, citus_local_table_1.b FROM (SELECT intermediate_result.a, intermediate_result.b FROM read_intermediate_result('XXX_1'::text, 'binary'::citus_copy_format) intermediate_result(a integer, b integer)) citus_local_table_1) citus_local_table USING (a)) ORDER BY nullkey_c1_t1.a, nullkey_c1_t1.b, citus_local_table.b
DEBUG: Creating router plan
a | b | b
---------------------------------------------------------------------
1 | 1 | 1
1 | 2 | 1
2 | 2 | 2
2 | 2 | 2
3 | 3 | 3
3 | 4 | 3
4 | 4 | 4
5 | 5 | 5
6 | 6 | 6
7 | 7 | 7
8 | 8 | 8
(11 rows)
SET citus.local_table_join_policy TO 'prefer-distributed';
SELECT * FROM nullkey_c1_t1 JOIN citus_local_table USING(a) ORDER BY 1,2,3;
DEBUG: Local tables cannot be used in distributed queries.
DEBUG: Wrapping relation "nullkey_c1_t1" to a subquery
DEBUG: Distributed planning for a fast-path router query
DEBUG: Creating router plan
DEBUG: generating subplan XXX_1 for subquery SELECT a, b FROM query_single_shard_table.nullkey_c1_t1 WHERE true
DEBUG: Plan XXX query after replacing subqueries and CTEs: SELECT nullkey_c1_t1.a, nullkey_c1_t1.b, citus_local_table.b FROM ((SELECT nullkey_c1_t1_1.a, nullkey_c1_t1_1.b FROM (SELECT intermediate_result.a, intermediate_result.b FROM read_intermediate_result('XXX_1'::text, 'binary'::citus_copy_format) intermediate_result(a integer, b integer)) nullkey_c1_t1_1) nullkey_c1_t1 JOIN query_single_shard_table.citus_local_table USING (a)) ORDER BY nullkey_c1_t1.a, nullkey_c1_t1.b, citus_local_table.b
DEBUG: Creating router plan
a | b | b
---------------------------------------------------------------------
1 | 1 | 1
1 | 2 | 1
2 | 2 | 2
2 | 2 | 2
3 | 3 | 3
3 | 4 | 3
4 | 4 | 4
5 | 5 | 5
6 | 6 | 6
7 | 7 | 7
8 | 8 | 8
(11 rows)
RESET citus.local_table_join_policy;
-- join with a distributed table
SET citus.enable_repartition_joins TO ON;
SET client_min_messages TO DEBUG1;
SELECT * FROM distributed_table d1 JOIN nullkey_c1_t1 USING(a) ORDER BY 1,2,3;
a | b | b
---------------------------------------------------------------------
3 | 3 | 3
3 | 3 | 4
4 | 4 | 4
5 | 5 | 5
6 | 6 | 6
7 | 7 | 7
8 | 8 | 8
(7 rows)
SELECT * FROM (SELECT * FROM distributed_table) d1 JOIN nullkey_c1_t1 USING(a) ORDER BY 1,2,3;
DEBUG: generating subplan XXX_1 for subquery SELECT a, b FROM query_single_shard_table.distributed_table
DEBUG: Plan XXX query after replacing subqueries and CTEs: SELECT d1.a, d1.b, nullkey_c1_t1.b FROM ((SELECT intermediate_result.a, intermediate_result.b FROM read_intermediate_result('XXX_1'::text, 'binary'::citus_copy_format) intermediate_result(a integer, b integer)) d1 JOIN query_single_shard_table.nullkey_c1_t1 USING (a)) ORDER BY d1.a, d1.b, nullkey_c1_t1.b
a | b | b
---------------------------------------------------------------------
3 | 3 | 3
3 | 3 | 4
4 | 4 | 4
5 | 5 | 5
6 | 6 | 6
7 | 7 | 7
8 | 8 | 8
(7 rows)
SELECT * FROM nullkey_c1_t1 JOIN (SELECT * FROM distributed_table) d1 USING(a) ORDER BY 1,2,3;
DEBUG: generating subplan XXX_1 for subquery SELECT a, b FROM query_single_shard_table.distributed_table
DEBUG: Plan XXX query after replacing subqueries and CTEs: SELECT nullkey_c1_t1.a, nullkey_c1_t1.b, d1.b FROM (query_single_shard_table.nullkey_c1_t1 JOIN (SELECT intermediate_result.a, intermediate_result.b FROM read_intermediate_result('XXX_1'::text, 'binary'::citus_copy_format) intermediate_result(a integer, b integer)) d1 USING (a)) ORDER BY nullkey_c1_t1.a, nullkey_c1_t1.b, d1.b
a | b | b
---------------------------------------------------------------------
3 | 3 | 3
3 | 4 | 3
4 | 4 | 4
5 | 5 | 5
6 | 6 | 6
7 | 7 | 7
8 | 8 | 8
(7 rows)
SELECT * FROM distributed_table d1 JOIN (SELECT * FROM nullkey_c1_t1) nullkey_c1_t1 USING(a) ORDER BY 1,2,3;
DEBUG: generating subplan XXX_1 for subquery SELECT a, b FROM query_single_shard_table.nullkey_c1_t1
DEBUG: Plan XXX query after replacing subqueries and CTEs: SELECT d1.a, d1.b, nullkey_c1_t1.b FROM (query_single_shard_table.distributed_table d1 JOIN (SELECT intermediate_result.a, intermediate_result.b FROM read_intermediate_result('XXX_1'::text, 'binary'::citus_copy_format) intermediate_result(a integer, b integer)) nullkey_c1_t1 USING (a)) ORDER BY d1.a, d1.b, nullkey_c1_t1.b
a | b | b
---------------------------------------------------------------------
3 | 3 | 3
3 | 3 | 4
4 | 4 | 4
5 | 5 | 5
6 | 6 | 6
7 | 7 | 7
8 | 8 | 8
(7 rows)
SELECT * FROM (SELECT * FROM nullkey_c1_t1) nullkey_c1_t1 JOIN distributed_table d1 USING(a) ORDER BY 1,2,3;
DEBUG: generating subplan XXX_1 for subquery SELECT a, b FROM query_single_shard_table.nullkey_c1_t1
DEBUG: Plan XXX query after replacing subqueries and CTEs: SELECT nullkey_c1_t1.a, nullkey_c1_t1.b, d1.b FROM ((SELECT intermediate_result.a, intermediate_result.b FROM read_intermediate_result('XXX_1'::text, 'binary'::citus_copy_format) intermediate_result(a integer, b integer)) nullkey_c1_t1 JOIN query_single_shard_table.distributed_table d1 USING (a)) ORDER BY nullkey_c1_t1.a, nullkey_c1_t1.b, d1.b
a | b | b
---------------------------------------------------------------------
3 | 3 | 3
3 | 4 | 3
4 | 4 | 4
5 | 5 | 5
6 | 6 | 6
7 | 7 | 7
8 | 8 | 8
(7 rows)
-- test joins with non-colocated distributed tables, by using subqueries
SELECT * FROM nullkey_c1_t1 t1 JOIN (SELECT * FROM distributed_table) t2 USING (a) JOIN (SELECT * FROM nullkey_c1_t2) t3 USING (a) ORDER BY 1,2,3,4 LIMIT 1;
DEBUG: generating subplan XXX_1 for subquery SELECT a, b FROM query_single_shard_table.distributed_table
DEBUG: Plan XXX query after replacing subqueries and CTEs: SELECT t1.a, t1.b, t2.b, t3.b FROM ((query_single_shard_table.nullkey_c1_t1 t1 JOIN (SELECT intermediate_result.a, intermediate_result.b FROM read_intermediate_result('XXX_1'::text, 'binary'::citus_copy_format) intermediate_result(a integer, b integer)) t2 USING (a)) JOIN (SELECT nullkey_c1_t2.a, nullkey_c1_t2.b FROM query_single_shard_table.nullkey_c1_t2) t3 USING (a)) ORDER BY t1.a, t1.b, t2.b, t3.b LIMIT 1
a | b | b | b
---------------------------------------------------------------------
3 | 3 | 3 | 3
(1 row)
SELECT * FROM (SELECT * FROM nullkey_c1_t1) t1 JOIN nullkey_c2_t1 t2 USING (a) JOIN (SELECT * FROM nullkey_c1_t2) t3 USING (a) ORDER BY 1,2,3,4 LIMIT 1;
DEBUG: push down of limit count: 1
a | b | b | b
---------------------------------------------------------------------
1 | 1 | 0 | 3
(1 row)
SELECT * FROM distributed_table t1 JOIN (SELECT * FROM nullkey_c1_t1) t2 USING (a) JOIN (SELECT b as a FROM distributed_table) t3 USING (a) ORDER BY 1,2,3 LIMIT 1;
DEBUG: generating subplan XXX_1 for subquery SELECT a, b FROM query_single_shard_table.nullkey_c1_t1
DEBUG: generating subplan XXX_2 for subquery SELECT b AS a FROM query_single_shard_table.distributed_table
DEBUG: Plan XXX query after replacing subqueries and CTEs: SELECT t1.a, t1.b, t2.b FROM ((query_single_shard_table.distributed_table t1 JOIN (SELECT intermediate_result.a, intermediate_result.b FROM read_intermediate_result('XXX_1'::text, 'binary'::citus_copy_format) intermediate_result(a integer, b integer)) t2 USING (a)) JOIN (SELECT intermediate_result.a FROM read_intermediate_result('XXX_2'::text, 'binary'::citus_copy_format) intermediate_result(a integer)) t3 USING (a)) ORDER BY t1.a, t1.b, t2.b LIMIT 1
DEBUG: push down of limit count: 1
a | b | b
---------------------------------------------------------------------
3 | 3 | 3
(1 row)
SELECT * FROM (SELECT * FROM nullkey_c2_t1) t1 JOIN nullkey_c1_t1 t2 USING (a) JOIN (SELECT * FROM nullkey_c2_t1) t3 USING (a) ORDER BY 1,2,3,4 LIMIT 1;
DEBUG: push down of limit count: 1
a | b | b | b
---------------------------------------------------------------------
1 | 0 | 1 | 0
(1 row)
SELECT * FROM nullkey_c1_t1 t1 JOIN (SELECT * FROM distributed_table) t2 USING (a) JOIN (SELECT * FROM distributed_table) t3 USING (a) ORDER BY 1,2,3,4 LIMIT 1;
DEBUG: push down of limit count: 1
a | b | b | b
---------------------------------------------------------------------
3 | 3 | 3 | 3
(1 row)
SELECT * FROM (SELECT * FROM nullkey_c1_t1) t1 JOIN nullkey_c2_t1 t2 USING (a) JOIN (SELECT * FROM nullkey_c2_t1) t3 USING (a) ORDER BY 1,2,3,4 LIMIT 1;
DEBUG: push down of limit count: 1
a | b | b | b
---------------------------------------------------------------------
1 | 1 | 0 | 0
(1 row)
SELECT * FROM distributed_table t1 JOIN (SELECT * FROM nullkey_c1_t1) t2 USING (a) JOIN (SELECT * FROM nullkey_c1_t1) t3 USING (a) ORDER BY 1,2,3,4 LIMIT 1;
DEBUG: generating subplan XXX_1 for subquery SELECT a, b FROM query_single_shard_table.nullkey_c1_t1
DEBUG: generating subplan XXX_2 for subquery SELECT a, b FROM query_single_shard_table.nullkey_c1_t1
DEBUG: Plan XXX query after replacing subqueries and CTEs: SELECT t1.a, t1.b, t2.b, t3.b FROM ((query_single_shard_table.distributed_table t1 JOIN (SELECT intermediate_result.a, intermediate_result.b FROM read_intermediate_result('XXX_1'::text, 'binary'::citus_copy_format) intermediate_result(a integer, b integer)) t2 USING (a)) JOIN (SELECT intermediate_result.a, intermediate_result.b FROM read_intermediate_result('XXX_2'::text, 'binary'::citus_copy_format) intermediate_result(a integer, b integer)) t3 USING (a)) ORDER BY t1.a, t1.b, t2.b, t3.b LIMIT 1
DEBUG: push down of limit count: 1
a | b | b | b
---------------------------------------------------------------------
3 | 3 | 3 | 3
(1 row)
SELECT * FROM (SELECT * FROM nullkey_c2_t1) t1 JOIN nullkey_c1_t1 t2 USING (a) JOIN (SELECT * FROM nullkey_c1_t1) t3 USING (a) ORDER BY 1,2,3,4 LIMIT 1;
DEBUG: push down of limit count: 1
a | b | b | b
---------------------------------------------------------------------
1 | 0 | 1 | 1
(1 row)
SELECT * FROM nullkey_c1_t1 t1 JOIN (SELECT * FROM nullkey_c1_t1) t2 USING (a) JOIN distributed_table t3 USING (a) ORDER BY 1,2,3,4 LIMIT 1;
DEBUG: push down of limit count: 1
a | b | b | b
---------------------------------------------------------------------
3 | 3 | 3 | 3
(1 row)
SELECT * FROM nullkey_c1_t1 t1 JOIN nullkey_c1_t1 t2 USING (a) JOIN nullkey_c2_t1 t3 USING (a) ORDER BY 1,2,3,4 LIMIT 1;
DEBUG: push down of limit count: 1
a | b | b | b
---------------------------------------------------------------------
1 | 1 | 1 | 0
(1 row)
SELECT * FROM (SELECT * FROM distributed_table) t1 JOIN distributed_table t2 USING (a) JOIN (SELECT * FROM nullkey_c1_t1) t3 USING (a) ORDER BY 1,2,3,4 LIMIT 1;
DEBUG: push down of limit count: 1
a | b | b | b
---------------------------------------------------------------------
3 | 3 | 3 | 3
(1 row)
SELECT * FROM (SELECT * FROM nullkey_c2_t1) t1 JOIN nullkey_c2_t1 t2 USING (a) JOIN (SELECT * FROM nullkey_c1_t1) t3 USING (a) ORDER BY 1,2,3,4 LIMIT 1;
DEBUG: push down of limit count: 1
a | b | b | b
---------------------------------------------------------------------
1 | 0 | 0 | 1
(1 row)
SELECT COUNT(*) FROM nullkey_c1_t1 t1
JOIN LATERAL (
SELECT * FROM distributed_table t2 WHERE t2.b > t1.a
) q USING(a);
count
---------------------------------------------------------------------
0
(1 row)
SELECT COUNT(*) FROM nullkey_c1_t1 t1
JOIN LATERAL (
SELECT *, random() FROM distributed_table t2 WHERE t2.b > t1.a
) q USING(a);
ERROR: complex joins are only supported when all distributed tables are co-located and joined on their distribution columns
SELECT COUNT(*) FROM distributed_table t1
JOIN LATERAL (
SELECT * FROM nullkey_c1_t1 t2 WHERE t2.b > t1.a
) q USING(a);
count
---------------------------------------------------------------------
1
(1 row)
SET client_min_messages TO DEBUG2;
SET citus.enable_repartition_joins TO OFF;
-- outer joins with different table types
SELECT COUNT(*) FROM nullkey_c1_t1 LEFT JOIN reference_table USING(a);
DEBUG: Creating router plan
count
---------------------------------------------------------------------
11
(1 row)
SELECT COUNT(*) FROM reference_table LEFT JOIN nullkey_c1_t1 USING(a);
DEBUG: Creating router plan
count
---------------------------------------------------------------------
9
(1 row)
SELECT COUNT(*) FROM nullkey_c1_t1 LEFT JOIN citus_local_table USING(a);
DEBUG: Local tables cannot be used in distributed queries.
DEBUG: Wrapping relation "citus_local_table" to a subquery
DEBUG: Distributed planning for a fast-path router query
DEBUG: Creating router plan
DEBUG: generating subplan XXX_1 for subquery SELECT a FROM query_single_shard_table.citus_local_table WHERE true
DEBUG: Plan XXX query after replacing subqueries and CTEs: SELECT count(*) AS count FROM (query_single_shard_table.nullkey_c1_t1 LEFT JOIN (SELECT citus_local_table_1.a, NULL::integer AS b FROM (SELECT intermediate_result.a FROM read_intermediate_result('XXX_1'::text, 'binary'::citus_copy_format) intermediate_result(a integer)) citus_local_table_1) citus_local_table USING (a))
DEBUG: Creating router plan
count
---------------------------------------------------------------------
11
(1 row)
SELECT COUNT(*) FROM citus_local_table LEFT JOIN nullkey_c1_t1 USING(a);
DEBUG: Local tables cannot be used in distributed queries.
DEBUG: Wrapping relation "citus_local_table" to a subquery
DEBUG: Distributed planning for a fast-path router query
DEBUG: Creating router plan
DEBUG: generating subplan XXX_1 for subquery SELECT a FROM query_single_shard_table.citus_local_table WHERE true
DEBUG: recursively planning right side of the left join since the outer side is a recurring rel
DEBUG: recursively planning distributed relation "nullkey_c1_t1" since it is part of a distributed join node that is outer joined with a recurring rel
DEBUG: Wrapping relation "nullkey_c1_t1" to a subquery
DEBUG: Distributed planning for a fast-path router query
DEBUG: Creating router plan
DEBUG: generating subplan XXX_2 for subquery SELECT a FROM query_single_shard_table.nullkey_c1_t1 WHERE true
DEBUG: Plan XXX query after replacing subqueries and CTEs: SELECT count(*) AS count FROM ((SELECT citus_local_table_1.a, NULL::integer AS b FROM (SELECT intermediate_result.a FROM read_intermediate_result('XXX_1'::text, 'binary'::citus_copy_format) intermediate_result(a integer)) citus_local_table_1) citus_local_table LEFT JOIN (SELECT nullkey_c1_t1_1.a, NULL::integer AS b FROM (SELECT intermediate_result.a FROM read_intermediate_result('XXX_2'::text, 'binary'::citus_copy_format) intermediate_result(a integer)) nullkey_c1_t1_1) nullkey_c1_t1 USING (a))
DEBUG: Creating router plan
count
---------------------------------------------------------------------
14
(1 row)
SELECT COUNT(*) FROM nullkey_c1_t1 LEFT JOIN postgres_local_table USING(a);
DEBUG: Local tables cannot be used in distributed queries.
DEBUG: Wrapping relation "postgres_local_table" to a subquery
DEBUG: generating subplan XXX_1 for subquery SELECT a FROM query_single_shard_table.postgres_local_table WHERE true
DEBUG: Plan XXX query after replacing subqueries and CTEs: SELECT count(*) AS count FROM (query_single_shard_table.nullkey_c1_t1 LEFT JOIN (SELECT postgres_local_table_1.a, NULL::integer AS b FROM (SELECT intermediate_result.a FROM read_intermediate_result('XXX_1'::text, 'binary'::citus_copy_format) intermediate_result(a integer)) postgres_local_table_1) postgres_local_table USING (a))
DEBUG: Creating router plan
count
---------------------------------------------------------------------
11
(1 row)
SELECT COUNT(*) FROM postgres_local_table LEFT JOIN nullkey_c1_t1 USING(a);
DEBUG: Local tables cannot be used in distributed queries.
