-- -- PG18 -- SHOW server_version \gset SELECT substring(:'server_version', '\d+')::int >= 18 AS server_version_ge_18 \gset -- test invalid statistics -- behavior is same among PG versions, error message differs -- relevant PG18 commit: 3eea4dc2c7, 38883916e CREATE STATISTICS tst ON a FROM (VALUES (x)) AS foo; ERROR: CREATE STATISTICS only supports relation names in the FROM clause CREATE FUNCTION tftest(int) returns table(a int, b int) as $$ SELECT $1, $1+i FROM generate_series(1,5) g(i); $$ LANGUAGE sql IMMUTABLE STRICT; CREATE STATISTICS alt_stat2 ON a FROM tftest(1); ERROR: CREATE STATISTICS only supports relation names in the FROM clause DROP FUNCTION tftest; \if :server_version_ge_18 \else \q \endif -- PG18-specific tests go here. -- -- Purpose: Verify PG18 behavior that NOT NULL constraints are materialized -- as pg_constraint rows with contype = 'n' on both coordinator and -- worker shards. Also confirm our helper view (table_checks) does -- NOT surface NOT NULL entries. -- https://github.com/postgres/postgres/commit/14e87ffa5c543b5f30ead7413084c25f7735039f CREATE SCHEMA pg18_nn; SET search_path TO pg18_nn; -- Local control table DROP TABLE IF EXISTS nn_local CASCADE; NOTICE: table "nn_local" does not exist, skipping CREATE TABLE nn_local( a int NOT NULL, b int, c text NOT NULL ); -- Distributed table DROP TABLE IF EXISTS nn_dist CASCADE; NOTICE: table "nn_dist" does not exist, skipping CREATE TABLE nn_dist( a int NOT NULL, b int, c text NOT NULL ); SELECT create_distributed_table('nn_dist', 'a'); create_distributed_table --------------------------------------------------------------------- (1 row) -- Coordinator: count NOT NULL constraint rows SELECT 'local_n_count' AS label, contype, count(*) FROM pg_constraint WHERE conrelid = 'pg18_nn.nn_local'::regclass GROUP BY contype ORDER BY contype; label | contype | count --------------------------------------------------------------------- local_n_count | n | 2 (1 row) SELECT 'dist_n_count' AS label, contype, count(*) FROM pg_constraint WHERE conrelid = 'pg18_nn.nn_dist'::regclass GROUP BY contype ORDER BY contype; label | contype | count --------------------------------------------------------------------- dist_n_count | n | 2 (1 row) -- Our helper view should exclude NOT NULL SELECT 'table_checks_local_count' AS label, count(*) FROM public.table_checks WHERE relid = 'pg18_nn.nn_local'::regclass; label | count --------------------------------------------------------------------- table_checks_local_count | 0 (1 row) SELECT 'table_checks_dist_count' AS label, count(*) FROM public.table_checks WHERE relid = 'pg18_nn.nn_dist'::regclass; label | count --------------------------------------------------------------------- table_checks_dist_count | 0 (1 row) -- Add a real CHECK to ensure table_checks still reports real checks ALTER TABLE nn_dist ADD CONSTRAINT nn_dist_check CHECK (b IS DISTINCT FROM 42); SELECT 'table_checks_dist_with_real_check' AS label, count(*) FROM public.table_checks WHERE relid = 'pg18_nn.nn_dist'::regclass; label | count --------------------------------------------------------------------- table_checks_dist_with_real_check | 1 (1 row) -- === Worker checks === \c - - - :worker_1_port SET client_min_messages TO WARNING; SET search_path TO pg18_nn; -- Pick one heap shard of nn_dist in our schema SELECT format('%I.%I', n.nspname, c.relname) AS shard_regclass FROM pg_class c JOIN pg_namespace n ON n.oid = c.relnamespace WHERE n.nspname = 'pg18_nn' AND c.relname LIKE 'nn_dist_%' AND c.relkind = 'r' ORDER BY c.relname LIMIT 1 \gset -- Expect: 2 NOT NULL rows (a,c) + 1 CHECK row on the shard SELECT 'worker_shard_n_count' AS label, contype, count(*) FROM pg_constraint WHERE conrelid = :'shard_regclass'::regclass GROUP BY contype ORDER BY contype; label | contype | count --------------------------------------------------------------------- worker_shard_n_count | c | 1 worker_shard_n_count | n | 2 (2 rows) -- table_checks on shard should hide NOT NULL SELECT 'table_checks_worker_shard_count' AS label, count(*) FROM public.table_checks WHERE relid = :'shard_regclass'::regclass; label | count --------------------------------------------------------------------- table_checks_worker_shard_count | 1 (1 row) -- Drop one NOT NULL on coordinator; verify propagation \c - - - :master_port SET search_path TO pg18_nn; ALTER TABLE nn_dist ALTER COLUMN c DROP NOT NULL; -- Re-check on worker: NOT NULL count should drop to 1 \c - - - :worker_1_port SET search_path TO pg18_nn; SELECT 'worker_shard_n_after_drop' AS label, contype, count(*) FROM pg_constraint WHERE conrelid = :'shard_regclass'::regclass GROUP BY contype ORDER BY contype; label | contype | count --------------------------------------------------------------------- worker_shard_n_after_drop | c | 1 worker_shard_n_after_drop | n | 1 (2 rows) -- And on coordinator \c - - - :master_port SET search_path TO pg18_nn; SELECT 'dist_n_after_drop' AS label, contype, count(*) FROM pg_constraint WHERE conrelid = 'pg18_nn.nn_dist'::regclass GROUP BY contype ORDER BY contype; label | contype | count --------------------------------------------------------------------- dist_n_after_drop | c | 1 dist_n_after_drop | n | 1 (2 rows) -- Purpose: test self join elimination for distributed, citus local and local tables. -- CREATE TABLE sje_d1 (id bigserial PRIMARY KEY, name text, created_at timestamptz DEFAULT now()); CREATE TABLE sje_d2 (id bigserial PRIMARY KEY, name text, created_at timestamptz DEFAULT now()); CREATE TABLE sje_local (id bigserial PRIMARY KEY, title text); SET citus.next_shard_id TO 4754000; SELECT create_distributed_table('sje_d1', 'id'); create_distributed_table --------------------------------------------------------------------- (1 row) SELECT create_distributed_table('sje_d2', 'id'); create_distributed_table --------------------------------------------------------------------- (1 row) INSERT INTO sje_d1 SELECT i, i::text, now() FROM generate_series(0,100)i; INSERT INTO sje_d2 SELECT i, i::text, now() FROM generate_series(0,100)i; INSERT INTO sje_local SELECT i, i::text FROM generate_series(0,100)i; -- Self-join elimination is applied when distributed tables are involved -- The query plan has only one join EXPLAIN (costs off) select count(1) from sje_d1 INNER JOIN sje_d2 u1 USING (id) INNER JOIN sje_d2 u2 USING (id) INNER JOIN sje_d2 u3 USING (id) INNER JOIN sje_d2 u4 USING (id) INNER JOIN sje_d2 u5 USING (id) INNER JOIN sje_d2 u6 USING (id); QUERY PLAN --------------------------------------------------------------------- Aggregate -> Custom Scan (Citus Adaptive) Task Count: 4 Tasks Shown: One of 4 -> Task Node: host=localhost port=xxxxx dbname=regression -> Aggregate -> Hash Join Hash Cond: (sje_d1.id = u6.id) -> Seq Scan on sje_d1_4754000 sje_d1 -> Hash -> Seq Scan on sje_d2_4754004 u6 (12 rows) select count(1) from sje_d1 INNER JOIN sje_d2 u1 USING (id) INNER JOIN sje_d2 u2 USING (id) INNER JOIN sje_d2 u3 USING (id) INNER JOIN sje_d2 u4 USING (id) INNER JOIN sje_d2 u5 USING (id) INNER JOIN sje_d2 u6 USING (id); count --------------------------------------------------------------------- 101 (1 row) -- Self-join elimination applied to from list join EXPLAIN (costs off) SELECT count(1) from sje_d1 d1, sje_d2 u1, sje_d2 u2, sje_d2 u3 WHERE d1.id = u1.id and u1.id = u2.id and u3.id = d1.id; QUERY PLAN --------------------------------------------------------------------- Aggregate -> Custom Scan (Citus Adaptive) Task Count: 4 Tasks Shown: One of 4 -> Task Node: host=localhost port=xxxxx dbname=regression -> Aggregate -> Hash Join Hash Cond: (d1.id = u3.id) -> Seq Scan on sje_d1_4754000 d1 -> Hash -> Seq Scan on sje_d2_4754004 u3 (12 rows) SELECT count(1) from sje_d1 d1, sje_d2 u1, sje_d2 u2, sje_d2 u3 WHERE d1.id = u1.id and u1.id = u2.id and u3.id = d1.id; count --------------------------------------------------------------------- 101 (1 row) -- Self-join elimination is not applied when a local table is involved -- This is a limitation that will be resolved in citus 14 EXPLAIN (costs off) select count(1) from sje_d1 INNER JOIN sje_local u1 USING (id) INNER JOIN sje_local u2 USING (id) INNER JOIN sje_local u3 USING (id) INNER JOIN sje_local u4 USING (id) INNER JOIN sje_local u5 USING (id) INNER JOIN sje_local u6 USING (id); QUERY PLAN --------------------------------------------------------------------- Aggregate -> Custom Scan (Citus Adaptive) -> Distributed Subplan XXX_1 -> Seq Scan on sje_local u1 -> Distributed Subplan XXX_2 -> Seq Scan on sje_local u2 -> Distributed Subplan XXX_3 -> Seq Scan on sje_local u3 -> Distributed Subplan XXX_4 -> Seq Scan on sje_local u4 -> Distributed Subplan XXX_5 -> Seq Scan on sje_local u5 -> Distributed Subplan XXX_6 -> Seq Scan on sje_local u6 Task Count: 4 Tasks Shown: One of 4 -> Task Node: host=localhost port=xxxxx dbname=regression -> Aggregate -> Hash Join Hash Cond: (intermediate_result_5.id = sje_d1.id) -> Function Scan on read_intermediate_result intermediate_result_5 -> Hash -> Hash Join Hash Cond: (intermediate_result_4.id = sje_d1.id) -> Function Scan on read_intermediate_result intermediate_result_4 -> Hash -> Hash Join Hash Cond: (intermediate_result_3.id = sje_d1.id) -> Function Scan on read_intermediate_result intermediate_result_3 -> Hash -> Hash Join Hash Cond: (intermediate_result_2.id = sje_d1.id) -> Function Scan on read_intermediate_result intermediate_result_2 -> Hash -> Hash Join Hash Cond: (intermediate_result_1.id = sje_d1.id) -> Function Scan on read_intermediate_result intermediate_result_1 -> Hash -> Hash Join Hash Cond: (intermediate_result.id = sje_d1.id) -> Function Scan on read_intermediate_result intermediate_result -> Hash -> Seq Scan on sje_d1_4754000 sje_d1 (44 rows) select count(1) from sje_d1 INNER JOIN sje_local u1 USING (id) INNER JOIN sje_local u2 USING (id) INNER JOIN sje_local u3 USING (id) INNER JOIN sje_local u4 USING (id) INNER JOIN sje_local u5 USING (id) INNER JOIN sje_local u6 USING (id); count --------------------------------------------------------------------- 101 (1 row) -- to test USING vs ON equivalence EXPLAIN (costs off) SELECT count(1) FROM sje_d1 d JOIN sje_d2 u1 ON (d.id = u1.id) JOIN sje_d2 u2 ON (u1.id = u2.id); QUERY PLAN --------------------------------------------------------------------- Aggregate -> Custom Scan (Citus Adaptive) Task Count: 4 Tasks Shown: One of 4 -> Task Node: host=localhost port=xxxxx dbname=regression -> Aggregate -> Hash Join Hash Cond: (d.id = u2.id) -> Seq Scan on sje_d1_4754000 d -> Hash -> Seq Scan on sje_d2_4754004 u2 (12 rows) SELECT count(1) FROM sje_d1 d JOIN sje_d2 u1 ON (d.id = u1.id) JOIN sje_d2 u2 ON (u1.id = u2.id); count --------------------------------------------------------------------- 101 (1 row) -- Null-introducing join can have SJE EXPLAIN (costs off) SELECT count(*) FROM sje_d1 d LEFT JOIN sje_d2 u1 USING (id) LEFT JOIN sje_d2 u2 USING (id); QUERY PLAN --------------------------------------------------------------------- Aggregate -> Custom Scan (Citus Adaptive) Task Count: 4 Tasks Shown: One of 4 -> Task Node: host=localhost port=xxxxx dbname=regression -> Aggregate -> Seq Scan on sje_d1_4754000 d (8 rows) SELECT count(*) FROM sje_d1 d LEFT JOIN sje_d2 u1 USING (id) LEFT JOIN sje_d2 u2 USING (id); count --------------------------------------------------------------------- 101 (1 row) -- prepared statement PREPARE sje_p(int,int) AS SELECT count(1) FROM sje_d1 d JOIN sje_d2 u1 USING (id) JOIN sje_d2 u2 USING (id) WHERE d.id BETWEEN $1 AND $2; EXPLAIN (costs off) EXECUTE sje_p(10,20); QUERY PLAN --------------------------------------------------------------------- Aggregate -> Custom Scan (Citus Adaptive) Task Count: 4 Tasks Shown: One of 4 -> Task Node: host=localhost port=xxxxx dbname=regression -> Aggregate -> Hash Join Hash Cond: (u2.id = d.id) -> Seq Scan on sje_d2_4754004 u2 -> Hash -> Bitmap Heap Scan on sje_d1_4754000 d Recheck Cond: ((id >= 10) AND (id <= 20)) -> Bitmap Index Scan on sje_d1_pkey_4754000 Index Cond: ((id >= 10) AND (id <= 20)) (15 rows) EXECUTE sje_p(10,20); count --------------------------------------------------------------------- 11 (1 row) -- cte EXPLAIN (costs off) WITH z AS (SELECT id FROM sje_d2 WHERE id % 2 = 0) SELECT count(1) FROM sje_d1 d JOIN z USING (id) JOIN sje_d2 u2 USING (id); QUERY PLAN --------------------------------------------------------------------- Aggregate -> Custom Scan (Citus Adaptive) Task Count: 4 Tasks Shown: One of 4 -> Task Node: host=localhost port=xxxxx dbname=regression -> Aggregate -> Hash Join Hash Cond: (d.id = u2.id) -> Seq Scan on sje_d1_4754000 d -> Hash -> Seq Scan on sje_d2_4754004 u2 Filter: ((id % '2'::bigint) = 0) (13 rows) WITH z AS (SELECT id FROM sje_d2 WHERE id % 2 = 0) SELECT count(1) FROM sje_d1 d JOIN z USING (id) JOIN sje_d2 u2 USING (id); count --------------------------------------------------------------------- 51 (1 row) -- PG18 Feature: JSON functionality - JSON_TABLE has COLUMNS clause for -- extracting multiple fields from JSON documents. -- PG18 commit: https://github.com/postgres/postgres/commit/bb766cd CREATE TABLE pg18_json_test (id serial PRIMARY KEY, data JSON); INSERT INTO pg18_json_test (data) VALUES ('{ "user": {"name": "Alice", "age": 30, "city": "San Diego"} }'), ('{ "user": {"name": "Bob", "age": 25, "city": "Los Angeles"} }'), ('{ "user": {"name": "Charlie", "age": 35, "city": "Los Angeles"} }'), ('{ "user": {"name": "Diana", "age": 28, "city": "Seattle"} } '), ('{ "user": {"name": "Evan", "age": 40, "city": "Portland"} } '), ('{ "user": {"name": "Ethan", "age": 32, "city": "Seattle"} } '), ('{ "user": {"name": "Fiona", "age": 27, "city": "Seattle"} } '), ('{ "user": {"name": "George", "age": 29, "city": "San Francisco"} } '), ('{ "user": {"name": "Hannah", "age": 33, "city": "Seattle"} } '), ('{ "user": {"name": "Ian", "age": 26, "city": "Portland"} } '), ('{ "user": {"name": "Jane", "age": 38, "city": "San Francisco"} } '); SELECT jt.name, jt.age FROM pg18_json_test, JSON_TABLE( data, '$.user' COLUMNS ( age INT PATH '$.age', name TEXT PATH '$.name' ) ) AS jt WHERE jt.age between 25 and 35 ORDER BY jt.age, jt.name; name | age --------------------------------------------------------------------- Bob | 25 Ian | 26 Fiona | 27 Diana | 28 George | 29 Alice | 30 Ethan | 32 Hannah | 33 Charlie | 35 (9 rows) SELECT jt.city, count(1) FROM pg18_json_test, JSON_TABLE( data, '$.user' COLUMNS ( city TEXT PATH '$.city' ) ) AS jt GROUP BY jt.city ORDER BY count(1) DESC; city | count --------------------------------------------------------------------- Seattle | 4 San Francisco | 2 Portland | 2 Los Angeles | 2 San Diego | 1 (5 rows) -- Make it distributed and repeat the queries SELECT create_distributed_table('pg18_json_test', 'id'); 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($$pg18_nn.pg18_json_test$$) create_distributed_table --------------------------------------------------------------------- (1 row) SELECT jt.name, jt.age FROM pg18_json_test, JSON_TABLE( data, '$.user' COLUMNS ( age INT PATH '$.age', name TEXT PATH '$.name' ) ) AS jt WHERE jt.age between 25 and 35 ORDER BY jt.age, jt.name; name | age --------------------------------------------------------------------- Bob | 25 Ian | 26 Fiona | 27 Diana | 28 George | 29 Alice | 30 Ethan | 32 Hannah | 33 Charlie | 35 (9 rows) SELECT jt.city, count(1) FROM pg18_json_test, JSON_TABLE( data, '$.user' COLUMNS ( city TEXT PATH '$.city' ) ) AS jt GROUP BY jt.city ORDER BY count(1) DESC; city | count --------------------------------------------------------------------- Seattle | 4 Portland | 2 Los Angeles | 2 San Francisco | 2 San Diego | 1 (5 rows) -- PG18 Feature: WITHOUT OVERLAPS can appear in PRIMARY KEY and UNIQUE constraints. -- PG18 commit: https://github.com/postgres/postgres/commit/fc0438b4e CREATE TABLE temporal_rng ( -- Since we can't depend on having btree_gist here, -- use an int4range instead of an int. -- (The rangetypes regression test uses the same trick.) id int4range, valid_at daterange, CONSTRAINT temporal_rng_pk PRIMARY KEY (id, valid_at WITHOUT OVERLAPS) ); SELECT create_distributed_table('temporal_rng', 'id'); create_distributed_table --------------------------------------------------------------------- (1 row) -- okay: INSERT INTO temporal_rng (id, valid_at) VALUES ('[1,2)', daterange('2018-01-02', '2018-02-03')); INSERT INTO temporal_rng (id, valid_at) VALUES ('[1,2)', daterange('2018-03-03', '2018-04-04')); INSERT INTO temporal_rng (id, valid_at) VALUES ('[2,3)', daterange('2018-01-01', '2018-01-05')); INSERT INTO temporal_rng (id, valid_at) VALUES ('[3,4)', daterange('2018-01-01', NULL)); -- should fail: INSERT INTO temporal_rng (id, valid_at) VALUES ('[1,2)', daterange('2018-01-01', '2018-01-05')); ERROR: conflicting key value violates exclusion constraint "temporal_rng_pk_4754013" DETAIL: Key (id, valid_at)=([1,2), [2018-01-01,2018-01-05)) conflicts with existing key (id, valid_at)=([1,2), [2018-01-02,2018-02-03)). CONTEXT: while executing command on localhost:xxxxx -- NULLs are not allowed in the shard key: INSERT INTO temporal_rng (id, valid_at) VALUES (NULL, daterange('2018-01-01', '2018-01-05')); ERROR: cannot perform an INSERT with NULL in the partition column INSERT INTO temporal_rng (id, valid_at) VALUES ('[3,4)', NULL); ERROR: null value in column "valid_at" violates not-null constraint DETAIL: Failing row contains ([3,4), null). CONTEXT: while executing command on localhost:xxxxx -- rejects empty: INSERT INTO temporal_rng (id, valid_at) VALUES ('[3,4)', 'empty'); ERROR: empty WITHOUT OVERLAPS value found in column "valid_at" in relation "temporal_rng_4754012" CONTEXT: while executing command on localhost:xxxxx SELECT * FROM temporal_rng ORDER BY id, valid_at; id | valid_at --------------------------------------------------------------------- [1,2) | [01-02-2018,02-03-2018) [1,2) | [03-03-2018,04-04-2018) [2,3) | [01-01-2018,01-05-2018) [3,4) | [01-01-2018,) (4 rows) -- Repeat with UNIQUE constraint CREATE TABLE temporal_rng_uq ( -- Since we can't depend on having btree_gist here, -- use an int4range instead of an int. id int4range, valid_at daterange, CONSTRAINT temporal_rng_uq_uk UNIQUE (id, valid_at WITHOUT OVERLAPS) ); SELECT create_distributed_table('temporal_rng_uq', 'id'); create_distributed_table --------------------------------------------------------------------- (1 row) -- okay: INSERT INTO temporal_rng_uq (id, valid_at) VALUES ('[1,2)', daterange('2018-01-02', '2018-02-03')); INSERT INTO temporal_rng_uq (id, valid_at) VALUES ('[1,2)', daterange('2018-03-03', '2018-04-04')); INSERT INTO temporal_rng_uq (id, valid_at) VALUES ('[2,3)', daterange('2018-01-01', '2018-01-05')); INSERT INTO temporal_rng_uq (id, valid_at) VALUES ('[3,4)', daterange('2018-01-01', NULL)); -- should fail: INSERT INTO temporal_rng_uq (id, valid_at) VALUES ('[1,2)', daterange('2018-01-01', '2018-01-05')); ERROR: conflicting key value violates exclusion constraint "temporal_rng_uq_uk_4754017" DETAIL: Key (id, valid_at)=([1,2), [2018-01-01,2018-01-05)) conflicts with existing key (id, valid_at)=([1,2), [2018-01-02,2018-02-03)). CONTEXT: while executing command on localhost:xxxxx -- NULLs are not allowed in the shard key: INSERT INTO temporal_rng_uq (id, valid_at) VALUES (NULL, daterange('2018-01-01', '2018-01-05')); ERROR: cannot perform an INSERT with NULL in the partition column INSERT INTO temporal_rng_uq (id, valid_at) VALUES ('[3,4)', NULL); -- rejects empty: INSERT INTO temporal_rng_uq (id, valid_at) VALUES ('[3,4)', 'empty'); ERROR: empty WITHOUT OVERLAPS value found in column "valid_at" in relation "temporal_rng_uq_4754016" CONTEXT: while executing command on localhost:xxxxx SELECT * FROM temporal_rng_uq ORDER BY id, valid_at; id | valid_at --------------------------------------------------------------------- [1,2) | [01-02-2018,02-03-2018) [1,2) | [03-03-2018,04-04-2018) [2,3) | [01-01-2018,01-05-2018) [3,4) | [01-01-2018,) [3,4) | (5 rows) DROP TABLE temporal_rng CASCADE; DROP TABLE temporal_rng_uq CASCADE; -- Repeat the tests with the PRIMARY KEY and UNIQUE constraints added -- after the table is created and distributed. INSERTs produce the -- same results as before. CREATE TABLE temporal_rng ( -- Since we can't depend on having btree_gist here, -- use an int4range instead of an int. -- (The rangetypes regression test uses the same trick.) id int4range, valid_at daterange ); SELECT create_distributed_table('temporal_rng', 'id'); create_distributed_table --------------------------------------------------------------------- (1 row) -- okay: INSERT INTO temporal_rng (id, valid_at) VALUES ('[1,2)', daterange('2018-01-02', '2018-02-03')); INSERT INTO temporal_rng (id, valid_at) VALUES ('[1,2)', daterange('2018-03-03', '2018-04-04')); INSERT INTO temporal_rng (id, valid_at) VALUES ('[2,3)', daterange('2018-01-01', '2018-01-05')); INSERT INTO temporal_rng (id, valid_at) VALUES ('[3,4)', daterange('2018-01-01', NULL)); ALTER TABLE temporal_rng ADD CONSTRAINT temporal_rng_pk PRIMARY KEY (id, valid_at WITHOUT OVERLAPS); -- should fail: INSERT INTO temporal_rng (id, valid_at) VALUES ('[1,2)', daterange('2018-01-01', '2018-01-05')); ERROR: conflicting key value violates exclusion constraint "temporal_rng_pk_4754021" DETAIL: Key (id, valid_at)=([1,2), [2018-01-01,2018-01-05)) conflicts with existing key (id, valid_at)=([1,2), [2018-01-02,2018-02-03)). CONTEXT: while executing command on localhost:xxxxx -- NULLs are not allowed in the shard key: INSERT INTO temporal_rng (id, valid_at) VALUES (NULL, daterange('2018-01-01', '2018-01-05')); ERROR: cannot perform an INSERT with NULL in the partition column INSERT INTO temporal_rng (id, valid_at) VALUES ('[3,4)', NULL); ERROR: null value in column "valid_at" violates not-null constraint DETAIL: Failing row contains ([3,4), null). CONTEXT: while executing command on localhost:xxxxx -- rejects empty: INSERT INTO temporal_rng (id, valid_at) VALUES ('[3,4)', 'empty'); ERROR: empty WITHOUT OVERLAPS value found in column "valid_at" in relation "temporal_rng_4754020" CONTEXT: while executing command on localhost:xxxxx SELECT * FROM temporal_rng ORDER BY id, valid_at; id | valid_at --------------------------------------------------------------------- [1,2) | [01-02-2018,02-03-2018) [1,2) | [03-03-2018,04-04-2018) [2,3) | [01-01-2018,01-05-2018) [3,4) | [01-01-2018,) (4 rows) -- Repeat with UNIQUE constraint CREATE TABLE temporal_rng_uq ( -- Since we can't depend on having btree_gist here, -- use an int4range instead of an int. id int4range, valid_at daterange ); SELECT create_distributed_table('temporal_rng_uq', 'id'); create_distributed_table --------------------------------------------------------------------- (1 row) -- okay: INSERT INTO temporal_rng_uq (id, valid_at) VALUES ('[1,2)', daterange('2018-01-02', '2018-02-03')); INSERT INTO temporal_rng_uq (id, valid_at) VALUES ('[1,2)', daterange('2018-03-03', '2018-04-04')); INSERT INTO temporal_rng_uq (id, valid_at) VALUES ('[2,3)', daterange('2018-01-01', '2018-01-05')); INSERT INTO temporal_rng_uq (id, valid_at) VALUES ('[3,4)', daterange('2018-01-01', NULL)); ALTER TABLE temporal_rng_uq ADD CONSTRAINT temporal_rng_uq_uk UNIQUE (id, valid_at WITHOUT OVERLAPS); -- should fail: INSERT INTO temporal_rng_uq (id, valid_at) VALUES ('[1,2)', daterange('2018-01-01', '2018-01-05')); ERROR: conflicting key value violates exclusion constraint "temporal_rng_uq_uk_4754025" DETAIL: Key (id, valid_at)=([1,2), [2018-01-01,2018-01-05)) conflicts with existing key (id, valid_at)=([1,2), [2018-01-02,2018-02-03)). CONTEXT: while executing command on localhost:xxxxx -- NULLs are not allowed in the shard key: INSERT INTO temporal_rng_uq (id, valid_at) VALUES (NULL, daterange('2018-01-01', '2018-01-05')); ERROR: cannot perform an INSERT with NULL in the partition column INSERT INTO temporal_rng_uq (id, valid_at) VALUES ('[3,4)', NULL); -- rejects empty: INSERT INTO temporal_rng_uq (id, valid_at) VALUES ('[3,4)', 'empty'); ERROR: empty WITHOUT OVERLAPS value found in column "valid_at" in relation "temporal_rng_uq_4754024" CONTEXT: while executing command on localhost:xxxxx SELECT * FROM temporal_rng_uq ORDER BY id, valid_at; id | valid_at --------------------------------------------------------------------- [1,2) | [01-02-2018,02-03-2018) [1,2) | [03-03-2018,04-04-2018) [2,3) | [01-01-2018,01-05-2018) [3,4) | [01-01-2018,) [3,4) | (5 rows) -- PG18 Feature: RETURNING old and new values in DML statements -- PG18 commit: https://github.com/postgres/postgres/commit/80feb727c CREATE TABLE users (id SERIAL PRIMARY KEY, email text, category int); INSERT INTO users (email, category) SELECT 'xxx@foo.com', i % 10 from generate_series (1,100) t(i); SELECT create_distributed_table('users','id'); 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($$pg18_nn.users$$) create_distributed_table --------------------------------------------------------------------- (1 row) UPDATE users SET email = 'colm@planet.com' WHERE