DEBUG: Wrapping relation "postgres_local_table" to a subquery
DEBUG: generating subplan XXX_1 for subquery SELECT a FROM query_single_shard_table.postgres_local_table WHERE true
DEBUG: recursively planning right side of the left join since the outer side is a recurring rel
DEBUG: recursively planning distributed relation "nullkey_c1_t1" since it is part of a distributed join node that is outer joined with a recurring rel
DEBUG: Wrapping relation "nullkey_c1_t1" to a subquery
DEBUG: Distributed planning for a fast-path router query
DEBUG: Creating router plan
DEBUG: generating subplan XXX_2 for subquery SELECT a FROM query_single_shard_table.nullkey_c1_t1 WHERE true
DEBUG: Plan XXX query after replacing subqueries and CTEs: SELECT count(*) AS count FROM ((SELECT postgres_local_table_1.a, NULL::integer AS b FROM (SELECT intermediate_result.a FROM read_intermediate_result('XXX_1'::text, 'binary'::citus_copy_format) intermediate_result(a integer)) postgres_local_table_1) postgres_local_table LEFT JOIN (SELECT nullkey_c1_t1_1.a, NULL::integer AS b FROM (SELECT intermediate_result.a FROM read_intermediate_result('XXX_2'::text, 'binary'::citus_copy_format) intermediate_result(a integer)) nullkey_c1_t1_1) nullkey_c1_t1 USING (a))
DEBUG: Creating router plan
count
---------------------------------------------------------------------
6
(1 row)
SELECT COUNT(*) FROM nullkey_c1_t1 FULL JOIN citus_local_table USING(a);
DEBUG: Local tables cannot be used in distributed queries.
DEBUG: Wrapping relation "citus_local_table" to a subquery
DEBUG: Distributed planning for a fast-path router query
DEBUG: Creating router plan
DEBUG: generating subplan XXX_1 for subquery SELECT a FROM query_single_shard_table.citus_local_table WHERE true
DEBUG: recursively planning left side of the full join since the other side is a recurring rel
DEBUG: recursively planning distributed relation "nullkey_c1_t1" since it is part of a distributed join node that is outer joined with a recurring rel
DEBUG: Wrapping relation "nullkey_c1_t1" to a subquery
DEBUG: Distributed planning for a fast-path router query
DEBUG: Creating router plan
DEBUG: generating subplan XXX_2 for subquery SELECT a FROM query_single_shard_table.nullkey_c1_t1 WHERE true
DEBUG: Plan XXX query after replacing subqueries and CTEs: SELECT count(*) AS count FROM ((SELECT nullkey_c1_t1_1.a, NULL::integer AS b FROM (SELECT intermediate_result.a FROM read_intermediate_result('XXX_2'::text, 'binary'::citus_copy_format) intermediate_result(a integer)) nullkey_c1_t1_1) nullkey_c1_t1 FULL JOIN (SELECT citus_local_table_1.a, NULL::integer AS b FROM (SELECT intermediate_result.a FROM read_intermediate_result('XXX_1'::text, 'binary'::citus_copy_format) intermediate_result(a integer)) citus_local_table_1) citus_local_table USING (a))
DEBUG: Creating router plan
count
---------------------------------------------------------------------
14
(1 row)
SELECT COUNT(*) FROM nullkey_c1_t1 FULL JOIN postgres_local_table USING(a);
DEBUG: Local tables cannot be used in distributed queries.
DEBUG: Wrapping relation "postgres_local_table" to a subquery
DEBUG: generating subplan XXX_1 for subquery SELECT a FROM query_single_shard_table.postgres_local_table WHERE true
DEBUG: recursively planning left side of the full join since the other side is a recurring rel
DEBUG: recursively planning distributed relation "nullkey_c1_t1" since it is part of a distributed join node that is outer joined with a recurring rel
DEBUG: Wrapping relation "nullkey_c1_t1" to a subquery
DEBUG: Distributed planning for a fast-path router query
DEBUG: Creating router plan
DEBUG: generating subplan XXX_2 for subquery SELECT a FROM query_single_shard_table.nullkey_c1_t1 WHERE true
DEBUG: Plan XXX query after replacing subqueries and CTEs: SELECT count(*) AS count FROM ((SELECT nullkey_c1_t1_1.a, NULL::integer AS b FROM (SELECT intermediate_result.a FROM read_intermediate_result('XXX_2'::text, 'binary'::citus_copy_format) intermediate_result(a integer)) nullkey_c1_t1_1) nullkey_c1_t1 FULL JOIN (SELECT postgres_local_table_1.a, NULL::integer AS b FROM (SELECT intermediate_result.a FROM read_intermediate_result('XXX_1'::text, 'binary'::citus_copy_format) intermediate_result(a integer)) postgres_local_table_1) postgres_local_table USING (a))
DEBUG: Creating router plan
count
---------------------------------------------------------------------
13
(1 row)
SELECT COUNT(*) FROM nullkey_c1_t1 FULL JOIN reference_table USING(a);
DEBUG: Creating router plan
count
---------------------------------------------------------------------
12
(1 row)
SET citus.enable_repartition_joins TO ON;
SET client_min_messages TO DEBUG1;
SELECT COUNT(*) FROM nullkey_c1_t1 JOIN append_table USING(a);
count
---------------------------------------------------------------------
2
(1 row)
SELECT COUNT(*) FROM nullkey_c1_t1 JOIN range_table USING(a);
count
---------------------------------------------------------------------
9
(1 row)
SET client_min_messages TO DEBUG2;
SET citus.enable_repartition_joins TO OFF;
SET citus.enable_non_colocated_router_query_pushdown TO ON;
SELECT COUNT(*) FROM nullkey_c1_t1 JOIN range_table USING(a) WHERE range_table.a = 20;
DEBUG: Creating router plan
count
---------------------------------------------------------------------
0
(1 row)
SET citus.enable_non_colocated_router_query_pushdown TO OFF;
SET citus.enable_repartition_joins TO ON;
SET client_min_messages TO DEBUG1;
SELECT COUNT(*) FROM nullkey_c1_t1 JOIN range_table USING(a) WHERE range_table.a = 20;
count
---------------------------------------------------------------------
0
(1 row)
SET client_min_messages TO DEBUG2;
SET citus.enable_repartition_joins TO OFF;
RESET citus.enable_non_colocated_router_query_pushdown;
-- lateral / semi / anti joins with different table types
-- with a reference table
SELECT COUNT(*) FROM nullkey_c1_t1 t1
LEFT JOIN LATERAL (
SELECT * FROM reference_table t2 WHERE t2.b > t1.a
) q USING(a);
DEBUG: Creating router plan
count
---------------------------------------------------------------------
11
(1 row)
SELECT COUNT(*) FROM nullkey_c1_t1 t1
WHERE EXISTS (
SELECT * FROM reference_table t2 WHERE t2.b > t1.a
);
DEBUG: Creating router plan
count
---------------------------------------------------------------------
7
(1 row)
SELECT COUNT(*) FROM nullkey_c1_t1 t1
WHERE NOT EXISTS (
SELECT * FROM reference_table t2 WHERE t2.b > t1.a
);
DEBUG: Creating router plan
count
---------------------------------------------------------------------
4
(1 row)
SELECT COUNT(*) FROM nullkey_c1_t1 t1
WHERE t1.b IN (
SELECT b+1 FROM reference_table t2 WHERE t2.b = t1.a
);
DEBUG: Creating router plan
count
---------------------------------------------------------------------
2
(1 row)
SELECT COUNT(*) FROM nullkey_c1_t1 t1
WHERE t1.b NOT IN (
SELECT a FROM reference_table t2 WHERE t2.b > t1.a
);
DEBUG: Creating router plan
count
---------------------------------------------------------------------
9
(1 row)
SELECT COUNT(*) FROM nullkey_c1_t1 t1
JOIN LATERAL (
SELECT * FROM reference_table t2 WHERE t2.b > t1.a
) q USING(a);
DEBUG: Creating router plan
count
---------------------------------------------------------------------
0
(1 row)
SELECT COUNT(*) FROM reference_table t1
LEFT JOIN LATERAL (
SELECT * FROM nullkey_c1_t1 t2 WHERE t2.b > t1.a
) q USING(a);
DEBUG: Creating router plan
count
---------------------------------------------------------------------
6
(1 row)
SELECT COUNT(*) FROM reference_table t1
WHERE EXISTS (
SELECT * FROM nullkey_c1_t1 t2 WHERE t2.b > t1.a
);
DEBUG: Creating router plan
count
---------------------------------------------------------------------
6
(1 row)
SELECT COUNT(*) FROM reference_table t1
WHERE t1.b IN (
SELECT b+1 FROM nullkey_c1_t1 t2 WHERE t2.b = t1.a
);
DEBUG: Creating router plan
count
---------------------------------------------------------------------
0
(1 row)
SELECT COUNT(*) FROM reference_table t1
WHERE t1.b NOT IN (
SELECT a FROM nullkey_c1_t1 t2 WHERE t2.b > t1.a
);
DEBUG: Creating router plan
count
---------------------------------------------------------------------
4
(1 row)
SELECT COUNT(*) FROM reference_table t1
JOIN LATERAL (
SELECT * FROM nullkey_c1_t1 t2 WHERE t2.b > t1.a
) q USING(a);
DEBUG: Creating router plan
count
---------------------------------------------------------------------
2
(1 row)
-- with a distributed table
SELECT COUNT(*) FROM nullkey_c1_t1 t1
LEFT JOIN LATERAL (
SELECT * FROM distributed_table t2 WHERE t2.b > t1.a
) q USING(a);
DEBUG: Router planner cannot handle multi-shard select queries
DEBUG: skipping recursive planning for the subquery since it contains references to outer queries
ERROR: complex joins are only supported when all distributed tables are co-located and joined on their distribution columns
SELECT COUNT(*) FROM nullkey_c1_t1 t1
WHERE EXISTS (
SELECT * FROM distributed_table t2 WHERE t2.b > t1.a
);
DEBUG: Router planner cannot handle multi-shard select queries
DEBUG: skipping recursive planning for the subquery since it contains references to outer queries
ERROR: complex joins are only supported when all distributed tables are co-located and joined on their distribution columns
SELECT COUNT(*) FROM nullkey_c1_t1 t1
WHERE NOT EXISTS (
SELECT * FROM distributed_table t2 WHERE t2.b > t1.a
);
DEBUG: Router planner cannot handle multi-shard select queries
DEBUG: skipping recursive planning for the subquery since it contains references to outer queries
ERROR: complex joins are only supported when all distributed tables are co-located and joined on their distribution columns
SELECT COUNT(*) FROM nullkey_c1_t1 t1
WHERE t1.b IN (
SELECT b+1 FROM distributed_table t2 WHERE t2.b = t1.a
);
DEBUG: Router planner cannot handle multi-shard select queries
DEBUG: skipping recursive planning for the subquery since it contains references to outer queries
ERROR: complex joins are only supported when all distributed tables are co-located and joined on their distribution columns
SELECT COUNT(*) FROM nullkey_c1_t1 t1
WHERE t1.b NOT IN (
SELECT a FROM distributed_table t2 WHERE t2.b > t1.a
);
DEBUG: Router planner cannot handle multi-shard select queries
DEBUG: skipping recursive planning for the subquery since it contains references to outer queries
ERROR: complex joins are only supported when all distributed tables are co-located and joined on their distribution columns
SELECT COUNT(*) FROM distributed_table t1
LEFT JOIN LATERAL (
SELECT * FROM nullkey_c1_t1 t2 WHERE t2.b > t1.a
) q USING(a);
DEBUG: Router planner cannot handle multi-shard select queries
DEBUG: skipping recursive planning for the subquery since it contains references to outer queries
ERROR: complex joins are only supported when all distributed tables are co-located and joined on their distribution columns
SELECT COUNT(*) FROM distributed_table t1
WHERE EXISTS (
SELECT * FROM nullkey_c1_t1 t2 WHERE t2.b > t1.a
);
DEBUG: Router planner cannot handle multi-shard select queries
DEBUG: skipping recursive planning for the subquery since it contains references to outer queries
ERROR: complex joins are only supported when all distributed tables are co-located and joined on their distribution columns
SELECT COUNT(*) FROM distributed_table t1
WHERE t1.b IN (
SELECT b+1 FROM nullkey_c1_t1 t2 WHERE t2.b = t1.a
);
DEBUG: Router planner cannot handle multi-shard select queries
DEBUG: skipping recursive planning for the subquery since it contains references to outer queries
ERROR: complex joins are only supported when all distributed tables are co-located and joined on their distribution columns
SELECT COUNT(*) FROM distributed_table t1
WHERE t1.b NOT IN (
SELECT a FROM nullkey_c1_t1 t2 WHERE t2.b > t1.a
);
DEBUG: Router planner cannot handle multi-shard select queries
DEBUG: skipping recursive planning for the subquery since it contains references to outer queries
ERROR: complex joins are only supported when all distributed tables are co-located and joined on their distribution columns
-- with postgres / citus local tables
SELECT COUNT(*) FROM nullkey_c1_t1 t1
LEFT JOIN LATERAL (
SELECT * FROM citus_local_table t2 WHERE t2.b > t1.a
) q USING(a);
DEBUG: Local tables cannot be used in distributed queries.
DEBUG: skipping recursive planning for the subquery since it contains references to outer queries
ERROR: direct joins between distributed and local tables are not supported
SELECT COUNT(*) FROM nullkey_c1_t1 t1
WHERE EXISTS (
SELECT * FROM citus_local_table t2 WHERE t2.b > t1.a
);
DEBUG: Local tables cannot be used in distributed queries.
DEBUG: skipping recursive planning for the subquery since it contains references to outer queries
ERROR: direct joins between distributed and local tables are not supported
SELECT COUNT(*) FROM nullkey_c1_t1 t1
WHERE NOT EXISTS (
SELECT * FROM citus_local_table t2 WHERE t2.b > t1.a
);
DEBUG: Local tables cannot be used in distributed queries.
DEBUG: skipping recursive planning for the subquery since it contains references to outer queries
ERROR: direct joins between distributed and local tables are not supported
SELECT COUNT(*) FROM nullkey_c1_t1 t1
WHERE t1.b + random() IN (
SELECT b+1 FROM citus_local_table t2 WHERE t2.b = t1.a
);
DEBUG: router planner does not support queries that reference non-colocated distributed tables
DEBUG: skipping recursive planning for the subquery since it contains references to outer queries
ERROR: direct joins between distributed and local tables are not supported
SELECT COUNT(*) FROM nullkey_c1_t1 t1
WHERE t1.b NOT IN (
SELECT a FROM citus_local_table t2 WHERE t2.b > t1.a
);
DEBUG: router planner does not support queries that reference non-colocated distributed tables
DEBUG: skipping recursive planning for the subquery since it contains references to outer queries
ERROR: direct joins between distributed and local tables are not supported
SELECT COUNT(*) FROM nullkey_c1_t1 t1
JOIN LATERAL (
SELECT * FROM citus_local_table t2 WHERE t2.b > t1.a
) q USING(a);
DEBUG: Local tables cannot be used in distributed queries.
DEBUG: skipping recursive planning for the subquery since it contains references to outer queries
ERROR: direct joins between distributed and local tables are not supported
-- The following and a few other tests in this file unnecessarily go through
-- recursive planning. This is because we recursive plan distributed tables
-- when they are referred in the inner side of an outer join, if the outer
-- side is a recurring rel. In future, we can optimize that such that we
-- can skip recursively planning the single-shard table because such a join
-- wouldn't result in returning recurring tuples.
--
-- And specifically for the tests that contains a sublink (as below), things
-- get even more interesting. We try to recursively plan the single-shard
-- table but we cannot do so due to the sublink. However, the final query
-- can go through router planner and hence is supported.
SELECT COUNT(*) FROM citus_local_table t1
LEFT JOIN LATERAL (
SELECT * FROM nullkey_c1_t1 t2 WHERE t2.b > t1.a
) q USING(a);
DEBUG: Local tables cannot be used in distributed queries.
DEBUG: Wrapping relation "citus_local_table" "t1" to a subquery
DEBUG: Distributed planning for a fast-path router query
DEBUG: Creating router plan
DEBUG: generating subplan XXX_1 for subquery SELECT a FROM query_single_shard_table.citus_local_table t1 WHERE true
DEBUG: recursively planning right side of the left join since the outer side is a recurring rel
DEBUG: recursively planning the distributed subquery since it is part of a distributed join node that is outer joined with a recurring rel
DEBUG: skipping recursive planning for the subquery since it contains references to outer queries
DEBUG: Plan XXX query after replacing subqueries and CTEs: SELECT count(*) AS count FROM ((SELECT t1_1.a, NULL::integer AS b FROM (SELECT intermediate_result.a FROM read_intermediate_result('XXX_1'::text, 'binary'::citus_copy_format) intermediate_result(a integer)) t1_1) t1 LEFT JOIN LATERAL (SELECT t2.a, t2.b FROM query_single_shard_table.nullkey_c1_t1 t2 WHERE (t2.b OPERATOR(pg_catalog.>) t1.a)) q USING (a))
DEBUG: Creating router plan
count
---------------------------------------------------------------------
11
(1 row)
SELECT COUNT(*) FROM postgres_local_table t1
LEFT JOIN LATERAL (
SELECT * FROM nullkey_c1_t1 t2 WHERE t2.b > t1.a
) q USING(a);
DEBUG: Local tables cannot be used in distributed queries.
DEBUG: Wrapping relation "postgres_local_table" "t1" to a subquery
DEBUG: generating subplan XXX_1 for subquery SELECT a FROM query_single_shard_table.postgres_local_table t1 WHERE true
DEBUG: recursively planning right side of the left join since the outer side is a recurring rel
DEBUG: recursively planning the distributed subquery since it is part of a distributed join node that is outer joined with a recurring rel
DEBUG: skipping recursive planning for the subquery since it contains references to outer queries
DEBUG: Plan XXX query after replacing subqueries and CTEs: SELECT count(*) AS count FROM ((SELECT t1_1.a, NULL::integer AS b FROM (SELECT intermediate_result.a FROM read_intermediate_result('XXX_1'::text, 'binary'::citus_copy_format) intermediate_result(a integer)) t1_1) t1 LEFT JOIN LATERAL (SELECT t2.a, t2.b FROM query_single_shard_table.nullkey_c1_t1 t2 WHERE (t2.b OPERATOR(pg_catalog.>) t1.a)) q USING (a))
DEBUG: Creating router plan
count
---------------------------------------------------------------------
6
(1 row)
SELECT COUNT(*) FROM citus_local_table t1
WHERE EXISTS (
SELECT * FROM nullkey_c1_t1 t2 WHERE t2.b > t1.a
);
DEBUG: Local tables cannot be used in distributed queries.
DEBUG: skipping recursive planning for the subquery since it contains references to outer queries
ERROR: direct joins between distributed and local tables are not supported
SELECT COUNT(*) FROM citus_local_table t1
WHERE t1.b + random() IN (
SELECT b+1 FROM nullkey_c1_t1 t2 WHERE t2.b = t1.a
);
DEBUG: router planner does not support queries that reference non-colocated distributed tables
DEBUG: skipping recursive planning for the subquery since it contains references to outer queries
ERROR: direct joins between distributed and local tables are not supported
SELECT COUNT(*) FROM citus_local_table t1
WHERE t1.b NOT IN (
SELECT a FROM nullkey_c1_t1 t2 WHERE t2.b > t1.a
);
DEBUG: router planner does not support queries that reference non-colocated distributed tables
DEBUG: skipping recursive planning for the subquery since it contains references to outer queries
ERROR: direct joins between distributed and local tables are not supported
SELECT COUNT(*) FROM citus_local_table t1
JOIN LATERAL (
SELECT * FROM nullkey_c1_t1 t2 WHERE t2.b > t1.a
) q USING(a);
DEBUG: Local tables cannot be used in distributed queries.
DEBUG: Wrapping relation "citus_local_table" "t1" to a subquery
DEBUG: Distributed planning for a fast-path router query
DEBUG: Creating router plan
DEBUG: generating subplan XXX_1 for subquery SELECT a FROM query_single_shard_table.citus_local_table t1 WHERE true
DEBUG: Plan XXX query after replacing subqueries and CTEs: SELECT count(*) AS count FROM ((SELECT t1_1.a, NULL::integer AS b FROM (SELECT intermediate_result.a FROM read_intermediate_result('XXX_1'::text, 'binary'::citus_copy_format) intermediate_result(a integer)) t1_1) t1 JOIN LATERAL (SELECT t2.a, t2.b FROM query_single_shard_table.nullkey_c1_t1 t2 WHERE (t2.b OPERATOR(pg_catalog.>) t1.a)) q USING (a))
DEBUG: Creating router plan
count
---------------------------------------------------------------------
2
(1 row)
SELECT COUNT(*) FROM nullkey_c1_t1 t1
LEFT JOIN LATERAL (
SELECT * FROM postgres_local_table t2 WHERE t2.b > t1.a
) q USING(a);
DEBUG: Local tables cannot be used in distributed queries.
DEBUG: skipping recursive planning for the subquery since it contains references to outer queries
ERROR: direct joins between distributed and local tables are not supported
HINT: Use CTE's or subqueries to select from local tables and use them in joins
SELECT COUNT(*) FROM nullkey_c1_t1 t1
WHERE EXISTS (
SELECT * FROM postgres_local_table t2 WHERE t2.b > t1.a
);
DEBUG: Local tables cannot be used in distributed queries.
DEBUG: skipping recursive planning for the subquery since it contains references to outer queries
ERROR: direct joins between distributed and local tables are not supported
HINT: Use CTE's or subqueries to select from local tables and use them in joins
SELECT COUNT(*) FROM nullkey_c1_t1 t1
WHERE NOT EXISTS (
SELECT * FROM postgres_local_table t2 WHERE t2.b > t1.a
);
DEBUG: Local tables cannot be used in distributed queries.
DEBUG: skipping recursive planning for the subquery since it contains references to outer queries
ERROR: direct joins between distributed and local tables are not supported
HINT: Use CTE's or subqueries to select from local tables and use them in joins
SELECT COUNT(*) FROM nullkey_c1_t1 t1
WHERE t1.b + random() IN (
SELECT b+1 FROM postgres_local_table t2 WHERE t2.b = t1.a
);
DEBUG: found no worker with all shard placements
DEBUG: skipping recursive planning for the subquery since it contains references to outer queries
ERROR: direct joins between distributed and local tables are not supported
HINT: Use CTE's or subqueries to select from local tables and use them in joins
SELECT COUNT(*) FROM nullkey_c1_t1 t1
WHERE t1.b NOT IN (
SELECT a FROM postgres_local_table t2 WHERE t2.b > t1.a
);
DEBUG: found no worker with all shard placements
DEBUG: skipping recursive planning for the subquery since it contains references to outer queries
ERROR: direct joins between distributed and local tables are not supported
HINT: Use CTE's or subqueries to select from local tables and use them in joins
SELECT COUNT(*) FROM nullkey_c1_t1 t1
JOIN LATERAL (
SELECT * FROM postgres_local_table t2 WHERE t2.b > t1.a
) q USING(a);
DEBUG: Local tables cannot be used in distributed queries.
DEBUG: skipping recursive planning for the subquery since it contains references to outer queries
ERROR: direct joins between distributed and local tables are not supported
HINT: Use CTE's or subqueries to select from local tables and use them in joins
SELECT COUNT(*) FROM postgres_local_table t1
WHERE EXISTS (
SELECT * FROM nullkey_c1_t1 t2 WHERE t2.b > t1.a
);
DEBUG: Local tables cannot be used in distributed queries.
DEBUG: skipping recursive planning for the subquery since it contains references to outer queries
ERROR: direct joins between distributed and local tables are not supported
HINT: Use CTE's or subqueries to select from local tables and use them in joins
SELECT COUNT(*) FROM postgres_local_table t1
WHERE t1.b + random() IN (
SELECT b+1 FROM nullkey_c1_t1 t2 WHERE t2.b = t1.a
);
DEBUG: found no worker with all shard placements
DEBUG: skipping recursive planning for the subquery since it contains references to outer queries
ERROR: direct joins between distributed and local tables are not supported
HINT: Use CTE's or subqueries to select from local tables and use them in joins
SELECT COUNT(*) FROM postgres_local_table t1
WHERE t1.b NOT IN (
SELECT a FROM nullkey_c1_t1 t2 WHERE t2.b > t1.a
);
DEBUG: found no worker with all shard placements
DEBUG: skipping recursive planning for the subquery since it contains references to outer queries
ERROR: direct joins between distributed and local tables are not supported
HINT: Use CTE's or subqueries to select from local tables and use them in joins
SELECT COUNT(*) FROM postgres_local_table t1
JOIN LATERAL (
SELECT * FROM nullkey_c1_t1 t2 WHERE t2.b > t1.a
) q USING(a);
DEBUG: Local tables cannot be used in distributed queries.
DEBUG: Wrapping relation "postgres_local_table" "t1" to a subquery
DEBUG: generating subplan XXX_1 for subquery SELECT a FROM query_single_shard_table.postgres_local_table t1 WHERE true
DEBUG: Plan XXX query after replacing subqueries and CTEs: SELECT count(*) AS count FROM ((SELECT t1_1.a, NULL::integer AS b FROM (SELECT intermediate_result.a FROM read_intermediate_result('XXX_1'::text, 'binary'::citus_copy_format) intermediate_result(a integer)) t1_1) t1 JOIN LATERAL (SELECT t2.a, t2.b FROM query_single_shard_table.nullkey_c1_t1 t2 WHERE (t2.b OPERATOR(pg_catalog.>) t1.a)) q USING (a))
DEBUG: Creating router plan
count
---------------------------------------------------------------------
0
(1 row)
-- insert .. select
-- between two colocated single-shard tables
-- The target list of "distributed statement"s that we send to workers
-- differ(*) in Postgres versions < 15. For this reason, we temporarily
-- disable debug messages here and run the EXPLAIN'ed version of the
-- command.
--
-- (*): < SELECT a, b > vs < SELECT table_name.a, table_name.b >
SET client_min_messages TO WARNING;
EXPLAIN (ANALYZE TRUE, TIMING FALSE, COSTS FALSE, SUMMARY FALSE, VERBOSE FALSE)
INSERT INTO nullkey_c1_t1 SELECT * FROM nullkey_c1_t2;
QUERY PLAN
---------------------------------------------------------------------
Custom Scan (Citus Adaptive) (actual rows=0 loops=1)
Task Count: 1
Tasks Shown: All
-> Task
Node: host=localhost port=xxxxx dbname=regression
-> Insert on nullkey_c1_t1_1620000 citus_table_alias (actual rows=0 loops=1)
-> Seq Scan on nullkey_c1_t2_1620001 nullkey_c1_t2 (actual rows=10 loops=1)
(7 rows)
SET client_min_messages TO DEBUG2;
-- between two non-colocated single-shard tables
INSERT INTO nullkey_c1_t1 SELECT * FROM nullkey_c2_t1;
DEBUG: Creating router plan
DEBUG: INSERT target relation and all source relations of the SELECT must be colocated in distributed INSERT ... SELECT
DEBUG: Distributed planning for a fast-path router query
DEBUG: Creating router plan
DEBUG: Collecting INSERT ... SELECT results on coordinator
-- between a single-shard table and a table of different type
SET client_min_messages TO WARNING;
EXPLAIN (ANALYZE TRUE, TIMING FALSE, COSTS FALSE, SUMMARY FALSE, VERBOSE FALSE)
INSERT INTO nullkey_c1_t1 SELECT * FROM reference_table;
QUERY PLAN
---------------------------------------------------------------------
Custom Scan (Citus Adaptive) (actual rows=0 loops=1)
Task Count: 1
Tasks Shown: All
-> Task
Node: host=localhost port=xxxxx dbname=regression
-> Insert on nullkey_c1_t1_1620000 citus_table_alias (actual rows=0 loops=1)
-> Seq Scan on reference_table_1620005 reference_table (actual rows=6 loops=1)
(7 rows)
SET client_min_messages TO DEBUG2;
INSERT INTO nullkey_c1_t1 SELECT * FROM distributed_table;
DEBUG: INSERT target relation and all source relations of the SELECT must be colocated in distributed INSERT ... SELECT
DEBUG: Router planner cannot handle multi-shard select queries
DEBUG: Collecting INSERT ... SELECT results on coordinator
INSERT INTO nullkey_c1_t1 SELECT * FROM citus_local_table;
DEBUG: Creating router plan
DEBUG: distributed INSERT ... SELECT cannot select from a local relation when inserting into a distributed table
DEBUG: Distributed planning for a fast-path router query
DEBUG: Creating router plan
DEBUG: Collecting INSERT ... SELECT results on coordinator
INSERT INTO nullkey_c1_t1 SELECT * FROM postgres_local_table;
DEBUG: Creating router plan
DEBUG: distributed INSERT ... SELECT can only select from distributed tables
DEBUG: Collecting INSERT ... SELECT results on coordinator
INSERT INTO reference_table SELECT * FROM nullkey_c1_t1;
DEBUG: Creating router plan
DEBUG: only reference tables may be queried when targeting a reference table with distributed INSERT ... SELECT
DEBUG: Distributed planning for a fast-path router query
DEBUG: Creating router plan
DEBUG: Collecting INSERT ... SELECT results on coordinator
INSERT INTO distributed_table SELECT * FROM nullkey_c1_t1;
DEBUG: Creating router plan
DEBUG: cannot perform distributed INSERT INTO ... SELECT because the partition columns in the source table and subquery do not match
DETAIL: The target table's partition column should correspond to a partition column in the subquery.
DEBUG: Distributed planning for a fast-path router query
DEBUG: Creating router plan
DEBUG: Collecting INSERT ... SELECT results on coordinator
INSERT INTO citus_local_table SELECT * FROM nullkey_c1_t1;
DEBUG: Creating router plan
DEBUG: distributed INSERT ... SELECT cannot insert into a local table that is added to metadata
DEBUG: Distributed planning for a fast-path router query
DEBUG: Creating router plan
DEBUG: Collecting INSERT ... SELECT results on coordinator
INSERT INTO postgres_local_table SELECT * FROM nullkey_c1_t1;
DEBUG: Creating router plan
-- test subquery
SELECT count(*) FROM
(
SELECT * FROM (SELECT * FROM nullkey_c1_t2) as subquery_inner
) AS subquery_top;
DEBUG: Creating router plan
count
---------------------------------------------------------------------
10
(1 row)
-- test cte inlining
WITH cte_nullkey_c1_t1 AS (SELECT * FROM nullkey_c1_t1),
cte_postgres_local_table AS (SELECT * FROM postgres_local_table),
cte_distributed_table AS (SELECT * FROM distributed_table)
SELECT COUNT(*) FROM cte_distributed_table, cte_nullkey_c1_t1, cte_postgres_local_table
WHERE cte_nullkey_c1_t1.a > 3 AND cte_distributed_table.a < 5;
DEBUG: CTE cte_nullkey_c1_t1 is going to be inlined via distributed planning
DEBUG: CTE cte_postgres_local_table is going to be inlined via distributed planning
DEBUG: CTE cte_distributed_table is going to be inlined via distributed planning
DEBUG: Local tables cannot be used in distributed queries.
DEBUG: generating subplan XXX_1 for subquery SELECT a, b FROM query_single_shard_table.postgres_local_table
DEBUG: Distributed planning for a fast-path router query
DEBUG: Creating router plan
DEBUG: generating subplan XXX_2 for subquery SELECT a, b FROM query_single_shard_table.nullkey_c1_t1
DEBUG: Plan XXX query after replacing subqueries and CTEs: SELECT count(*) AS count FROM (SELECT distributed_table.a, distributed_table.b FROM query_single_shard_table.distributed_table) cte_distributed_table, (SELECT intermediate_result.a, intermediate_result.b FROM read_intermediate_result('XXX_2'::text, 'binary'::citus_copy_format) intermediate_result(a integer, b integer)) cte_nullkey_c1_t1, (SELECT intermediate_result.a, intermediate_result.b FROM read_intermediate_result('XXX_1'::text, 'binary'::citus_copy_format) intermediate_result(a integer, b integer)) cte_postgres_local_table WHERE ((cte_nullkey_c1_t1.a OPERATOR(pg_catalog.>) 3) AND (cte_distributed_table.a OPERATOR(pg_catalog.<) 5))
DEBUG: Router planner cannot handle multi-shard select queries
count
---------------------------------------------------------------------
78144
(1 row)
-- test recursive ctes
WITH level_0 AS (
WITH level_1 AS (
WITH RECURSIVE level_2_recursive(x) AS (
VALUES (1)
UNION ALL
SELECT a + 1 FROM nullkey_c1_t1 JOIN level_2_recursive ON (a = x) WHERE a < 2
)
SELECT * FROM level_2_recursive RIGHT JOIN reference_table ON (level_2_recursive.x = reference_table.a)
)
SELECT * FROM level_1
)
SELECT COUNT(*) FROM level_0;
DEBUG: CTE level_0 is going to be inlined via distributed planning
DEBUG: CTE level_1 is going to be inlined via distributed planning
DEBUG: Creating router plan
count
---------------------------------------------------------------------
106
(1 row)
WITH level_0 AS (
WITH level_1 AS (
WITH RECURSIVE level_2_recursive(x) AS (
VALUES (1)
UNION ALL
SELECT a + 1 FROM nullkey_c1_t1 JOIN level_2_recursive ON (a = x) WHERE a < 100
)
SELECT * FROM level_2_recursive JOIN distributed_table ON (level_2_recursive.x = distributed_table.a)
)
SELECT * FROM level_1
)
SELECT COUNT(*) FROM level_0;
DEBUG: CTE level_0 is going to be inlined via distributed planning
DEBUG: CTE level_1 is going to be inlined via distributed planning
DEBUG: Router planner cannot handle multi-shard select queries
DEBUG: Router planner cannot handle multi-shard select queries
DEBUG: generating subplan XXX_1 for CTE level_0: WITH level_1 AS (WITH RECURSIVE level_2_recursive(x) AS (VALUES (1) UNION ALL SELECT (nullkey_c1_t1.a OPERATOR(pg_catalog.+) 1) FROM (query_single_shard_table.nullkey_c1_t1 JOIN level_2_recursive level_2_recursive_1 ON ((nullkey_c1_t1.a OPERATOR(pg_catalog.=) level_2_recursive_1.x))) WHERE (nullkey_c1_t1.a OPERATOR(pg_catalog.<) 100)) SELECT level_2_recursive.x, distributed_table.a, distributed_table.b FROM (level_2_recursive JOIN query_single_shard_table.distributed_table ON ((level_2_recursive.x OPERATOR(pg_catalog.=) distributed_table.a)))) SELECT x, a, b FROM level_1
DEBUG: CTE level_1 is going to be inlined via distributed planning
DEBUG: Router planner cannot handle multi-shard select queries
DEBUG: Router planner cannot handle multi-shard select queries
DEBUG: generating subplan XXX_1 for CTE level_1: WITH RECURSIVE level_2_recursive(x) AS (VALUES (1) UNION ALL SELECT (nullkey_c1_t1.a OPERATOR(pg_catalog.+) 1) FROM (query_single_shard_table.nullkey_c1_t1 JOIN level_2_recursive level_2_recursive_1 ON ((nullkey_c1_t1.a OPERATOR(pg_catalog.=) level_2_recursive_1.x))) WHERE (nullkey_c1_t1.a OPERATOR(pg_catalog.<) 100)) SELECT level_2_recursive.x, distributed_table.a, distributed_table.b FROM (level_2_recursive JOIN query_single_shard_table.distributed_table ON ((level_2_recursive.x OPERATOR(pg_catalog.=) distributed_table.a)))
DEBUG: Router planner cannot handle multi-shard select queries
ERROR: recursive CTEs are only supported when they contain a filter on the distribution column
-- grouping set
SELECT
id, substring(title, 2, 1) AS subtitle, count(*)
FROM articles_hash
WHERE author_id = 1 or author_id = 2
GROUP BY GROUPING SETS ((id),(subtitle))
ORDER BY id, subtitle;
DEBUG: Distributed planning for a fast-path router query
DEBUG: Creating router plan
id | subtitle | count
---------------------------------------------------------------------
42 | | 1
| u | 1
(2 rows)
-- subquery in SELECT clause
SELECT a.title AS name, (SELECT a2.id FROM articles_hash a2 WHERE a.id = a2.id LIMIT 1)
AS special_price FROM articles_hash a
ORDER BY 1,2;
DEBUG: Creating router plan
name | special_price
---------------------------------------------------------------------
adversa | 15
aerophyte | 28
affixal | 43
altdorfer | 4
amateur | 29
anjanette | 50
anyone | 49
aruru | 5
aseyev | 13
assembly | 18
aubergiste | 19
ausable | 42
(12 rows)
-- test having clause
SELECT COUNT(*), b FROM nullkey_c1_t1 GROUP BY 2
HAVING (SELECT COUNT(*) FROM nullkey_c1_t2) > 0
ORDER BY 1,2;
DEBUG: Creating router plan
count | b
---------------------------------------------------------------------
2 | 9
2 | 10
3 | 0
4 | 1
4 | 8
6 | 6
6 | 7
8 | 3
8 | 4
8 | 5
9 | 2
(11 rows)
SELECT COUNT(*), b FROM nullkey_c1_t1 GROUP BY 2
HAVING (SELECT COUNT(*) FROM nullkey_c2_t1) > 0
ORDER BY 1,2;
DEBUG: router planner does not support queries that reference non-colocated distributed tables
DEBUG: Distributed planning for a fast-path router query
DEBUG: Creating router plan
DEBUG: generating subplan XXX_1 for subquery SELECT count(*) AS count FROM query_single_shard_table.nullkey_c2_t1
DEBUG: Plan XXX query after replacing subqueries and CTEs: SELECT count(*) AS count, b FROM query_single_shard_table.nullkey_c1_t1 GROUP BY b HAVING ((SELECT intermediate_result.count FROM read_intermediate_result('XXX_1'::text, 'binary'::citus_copy_format) intermediate_result(count bigint)) OPERATOR(pg_catalog.>) 0) ORDER BY (count(*)), b
DEBUG: Creating router plan
count | b
---------------------------------------------------------------------
2 | 9
2 | 10
3 | 0
4 | 1
4 | 8
6 | 6
6 | 7
8 | 3
8 | 4
8 | 5
9 | 2
(11 rows)
SELECT COUNT(*), b FROM nullkey_c1_t1 GROUP BY 2
HAVING (SELECT COUNT(*) FROM distributed_table) > 0
ORDER BY 1,2;
DEBUG: Router planner cannot handle multi-shard select queries
DEBUG: Router planner cannot handle multi-shard select queries
DEBUG: generating subplan XXX_1 for subquery SELECT count(*) AS count FROM query_single_shard_table.distributed_table
DEBUG: Plan XXX query after replacing subqueries and CTEs: SELECT count(*) AS count, b FROM query_single_shard_table.nullkey_c1_t1 GROUP BY b HAVING ((SELECT intermediate_result.count FROM read_intermediate_result('XXX_1'::text, 'binary'::citus_copy_format) intermediate_result(count bigint)) OPERATOR(pg_catalog.>) 0) ORDER BY (count(*)), b
DEBUG: Creating router plan
count | b
---------------------------------------------------------------------
2 | 9
2 | 10
3 | 0
4 | 1
4 | 8
6 | 6
6 | 7
8 | 3
8 | 4
8 | 5
9 | 2
(11 rows)
SELECT COUNT(*), b FROM nullkey_c1_t1 t4 GROUP BY 2
HAVING (
SELECT COUNT(*) FROM nullkey_c1_t1 t1 JOIN (SELECT * FROM nullkey_c1_t2) t2 USING (a) JOIN (SELECT * FROM nullkey_c1_t2) t3 USING (a)
WHERE t2.b > t4.b
) > 5
ORDER BY 1,2;
DEBUG: Creating router plan
count | b
---------------------------------------------------------------------
3 | 0
4 | 1
6 | 6
8 | 3
8 | 4
8 | 5
9 | 2
(7 rows)
SELECT COUNT(*), b FROM distributed_table t4 GROUP BY 2
HAVING (
SELECT COUNT(*) FROM nullkey_c1_t1 t1 JOIN (SELECT * FROM distributed_table) t2 USING (a) JOIN (SELECT * FROM nullkey_c1_t2) t3 USING (a)
WHERE t2.b > t4.b
) > 5
ORDER BY 1,2;
DEBUG: Router planner cannot handle multi-shard select queries
DEBUG: Router planner cannot handle multi-shard select queries
DEBUG: generating subplan XXX_1 for subquery SELECT a, b FROM query_single_shard_table.distributed_table
ERROR: Subqueries in HAVING cannot refer to outer query
-- test prepared statements
-- prepare queries can be router plannable
PREPARE author_1_articles as
SELECT *
FROM articles_hash
WHERE author_id = 1;
EXECUTE author_1_articles;
DEBUG: Distributed planning for a fast-path router query
DEBUG: Creating router plan
id | author_id | title | word_count
---------------------------------------------------------------------
(0 rows)
EXECUTE author_1_articles;
id | author_id | title | word_count
---------------------------------------------------------------------
(0 rows)
EXECUTE author_1_articles;
id | author_id | title | word_count
---------------------------------------------------------------------
(0 rows)
EXECUTE author_1_articles;
id | author_id | title | word_count
---------------------------------------------------------------------
(0 rows)
EXECUTE author_1_articles;
id | author_id | title | word_count
---------------------------------------------------------------------
(0 rows)
EXECUTE author_1_articles;
id | author_id | title | word_count
---------------------------------------------------------------------
(0 rows)
-- parametric prepare queries can be router plannable
PREPARE author_articles(int) as
SELECT *
FROM articles_hash
WHERE author_id = $1;
EXECUTE author_articles(1);
DEBUG: Distributed planning for a fast-path router query
DEBUG: Creating router plan
id | author_id | title | word_count
---------------------------------------------------------------------
(0 rows)
EXECUTE author_articles(1);
DEBUG: Distributed planning for a fast-path router query
DEBUG: Creating router plan
id | author_id | title | word_count
---------------------------------------------------------------------
(0 rows)
EXECUTE author_articles(1);
DEBUG: Distributed planning for a fast-path router query
DEBUG: Creating router plan
id | author_id | title | word_count
---------------------------------------------------------------------
(0 rows)
EXECUTE author_articles(1);
DEBUG: Distributed planning for a fast-path router query
DEBUG: Creating router plan
id | author_id | title | word_count
---------------------------------------------------------------------
(0 rows)
EXECUTE author_articles(1);
DEBUG: Distributed planning for a fast-path router query
DEBUG: Creating router plan
id | author_id | title | word_count
---------------------------------------------------------------------
(0 rows)
EXECUTE author_articles(1);
DEBUG: Distributed planning for a fast-path router query
DEBUG: Creating router plan
id | author_id | title | word_count
---------------------------------------------------------------------
(0 rows)
EXECUTE author_articles(NULL);
id | author_id | title | word_count
---------------------------------------------------------------------
(0 rows)
EXECUTE author_articles(NULL);
id | author_id | title | word_count
---------------------------------------------------------------------
(0 rows)
EXECUTE author_articles(NULL);
id | author_id | title | word_count
---------------------------------------------------------------------
(0 rows)
EXECUTE author_articles(NULL);
id | author_id | title | word_count
---------------------------------------------------------------------
(0 rows)
EXECUTE author_articles(NULL);
id | author_id | title | word_count
---------------------------------------------------------------------
(0 rows)
EXECUTE author_articles(NULL);
id | author_id | title | word_count
---------------------------------------------------------------------
(0 rows)
EXECUTE author_articles(NULL);
id | author_id | title | word_count
---------------------------------------------------------------------
(0 rows)
PREPARE author_articles_update(int) AS
UPDATE articles_hash SET title = 'test' WHERE author_id = $1;
EXECUTE author_articles_update(NULL);
DEBUG: Distributed planning for a fast-path router query
DEBUG: Creating router plan
EXECUTE author_articles_update(NULL);
DEBUG: Distributed planning for a fast-path router query
DEBUG: Creating router plan
EXECUTE author_articles_update(NULL);
DEBUG: Distributed planning for a fast-path router query
DEBUG: Creating router plan
EXECUTE author_articles_update(NULL);
DEBUG: Distributed planning for a fast-path router query
DEBUG: Creating router plan
EXECUTE author_articles_update(NULL);
DEBUG: Distributed planning for a fast-path router query
DEBUG: Creating router plan
EXECUTE author_articles_update(NULL);
DEBUG: Distributed planning for a fast-path router query
DEBUG: Creating router plan
EXECUTE author_articles_update(NULL);
-- More tests with insert .. select.