id = 1 RETURNING OLD.email AS previous_email, NEW.email AS current_email; previous_email | current_email --------------------------------------------------------------------- xxx@foo.com | colm@planet.com (1 row) SELECT * FROM users WHERE id = 1 ORDER BY id; id | email | category --------------------------------------------------------------------- 1 | colm@planet.com | 1 (1 row) UPDATE users SET email = 'tim@arctic.net' WHERE id = 22 RETURNING OLD.email AS previous_email, NEW.email AS current_email; previous_email | current_email --------------------------------------------------------------------- xxx@foo.com | tim@arctic.net (1 row) UPDATE users SET email = 'john@farm.ie' WHERE id = 33 RETURNING OLD.email AS previous_email, NEW.email AS current_email; previous_email | current_email --------------------------------------------------------------------- xxx@foo.com | john@farm.ie (1 row) SELECT * FROM users WHERE id = 22 ORDER BY id; id | email | category --------------------------------------------------------------------- 22 | tim@arctic.net | 2 (1 row) SELECT * FROM users WHERE email not like 'xxx@%' ORDER BY id; id | email | category --------------------------------------------------------------------- 1 | colm@planet.com | 1 22 | tim@arctic.net | 2 33 | john@farm.ie | 3 (3 rows) -- NULL values creep into the email column.. INSERT INTO users (email, category) VALUES (null, 5) RETURNING OLD.email AS previous_email, NEW.email AS current_email; previous_email | current_email --------------------------------------------------------------------- | (1 row) UPDATE users SET email = NULL WHERE id = 79 RETURNING OLD.email AS previous_email, NEW.email AS current_email; previous_email | current_email --------------------------------------------------------------------- xxx@foo.com | (1 row) -- Now add a NOT NULL constraint on email, but do -- not apply it to existing rows yet. ALTER TABLE users ADD CONSTRAINT users_email_not_null CHECK (email IS NOT NULL) NOT VALID; UPDATE users SET email = NULL WHERE id = 50 RETURNING OLD.email AS previous_email, NEW.email AS current_email; ERROR: new row for relation "users_4754028" violates check constraint "users_email_not_null_4754028" DETAIL: Failing row contains (50, null, 0). CONTEXT: while executing command on localhost:xxxxx -- Validation should fail due to existing NULLs ALTER TABLE users VALIDATE CONSTRAINT users_email_not_null; ERROR: check constraint "users_email_not_null_4754028" of relation "users_4754028" is violated by some row CONTEXT: while executing command on localhost:xxxxx -- Fix NULL emails to a default value UPDATE users SET email = 'xxx@foo.com' WHERE email IS NULL RETURNING OLD.email AS previous_email, NEW.email AS current_email; previous_email | current_email --------------------------------------------------------------------- | xxx@foo.com | xxx@foo.com (2 rows) -- Validation should now succeed ALTER TABLE users VALIDATE CONSTRAINT users_email_not_null; -- And prevent future NULLs INSERT INTO users (email, category) VALUES (null, 10) RETURNING OLD.email AS previous_email, NEW.email AS current_email; ERROR: new row for relation "users_4754030" violates check constraint "users_email_not_null_4754030" DETAIL: Failing row contains (102, null, 10). CONTEXT: while executing command on localhost:xxxxx -- PG18 Feature: support for LIKE in CREATE FOREIGN TABLE -- PG18 commit: https://github.com/postgres/postgres/commit/302cf1575 SET citus.use_citus_managed_tables TO ON; CREATE EXTENSION postgres_fdw; CREATE SERVER foreign_server FOREIGN DATA WRAPPER postgres_fdw OPTIONS (host 'localhost', port :'master_port', dbname 'regression'); CREATE USER MAPPING FOR CURRENT_USER SERVER foreign_server OPTIONS (user 'postgres'); CREATE TABLE ctl_table(a int PRIMARY KEY, b varchar COMPRESSION pglz, c int GENERATED ALWAYS AS (a * 2) STORED, d bigint GENERATED ALWAYS AS IDENTITY, e int DEFAULT 1); CREATE INDEX ctl_table_ab_key ON ctl_table(a, b); COMMENT ON COLUMN ctl_table.b IS 'Column b'; CREATE STATISTICS ctl_table_stat ON a,b FROM ctl_table; INSERT INTO ctl_table VALUES (1, 'first'), (2, 'second'), (3, 'third'), (4, 'fourth'); -- Test EXCLUDING ALL CREATE FOREIGN TABLE ctl_ft1(LIKE ctl_table EXCLUDING ALL) SERVER foreign_server OPTIONS (schema_name 'pg18_nn', table_name 'ctl_table'); -- Test INCLUDING ALL CREATE FOREIGN TABLE ctl_ft2(LIKE ctl_table INCLUDING ALL) SERVER foreign_server OPTIONS (schema_name 'pg18_nn', table_name 'ctl_table'); -- check that the foreign tables are citus local table SELECT partmethod, repmodel FROM pg_dist_partition WHERE logicalrelid IN ('ctl_ft1'::regclass, 'ctl_ft2'::regclass) ORDER BY logicalrelid; partmethod | repmodel --------------------------------------------------------------------- n | s n | s (2 rows) -- we can query the foreign tables EXPLAIN (VERBOSE, COSTS OFF) SELECT * FROM ctl_ft1 ORDER BY a; QUERY PLAN --------------------------------------------------------------------- Custom Scan (Citus Adaptive) Output: remote_scan.a, remote_scan.b, remote_scan.c, remote_scan.d, remote_scan.e Task Count: 1 Tasks Shown: All -> Task Query: SELECT a, b, c, d, e FROM pg18_nn.ctl_ft1_4754033 ctl_ft1 ORDER BY a Node: host=localhost port=xxxxx dbname=regression -> Foreign Scan on pg18_nn.ctl_ft1_4754033 ctl_ft1 Output: a, b, c, d, e Remote SQL: SELECT a, b, c, d, e FROM pg18_nn.ctl_table ORDER BY a ASC NULLS LAST (10 rows) SELECT * FROM ctl_ft1 ORDER BY a; a | b | c | d | e --------------------------------------------------------------------- 1 | first | 2 | 1 | 1 2 | second | 4 | 2 | 1 3 | third | 6 | 3 | 1 4 | fourth | 8 | 4 | 1 (4 rows) EXPLAIN (VERBOSE, COSTS OFF) SELECT * FROM ctl_ft2 ORDER BY a; QUERY PLAN --------------------------------------------------------------------- Custom Scan (Citus Adaptive) Output: remote_scan.a, remote_scan.b, remote_scan.c, remote_scan.d, remote_scan.e Task Count: 1 Tasks Shown: All -> Task Query: SELECT a, b, c, d, e FROM pg18_nn.ctl_ft2_4754034 ctl_ft2 ORDER BY a Node: host=localhost port=xxxxx dbname=regression -> Foreign Scan on pg18_nn.ctl_ft2_4754034 ctl_ft2 Output: a, b, c, d, e Remote SQL: SELECT a, b, c, d, e FROM pg18_nn.ctl_table ORDER BY a ASC NULLS LAST (10 rows) SELECT * FROM ctl_ft2 ORDER BY a; a | b | c | d | e --------------------------------------------------------------------- 1 | first | 2 | 1 | 1 2 | second | 4 | 2 | 1 3 | third | 6 | 3 | 1 4 | fourth | 8 | 4 | 1 (4 rows) -- Clean up foreign table test RESET citus.use_citus_managed_tables; SELECT undistribute_table('ctl_ft1'); NOTICE: creating a new table for pg18_nn.ctl_ft1 NOTICE: dropping the old pg18_nn.ctl_ft1 NOTICE: renaming the new table to pg18_nn.ctl_ft1 undistribute_table --------------------------------------------------------------------- (1 row) SELECT undistribute_table('ctl_ft2'); NOTICE: creating a new table for pg18_nn.ctl_ft2 NOTICE: dropping the old pg18_nn.ctl_ft2 NOTICE: renaming the new table to pg18_nn.ctl_ft2 undistribute_table --------------------------------------------------------------------- (1 row) DROP SERVER foreign_server CASCADE; NOTICE: drop cascades to 3 other objects DETAIL: drop cascades to user mapping for postgres on server foreign_server drop cascades to foreign table ctl_ft1 drop cascades to foreign table ctl_ft2 -- PG18 Feature: PERIOD clause in foreign key constraint definitions. -- PG18 commit: https://github.com/postgres/postgres/commit/89f908a6d -- This test verifies that the PG18 tests apply to Citus tables CREATE EXTENSION btree_gist; -- needed for range type indexing CREATE TABLE temporal_test ( id integer, valid_at daterange, CONSTRAINT temporal_test_pk PRIMARY KEY (id, valid_at WITHOUT OVERLAPS) ); SET citus.shard_count TO 4; SELECT create_reference_table( 'temporal_test'); create_reference_table --------------------------------------------------------------------- (1 row) INSERT INTO temporal_test VALUES (1, '[2000-01-01,2001-01-01)'); -- same key, doesn't overlap: INSERT INTO temporal_test VALUES (1, '[2001-01-01,2002-01-01)'); -- overlaps but different key: INSERT INTO temporal_test VALUES (2, '[2000-01-01,2001-01-01)'); -- should fail: INSERT INTO temporal_test VALUES (1, '[2000-06-01,2001-01-01)'); ERROR: conflicting key value violates exclusion constraint "temporal_test_pk_4754035" DETAIL: Key (id, valid_at)=(1, [2000-06-01,2001-01-01)) conflicts with existing key (id, valid_at)=(1, [2000-01-01,2001-01-01)). CONTEXT: while executing command on localhost:xxxxx -- Required for foreign key constraint on distributed table SET citus.shard_replication_factor TO 1; -- Create and distribute a table with temporal foreign key constraints CREATE TABLE temporal_fk_rng2rng ( id integer, valid_at daterange, parent_id integer, CONSTRAINT temporal_fk_rng2rng_pk PRIMARY KEY (id, valid_at WITHOUT OVERLAPS) ); SELECT create_distributed_table( 'temporal_fk_rng2rng', 'id'); create_distributed_table --------------------------------------------------------------------- (1 row) -- -- Add foreign key constraint with PERIOD clause -- This is propagated to worker shards ALTER TABLE temporal_fk_rng2rng ADD CONSTRAINT temporal_fk_rng2rng_fk FOREIGN KEY (parent_id, PERIOD valid_at) REFERENCES temporal_test (id, PERIOD valid_at); INSERT INTO temporal_fk_rng2rng VALUES (1, '[2000-01-01,2001-01-01)', 1); -- okay spanning two parent records: INSERT INTO temporal_fk_rng2rng VALUES (2, '[2000-01-01,2002-01-01)', 1); -- key is missing INSERT INTO temporal_fk_rng2rng VALUES (3, '[2000-01-01,2001-01-01)', 3); ERROR: insert or update on table "temporal_fk_rng2rng_4754037" violates foreign key constraint "temporal_fk_rng2rng_fk_4754037" DETAIL: Key (parent_id, valid_at)=(3, [2000-01-01,2001-01-01)) is not present in table "temporal_test_4754035". CONTEXT: while executing command on localhost:xxxxx -- key exist but is outside range INSERT INTO temporal_fk_rng2rng VALUES (4, '[2001-01-01,2002-01-01)', 2); ERROR: insert or update on table "temporal_fk_rng2rng_4754037" violates foreign key constraint "temporal_fk_rng2rng_fk_4754037" DETAIL: Key (parent_id, valid_at)=(2, [2001-01-01,2002-01-01)) is not present in table "temporal_test_4754035". CONTEXT: while executing command on localhost:xxxxx -- key exist but is partly outside range INSERT INTO temporal_fk_rng2rng VALUES (5, '[2000-01-01,2002-01-01)', 2); ERROR: insert or update on table "temporal_fk_rng2rng_4754036" violates foreign key constraint "temporal_fk_rng2rng_fk_4754036" DETAIL: Key (parent_id, valid_at)=(2, [2000-01-01,2002-01-01)) is not present in table "temporal_test_4754035". CONTEXT: while executing command on localhost:xxxxx -- PG18 Feature: REJECT_LIMIT option for COPY errors -- PG18 commit: https://github.com/postgres/postgres/commit/4ac2a9bec -- Citus does not support COPY with ON_ERROR so just need to -- ensure the appropriate error is returned. CREATE TABLE check_ign_err (n int, m int[], k int); SELECT create_distributed_table('check_ign_err', 'n'); create_distributed_table --------------------------------------------------------------------- (1 row) COPY check_ign_err FROM STDIN WITH (on_error stop, reject_limit 5); ERROR: Citus does not support COPY FROM with ON_ERROR option. COPY check_ign_err FROM STDIN WITH (ON_ERROR ignore, REJECT_LIMIT 100); ERROR: Citus does not support COPY FROM with ON_ERROR option. COPY check_ign_err FROM STDIN WITH (on_error ignore, log_verbosity verbose, reject_limit 50); ERROR: Citus does not support COPY FROM with ON_ERROR option. COPY check_ign_err FROM STDIN WITH (reject_limt 77, log_verbosity verbose, on_error ignore); ERROR: Citus does not support COPY FROM with ON_ERROR option. -- PG requires on_error when reject_limit is specified COPY check_ign_err FROM STDIN WITH (reject_limit 100); ERROR: COPY REJECT_LIMIT requires ON_ERROR to be set to IGNORE -- PG18 Feature: COPY TABLE TO on a materialized view -- PG18 commit: https://github.com/postgres/postgres/commit/534874fac -- This does not work in Citus as a materialized view cannot be distributed. -- So just verify that the appropriate error is raised. CREATE MATERIALIZED VIEW copytest_mv AS SELECT i as id, md5(i::text) as hashval FROM generate_series(1,100) i; -- Attempting to make it distributed should fail with appropriate error as -- Citus does not yet support materialized views. SELECT create_distributed_table('copytest_mv', 'id'); ERROR: copytest_mv is not a regular, foreign or partitioned table -- After that, any command on the materialized view is outside Citus support. -- PG18 