--
-- The target list of "distributed statement"s that we send to workers
-- might differ(*) in Postgres versions < 15 and they are reported when
-- "log level >= DEBUG2". For this reason, we set log level to DEBUG1 to
-- avoid reporting them.
--
-- DEBUG1 still allows reporting the reason why given INSERT .. SELECT
-- query is not distributed / requires pull-to-coordinator.
SET client_min_messages TO DEBUG1;
INSERT INTO bigserial_test (x, y) SELECT x, y FROM bigserial_test;
DEBUG: volatile functions are not allowed in distributed INSERT ... SELECT queries
DEBUG: Collecting INSERT ... SELECT results on coordinator
INSERT INTO bigserial_test (x, y) SELECT a, a FROM reference_table;
DEBUG: volatile functions are not allowed in distributed INSERT ... SELECT queries
DEBUG: Collecting INSERT ... SELECT results on coordinator
INSERT INTO agg_events
(user_id)
SELECT f2.id FROM
(SELECT
id
FROM (SELECT users_ref_table.user_id AS id
FROM raw_events_first,
users_ref_table
WHERE raw_events_first.user_id = users_ref_table.user_id ) AS foo) as f
INNER JOIN
(SELECT v4,
v1,
id
FROM (SELECT SUM(raw_events_second.value_4) AS v4,
SUM(raw_events_first.value_1) AS v1,
raw_events_second.user_id AS id
FROM raw_events_first,
raw_events_second
WHERE raw_events_first.user_id = raw_events_second.user_id
GROUP BY raw_events_second.user_id
HAVING SUM(raw_events_second.value_4) > 1000) AS foo2 ) as f2
ON (f.id = f2.id)
WHERE f.id IN (SELECT user_id
FROM raw_events_second);
-- upsert with returning
INSERT INTO agg_events AS ae
(
user_id,
value_1_agg,
agg_time
)
SELECT user_id,
value_1,
time
FROM raw_events_first
ON conflict (user_id, value_1_agg)
DO UPDATE
SET agg_time = EXCLUDED.agg_time
WHERE ae.agg_time < EXCLUDED.agg_time
RETURNING user_id, value_1_agg;
user_id | value_1_agg
---------------------------------------------------------------------
1 | 10
2 | 20
3 | 30
4 | 40
5 | 50
6 | 60
(6 rows)
-- using a left join
INSERT INTO agg_events (user_id)
SELECT
raw_events_first.user_id
FROM
raw_events_first LEFT JOIN raw_events_second ON raw_events_first.user_id = raw_events_second.user_id
WHERE raw_events_second.user_id = 10 OR raw_events_second.user_id = 11;
INSERT INTO agg_events (user_id)
SELECT
users_ref_table.user_id
FROM
users_ref_table LEFT JOIN raw_events_second ON users_ref_table.user_id = raw_events_second.user_id
WHERE raw_events_second.user_id = 10 OR raw_events_second.user_id = 11;
INSERT INTO agg_events (user_id)
SELECT COALESCE(raw_events_first.user_id, users_ref_table.user_id)
FROM raw_events_first
RIGHT JOIN (users_ref_table LEFT JOIN raw_events_second ON users_ref_table.user_id = raw_events_second.user_id)
ON raw_events_first.user_id = users_ref_table.user_id;
-- using a full join
INSERT INTO agg_events (user_id, value_1_agg)
SELECT t1.user_id AS col1,
t2.user_id AS col2
FROM raw_events_first t1
FULL JOIN raw_events_second t2
ON t1.user_id = t2.user_id;
-- using semi join
INSERT INTO raw_events_second
(user_id)
SELECT user_id
FROM raw_events_first
WHERE user_id IN (SELECT raw_events_second.user_id
FROM raw_events_second, raw_events_first
WHERE raw_events_second.user_id = raw_events_first.user_id AND raw_events_first.user_id = 200);
-- using lateral join
INSERT INTO raw_events_second
(user_id)
SELECT user_id
FROM raw_events_first
WHERE NOT EXISTS (SELECT 1
FROM raw_events_second
WHERE raw_events_second.user_id =raw_events_first.user_id);
INSERT INTO raw_events_second
(user_id)
SELECT user_id
FROM users_ref_table
WHERE NOT EXISTS (SELECT 1
FROM raw_events_second
WHERE raw_events_second.user_id = users_ref_table.user_id);
-- using inner join
INSERT INTO agg_events (user_id)
SELECT raw_events_first.user_id
FROM raw_events_first INNER JOIN raw_events_second ON raw_events_first.user_id = raw_events_second.value_1
WHERE raw_events_first.value_1 IN (10, 11,12) OR raw_events_second.user_id IN (1,2,3,4);
INSERT INTO agg_events (user_id)
SELECT raw_events_first.user_id
FROM raw_events_first INNER JOIN users_ref_table ON raw_events_first.user_id = users_ref_table.user_id
WHERE raw_events_first.value_1 IN (10, 11,12) OR users_ref_table.user_id IN (1,2,3,4);
-- limit / offset clause
INSERT INTO agg_events (user_id) SELECT raw_events_first.user_id FROM raw_events_first LIMIT 1;
INSERT INTO agg_events (user_id) SELECT raw_events_first.user_id FROM raw_events_first OFFSET 1;
INSERT INTO agg_events (user_id) SELECT users_ref_table.user_id FROM users_ref_table LIMIT 1;
-- using a materialized cte
WITH cte AS MATERIALIZED
(SELECT max(value_1)+1 as v1_agg, user_id FROM raw_events_first GROUP BY user_id)
INSERT INTO agg_events (value_1_agg, user_id)
SELECT v1_agg, user_id FROM cte;
DEBUG: distributed INSERT ... SELECT can only select from distributed tables
DEBUG: Collecting INSERT ... SELECT results on coordinator
INSERT INTO raw_events_second
WITH cte AS MATERIALIZED (SELECT * FROM raw_events_first)
SELECT user_id * 1000, time, value_1, value_2, value_3, value_4 FROM cte;
INSERT INTO raw_events_second (user_id)
WITH cte AS MATERIALIZED (SELECT * FROM users_ref_table)
SELECT user_id FROM cte;
-- using a regular cte
WITH cte AS (SELECT * FROM raw_events_first)
INSERT INTO raw_events_second
SELECT user_id * 7000, time, value_1, value_2, value_3, value_4 FROM cte;
DEBUG: distributed INSERT ... SELECT can only select from distributed tables
DEBUG: CTE cte is going to be inlined via distributed planning
DEBUG: Collecting INSERT ... SELECT results on coordinator
INSERT INTO raw_events_second
WITH cte AS (SELECT * FROM raw_events_first)
SELECT * FROM cte;
DEBUG: CTE cte is going to be inlined via distributed planning
INSERT INTO agg_events
WITH sub_cte AS (SELECT 1)
SELECT
raw_events_first.user_id, (SELECT * FROM sub_cte)
FROM
raw_events_first;
DEBUG: CTE sub_cte is going to be inlined via distributed planning
-- we still support complex joins via INSERT's cte list ..
WITH cte AS (
SELECT DISTINCT(reference_table.a) AS a, 1 AS b
FROM distributed_table RIGHT JOIN reference_table USING (a)
)
INSERT INTO raw_events_second (user_id, value_1)
SELECT (a+5)*-1, b FROM cte;
DEBUG: distributed INSERT ... SELECT can only select from distributed tables
DEBUG: CTE cte is going to be inlined via distributed planning
DEBUG: recursively planning left side of the right join since the outer side is a recurring rel
DEBUG: recursively planning distributed relation "distributed_table" since it is part of a distributed join node that is outer joined with a recurring rel
DEBUG: Wrapping relation "distributed_table" to a subquery
DEBUG: generating subplan XXX_1 for subquery SELECT a FROM query_single_shard_table.distributed_table WHERE true
DEBUG: Plan XXX query after replacing subqueries and CTEs: SELECT "?column?" AS user_id, b AS value_1 FROM (SELECT ((cte.a OPERATOR(pg_catalog.+) 5) OPERATOR(pg_catalog.*) '-1'::integer), cte.b FROM (SELECT DISTINCT reference_table.a, 1 AS b FROM ((SELECT distributed_table_1.a, NULL::integer AS b FROM (SELECT intermediate_result.a FROM read_intermediate_result('XXX_1'::text, 'binary'::citus_copy_format) intermediate_result(a integer)) distributed_table_1) distributed_table RIGHT JOIN query_single_shard_table.reference_table USING (a))) cte) citus_insert_select_subquery("?column?", b)
DEBUG: Collecting INSERT ... SELECT results on coordinator
-- .. and via SELECT's cte list too
INSERT INTO raw_events_second (user_id, value_1)
WITH cte AS (
SELECT DISTINCT(reference_table.a) AS a, 1 AS b
FROM distributed_table RIGHT JOIN reference_table USING (a)
)
SELECT (a+5)*2, b FROM cte;
DEBUG: CTE cte is going to be inlined via distributed planning
DEBUG: cannot perform a lateral outer join when a distributed subquery references a reference table
DEBUG: recursively planning left side of the right join since the outer side is a recurring rel
DEBUG: recursively planning distributed relation "distributed_table" since it is part of a distributed join node that is outer joined with a recurring rel
DEBUG: Wrapping relation "distributed_table" to a subquery
DEBUG: generating subplan XXX_1 for subquery SELECT a FROM query_single_shard_table.distributed_table WHERE true
DEBUG: Plan XXX query after replacing subqueries and CTEs: SELECT ((a OPERATOR(pg_catalog.+) 5) OPERATOR(pg_catalog.*) 2) AS user_id, b AS value_1 FROM (SELECT DISTINCT reference_table.a, 1 AS b FROM ((SELECT distributed_table_1.a, NULL::integer AS b FROM (SELECT intermediate_result.a FROM read_intermediate_result('XXX_1'::text, 'binary'::citus_copy_format) intermediate_result(a integer)) distributed_table_1) distributed_table RIGHT JOIN query_single_shard_table.reference_table USING (a))) cte
DEBUG: Collecting INSERT ... SELECT results on coordinator
-- using set operations
INSERT INTO
raw_events_first(user_id)
(SELECT user_id FROM raw_events_first) INTERSECT
(SELECT user_id FROM raw_events_first);
INSERT INTO
raw_events_first(user_id)
(SELECT user_id FROM users_ref_table) INTERSECT
(SELECT user_id FROM raw_events_first);
-- group by clause inside subquery
INSERT INTO agg_events
(user_id)
SELECT f2.id FROM
(SELECT
id
FROM (SELECT raw_events_second.user_id AS id
FROM raw_events_first,
raw_events_second
WHERE raw_events_first.user_id = raw_events_second.user_id ) AS foo) as f
INNER JOIN
(SELECT v4,
v1,
id
FROM (SELECT SUM(raw_events_second.value_4) AS v4,
SUM(raw_events_first.value_1) AS v1,
raw_events_second.user_id AS id
FROM raw_events_first,
raw_events_second
WHERE raw_events_first.user_id = raw_events_second.user_id
GROUP BY raw_events_second.user_id
HAVING SUM(raw_events_second.value_4) > 1000) AS foo2 ) as f2
ON (f.id = f2.id)
WHERE f.id IN (SELECT user_id
FROM raw_events_second);
-- group by clause inside lateral subquery
INSERT INTO agg_events (user_id, value_4_agg)
SELECT
averages.user_id, avg(averages.value_4)
FROM
(SELECT
t1.user_id
FROM
raw_events_second t1 JOIN raw_events_second t2 on (t1.user_id = t2.user_id)
) reference_ids
JOIN LATERAL
(SELECT
user_id, value_4
FROM
raw_events_first) as averages ON averages.value_4 = reference_ids.user_id
GROUP BY averages.user_id;
-- using aggregates
INSERT INTO agg_events
(value_3_agg,
value_4_agg,
value_1_agg,
value_2_agg,
user_id)
SELECT SUM(value_3),
Count(value_4),
user_id,
SUM(value_1),
Avg(value_2)
FROM raw_events_first
GROUP BY user_id;
INSERT INTO agg_events (value_3_agg, value_1_agg)
SELECT AVG(user_id), SUM(user_id)
FROM users_ref_table
GROUP BY user_id;
-- using generate_series
INSERT INTO raw_events_first (user_id, value_1, value_2)
SELECT s, s, s FROM generate_series(1, 5) s;
DEBUG: distributed INSERT ... SELECT can only select from distributed tables
DEBUG: Collecting INSERT ... SELECT results on coordinator
CREATE SEQUENCE insert_select_test_seq;
-- nextval() expression in select's targetlist
INSERT INTO raw_events_first (user_id, value_1, value_2)
SELECT s, nextval('insert_select_test_seq'), (random()*10)::int
FROM generate_series(100, 105) s;
DEBUG: distributed INSERT ... SELECT can only select from distributed tables
DEBUG: Collecting INSERT ... SELECT results on coordinator
-- non-immutable function
INSERT INTO modify_fast_path (key, value_1) VALUES (2,1) RETURNING value_1, random() * key;
ERROR: non-IMMUTABLE functions are not allowed in the RETURNING clause
SET client_min_messages TO DEBUG2;
-- update / delete
UPDATE nullkey_c1_t1 SET a = 1 WHERE b = 5;
DEBUG: Distributed planning for a fast-path router query
DEBUG: Creating router plan
UPDATE nullkey_c1_t1 SET a = 1 WHERE a = 5;
DEBUG: Distributed planning for a fast-path router query
DEBUG: Creating router plan
UPDATE nullkey_c1_t1 SET a = random();
DEBUG: functions used in UPDATE queries on distributed tables must not be VOLATILE
ERROR: functions used in UPDATE queries on distributed tables must not be VOLATILE
UPDATE nullkey_c1_t1 SET a = 1 WHERE a = random();
DEBUG: functions used in the WHERE/ON/WHEN clause of modification queries on distributed tables must not be VOLATILE
ERROR: functions used in the WHERE/ON/WHEN clause of modification queries on distributed tables must not be VOLATILE
DELETE FROM nullkey_c1_t1 WHERE b = 5;
DEBUG: Distributed planning for a fast-path router query
DEBUG: Creating router plan
DELETE FROM nullkey_c1_t1 WHERE a = random();
DEBUG: functions used in the WHERE/ON/WHEN clause of modification queries on distributed tables must not be VOLATILE
ERROR: functions used in the WHERE/ON/WHEN clause of modification queries on distributed tables must not be VOLATILE
-- simple update queries between different table types / colocated tables
UPDATE nullkey_c1_t1 SET b = 5 FROM nullkey_c1_t2 WHERE nullkey_c1_t1.b = nullkey_c1_t2.b;
DEBUG: Creating router plan
UPDATE nullkey_c1_t1 SET b = 5 FROM nullkey_c2_t1 WHERE nullkey_c1_t1.b = nullkey_c2_t1.b;
DEBUG: router planner does not support queries that reference non-colocated distributed tables
ERROR: router planner does not support queries that reference non-colocated distributed tables
UPDATE nullkey_c1_t1 SET b = 5 FROM reference_table WHERE nullkey_c1_t1.b = reference_table.b;
DEBUG: Creating router plan
UPDATE nullkey_c1_t1 SET b = 5 FROM distributed_table WHERE nullkey_c1_t1.b = distributed_table.b;
DEBUG: complex joins are only supported when all distributed tables are co-located and joined on their distribution columns
ERROR: complex joins are only supported when all distributed tables are co-located and joined on their distribution columns
UPDATE nullkey_c1_t1 SET b = 5 FROM distributed_table WHERE nullkey_c1_t1.b = distributed_table.a;
DEBUG: complex joins are only supported when all distributed tables are co-located and joined on their distribution columns
ERROR: complex joins are only supported when all distributed tables are co-located and joined on their distribution columns
UPDATE nullkey_c1_t1 SET b = 5 FROM citus_local_table WHERE nullkey_c1_t1.b = citus_local_table.b;
DEBUG: local table citus_local_table cannot be joined with these distributed tables
DEBUG: Wrapping relation "citus_local_table" to a subquery
DEBUG: Distributed planning for a fast-path router query
DEBUG: Creating router plan