Feature: text search with nondeterministic collations -- PG18 commit: https://github.com/postgres/postgres/commit/329304c90 -- This test verifies that the PG18 tests apply to Citus tables; Citus -- just passes through the collation info and text search queries to -- worker shards. CREATE COLLATION ignore_accents (provider = icu, locale = '@colStrength=primary;colCaseLevel=yes', deterministic = false); NOTICE: using standard form "und-u-kc-ks-level1" for ICU locale "@colStrength=primary;colCaseLevel=yes" -- nondeterministic collations CREATE COLLATION ctest_det (provider = icu, locale = '', deterministic = true); NOTICE: using standard form "und" for ICU locale "" CREATE COLLATION ctest_nondet (provider = icu, locale = '', deterministic = false); NOTICE: using standard form "und" for ICU locale "" CREATE TABLE strtest1 (a int, b text); SELECT create_distributed_table('strtest1', 'a'); create_distributed_table --------------------------------------------------------------------- (1 row) INSERT INTO strtest1 VALUES (1, U&'zy\00E4bc'); INSERT INTO strtest1 VALUES (2, U&'zy\0061\0308bc'); INSERT INTO strtest1 VALUES (3, U&'ab\00E4cd'); INSERT INTO strtest1 VALUES (4, U&'ab\0061\0308cd'); INSERT INTO strtest1 VALUES (5, U&'ab\00E4cd'); INSERT INTO strtest1 VALUES (6, U&'ab\0061\0308cd'); INSERT INTO strtest1 VALUES (7, U&'ab\00E4cd'); SELECT * FROM strtest1 WHERE b = 'zyäbc' COLLATE ctest_det ORDER BY a; a | b --------------------------------------------------------------------- 1 | zyäbc (1 row) SELECT * FROM strtest1 WHERE b = 'zyäbc' COLLATE ctest_nondet ORDER BY a; a | b --------------------------------------------------------------------- 1 | zyäbc 2 | zyäbc (2 rows) SELECT strpos(b COLLATE ctest_det, 'bc') FROM strtest1 ORDER BY a; strpos --------------------------------------------------------------------- 4 5 0 0 0 0 0 (7 rows) SELECT strpos(b COLLATE ctest_nondet, 'bc') FROM strtest1 ORDER BY a; strpos --------------------------------------------------------------------- 4 5 0 0 0 0 0 (7 rows) SELECT replace(b COLLATE ctest_det, U&'\00E4b', 'X') FROM strtest1 ORDER BY a; replace --------------------------------------------------------------------- zyXc zyäbc abäcd abäcd abäcd abäcd abäcd (7 rows) SELECT replace(b COLLATE ctest_nondet, U&'\00E4b', 'X') FROM strtest1 ORDER BY a; replace --------------------------------------------------------------------- zyXc zyXc abäcd abäcd abäcd abäcd abäcd (7 rows) SELECT a, split_part(b COLLATE ctest_det, U&'\00E4b', 2) FROM strtest1 ORDER BY a; a | split_part --------------------------------------------------------------------- 1 | c 2 | 3 | 4 | 5 | 6 | 7 | (7 rows) SELECT a, split_part(b COLLATE ctest_nondet, U&'\00E4b', 2) FROM strtest1 ORDER BY a; a | split_part --------------------------------------------------------------------- 1 | c 2 | c 3 | 4 | 5 | 6 | 7 | (7 rows) SELECT a, split_part(b COLLATE ctest_det, U&'\00E4b', -1) FROM strtest1 ORDER BY a; a | split_part --------------------------------------------------------------------- 1 | c 2 | zyäbc 3 | abäcd 4 | abäcd 5 | abäcd 6 | abäcd 7 | abäcd (7 rows) SELECT a, split_part(b COLLATE ctest_nondet, U&'\00E4b', -1) FROM strtest1 ORDER BY a; a | split_part --------------------------------------------------------------------- 1 | c 2 | c 3 | abäcd 4 | abäcd 5 | abäcd 6 | abäcd 7 | abäcd (7 rows) SELECT a, string_to_array(b COLLATE ctest_det, U&'\00E4b') FROM strtest1 ORDER BY a; a | string_to_array --------------------------------------------------------------------- 1 | {zy,c} 2 | {zyäbc} 3 | {abäcd} 4 | {abäcd} 5 | {abäcd} 6 | {abäcd} 7 | {abäcd} (7 rows) SELECT a, string_to_array(b COLLATE ctest_nondet, U&'\00E4b') FROM strtest1 ORDER BY a; a | string_to_array --------------------------------------------------------------------- 1 | {zy,c} 2 | {zy,c} 3 | {abäcd} 4 | {abäcd} 5 | {abäcd} 6 | {abäcd} 7 | {abäcd} (7 rows) SELECT * FROM strtest1 WHERE b LIKE 'zyäbc' COLLATE ctest_det ORDER BY a; a | b --------------------------------------------------------------------- 1 | zyäbc (1 row) SELECT * FROM strtest1 WHERE b LIKE 'zyäbc' COLLATE ctest_nondet ORDER BY a; a | b --------------------------------------------------------------------- 1 | zyäbc 2 | zyäbc (2 rows) CREATE TABLE strtest2 (a int, b text); SELECT create_distributed_table('strtest2', 'a'); create_distributed_table --------------------------------------------------------------------- (1 row) INSERT INTO strtest2 VALUES (1, 'cote'), (2, 'côte'), (3, 'coté'), (4, 'côté'); CREATE TABLE strtest2nfd (a int, b text); SELECT create_distributed_table('strtest2nfd', 'a'); create_distributed_table --------------------------------------------------------------------- (1 row) INSERT INTO strtest2nfd VALUES (1, 'cote'), (2, 'côte'), (3, 'coté'), (4, 'côté'); UPDATE strtest2nfd SET b = normalize(b, nfd); -- This shows why replace should be greedy. Otherwise, in the NFD -- case, the match would stop before the decomposed accents, which -- would leave the accents in the results. SELECT a, b, replace(b COLLATE ignore_accents, 'co', 'ma') FROM strtest2 ORDER BY a, b; a | b | replace --------------------------------------------------------------------- 1 | cote | mate 2 | côte | mate 3 | coté | maté 4 | côté | maté (4 rows) SELECT a, b, replace(b COLLATE ignore_accents, 'co', 'ma') FROM strtest2nfd ORDER BY a, b; a | b | replace --------------------------------------------------------------------- 1 | cote | mate 2 | côte | mate 3 | coté | maté 4 | côté | maté (4 rows) -- PG18 Feature: LIKE support for non-deterministic collations -- PG18 commit: https://github.com/postgres/postgres/commit/85b7efa1c -- As with non-deterministic collation text search, we verify that -- LIKE with non-deterministic collation is passed through by Citus -- and expected results are returned by the queries. INSERT INTO strtest1 VALUES (8, U&'abc'); INSERT INTO strtest1 VALUES (9, 'abc'); SELECT a, b FROM strtest1 WHERE b LIKE 'abc' COLLATE ctest_det ORDER BY a; a | b --------------------------------------------------------------------- 8 | abc 9 | abc (2 rows) SELECT a, b FROM strtest1 WHERE b LIKE 'a\bc' COLLATE ctest_det ORDER BY a; a | b --------------------------------------------------------------------- 