DEBUG: generating subplan XXX_1 for subquery SELECT b FROM query_single_shard_table.citus_local_table WHERE true
DEBUG: Plan XXX query after replacing subqueries and CTEs: UPDATE query_single_shard_table.nullkey_c1_t1 SET b = 5 FROM (SELECT NULL::integer AS a, citus_local_table_1.b FROM (SELECT intermediate_result.b FROM read_intermediate_result('XXX_1'::text, 'binary'::citus_copy_format) intermediate_result(b integer)) citus_local_table_1) citus_local_table WHERE (nullkey_c1_t1.b OPERATOR(pg_catalog.=) citus_local_table.b)
DEBUG: Creating router plan
UPDATE nullkey_c1_t1 SET b = 5 FROM postgres_local_table WHERE nullkey_c1_t1.b = postgres_local_table.b;
DEBUG: relation postgres_local_table is not distributed
DEBUG: Wrapping relation "postgres_local_table" to a subquery
DEBUG: generating subplan XXX_1 for subquery SELECT b FROM query_single_shard_table.postgres_local_table WHERE true
DEBUG: Plan XXX query after replacing subqueries and CTEs: UPDATE query_single_shard_table.nullkey_c1_t1 SET b = 5 FROM (SELECT NULL::integer AS a, postgres_local_table_1.b FROM (SELECT intermediate_result.b FROM read_intermediate_result('XXX_1'::text, 'binary'::citus_copy_format) intermediate_result(b integer)) postgres_local_table_1) postgres_local_table WHERE (nullkey_c1_t1.b OPERATOR(pg_catalog.=) postgres_local_table.b)
DEBUG: Creating router plan
UPDATE reference_table SET b = 5 FROM nullkey_c1_t1 WHERE nullkey_c1_t1.b = reference_table.b;
ERROR: cannot perform select on a distributed table and modify a reference table
UPDATE distributed_table SET b = 5 FROM nullkey_c1_t1 WHERE nullkey_c1_t1.b = distributed_table.b;
DEBUG: complex joins are only supported when all distributed tables are co-located and joined on their distribution columns
ERROR: complex joins are only supported when all distributed tables are co-located and joined on their distribution columns
UPDATE distributed_table SET b = 5 FROM nullkey_c1_t1 WHERE nullkey_c1_t1.b = distributed_table.a;
DEBUG: complex joins are only supported when all distributed tables are co-located and joined on their distribution columns
ERROR: complex joins are only supported when all distributed tables are co-located and joined on their distribution columns
UPDATE citus_local_table SET b = 5 FROM nullkey_c1_t1 WHERE nullkey_c1_t1.b = citus_local_table.b;
DEBUG: local table citus_local_table cannot be joined with these distributed tables
DEBUG: Wrapping relation "nullkey_c1_t1" to a subquery
DEBUG: Distributed planning for a fast-path router query
DEBUG: Creating router plan
DEBUG: generating subplan XXX_1 for subquery SELECT b FROM query_single_shard_table.nullkey_c1_t1 WHERE true
DEBUG: Plan XXX query after replacing subqueries and CTEs: UPDATE query_single_shard_table.citus_local_table SET b = 5 FROM (SELECT NULL::integer AS a, nullkey_c1_t1_1.b FROM (SELECT intermediate_result.b FROM read_intermediate_result('XXX_1'::text, 'binary'::citus_copy_format) intermediate_result(b integer)) nullkey_c1_t1_1) nullkey_c1_t1 WHERE (nullkey_c1_t1.b OPERATOR(pg_catalog.=) citus_local_table.b)
DEBUG: Creating router plan
UPDATE postgres_local_table SET b = 5 FROM nullkey_c1_t1 WHERE nullkey_c1_t1.b = postgres_local_table.b;
DEBUG: relation postgres_local_table is not distributed
DEBUG: Wrapping relation "nullkey_c1_t1" to a subquery
DEBUG: Distributed planning for a fast-path router query
DEBUG: Creating router plan
DEBUG: generating subplan XXX_1 for subquery SELECT b FROM query_single_shard_table.nullkey_c1_t1 WHERE true
DEBUG: Plan XXX query after replacing subqueries and CTEs: UPDATE query_single_shard_table.postgres_local_table SET b = 5 FROM (SELECT NULL::integer AS a, nullkey_c1_t1_1.b FROM (SELECT intermediate_result.b FROM read_intermediate_result('XXX_1'::text, 'binary'::citus_copy_format) intermediate_result(b integer)) nullkey_c1_t1_1) nullkey_c1_t1 WHERE (nullkey_c1_t1.b OPERATOR(pg_catalog.=) postgres_local_table.b)
DEBUG: Creating router plan
-- simple delete queries between different table types / colocated tables
DELETE FROM nullkey_c1_t1 USING nullkey_c1_t2 WHERE nullkey_c1_t1.b = nullkey_c1_t2.b;
DEBUG: Creating router plan
DELETE FROM nullkey_c1_t1 USING nullkey_c2_t1 WHERE nullkey_c1_t1.b = nullkey_c2_t1.b;
DEBUG: router planner does not support queries that reference non-colocated distributed tables
ERROR: router planner does not support queries that reference non-colocated distributed tables
DELETE FROM nullkey_c1_t1 USING reference_table WHERE nullkey_c1_t1.b = reference_table.b;
DEBUG: Creating router plan
DELETE FROM nullkey_c1_t1 USING distributed_table WHERE nullkey_c1_t1.b = distributed_table.b;
DEBUG: complex joins are only supported when all distributed tables are co-located and joined on their distribution columns
ERROR: complex joins are only supported when all distributed tables are co-located and joined on their distribution columns
DELETE FROM nullkey_c1_t1 USING distributed_table WHERE nullkey_c1_t1.b = distributed_table.a;
DEBUG: complex joins are only supported when all distributed tables are co-located and joined on their distribution columns
ERROR: complex joins are only supported when all distributed tables are co-located and joined on their distribution columns
DELETE FROM nullkey_c1_t1 USING citus_local_table WHERE nullkey_c1_t1.b = citus_local_table.b;
DEBUG: local table citus_local_table cannot be joined with these distributed tables
DEBUG: Wrapping relation "citus_local_table" to a subquery
DEBUG: Distributed planning for a fast-path router query
DEBUG: Creating router plan
DEBUG: generating subplan XXX_1 for subquery SELECT b FROM query_single_shard_table.citus_local_table WHERE true
DEBUG: Plan XXX query after replacing subqueries and CTEs: DELETE FROM query_single_shard_table.nullkey_c1_t1 USING (SELECT NULL::integer AS a, citus_local_table_1.b FROM (SELECT intermediate_result.b FROM read_intermediate_result('XXX_1'::text, 'binary'::citus_copy_format) intermediate_result(b integer)) citus_local_table_1) citus_local_table WHERE (nullkey_c1_t1.b OPERATOR(pg_catalog.=) citus_local_table.b)
DEBUG: Creating router plan
DELETE FROM nullkey_c1_t1 USING postgres_local_table WHERE nullkey_c1_t1.b = postgres_local_table.b;
DEBUG: relation postgres_local_table is not distributed
DEBUG: Wrapping relation "postgres_local_table" to a subquery
DEBUG: generating subplan XXX_1 for subquery SELECT b FROM query_single_shard_table.postgres_local_table WHERE true
DEBUG: Plan XXX query after replacing subqueries and CTEs: DELETE FROM query_single_shard_table.nullkey_c1_t1 USING (SELECT NULL::integer AS a, postgres_local_table_1.b FROM (SELECT intermediate_result.b FROM read_intermediate_result('XXX_1'::text, 'binary'::citus_copy_format) intermediate_result(b integer)) postgres_local_table_1) postgres_local_table WHERE (nullkey_c1_t1.b OPERATOR(pg_catalog.=) postgres_local_table.b)
DEBUG: Creating router plan
DELETE FROM reference_table USING nullkey_c1_t1 WHERE nullkey_c1_t1.b = reference_table.b;
ERROR: cannot perform select on a distributed table and modify a reference table
DELETE FROM distributed_table USING nullkey_c1_t1 WHERE nullkey_c1_t1.b = distributed_table.b;
DEBUG: complex joins are only supported when all distributed tables are co-located and joined on their distribution columns
ERROR: complex joins are only supported when all distributed tables are co-located and joined on their distribution columns
DELETE FROM distributed_table USING nullkey_c1_t1 WHERE nullkey_c1_t1.b = distributed_table.a;
DEBUG: complex joins are only supported when all distributed tables are co-located and joined on their distribution columns
ERROR: complex joins are only supported when all distributed tables are co-located and joined on their distribution columns
DELETE FROM citus_local_table USING nullkey_c1_t1 WHERE nullkey_c1_t1.b = citus_local_table.b;
DEBUG: local table citus_local_table cannot be joined with these distributed tables
DEBUG: Wrapping relation "nullkey_c1_t1" to a subquery
DEBUG: Distributed planning for a fast-path router query
DEBUG: Creating router plan
DEBUG: generating subplan XXX_1 for subquery SELECT b FROM query_single_shard_table.nullkey_c1_t1 WHERE true
DEBUG: Plan XXX query after replacing subqueries and CTEs: DELETE FROM query_single_shard_table.citus_local_table USING (SELECT NULL::integer AS a, nullkey_c1_t1_1.b FROM (SELECT intermediate_result.b FROM read_intermediate_result('XXX_1'::text, 'binary'::citus_copy_format) intermediate_result(b integer)) nullkey_c1_t1_1) nullkey_c1_t1 WHERE (nullkey_c1_t1.b OPERATOR(pg_catalog.=) citus_local_table.b)
DEBUG: Creating router plan
DELETE FROM postgres_local_table USING nullkey_c1_t1 WHERE nullkey_c1_t1.b = postgres_local_table.b;
DEBUG: relation postgres_local_table is not distributed
DEBUG: Wrapping relation "nullkey_c1_t1" to a subquery
DEBUG: Distributed planning for a fast-path router query
DEBUG: Creating router plan
DEBUG: generating subplan XXX_1 for subquery SELECT b FROM query_single_shard_table.nullkey_c1_t1 WHERE true
DEBUG: Plan XXX query after replacing subqueries and CTEs: DELETE FROM query_single_shard_table.postgres_local_table USING (SELECT NULL::integer AS a, nullkey_c1_t1_1.b FROM (SELECT intermediate_result.b FROM read_intermediate_result('XXX_1'::text, 'binary'::citus_copy_format) intermediate_result(b integer)) nullkey_c1_t1_1) nullkey_c1_t1 WHERE (nullkey_c1_t1.b OPERATOR(pg_catalog.=) postgres_local_table.b)
DEBUG: Creating router plan
-- slightly more complex update queries
UPDATE nullkey_c1_t1 SET b = 5 WHERE nullkey_c1_t1.b IN (SELECT b FROM distributed_table);
DEBUG: complex joins are only supported when all distributed tables are co-located and joined on their distribution columns
DEBUG: Router planner cannot handle multi-shard select queries
DEBUG: generating subplan XXX_1 for subquery SELECT b FROM query_single_shard_table.distributed_table
DEBUG: Plan XXX query after replacing subqueries and CTEs: UPDATE query_single_shard_table.nullkey_c1_t1 SET b = 5 WHERE (b OPERATOR(pg_catalog.=) ANY (SELECT intermediate_result.b FROM read_intermediate_result('XXX_1'::text, 'binary'::citus_copy_format) intermediate_result(b integer)))
DEBUG: Creating router plan
WITH cte AS materialized(
SELECT * FROM distributed_table
)
UPDATE nullkey_c1_t1 SET b = 5 FROM cte WHERE nullkey_c1_t1.b = cte.a;
DEBUG: complex joins are only supported when all distributed tables are co-located and joined on their distribution columns
DEBUG: generating subplan XXX_1 for CTE cte: SELECT a, b FROM query_single_shard_table.distributed_table
DEBUG: Router planner cannot handle multi-shard select queries
DEBUG: Plan XXX query after replacing subqueries and CTEs: UPDATE query_single_shard_table.nullkey_c1_t1 SET b = 5 FROM (SELECT intermediate_result.a, intermediate_result.b FROM read_intermediate_result('XXX_1'::text, 'binary'::citus_copy_format) intermediate_result(a integer, b integer)) cte WHERE (nullkey_c1_t1.b OPERATOR(pg_catalog.=) cte.a)
DEBUG: Creating router plan
WITH cte AS (
SELECT reference_table.a AS a, 1 AS b
FROM distributed_table RIGHT JOIN reference_table USING (a)
)
UPDATE nullkey_c1_t1 SET b = 5 WHERE nullkey_c1_t1.b IN (SELECT b FROM cte);
DEBUG: complex joins are only supported when all distributed tables are co-located and joined on their distribution columns
DEBUG: generating subplan XXX_1 for CTE cte: SELECT reference_table.a, 1 AS b FROM (query_single_shard_table.distributed_table RIGHT JOIN query_single_shard_table.reference_table USING (a))
DEBUG: Router planner cannot handle multi-shard select queries
DEBUG: recursively planning left side of the right join since the outer side is a recurring rel
DEBUG: recursively planning distributed relation "distributed_table" since it is part of a distributed join node that is outer joined with a recurring rel
DEBUG: Wrapping relation "distributed_table" to a subquery
DEBUG: Router planner cannot handle multi-shard select queries
DEBUG: generating subplan XXX_1 for subquery SELECT a FROM query_single_shard_table.distributed_table WHERE true
DEBUG: Plan XXX query after replacing subqueries and CTEs: SELECT reference_table.a, 1 AS b FROM ((SELECT distributed_table_1.a, NULL::integer AS b FROM (SELECT intermediate_result.a FROM read_intermediate_result('XXX_1'::text, 'binary'::citus_copy_format) intermediate_result(a integer)) distributed_table_1) distributed_table RIGHT JOIN query_single_shard_table.reference_table USING (a))
DEBUG: Creating router plan
DEBUG: Plan XXX query after replacing subqueries and CTEs: UPDATE query_single_shard_table.nullkey_c1_t1 SET b = 5 WHERE (b OPERATOR(pg_catalog.=) ANY (SELECT cte.b FROM (SELECT intermediate_result.a, intermediate_result.b FROM read_intermediate_result('XXX_1'::text, 'binary'::citus_copy_format) intermediate_result(a integer, b integer)) cte))
DEBUG: Creating router plan
UPDATE nullkey_c1_t1 SET b = 5 FROM reference_table WHERE EXISTS (
SELECT 1 FROM reference_table LEFT JOIN nullkey_c1_t1 USING (a) WHERE nullkey_c1_t1.b IS NULL
);
DEBUG: Creating router plan
UPDATE nullkey_c1_t1 tx SET b = (
SELECT nullkey_c1_t2.b FROM nullkey_c1_t2 JOIN nullkey_c1_t1 ON (nullkey_c1_t1.a != nullkey_c1_t2.a) WHERE nullkey_c1_t1.a = tx.a ORDER BY 1 LIMIT 1
);
DEBUG: Creating router plan
UPDATE nullkey_c1_t1 tx SET b = t2.b FROM nullkey_c1_t1 t1 JOIN nullkey_c1_t2 t2 ON (t1.a = t2.a);
DEBUG: Creating router plan
WITH cte AS (
SELECT * FROM nullkey_c1_t2 ORDER BY 1,2 LIMIT 10
)
UPDATE nullkey_c1_t1 SET b = 5 WHERE nullkey_c1_t1.a IN (SELECT b FROM cte);
DEBUG: Creating router plan
UPDATE modify_fast_path SET value_1 = value_1 + 12 * value_1 WHERE key = 1;
DEBUG: Distributed planning for a fast-path router query
DEBUG: Creating router plan
UPDATE modify_fast_path SET value_1 = NULL WHERE value_1 = 15 AND (key = 1 OR value_2 = 'citus');
DEBUG: Distributed planning for a fast-path router query
DEBUG: Creating router plan
UPDATE modify_fast_path SET value_1 = 5 WHERE key = 2 RETURNING value_1 * 15, value_1::numeric * 16;
DEBUG: Distributed planning for a fast-path router query
DEBUG: Creating router plan
?column? | ?column?
---------------------------------------------------------------------
(0 rows)
UPDATE modify_fast_path
SET value_1 = 1
FROM modify_fast_path_reference
WHERE
modify_fast_path.key = modify_fast_path_reference.key AND
modify_fast_path.key = 1 AND
modify_fast_path_reference.key = 1;
DEBUG: Creating router plan
PREPARE p1 (int, int, int) AS
UPDATE modify_fast_path SET value_1 = value_1 + $1 WHERE key = $2 AND value_1 = $3;
EXECUTE p1(1,1,1);
DEBUG: Distributed planning for a fast-path router query
DEBUG: Creating router plan
EXECUTE p1(2,2,2);
DEBUG: Distributed planning for a fast-path router query
DEBUG: Creating router plan
EXECUTE p1(3,3,3);
DEBUG: Distributed planning for a fast-path router query
DEBUG: Creating router plan
EXECUTE p1(4,4,4);
DEBUG: Distributed planning for a fast-path router query
DEBUG: Creating router plan
EXECUTE p1(5,5,5);
DEBUG: Distributed planning for a fast-path router query
DEBUG: Creating router plan
EXECUTE p1(6,6,6);
DEBUG: Distributed planning for a fast-path router query
DEBUG: Creating router plan
EXECUTE p1(7,7,7);
PREPARE prepared_zero_shard_update(int) AS UPDATE modify_fast_path SET value_1 = 1 WHERE key = $1 AND false;
EXECUTE prepared_zero_shard_update(1);
DEBUG: Distributed planning for a fast-path router query
DEBUG: Creating router plan
EXECUTE prepared_zero_shard_update(2);
DEBUG: Distributed planning for a fast-path router query
DEBUG: Creating router plan
EXECUTE prepared_zero_shard_update(3);
DEBUG: Distributed planning for a fast-path router query
DEBUG: Creating router plan
EXECUTE prepared_zero_shard_update(4);
DEBUG: Distributed planning for a fast-path router query
DEBUG: Creating router plan
EXECUTE prepared_zero_shard_update(5);
DEBUG: Distributed planning for a fast-path router query
DEBUG: Creating router plan
EXECUTE prepared_zero_shard_update(6);
DEBUG: Distributed planning for a fast-path router query
DEBUG: Creating router plan
EXECUTE prepared_zero_shard_update(7);
-- slightly more complex delete queries
DELETE FROM nullkey_c1_t1 WHERE nullkey_c1_t1.b IN (SELECT b FROM distributed_table);
DEBUG: complex joins are only supported when all distributed tables are co-located and joined on their distribution columns
DEBUG: Router planner cannot handle multi-shard select queries
DEBUG: generating subplan XXX_1 for subquery SELECT b FROM query_single_shard_table.distributed_table
DEBUG: Plan XXX query after replacing subqueries and CTEs: DELETE FROM query_single_shard_table.nullkey_c1_t1 WHERE (b OPERATOR(pg_catalog.=) ANY (SELECT intermediate_result.b FROM read_intermediate_result('XXX_1'::text, 'binary'::citus_copy_format) intermediate_result(b integer)))
DEBUG: Creating router plan
WITH cte AS materialized(
SELECT * FROM distributed_table
)
DELETE FROM nullkey_c1_t1 USING cte WHERE nullkey_c1_t1.b = cte.a;
DEBUG: complex joins are only supported when all distributed tables are co-located and joined on their distribution columns
DEBUG: generating subplan XXX_1 for CTE cte: SELECT a, b FROM query_single_shard_table.distributed_table
DEBUG: Router planner cannot handle multi-shard select queries
DEBUG: Plan XXX query after replacing subqueries and CTEs: DELETE FROM query_single_shard_table.nullkey_c1_t1 USING (SELECT intermediate_result.a, intermediate_result.b FROM read_intermediate_result('XXX_1'::text, 'binary'::citus_copy_format) intermediate_result(a integer, b integer)) cte WHERE (nullkey_c1_t1.b OPERATOR(pg_catalog.=) cte.a)
DEBUG: Creating router plan
WITH cte AS (
SELECT reference_table.a AS a, 1 AS b
FROM distributed_table RIGHT JOIN reference_table USING (a)
)
DELETE FROM nullkey_c1_t1 WHERE nullkey_c1_t1.b IN (SELECT b FROM cte);
DEBUG: complex joins are only supported when all distributed tables are co-located and joined on their distribution columns
DEBUG: generating subplan XXX_1 for CTE cte: SELECT reference_table.a, 1 AS b FROM (query_single_shard_table.distributed_table RIGHT JOIN query_single_shard_table.reference_table USING (a))
DEBUG: Router planner cannot handle multi-shard select queries
DEBUG: recursively planning left side of the right join since the outer side is a recurring rel
DEBUG: recursively planning distributed relation "distributed_table" since it is part of a distributed join node that is outer joined with a recurring rel
DEBUG: Wrapping relation "distributed_table" to a subquery
DEBUG: Router planner cannot handle multi-shard select queries
DEBUG: generating subplan XXX_1 for subquery SELECT a FROM query_single_shard_table.distributed_table WHERE true
DEBUG: Plan XXX query after replacing subqueries and CTEs: SELECT reference_table.a, 1 AS b FROM ((SELECT distributed_table_1.a, NULL::integer AS b FROM (SELECT intermediate_result.a FROM read_intermediate_result('XXX_1'::text, 'binary'::citus_copy_format) intermediate_result(a integer)) distributed_table_1) distributed_table RIGHT JOIN query_single_shard_table.reference_table USING (a))
DEBUG: Creating router plan
DEBUG: Plan XXX query after replacing subqueries and CTEs: DELETE FROM query_single_shard_table.nullkey_c1_t1 WHERE (b OPERATOR(pg_catalog.=) ANY (SELECT cte.b FROM (SELECT intermediate_result.a, intermediate_result.b FROM read_intermediate_result('XXX_1'::text, 'binary'::citus_copy_format) intermediate_result(a integer, b integer)) cte))
DEBUG: Creating router plan
DELETE FROM nullkey_c1_t1 USING reference_table WHERE EXISTS (
SELECT 1 FROM reference_table LEFT JOIN nullkey_c1_t1 USING (a) WHERE nullkey_c1_t1.b IS NULL
);
DEBUG: Creating router plan
DELETE FROM nullkey_c1_t1 tx USING nullkey_c1_t1 t1 JOIN nullkey_c1_t2 t2 ON (t1.a = t2.a);
DEBUG: Creating router plan
WITH cte AS (
SELECT * FROM nullkey_c1_t2 ORDER BY 1,2 LIMIT 10
)
DELETE FROM nullkey_c1_t1 WHERE nullkey_c1_t1.a IN (SELECT b FROM cte);
DEBUG: Creating router plan
DELETE FROM modify_fast_path WHERE value_1 = 15 AND (key = 1 OR value_2 = 'citus');
DEBUG: Distributed planning for a fast-path router query
DEBUG: Creating router plan
DELETE FROM modify_fast_path WHERE key = 2 RETURNING value_1 * 15, value_1::numeric * 16;
DEBUG: Distributed planning for a fast-path router query
DEBUG: Creating router plan
?column? | ?column?
---------------------------------------------------------------------
(0 rows)
DELETE FROM modify_fast_path
USING modify_fast_path_reference
WHERE
modify_fast_path.key = modify_fast_path_reference.key AND
modify_fast_path.key = 1 AND
modify_fast_path_reference.key = 1;
DEBUG: Creating router plan
PREPARE p2 (int, int, int) AS
DELETE FROM modify_fast_path WHERE key = ($2)*$1 AND value_1 = $3;
EXECUTE p2(1,1,1);
DEBUG: Distributed planning for a fast-path router query
DEBUG: Creating router plan
EXECUTE p2(2,2,2);
DEBUG: Distributed planning for a fast-path router query
DEBUG: Creating router plan
EXECUTE p2(3,3,3);
DEBUG: Distributed planning for a fast-path router query
DEBUG: Creating router plan
EXECUTE p2(4,4,4);
DEBUG: Distributed planning for a fast-path router query
DEBUG: Creating router plan
EXECUTE p2(5,5,5);
DEBUG: Distributed planning for a fast-path router query
DEBUG: Creating router plan
EXECUTE p2(6,6,6);
DEBUG: Distributed planning for a fast-path router query
DEBUG: Creating router plan
EXECUTE p2(7,7,7);
PREPARE prepared_zero_shard_delete(int) AS DELETE FROM modify_fast_path WHERE key = $1 AND false;
EXECUTE prepared_zero_shard_delete(1);
DEBUG: Distributed planning for a fast-path router query
DEBUG: Creating router plan
EXECUTE prepared_zero_shard_delete(2);
DEBUG: Distributed planning for a fast-path router query
DEBUG: Creating router plan
EXECUTE prepared_zero_shard_delete(3);
DEBUG: Distributed planning for a fast-path router query
DEBUG: Creating router plan
EXECUTE prepared_zero_shard_delete(4);
DEBUG: Distributed planning for a fast-path router query
DEBUG: Creating router plan
EXECUTE prepared_zero_shard_delete(5);
DEBUG: Distributed planning for a fast-path router query
DEBUG: Creating router plan
EXECUTE prepared_zero_shard_delete(6);
DEBUG: Distributed planning for a fast-path router query
DEBUG: Creating router plan
EXECUTE prepared_zero_shard_delete(7);
-- test modifying ctes
WITH cte AS (
UPDATE modify_fast_path SET value_1 = value_1 + 1 WHERE key = 1 RETURNING *
)
SELECT * FROM cte;
DEBUG: Creating router plan
key | value_1 | value_2
---------------------------------------------------------------------
(0 rows)
WITH cte AS (
DELETE FROM modify_fast_path WHERE key = 1 RETURNING *
)
SELECT * FROM modify_fast_path;
DEBUG: Creating router plan
key | value_1 | value_2
---------------------------------------------------------------------
(0 rows)
WITH cte AS (
DELETE FROM modify_fast_path WHERE key = 1 RETURNING *
)
SELECT * FROM modify_fast_path_reference WHERE key IN (SELECT key FROM cte);
DEBUG: Creating router plan
key | value_1 | value_2
---------------------------------------------------------------------
(0 rows)
WITH cte AS (
DELETE FROM reference_table WHERE a = 1 RETURNING *
)
SELECT * FROM nullkey_c1_t1 WHERE a IN (SELECT a FROM cte);
DEBUG: cannot router plan modification of a non-distributed table
DEBUG: generating subplan XXX_1 for CTE cte: DELETE FROM query_single_shard_table.reference_table WHERE (a OPERATOR(pg_catalog.=) 1) RETURNING a, b
DEBUG: Distributed planning for a fast-path router query
DEBUG: Creating router plan
DEBUG: Plan XXX query after replacing subqueries and CTEs: SELECT a, b FROM query_single_shard_table.nullkey_c1_t1 WHERE (a OPERATOR(pg_catalog.=) ANY (SELECT cte.a FROM (SELECT intermediate_result.a, intermediate_result.b FROM read_intermediate_result('XXX_1'::text, 'binary'::citus_copy_format) intermediate_result(a integer, b integer)) cte))
DEBUG: Creating router plan
a | b
---------------------------------------------------------------------
(0 rows)
WITH cte AS (
DELETE FROM nullkey_c1_t1 WHERE a = 1 RETURNING *
)
SELECT * FROM nullkey_c1_t2 WHERE a IN (SELECT a FROM cte);
DEBUG: Creating router plan
a | b
---------------------------------------------------------------------
(0 rows)
WITH cte AS (
DELETE FROM nullkey_c1_t1 WHERE a = 1 RETURNING *
)
SELECT * FROM nullkey_c2_t1 WHERE a IN (SELECT a FROM cte);
DEBUG: router planner does not support queries that reference non-colocated distributed tables
DEBUG: generating subplan XXX_1 for CTE cte: DELETE FROM query_single_shard_table.nullkey_c1_t1 WHERE (a OPERATOR(pg_catalog.=) 1) RETURNING a, b
DEBUG: Distributed planning for a fast-path router query
DEBUG: Creating router plan
DEBUG: Plan XXX query after replacing subqueries and CTEs: SELECT a, b FROM query_single_shard_table.nullkey_c2_t1 WHERE (a OPERATOR(pg_catalog.=) ANY (SELECT cte.a FROM (SELECT intermediate_result.a, intermediate_result.b FROM read_intermediate_result('XXX_1'::text, 'binary'::citus_copy_format) intermediate_result(a integer, b integer)) cte))
DEBUG: Creating router plan
a | b
---------------------------------------------------------------------
(0 rows)
WITH cte AS (
DELETE FROM nullkey_c1_t1 WHERE a = 1 RETURNING *
)
SELECT * FROM distributed_table WHERE a IN (SELECT a FROM cte);
DEBUG: Router planner cannot handle multi-shard select queries
DEBUG: generating subplan XXX_1 for CTE cte: DELETE FROM query_single_shard_table.nullkey_c1_t1 WHERE (a OPERATOR(pg_catalog.=) 1) RETURNING a, b
DEBUG: Distributed planning for a fast-path router query
DEBUG: Creating router plan
DEBUG: Plan XXX query after replacing subqueries and CTEs: SELECT a, b FROM query_single_shard_table.distributed_table WHERE (a OPERATOR(pg_catalog.=) ANY (SELECT cte.a FROM (SELECT intermediate_result.a, intermediate_result.b FROM read_intermediate_result('XXX_1'::text, 'binary'::citus_copy_format) intermediate_result(a integer, b integer)) cte))
DEBUG: Router planner cannot handle multi-shard select queries
a | b
---------------------------------------------------------------------
(0 rows)
-- Below two queries fail very late when
-- citus.enable_non_colocated_router_query_pushdown is set to on.
SET citus.enable_non_colocated_router_query_pushdown TO ON;
WITH cte AS (
DELETE FROM distributed_table WHERE a = 1 RETURNING *
)
SELECT * FROM nullkey_c1_t1 WHERE a IN (SELECT a FROM cte);
DEBUG: Creating router plan
DEBUG: query has a single distribution column value: 1
ERROR: relation "query_single_shard_table.nullkey_c1_t1_1620000" does not exist
CONTEXT: while executing command on localhost:xxxxx
WITH cte AS (
DELETE FROM distributed_table WHERE a = 1 RETURNING *
)
SELECT * FROM nullkey_c1_t1 WHERE b IN (SELECT b FROM cte);
DEBUG: Creating router plan
DEBUG: query has a single distribution column value: 1
ERROR: relation "query_single_shard_table.nullkey_c1_t1_1620000" does not exist
CONTEXT: while executing command on localhost:xxxxx
SET citus.enable_non_colocated_router_query_pushdown TO OFF;
WITH cte AS (
DELETE FROM distributed_table WHERE a = 1 RETURNING *
)
SELECT * FROM nullkey_c1_t1 WHERE a IN (SELECT a FROM cte);
DEBUG: router planner does not support queries that reference non-colocated distributed tables
DEBUG: generating subplan XXX_1 for CTE cte: DELETE FROM query_single_shard_table.distributed_table WHERE (a OPERATOR(pg_catalog.=) 1) RETURNING a, b
DEBUG: Distributed planning for a fast-path router query
DEBUG: Creating router plan
DEBUG: query has a single distribution column value: 1
DEBUG: Plan XXX query after replacing subqueries and CTEs: SELECT a, b FROM query_single_shard_table.nullkey_c1_t1 WHERE (a OPERATOR(pg_catalog.=) ANY (SELECT cte.a FROM (SELECT intermediate_result.a, intermediate_result.b FROM read_intermediate_result('XXX_1'::text, 'binary'::citus_copy_format) intermediate_result(a integer, b integer)) cte))
DEBUG: Creating router plan
a | b
---------------------------------------------------------------------
(0 rows)
WITH cte AS (
DELETE FROM distributed_table WHERE a = 1 RETURNING *
)
SELECT * FROM nullkey_c1_t1 WHERE b IN (SELECT b FROM cte);
DEBUG: router planner does not support queries that reference non-colocated distributed tables
DEBUG: generating subplan XXX_1 for CTE cte: DELETE FROM query_single_shard_table.distributed_table WHERE (a OPERATOR(pg_catalog.=) 1) RETURNING a, b
DEBUG: Distributed planning for a fast-path router query
DEBUG: Creating router plan
DEBUG: query has a single distribution column value: 1
DEBUG: Plan XXX query after replacing subqueries and CTEs: SELECT a, b FROM query_single_shard_table.nullkey_c1_t1 WHERE (b OPERATOR(pg_catalog.=) ANY (SELECT cte.b FROM (SELECT intermediate_result.a, intermediate_result.b FROM read_intermediate_result('XXX_1'::text, 'binary'::citus_copy_format) intermediate_result(a integer, b integer)) cte))
DEBUG: Creating router plan
a | b
---------------------------------------------------------------------
(0 rows)
RESET citus.enable_non_colocated_router_query_pushdown;
WITH cte AS (
UPDATE modify_fast_path SET value_1 = value_1 + 1 WHERE key = 1 RETURNING *
)
UPDATE modify_fast_path SET value_1 = value_1 + 1 WHERE key = 1;
DEBUG: Creating router plan
WITH cte AS (
DELETE FROM modify_fast_path WHERE key = 1 RETURNING *
)
DELETE FROM modify_fast_path WHERE key = 1;
DEBUG: Creating router plan
-- test window functions
SELECT
user_id, avg(avg(value_3)) OVER (PARTITION BY user_id, MIN(value_2))
FROM
raw_events_first
GROUP BY
1
ORDER BY
2 DESC NULLS LAST, 1 DESC;
DEBUG: Distributed planning for a fast-path router query
DEBUG: Creating router plan
user_id | avg
---------------------------------------------------------------------
6 | 6000.1
5 | 5000.1
4 | 4000.1
3 | 3000.1
2 | 2000.1
1 | 1000.1
105 |
104 |
103 |
102 |
101 |
100 |
(12 rows)
SELECT
user_id, max(value_1) OVER (PARTITION BY user_id, MIN(value_2))
FROM (
SELECT
DISTINCT us.user_id, us.value_2, us.value_1, random() as r1
FROM
raw_events_first as us, raw_events_second
WHERE
us.user_id = raw_events_second.user_id
ORDER BY
user_id, value_2
) s
GROUP BY
1, value_1
ORDER BY
2 DESC, 1;
DEBUG: Creating router plan
user_id | max
---------------------------------------------------------------------
1 |
2 |
3 |
4 |
5 |
6 |
6 | 60
5 | 50
4 | 40
3 | 30
2 | 20
1 | 10
5 | 5
4 | 4
3 | 3
2 | 2
1 | 1
(17 rows)
SELECT
DISTINCT ON (raw_events_second.user_id, rnk) raw_events_second.user_id, rank() OVER my_win AS rnk
FROM
raw_events_second, raw_events_first
WHERE
raw_events_first.user_id = raw_events_second.user_id
WINDOW
my_win AS (PARTITION BY raw_events_second.user_id, raw_events_first.value_1 ORDER BY raw_events_second.time DESC)
ORDER BY
rnk DESC, 1 DESC
LIMIT 10;
DEBUG: Creating router plan
user_id | rnk
---------------------------------------------------------------------
6 | 2
5 | 2
4 | 2
3 | 2
2 | 2
1 | 2
6 | 1
5 | 1
4 | 1
3 | 1
(10 rows)
-- more tests with ctes and subqueries
-- CTEs are recursively planned, and subquery foo is also recursively planned.
-- Then the final plan becomes a router plan.
WITH cte AS MATERIALIZED (
WITH local_cte AS MATERIALIZED (
SELECT * FROM users_table_local
),
dist_cte AS MATERIALIZED (
SELECT user_id FROM colocated_events_table
)
SELECT dist_cte.user_id FROM local_cte JOIN dist_cte ON dist_cte.user_id=local_cte.user_id
)
SELECT count(*)
FROM cte,
(
SELECT DISTINCT users_table.user_id
FROM users_table, colocated_events_table
WHERE users_table.user_id = colocated_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;
DEBUG: Local tables cannot be used in distributed queries.
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 query_single_shard_table.users_table_local), dist_cte AS MATERIALIZED (SELECT colocated_events_table.user_id FROM query_single_shard_table.colocated_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: Local tables cannot be used in distributed queries.
DEBUG: generating subplan XXX_1 for CTE local_cte: SELECT user_id, "time", value_1, value_2, value_3, value_4 FROM query_single_shard_table.users_table_local
DEBUG: generating subplan XXX_2 for CTE dist_cte: SELECT user_id FROM query_single_shard_table.colocated_events_table
DEBUG: Distributed planning for a fast-path router query
DEBUG: Creating router plan
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: Creating router plan
DEBUG: Creating router plan
DEBUG: generating subplan XXX_2 for subquery SELECT DISTINCT users_table.user_id FROM query_single_shard_table.users_table, query_single_shard_table.colocated_events_table WHERE ((users_table.user_id OPERATOR(pg_catalog.=) colocated_events_table.user_id) AND (colocated_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 count(*) AS count FROM (SELECT intermediate_result.user_id FROM read_intermediate_result('XXX_1'::text, 'binary'::citus_copy_format) intermediate_result(user_id integer)) cte, (SELECT intermediate_result.user_id FROM read_intermediate_result('XXX_2'::text, 'binary'::citus_copy_format) intermediate_result(user_id integer)) foo WHERE (foo.user_id OPERATOR(pg_catalog.=) cte.user_id)
DEBUG: Creating router plan
count
---------------------------------------------------------------------
1644
(1 row)
-- CTEs are colocated, route entire query.
WITH cte1 AS (
SELECT * FROM users_table WHERE user_id = 1
), cte2 AS (
SELECT * FROM colocated_events_table WHERE user_id = 1
)
SELECT cte1.user_id, cte1.value_1, cte2.user_id, cte2.event_type
FROM cte1, cte2
ORDER BY cte1.user_id, cte1.value_1, cte2.user_id, cte2.event_type
LIMIT 5;
DEBUG: CTE cte1 is going to be inlined via distributed planning
DEBUG: CTE cte2 is going to be inlined via distributed planning
DEBUG: Creating router plan
user_id | value_1 | user_id | event_type
---------------------------------------------------------------------
1 | 1 | 1 | 0
1 | 1 | 1 | 0
1 | 1 | 1 | 1
1 | 1 | 1 | 1
1 | 1 | 1 | 2
(5 rows)
-- CTEs aren't colocated, CTEs become intermediate results.