8 | abc 9 | abc (2 rows) SELECT a, b FROM strtest1 WHERE b LIKE 'abc' COLLATE ctest_nondet ORDER BY a; a | b --------------------------------------------------------------------- 8 | abc 9 | abc (2 rows) SELECT a, b FROM strtest1 WHERE b LIKE 'a\bc' COLLATE ctest_nondet ORDER BY a; a | b --------------------------------------------------------------------- 8 | abc 9 | abc (2 rows) CREATE COLLATION case_insensitive (provider = icu, locale = '@colStrength=secondary', deterministic = false); NOTICE: using standard form "und-u-ks-level2" for ICU locale "@colStrength=secondary" SELECT a, b FROM strtest1 WHERE b LIKE 'ABC' COLLATE case_insensitive ORDER BY a; a | b --------------------------------------------------------------------- 8 | abc 9 | abc (2 rows) SELECT a, b FROM strtest1 WHERE b LIKE 'ABC%' COLLATE case_insensitive ORDER BY a; a | b --------------------------------------------------------------------- 8 | abc 9 | abc (2 rows) INSERT INTO strtest1 VALUES (10, U&'\00E4bc'); INSERT INTO strtest1 VALUES (12, U&'\0061\0308bc'); SELECT * FROM strtest1 WHERE b LIKE 'äbc' COLLATE ctest_det ORDER BY a; a | b --------------------------------------------------------------------- 10 | äbc (1 row) SELECT * FROM strtest1 WHERE b LIKE 'äbc' COLLATE ctest_nondet ORDER BY a; a | b --------------------------------------------------------------------- 10 | äbc 12 | äbc (2 rows) -- Tests with ignore_accents collation. Taken from -- PG18 regress tests and applied to a Citus table. INSERT INTO strtest1 VALUES (10, U&'\0061\0308bc'); INSERT INTO strtest1 VALUES (11, U&'\00E4bc'); INSERT INTO strtest1 VALUES (12, U&'cb\0061\0308'); INSERT INTO strtest1 VALUES (13, U&'\0308bc'); INSERT INTO strtest1 VALUES (14, 'foox'); SELECT a, b FROM strtest1 WHERE b LIKE U&'\00E4_c' COLLATE ignore_accents ORDER BY a, b; a | b --------------------------------------------------------------------- 8 | abc 9 | abc 10 | äbc 10 | äbc 11 | äbc 12 | äbc (6 rows) -- and in reverse: SELECT a, b FROM strtest1 WHERE b LIKE U&'\0061\0308_c' COLLATE ignore_accents ORDER BY a, b; a | b --------------------------------------------------------------------- 8 | abc 9 | abc 10 | äbc 10 | äbc 11 | äbc 12 | äbc (6 rows) -- inner % matches b: SELECT a, b FROM strtest1 WHERE b LIKE U&'\00E4%c' COLLATE ignore_accents ORDER BY a, b; a | b --------------------------------------------------------------------- 8 | abc 9 | abc 10 | äbc 10 | äbc 11 | äbc 12 | äbc (6 rows) -- inner %% matches b then zero: SELECT a, b FROM strtest1 WHERE b LIKE U&'\00E4%%c' COLLATE ignore_accents ORDER BY a, b; a | b --------------------------------------------------------------------- 8 | abc 9 | abc 10 | äbc 10 | äbc 11 | äbc 12 | äbc (6 rows) -- inner %% matches b then zero: SELECT a, b FROM strtest1 WHERE b LIKE U&'c%%\00E4' COLLATE ignore_accents ORDER BY a, b; a | b --------------------------------------------------------------------- 12 | cbä (1 row) -- trailing _ matches two codepoints that form one grapheme: SELECT a, b FROM strtest1 WHERE b LIKE U&'cb_' COLLATE ignore_accents ORDER BY a, b; a | b --------------------------------------------------------------------- (0 rows) -- trailing __ matches two codepoints that form one grapheme: SELECT a, b FROM strtest1 WHERE b LIKE U&'cb__' COLLATE ignore_accents ORDER BY a, b; a | b --------------------------------------------------------------------- 12 | cbä (1 row) -- leading % matches zero: SELECT a, b FROM strtest1 WHERE b LIKE U&'%\00E4bc' COLLATE ignore_accents ORDER BY a; a | b --------------------------------------------------------------------- 1 | zyäbc 2 | zyäbc 8 | abc 9 | abc 10 | äbc 10 | äbc 11 | äbc 12 | äbc (8 rows) -- leading % matches zero (with later %): SELECT a, b FROM strtest1 WHERE b LIKE U&'%\00E4%c' COLLATE ignore_accents ORDER BY a, b; a | b --------------------------------------------------------------------- 1 | zyäbc 2 | zyäbc 8 | abc 9 | abc 10 | äbc 10 | äbc 11 | äbc 12 | äbc (8 rows) -- trailing % matches zero: SELECT a, b FROM strtest1 WHERE b LIKE U&'\00E4bc%' COLLATE ignore_accents ORDER BY a, b; a | b --------------------------------------------------------------------- 8 | abc 9 | abc 10 | äbc 10 | äbc 11 | äbc 12 | äbc (6 rows) -- trailing % matches zero (with previous %): SELECT a, b FROM strtest1 WHERE b LIKE U&'\00E4%c%' COLLATE ignore_accents ORDER BY a, b; a | b --------------------------------------------------------------------- 3 | abäcd 4 | abäcd 5 | abäcd 6 | abäcd 7 | abäcd 8 | abc 9 | abc 10 | äbc 10 | äbc 11 | äbc 12 | äbc (11 rows) -- _ versus two codepoints that form one grapheme: SELECT a, b FROM strtest1 WHERE b LIKE U&'_bc' COLLATE ignore_accents ORDER BY a, b; a | b --------------------------------------------------------------------- 8 | abc 9 | abc 10 | äbc 10 | äbc 11 | äbc 12 | äbc 13 | ̈bc (7 rows) -- (actually this matches because) SELECT a, b FROM strtest1 WHERE b = 'bc' COLLATE ignore_accents ORDER BY a, b; a | b --------------------------------------------------------------------- 13 | ̈bc (1 row) -- __ matches two codepoints that form one grapheme: SELECT a, b FROM strtest1 WHERE b LIKE U&'__bc' COLLATE ignore_accents ORDER BY a, b; a | b --------------------------------------------------------------------- 10 | äbc 12 | äbc (2 rows) -- _ matches one codepoint that forms half a grapheme: SELECT a, b FROM strtest1 WHERE b LIKE U&'_\0308bc' COLLATE ignore_accents ORDER BY a, b; a | b --------------------------------------------------------------------- 8 | abc 9 | abc 10 | äbc 10 | äbc 11 | äbc 12 | äbc 13 | ̈bc (7 rows) -- doesn't match because \00e4 doesn't match only \0308 SELECT a, b FROM strtest1 WHERE b LIKE U&'_\00e4bc' COLLATE ignore_accents ORDER BY a, b; a | b --------------------------------------------------------------------- (0 rows) -- escape character at end of pattern SELECT a, b FROM strtest1 WHERE b LIKE 'foo\' COLLATE ignore_accents ORDER BY a, b; ERROR: LIKE pattern must not end with escape character CONTEXT: while executing command on localhost:xxxxx -- cleanup with minimum verbosity SET client_min_messages TO ERROR; RESET search_path; RESET citus.shard_count; RESET citus.shard_replication_factor; DROP SCHEMA pg18_nn CASCADE; RESET client_min_messages;