WITH cte1 AS MATERIALIZED (
SELECT * FROM users_table WHERE user_id = 1
), cte2 AS MATERIALIZED (
SELECT * FROM non_colocated_events_table WHERE user_id = 6
)
SELECT cte1.user_id, cte1.value_1, cte2.user_id, cte2.user_id
FROM cte1, cte2
ORDER BY cte1.user_id, cte1.value_1, cte2.user_id, cte2.event_type
LIMIT 5;
DEBUG: router planner does not support queries that reference non-colocated distributed tables
DEBUG: generating subplan XXX_1 for CTE cte1: SELECT user_id, "time", value_1, value_2, value_3, value_4 FROM query_single_shard_table.users_table WHERE (user_id OPERATOR(pg_catalog.=) 1)
DEBUG: Distributed planning for a fast-path router query
DEBUG: Creating router plan
DEBUG: generating subplan XXX_2 for CTE cte2: SELECT user_id, "time", event_type, value_2, value_3, value_4 FROM query_single_shard_table.non_colocated_events_table WHERE (user_id OPERATOR(pg_catalog.=) 6)
DEBUG: Distributed planning for a fast-path router query
DEBUG: Creating router plan
DEBUG: Plan XXX query after replacing subqueries and CTEs: SELECT cte1.user_id, cte1.value_1, cte2.user_id, cte2.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)) cte1, (SELECT intermediate_result.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, "time" timestamp without time zone, event_type integer, value_2 integer, value_3 double precision, value_4 bigint)) cte2 ORDER BY cte1.user_id, cte1.value_1, cte2.user_id, cte2.event_type LIMIT 5
DEBUG: Creating router plan
user_id | value_1 | user_id | user_id
---------------------------------------------------------------------
1 | 1 | 6 | 6
1 | 1 | 6 | 6
1 | 1 | 6 | 6
1 | 1 | 6 | 6
1 | 1 | 6 | 6
(5 rows)
-- users_table & colocated_users_table are colocated, route entire query.
WITH cte1 AS (
SELECT * FROM users_table WHERE user_id = 1
)
UPDATE colocated_users_table dt SET value = cte1.value_1
FROM cte1 WHERE cte1.user_id = dt.id AND dt.id = 1;
DEBUG: Creating router plan
-- users_table & non_colocated_users_table are not colocated, cte is recursive planned.
WITH cte1 AS (
SELECT * FROM users_table WHERE user_id = 1
)
UPDATE non_colocated_users_table dt SET value = cte1.value_1
FROM cte1 WHERE cte1.user_id = dt.id AND dt.id = 1;
DEBUG: router planner does not support queries that reference non-colocated distributed tables
DEBUG: generating subplan XXX_1 for CTE cte1: SELECT user_id, "time", value_1, value_2, value_3, value_4 FROM query_single_shard_table.users_table WHERE (user_id OPERATOR(pg_catalog.=) 1)
DEBUG: Distributed planning for a fast-path router query
DEBUG: Creating router plan
DEBUG: Plan XXX query after replacing subqueries and CTEs: UPDATE query_single_shard_table.non_colocated_users_table dt SET value = cte1.value_1 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)) cte1 WHERE ((cte1.user_id OPERATOR(pg_catalog.=) dt.id) AND (dt.id OPERATOR(pg_catalog.=) 1))
DEBUG: Creating router plan
-- All relations are not colocated, CTEs become intermediate results.
WITH cte1 AS MATERIALIZED (
SELECT * FROM users_table WHERE user_id = 1
), cte2 AS MATERIALIZED (
SELECT * FROM non_colocated_events_table WHERE user_id = 6
)
UPDATE non_colocated_users_table dt SET value = cte1.value_1 + cte2.event_type
FROM cte1, cte2 WHERE cte1.user_id = dt.id AND dt.id = 1;
DEBUG: router planner does not support queries that reference non-colocated distributed tables
DEBUG: generating subplan XXX_1 for CTE cte1: SELECT user_id, "time", value_1, value_2, value_3, value_4 FROM query_single_shard_table.users_table WHERE (user_id OPERATOR(pg_catalog.=) 1)
DEBUG: Distributed planning for a fast-path router query
DEBUG: Creating router plan
DEBUG: generating subplan XXX_2 for CTE cte2: SELECT user_id, "time", event_type, value_2, value_3, value_4 FROM query_single_shard_table.non_colocated_events_table WHERE (user_id OPERATOR(pg_catalog.=) 6)
DEBUG: Distributed planning for a fast-path router query
DEBUG: Creating router plan
DEBUG: Plan XXX query after replacing subqueries and CTEs: UPDATE query_single_shard_table.non_colocated_users_table dt SET value = (cte1.value_1 OPERATOR(pg_catalog.+) cte2.event_type) 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)) cte1, (SELECT intermediate_result.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, "time" timestamp without time zone, event_type integer, value_2 integer, value_3 double precision, value_4 bigint)) cte2 WHERE ((cte1.user_id OPERATOR(pg_catalog.=) dt.id) AND (dt.id OPERATOR(pg_catalog.=) 1))
DEBUG: Creating router plan
-- Volatile function calls should not be routed.
WITH cte1 AS MATERIALIZED (SELECT id, value FROM func())
UPDATE colocated_users_table dt SET value = cte1.value
FROM cte1 WHERE dt.id = 1;
DEBUG: Router planner doesn't support VOLATILE functions in common table expressions.
DEBUG: generating subplan XXX_1 for CTE cte1: SELECT id, value FROM query_single_shard_table.func() func(id, value)
DEBUG: Plan XXX query after replacing subqueries and CTEs: UPDATE query_single_shard_table.colocated_users_table dt SET value = cte1.value FROM (SELECT intermediate_result.id, intermediate_result.value FROM read_intermediate_result('XXX_1'::text, 'binary'::citus_copy_format) intermediate_result(id integer, value integer)) cte1 WHERE (dt.id OPERATOR(pg_catalog.=) 1)
DEBUG: Creating router plan
-- CTEs are recursively planned, and subquery foo is also recursively planned.
WITH cte AS MATERIALIZED (
WITH local_cte AS MATERIALIZED (
SELECT * FROM users_table_local
),
dist_cte AS MATERIALIZED (
SELECT user_id FROM colocated_events_table
)
SELECT dist_cte.user_id FROM local_cte JOIN dist_cte ON dist_cte.user_id=local_cte.user_id
)
SELECT count(*)
FROM
cte,
(
SELECT DISTINCT users_table.user_id
FROM users_table, colocated_events_table
WHERE users_table.user_id = colocated_events_table.user_id AND event_type IN (1,2,3,4)
ORDER BY 1 DESC LIMIT 5
) AS foo, colocated_events_table
WHERE foo.user_id = cte.user_id AND colocated_events_table.user_id = cte.user_id;
DEBUG: Local tables cannot be used in distributed queries.
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 query_single_shard_table.users_table_local), dist_cte AS MATERIALIZED (SELECT colocated_events_table.user_id FROM query_single_shard_table.colocated_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: Local tables cannot be used in distributed queries.
DEBUG: generating subplan XXX_1 for CTE local_cte: SELECT user_id, "time", value_1, value_2, value_3, value_4 FROM query_single_shard_table.users_table_local
DEBUG: generating subplan XXX_2 for CTE dist_cte: SELECT user_id FROM query_single_shard_table.colocated_events_table
DEBUG: Distributed planning for a fast-path router query
DEBUG: Creating router plan
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: Creating router plan
DEBUG: Creating router plan
DEBUG: generating subplan XXX_2 for subquery SELECT DISTINCT users_table.user_id FROM query_single_shard_table.users_table, query_single_shard_table.colocated_events_table WHERE ((users_table.user_id OPERATOR(pg_catalog.=) colocated_events_table.user_id) AND (colocated_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 count(*) AS count FROM (SELECT intermediate_result.user_id FROM read_intermediate_result('XXX_1'::text, 'binary'::citus_copy_format) intermediate_result(user_id integer)) cte, (SELECT intermediate_result.user_id FROM read_intermediate_result('XXX_2'::text, 'binary'::citus_copy_format) intermediate_result(user_id integer)) foo, query_single_shard_table.colocated_events_table WHERE ((foo.user_id OPERATOR(pg_catalog.=) cte.user_id) AND (colocated_events_table.user_id OPERATOR(pg_catalog.=) cte.user_id))
DEBUG: Creating router plan
count
---------------------------------------------------------------------
30608
(1 row)
-- CTEs are replaced and subquery in WHERE is also replaced.
WITH cte AS MATERIALIZED (
WITH local_cte AS MATERIALIZED (
SELECT * FROM users_table_local
),
dist_cte AS MATERIALIZED (
SELECT user_id FROM colocated_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 (
SELECT DISTINCT value_2 FROM users_table WHERE value_1 >= 1 AND value_1 <= 20 ORDER BY 1 LIMIT 5
)
ORDER BY 1 DESC;
DEBUG: Local tables cannot be used in distributed queries.
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 query_single_shard_table.users_table_local), dist_cte AS MATERIALIZED (SELECT colocated_events_table.user_id FROM query_single_shard_table.colocated_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: Local tables cannot be used in distributed queries.
DEBUG: generating subplan XXX_1 for CTE local_cte: SELECT user_id, "time", value_1, value_2, value_3, value_4 FROM query_single_shard_table.users_table_local
DEBUG: generating subplan XXX_2 for CTE dist_cte: SELECT user_id FROM query_single_shard_table.colocated_events_table
DEBUG: Distributed planning for a fast-path router query
DEBUG: Creating router plan
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: Creating router plan
DEBUG: Distributed planning for a fast-path router query
DEBUG: Creating router plan
DEBUG: generating subplan XXX_2 for subquery SELECT DISTINCT value_2 FROM query_single_shard_table.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 DISTINCT cte.user_id FROM query_single_shard_table.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 intermediate_result.value_2 FROM read_intermediate_result('XXX_2'::text, 'binary'::citus_copy_format) intermediate_result(value_2 integer)))) ORDER BY cte.user_id DESC
DEBUG: Creating router plan
user_id
---------------------------------------------------------------------
4
3
2
1
(4 rows)
-- Subquery in WHERE clause is planned recursively due to the recurring table
-- in FROM clause.
WITH cte AS MATERIALIZED (
WITH local_cte AS MATERIALIZED (
SELECT * FROM users_table_local
),
dist_cte AS MATERIALIZED (
SELECT user_id FROM colocated_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 cte
WHERE cte.user_id IN (SELECT DISTINCT user_id FROM users_table WHERE value_1 >= 1 AND value_1 <= 20)
ORDER BY 1 DESC;
DEBUG: Local tables cannot be used in distributed queries.
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 query_single_shard_table.users_table_local), dist_cte AS MATERIALIZED (SELECT colocated_events_table.user_id FROM query_single_shard_table.colocated_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: Local tables cannot be used in distributed queries.
DEBUG: generating subplan XXX_1 for CTE local_cte: SELECT user_id, "time", value_1, value_2, value_3, value_4 FROM query_single_shard_table.users_table_local
DEBUG: generating subplan XXX_2 for CTE dist_cte: SELECT user_id FROM query_single_shard_table.colocated_events_table
DEBUG: Distributed planning for a fast-path router query
DEBUG: Creating router plan
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: Creating router plan
DEBUG: Distributed planning for a fast-path router query
DEBUG: Creating router plan
DEBUG: generating subplan XXX_2 for subquery SELECT DISTINCT user_id FROM query_single_shard_table.users_table WHERE ((value_1 OPERATOR(pg_catalog.>=) 1) AND (value_1 OPERATOR(pg_catalog.<=) 20))
DEBUG: Plan XXX query after replacing subqueries and CTEs: SELECT DISTINCT user_id FROM (SELECT intermediate_result.user_id FROM read_intermediate_result('XXX_1'::text, 'binary'::citus_copy_format) intermediate_result(user_id integer)) cte WHERE (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))) ORDER BY user_id DESC
DEBUG: Creating router plan
user_id
---------------------------------------------------------------------
6
5
4
3
2
1
(6 rows)
-- CTEs inside a subquery and the final query becomes a router
-- query.
SELECT
user_id
FROM
(
WITH cte AS MATERIALIZED (
SELECT DISTINCT users_table.user_id
FROM users_table, colocated_events_table
WHERE users_table.user_id = colocated_events_table.user_id AND
event_type IN (1,2,3,4)
)
SELECT * FROM cte ORDER BY 1 DESC
) AS foo
ORDER BY 1 DESC;
DEBUG: Creating router plan
user_id
---------------------------------------------------------------------
6
5
4
3
2
1
(6 rows)
-- CTEs inside a deeper subquery and also the subquery that contains the CTE are
-- recursively planned.
SELECT DISTINCT bar.user_id
FROM
(
WITH cte AS MATERIALIZED (
SELECT DISTINCT users_table.user_id
FROM users_table, colocated_events_table
WHERE users_table.user_id = colocated_events_table.user_id AND event_type IN (1,2,3,4)
)
SELECT * FROM cte ORDER BY 1 DESC
) AS foo,
(
SELECT users_table.user_id, some_events.event_type
FROM
users_table,
(
WITH cte AS MATERIALIZED (
SELECT event_type, users_table.user_id
FROM users_table, colocated_events_table
WHERE users_table.user_id = colocated_events_table.user_id AND value_1 IN (1,2)
) SELECT * FROM cte ORDER BY 1 DESC
) AS some_events
WHERE users_table.user_id = some_events.user_id AND event_type IN (1,2,3,4)
ORDER BY 2,1 LIMIT 2
) AS bar
WHERE foo.user_id = bar.user_id
ORDER BY 1 DESC LIMIT 5;
DEBUG: Creating router plan
user_id
---------------------------------------------------------------------
1
(1 row)
-- Recursively plan subqueries inside the CTEs that contains LIMIT and OFFSET.
WITH cte AS MATERIALIZED (
WITH local_cte AS MATERIALIZED (
SELECT * FROM users_table_local
),
dist_cte AS MATERIALIZED (
SELECT
user_id
FROM
colocated_events_table,
(SELECT DISTINCT value_2 FROM users_table OFFSET 0) as foo
WHERE
colocated_events_table.user_id = foo.value_2 AND
colocated_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(*)
FROM
cte,
(
SELECT DISTINCT users_table.user_id
FROM users_table, colocated_events_table
WHERE users_table.user_id = colocated_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;
DEBUG: Local tables cannot be used in distributed queries.
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 query_single_shard_table.users_table_local), dist_cte AS MATERIALIZED (SELECT colocated_events_table.user_id FROM query_single_shard_table.colocated_events_table, (SELECT DISTINCT users_table.value_2 FROM query_single_shard_table.users_table OFFSET 0) foo WHERE ((colocated_events_table.user_id OPERATOR(pg_catalog.=) foo.value_2) AND (colocated_events_table.user_id OPERATOR(pg_catalog.=) ANY (SELECT DISTINCT users_table.value_1 FROM query_single_shard_table.users_table ORDER BY users_table.value_1 LIMIT 3)))) SELECT dist_cte.user_id FROM (local_cte JOIN dist_cte ON ((dist_cte.user_id OPERATOR(pg_catalog.=) local_cte.user_id)))
DEBUG: Local tables cannot be used in distributed queries.
DEBUG: generating subplan XXX_1 for CTE local_cte: SELECT user_id, "time", value_1, value_2, value_3, value_4 FROM query_single_shard_table.users_table_local
DEBUG: generating subplan XXX_2 for CTE dist_cte: SELECT colocated_events_table.user_id FROM query_single_shard_table.colocated_events_table, (SELECT DISTINCT users_table.value_2 FROM query_single_shard_table.users_table OFFSET 0) foo WHERE ((colocated_events_table.user_id OPERATOR(pg_catalog.=) foo.value_2) AND (colocated_events_table.user_id OPERATOR(pg_catalog.=) ANY (SELECT DISTINCT users_table.value_1 FROM query_single_shard_table.users_table ORDER BY users_table.value_1 LIMIT 3)))
DEBUG: Creating router plan
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: Creating router plan
DEBUG: Creating router plan
DEBUG: generating subplan XXX_2 for subquery SELECT DISTINCT users_table.user_id FROM query_single_shard_table.users_table, query_single_shard_table.colocated_events_table WHERE ((users_table.user_id OPERATOR(pg_catalog.=) colocated_events_table.user_id) AND (colocated_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 count(*) AS count FROM (SELECT intermediate_result.user_id FROM read_intermediate_result('XXX_1'::text, 'binary'::citus_copy_format) intermediate_result(user_id integer)) cte, (SELECT intermediate_result.user_id FROM read_intermediate_result('XXX_2'::text, 'binary'::citus_copy_format) intermediate_result(user_id integer)) foo WHERE (foo.user_id OPERATOR(pg_catalog.=) cte.user_id)
DEBUG: Creating router plan
count
---------------------------------------------------------------------
432
(1 row)
-- more tests with sublinks and subqueries in targetlist
SELECT event_type, (SELECT e.value_2 FROM users_reference_table WHERE user_id = 1 AND value_1 = 1), (SELECT e.value_2)
FROM non_colocated_events_table e
ORDER BY 1,2 LIMIT 1;
DEBUG: Creating router plan
event_type | value_2 | value_2
---------------------------------------------------------------------
0 | | 0
(1 row)
SELECT event_type, (SELECT time FROM users_table WHERE user_id = e.user_id ORDER BY time LIMIT 1)
FROM non_colocated_events_table e
ORDER BY 1,2 LIMIT 1;
DEBUG: router planner does not support queries that reference non-colocated distributed tables
DEBUG: push down of limit count: 1
ERROR: cannot push down this subquery
DETAIL: users_table and non_colocated_events_table are not colocated
SELECT event_type, (SELECT max(time) FROM users_table WHERE user_id = e.value_2)
FROM non_colocated_events_table e
ORDER BY 1,2 LIMIT 1;
DEBUG: router planner does not support queries that reference non-colocated distributed tables
DEBUG: push down of limit count: 1
ERROR: cannot push down this subquery
DETAIL: users_table and non_colocated_events_table are not colocated
SELECT event_type, (SELECT max(time) FROM users_table)
FROM non_colocated_events_table e
ORDER BY 1,2 LIMIT 1;
DEBUG: router planner does not support queries that reference non-colocated distributed tables
DEBUG: Distributed planning for a fast-path router query
DEBUG: Creating router plan
DEBUG: generating subplan XXX_1 for subquery SELECT max("time") AS max FROM query_single_shard_table.users_table
DEBUG: Plan XXX query after replacing subqueries and CTEs: SELECT event_type, (SELECT intermediate_result.max FROM read_intermediate_result('XXX_1'::text, 'binary'::citus_copy_format) intermediate_result(max timestamp without time zone)) AS max FROM query_single_shard_table.non_colocated_events_table e ORDER BY event_type, (SELECT intermediate_result.max FROM read_intermediate_result('XXX_1'::text, 'binary'::citus_copy_format) intermediate_result(max timestamp without time zone)) LIMIT 1
DEBUG: Creating router plan
event_type | max
---------------------------------------------------------------------
0 | Thu Nov 23 17:30:34.635085 2017
(1 row)
WITH cte_1 AS (SELECT max(time) FROM users_table)
SELECT event_type, (SELECT * FROM cte_1)
FROM non_colocated_events_table e
ORDER BY 1,2 LIMIT 1;
DEBUG: CTE cte_1 is going to be inlined via distributed planning
DEBUG: router planner does not support queries that reference non-colocated distributed tables
DEBUG: Distributed planning for a fast-path router query
DEBUG: Creating router plan
DEBUG: generating subplan XXX_1 for subquery SELECT max("time") AS max FROM query_single_shard_table.users_table
DEBUG: Plan XXX query after replacing subqueries and CTEs: SELECT event_type, (SELECT cte_1.max FROM (SELECT intermediate_result.max FROM read_intermediate_result('XXX_1'::text, 'binary'::citus_copy_format) intermediate_result(max timestamp without time zone)) cte_1) AS max FROM query_single_shard_table.non_colocated_events_table e ORDER BY event_type, (SELECT cte_1.max FROM (SELECT intermediate_result.max FROM read_intermediate_result('XXX_1'::text, 'binary'::citus_copy_format) intermediate_result(max timestamp without time zone)) cte_1) LIMIT 1
DEBUG: Creating router plan
event_type | max
---------------------------------------------------------------------
0 | Thu Nov 23 17:30:34.635085 2017
(1 row)
WITH cte_1 AS (SELECT max(time) FROM users_table)
SELECT event_type, (SELECT * FROM cte_1 LIMIT 1)
FROM non_colocated_events_table e
ORDER BY 1,2 LIMIT 1;
DEBUG: CTE cte_1 is going to be inlined via distributed planning
DEBUG: router planner does not support queries that reference non-colocated distributed tables
DEBUG: Distributed planning for a fast-path router query
DEBUG: Creating router plan
DEBUG: generating subplan XXX_1 for subquery SELECT max("time") AS max FROM query_single_shard_table.users_table
DEBUG: Creating router plan
DEBUG: generating subplan XXX_2 for subquery SELECT max FROM (SELECT intermediate_result.max FROM read_intermediate_result('XXX_1'::text, 'binary'::citus_copy_format) intermediate_result(max timestamp without time zone)) cte_1 LIMIT 1
DEBUG: Plan XXX query after replacing subqueries and CTEs: SELECT event_type, (SELECT intermediate_result.max FROM read_intermediate_result('XXX_2'::text, 'binary'::citus_copy_format) intermediate_result(max timestamp without time zone)) AS max FROM query_single_shard_table.non_colocated_events_table e ORDER BY event_type, (SELECT intermediate_result.max FROM read_intermediate_result('XXX_2'::text, 'binary'::citus_copy_format) intermediate_result(max timestamp without time zone)) LIMIT 1
DEBUG: Creating router plan
event_type | max
---------------------------------------------------------------------
0 | Thu Nov 23 17:30:34.635085 2017
(1 row)
WITH cte_1 AS (SELECT max(time) m FROM users_table)
SELECT count(*), (SELECT * FROM cte_1 c1 join cte_1 c2 using (m))
FROM non_colocated_events_table e
GROUP BY 2
ORDER BY 1,2 LIMIT 1;
DEBUG: router planner does not support queries that reference non-colocated distributed tables
DEBUG: generating subplan XXX_1 for CTE cte_1: SELECT max("time") AS m FROM query_single_shard_table.users_table
DEBUG: Distributed planning for a fast-path router query
DEBUG: Creating router plan
DEBUG: Plan XXX query after replacing subqueries and CTEs: SELECT count(*) AS count, (SELECT c1.m FROM ((SELECT intermediate_result.m FROM read_intermediate_result('XXX_1'::text, 'binary'::citus_copy_format) intermediate_result(m timestamp without time zone)) c1 JOIN (SELECT intermediate_result.m FROM read_intermediate_result('XXX_1'::text, 'binary'::citus_copy_format) intermediate_result(m timestamp without time zone)) c2 USING (m))) AS m FROM query_single_shard_table.non_colocated_events_table e GROUP BY (SELECT c1.m FROM ((SELECT intermediate_result.m FROM read_intermediate_result('XXX_1'::text, 'binary'::citus_copy_format) intermediate_result(m timestamp without time zone)) c1 JOIN (SELECT intermediate_result.m FROM read_intermediate_result('XXX_1'::text, 'binary'::citus_copy_format) intermediate_result(m timestamp without time zone)) c2 USING (m))) ORDER BY (count(*)), (SELECT c1.m FROM ((SELECT intermediate_result.m FROM read_intermediate_result('XXX_1'::text, 'binary'::citus_copy_format) intermediate_result(m timestamp without time zone)) c1 JOIN (SELECT intermediate_result.m FROM read_intermediate_result('XXX_1'::text, 'binary'::citus_copy_format) intermediate_result(m timestamp without time zone)) c2 USING (m))) LIMIT 1
DEBUG: Creating router plan
count | m
---------------------------------------------------------------------
101 | Thu Nov 23 17:30:34.635085 2017
(1 row)
WITH cte_1 AS (SELECT min(user_id) u, max(time) m FROM users_table)
SELECT count(*), (SELECT max(time) FROM users_table WHERE user_id = cte_1.u GROUP BY user_id)
FROM cte_1
GROUP BY 2
ORDER BY 1,2 LIMIT 1;
DEBUG: CTE cte_1 is going to be inlined via distributed planning
DEBUG: Creating router plan
count | max
---------------------------------------------------------------------
1 | Thu Nov 23 17:30:34.635085 2017
(1 row)
SELECT sum(e.user_id) + (SELECT max(value_3) FROM users_table WHERE user_id = e.user_id GROUP BY user_id)
FROM non_colocated_events_table e
GROUP BY e.user_id
ORDER BY 1 LIMIT 3;
DEBUG: router planner does not support queries that reference non-colocated distributed tables
ERROR: cannot push down subquery on the target list
DETAIL: Subqueries in the SELECT part of the query can only be pushed down if they happen before aggregates and window functions
SELECT e.user_id, sum((SELECT any_value(value_3) FROM users_reference_table WHERE user_id = e.user_id GROUP BY user_id)) OVER (PARTITION BY e.user_id)
FROM non_colocated_events_table e
ORDER BY 1, 2 LIMIT 3;
DEBUG: Creating router plan
user_id | sum
---------------------------------------------------------------------
1 |
1 |
1 |
(3 rows)
SELECT (SELECT (SELECT e.user_id + user_id) FROM users_table WHERE user_id = e.user_id GROUP BY user_id)
FROM non_colocated_events_table e
GROUP BY 1
ORDER BY 1 LIMIT 3;
DEBUG: router planner does not support queries that reference non-colocated distributed tables
DEBUG: push down of limit count: 3
ERROR: cannot push down this subquery
DETAIL: users_table and non_colocated_events_table are not colocated
SELECT (SELECT (SELECT e.user_id + user_id) FROM users_reference_table WHERE user_id = e.user_id GROUP BY user_id)
FROM non_colocated_events_table e
GROUP BY 1
ORDER BY 1 LIMIT 3;
DEBUG: Creating router plan
?column?
---------------------------------------------------------------------
(1 row)
WITH cte_1 AS (SELECT user_id FROM users_table ORDER BY 1 LIMIT 1)
SELECT (SELECT (SELECT e.user_id + user_id) FROM cte_1 WHERE user_id = e.user_id GROUP BY user_id)
FROM non_colocated_events_table e
GROUP BY 1
ORDER BY 1 LIMIT 3;
DEBUG: CTE cte_1 is going to be inlined via distributed planning
DEBUG: router planner does not support queries that reference non-colocated distributed tables
DEBUG: Distributed planning for a fast-path router query
DEBUG: Creating router plan
DEBUG: generating subplan XXX_1 for subquery SELECT user_id FROM query_single_shard_table.users_table ORDER BY user_id LIMIT 1
DEBUG: skipping recursive planning for the subquery since it contains references to outer queries
DEBUG: Plan XXX query after replacing subqueries and CTEs: SELECT (SELECT (SELECT (e.user_id OPERATOR(pg_catalog.+) cte_1.user_id)) FROM (SELECT intermediate_result.user_id FROM read_intermediate_result('XXX_1'::text, 'binary'::citus_copy_format) intermediate_result(user_id integer)) cte_1 WHERE (cte_1.user_id OPERATOR(pg_catalog.=) e.user_id) GROUP BY cte_1.user_id) FROM query_single_shard_table.non_colocated_events_table e GROUP BY (SELECT (SELECT (e.user_id OPERATOR(pg_catalog.+) cte_1.user_id)) FROM (SELECT intermediate_result.user_id FROM read_intermediate_result('XXX_1'::text, 'binary'::citus_copy_format) intermediate_result(user_id integer)) cte_1 WHERE (cte_1.user_id OPERATOR(pg_catalog.=) e.user_id) GROUP BY cte_1.user_id) ORDER BY (SELECT (SELECT (e.user_id OPERATOR(pg_catalog.+) cte_1.user_id)) FROM (SELECT intermediate_result.user_id FROM read_intermediate_result('XXX_1'::text, 'binary'::citus_copy_format) intermediate_result(user_id integer)) cte_1 WHERE (cte_1.user_id OPERATOR(pg_catalog.=) e.user_id) GROUP BY cte_1.user_id) LIMIT 3
DEBUG: Creating router plan
?column?
---------------------------------------------------------------------
2
(2 rows)
SELECT (SELECT (SELECT e.user_id + user_id) FROM (SELECT 1 AS user_id) s WHERE user_id = e.user_id GROUP BY user_id)
FROM non_colocated_events_table e
GROUP BY 1
ORDER BY 1 LIMIT 3;
DEBUG: Creating router plan
?column?
---------------------------------------------------------------------
2
(2 rows)
CREATE TEMP VIEW view_1 AS (SELECT user_id, value_2 FROM users_table WHERE user_id = 1 AND value_1 = 1 ORDER BY 1,2);
WARNING: "view view_1" has dependency on unsupported object "schema pg_temp_xxx"
DETAIL: "view view_1" will be created only locally
SELECT (SELECT value_2 FROM view_1 WHERE user_id = e.user_id GROUP BY value_2)
FROM non_colocated_events_table e
GROUP BY 1
ORDER BY 1 LIMIT 3;
DEBUG: router planner does not support queries that reference non-colocated distributed tables
DEBUG: push down of limit count: 3
ERROR: cannot push down this subquery
DETAIL: users_table and non_colocated_events_table are not colocated
SELECT
user_id, count(*)
FROM
non_colocated_events_table e1
GROUP BY user_id
HAVING
count(*) > (SELECT count(*) FROM (SELECT
(SELECT sum(user_id) FROM users_table WHERE user_id = u1.user_id GROUP BY user_id)
FROM users_table u1
GROUP BY user_id) as foo) ORDER BY 1 DESC;
DEBUG: router planner does not support queries that reference non-colocated distributed tables
DEBUG: Creating router plan
DEBUG: generating subplan XXX_1 for subquery SELECT count(*) AS count FROM (SELECT (SELECT sum(users_table.user_id) AS sum FROM query_single_shard_table.users_table WHERE (users_table.user_id OPERATOR(pg_catalog.=) u1.user_id) GROUP BY users_table.user_id) AS sum FROM query_single_shard_table.users_table u1 GROUP BY u1.user_id) foo
DEBUG: Plan XXX query after replacing subqueries and CTEs: SELECT user_id, count(*) AS count FROM query_single_shard_table.non_colocated_events_table e1 GROUP BY user_id HAVING (count(*) OPERATOR(pg_catalog.>) (SELECT intermediate_result.count FROM read_intermediate_result('XXX_1'::text, 'binary'::citus_copy_format) intermediate_result(count bigint))) ORDER BY user_id DESC
DEBUG: Creating router plan
user_id | count
---------------------------------------------------------------------
6 | 10
5 | 14
4 | 17
3 | 21
2 | 24
1 | 15
(6 rows)
SELECT count(*) FROM (SELECT
(SELECT user_id FROM users_table WHERE user_id = u1.user_id FOR UPDATE)
FROM users_table u1
GROUP BY user_id) as foo;
DEBUG: Creating router plan
count
---------------------------------------------------------------------
6
(1 row)
-- test single hash repartition join
SET citus.log_multi_join_order TO ON;
SET client_min_messages TO DEBUG1;
SET citus.enable_repartition_joins TO ON;
SET citus.enable_single_hash_repartition_joins TO ON;
SELECT count(*) FROM nullkey_c1_t1 JOIN distributed_table USING(a);
LOG: join order: [ "nullkey_c1_t1" ][ single hash partition join "distributed_table" ]
count
---------------------------------------------------------------------
0
(1 row)
select count(*) from nullkey_c1_t1 JOIN nullkey_c2_t2 USING(a);
LOG: join order: [ "nullkey_c1_t1" ][ dual partition join "nullkey_c2_t2" ]
count
---------------------------------------------------------------------
0
(1 row)
RESET citus.log_multi_join_order;
SET client_min_messages TO DEBUG2;
RESET citus.enable_repartition_joins;
RESET citus.enable_single_hash_repartition_joins;
SET client_min_messages TO DEBUG1;
SET citus.enable_repartition_joins TO ON;
SET citus.log_multi_join_order TO ON;
SELECT count(*), avg(avgsub.a)
FROM (
SELECT table_0.a
FROM reference_table AS table_0
INNER JOIN nullkey_c1_t1 AS table_1 USING (a)
INNER JOIN reference_table AS table_2 USING (a)
INNER JOIN nullkey_c2_t1 AS table_3 USING (a)
ORDER BY a LIMIT 7
) AS avgsub;
LOG: join order: [ "nullkey_c1_t1" ][ reference join "reference_table" ][ reference join "reference_table" ][ dual partition join "nullkey_c2_t1" ]
DEBUG: push down of limit count: 7
DEBUG: generating subplan XXX_1 for subquery SELECT table_0.a FROM (((query_single_shard_table.reference_table table_0 JOIN query_single_shard_table.nullkey_c1_t1 table_1 USING (a)) JOIN query_single_shard_table.reference_table table_2 USING (a)) JOIN query_single_shard_table.nullkey_c2_t1 table_3 USING (a)) ORDER BY table_0.a LIMIT 7
DEBUG: Plan XXX query after replacing subqueries and CTEs: SELECT count(*) AS count, avg(a) AS avg FROM (SELECT intermediate_result.a FROM read_intermediate_result('XXX_1'::text, 'binary'::citus_copy_format) intermediate_result(a integer)) avgsub
count | avg
---------------------------------------------------------------------
0 |
(1 row)
SET citus.enable_single_hash_repartition_joins TO ON;
-- We prefer dual-hash repartition join over single-hash repartition join
-- even if citus.enable_single_hash_repartition_joins is set to ON. This
-- happens because single shard tables don't have a shard key.
SELECT count(*), avg(avgsub.a)
FROM (
SELECT table_0.a
FROM reference_table AS table_0
INNER JOIN nullkey_c1_t1 AS table_1 USING (a)
INNER JOIN reference_table AS table_2 USING (a)
INNER JOIN nullkey_c2_t1 AS table_3 USING (a)
ORDER BY a LIMIT 7
) AS avgsub;
LOG: join order: [ "nullkey_c1_t1" ][ reference join "reference_table" ][ reference join "reference_table" ][ dual partition join "nullkey_c2_t1" ]
DEBUG: push down of limit count: 7
DEBUG: generating subplan XXX_1 for subquery SELECT table_0.a FROM (((query_single_shard_table.reference_table table_0 JOIN query_single_shard_table.nullkey_c1_t1 table_1 USING (a)) JOIN query_single_shard_table.reference_table table_2 USING (a)) JOIN query_single_shard_table.nullkey_c2_t1 table_3 USING (a)) ORDER BY table_0.a LIMIT 7
DEBUG: Plan XXX query after replacing subqueries and CTEs: SELECT count(*) AS count, avg(a) AS avg FROM (SELECT intermediate_result.a FROM read_intermediate_result('XXX_1'::text, 'binary'::citus_copy_format) intermediate_result(a integer)) avgsub
count | avg
---------------------------------------------------------------------
0 |
(1 row)
RESET citus.enable_single_hash_repartition_joins;
SET client_min_messages TO DEBUG2;
RESET citus.enable_repartition_joins;
RESET citus.log_multi_join_order;
SELECT count(*), avg(avgsub.a)
FROM (
SELECT table_0.a
FROM nullkey_c1_t1 AS table_0
RIGHT JOIN (
SELECT table_2.a FROM (
SELECT table_3.a FROM nullkey_c2_t1 AS table_3
ORDER BY a LIMIT 0
) AS table_2
INNER JOIN nullkey_c2_t1 AS table_4 USING (a)
WHERE table_4.a < 8
) AS table_1 USING (a)
) AS avgsub;
DEBUG: router planner does not support queries that reference non-colocated distributed tables
DEBUG: Distributed planning for a fast-path router query
DEBUG: Creating router plan
DEBUG: generating subplan XXX_1 for subquery SELECT a FROM query_single_shard_table.nullkey_c2_t1 table_3 ORDER BY a LIMIT 0
DEBUG: Plan XXX query after replacing subqueries and CTEs: SELECT count(*) AS count, avg(a) AS avg FROM (SELECT table_0.a FROM (query_single_shard_table.nullkey_c1_t1 table_0 RIGHT JOIN (SELECT table_2.a FROM ((SELECT intermediate_result.a FROM read_intermediate_result('XXX_1'::text, 'binary'::citus_copy_format) intermediate_result(a integer)) table_2 JOIN query_single_shard_table.nullkey_c2_t1 table_4 USING (a)) WHERE (table_4.a OPERATOR(pg_catalog.<) 8)) table_1 USING (a))) avgsub
DEBUG: router planner does not support queries that reference non-colocated distributed tables
ERROR: cannot perform a lateral outer join when a distributed subquery references complex subqueries, CTEs or local tables
-- test nested exec
CREATE FUNCTION dist_query_single_shard(p_key int)
RETURNS bigint
LANGUAGE plpgsql AS $$
DECLARE
result bigint;
BEGIN
SELECT count(*) INTO result FROM query_single_shard_table.nullkey_c1_t1 WHERE a = p_key;
RETURN result;
END;
$$;
DEBUG: switching to sequential query execution mode
DETAIL: A command for a distributed function is run. To make sure subsequent commands see the function correctly we need to make sure to use only one connection for all future commands
CREATE FUNCTION ref_query()
RETURNS bigint
LANGUAGE plpgsql AS $$
DECLARE
result bigint;
BEGIN
SELECT count(*) INTO result FROM query_single_shard_table.reference_table;
RETURN result;
END;
$$;
DEBUG: switching to sequential query execution mode
DETAIL: A command for a distributed function is run. To make sure subsequent commands see the function correctly we need to make sure to use only one connection for all future commands
SELECT dist_query_single_shard(count(*)::int) FROM nullkey_c1_t1;
DEBUG: Distributed planning for a fast-path router query
DEBUG: Creating router plan
ERROR: cannot execute a distributed query from a query on a shard
DETAIL: Executing a distributed query in a function call that may be pushed to a remote node can lead to incorrect results.
HINT: Avoid nesting of distributed queries or use alter user current_user set citus.allow_nested_distributed_execution to on to allow it with possible incorrectness.
CONTEXT: SQL statement "SELECT count(*) FROM query_single_shard_table.nullkey_c1_t1 WHERE a = p_key"
PL/pgSQL function query_single_shard_table.dist_query_single_shard(integer) line XX at SQL statement
while executing command on localhost:xxxxx
SELECT ref_query()+count(*) FROM nullkey_c1_t1;
DEBUG: Distributed planning for a fast-path router query
DEBUG: Creating router plan
ERROR: cannot execute a distributed query from a query on a shard
DETAIL: Executing a distributed query in a function call that may be pushed to a remote node can lead to incorrect results.
HINT: Avoid nesting of distributed queries or use alter user current_user set citus.allow_nested_distributed_execution to on to allow it with possible incorrectness.
CONTEXT: SQL statement "SELECT count(*) FROM query_single_shard_table.reference_table"
PL/pgSQL function query_single_shard_table.ref_query() line XX at SQL statement
while executing command on localhost:xxxxx
SET client_min_messages TO ERROR;
DROP SCHEMA query_single_shard_table CASCADE;