mirror of
https://github.com/timescale/timescaledb.git
synced 2025-05-23 14:39:15 +08:00
a3d8f9fecd
The current SELECT DISTINCT pushdown code assumed that the targetlist will always contain references to column attributes of the target table. So, even a simple "SELECT DISTINCT 1 from table;" causes a segmentation fault because the "varno" field is not assigned. Fix this oversight. Issue reported by @svenklemm Fixes timescale/timescaledb-private#920
142 lines
4.2 KiB
SQL
142 lines
4.2 KiB
SQL
-- This file and its contents are licensed under the Timescale License.
|
|
-- Please see the included NOTICE for copyright information and
|
|
-- LICENSE-TIMESCALE for a copy of the license.
|
|
|
|
\echo '%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%'
|
|
\echo '%%% RUNNING TESTS on table:' :TABLE_NAME
|
|
\echo '%%% PREFIX:' :PREFIX
|
|
\echo '%%% ORDER_BY_1:' :ORDER_BY_1
|
|
\echo '%%% ORDER_BY_1_2:' :ORDER_BY_1_2
|
|
\echo '%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%'
|
|
|
|
-- Test SkipScan with SELECT DISTINCT in multi-node environment
|
|
-- Ensure that a Unique plan gets chosen on the access node
|
|
SET enable_hashagg TO false;
|
|
\qecho Unique plan on access node for SELECT DISTINCT
|
|
:PREFIX
|
|
SELECT DISTINCT device_id
|
|
FROM :TABLE_NAME
|
|
:ORDER_BY_1
|
|
LIMIT 10;
|
|
RESET enable_hashagg;
|
|
|
|
SET timescaledb.enable_per_data_node_queries = true;
|
|
-- SELECT DISTINCT on expressions won't be pushed down
|
|
\qecho SELECT DISTINCT on expressions is not pushed down
|
|
:PREFIX
|
|
SELECT DISTINCT device_id*v1
|
|
FROM :TABLE_NAME
|
|
:ORDER_BY_1
|
|
LIMIT 10;
|
|
|
|
SET timescaledb.enable_remote_explain = ON;
|
|
-- SELECT DISTINCT on column with index should use SkipScan
|
|
\qecho SELECT DISTINCT on column with index uses SkipScan
|
|
:PREFIX
|
|
SELECT DISTINCT device_id
|
|
FROM :TABLE_NAME
|
|
:ORDER_BY_1
|
|
LIMIT 10;
|
|
|
|
-- SELECT DISTINCT with constants and NULLs in targetlist should use SkipScan
|
|
\qecho SELECT DISTINCT with constants and NULLs in targetlist uses SkipScan
|
|
:PREFIX
|
|
SELECT DISTINCT device_id, NULL, 'const1'
|
|
FROM :TABLE_NAME
|
|
:ORDER_BY_1
|
|
LIMIT 10;
|
|
|
|
-- SELECT DISTINCT with a mix of constants and columns should send only
|
|
-- the columns to the remote side. However SkipScan won't be used because
|
|
-- right now only single column is supported in SkipScans
|
|
\qecho SELECT DISTINCT only sends columns to the data nodes
|
|
:PREFIX
|
|
SELECT DISTINCT device_id, time, NULL, 'const1'
|
|
FROM :TABLE_NAME
|
|
:ORDER_BY_1_2
|
|
LIMIT 10;
|
|
|
|
-- SELECT DISTINCT will be pushed down in the attribute attno order. This
|
|
-- is ok because "DISTINCT SELECT col1, col2" returns the same values
|
|
-- (subject to ORDER BY clauses) as "DISTINCE SELECT col2, col1"
|
|
\qecho SELECT DISTINCE is pushed down in attribute attno order
|
|
:PREFIX
|
|
SELECT DISTINCT device_id, time
|
|
FROM :TABLE_NAME
|
|
:ORDER_BY_1_2
|
|
LIMIT 10;
|
|
|
|
-- SELECT DISTINCT ON on multiple columns will be pushed to the remote side.
|
|
-- However SkipScan won't be used since only one column is supported
|
|
\qecho SELECT DISTINCT ON multiple columns is pushed to data nodes
|
|
:PREFIX
|
|
SELECT DISTINCT ON (device_id, time) device_id, time
|
|
FROM :TABLE_NAME
|
|
:ORDER_BY_1_2
|
|
LIMIT 10;
|
|
|
|
-- Another variation with SELECT DISTINCT
|
|
\qecho SELECT DISTINCT within a sub-select
|
|
:PREFIX
|
|
SELECT device_id, time, 'const1' FROM (SELECT DISTINCT ON (device_id) device_id, time
|
|
FROM :TABLE_NAME
|
|
:ORDER_BY_1_2
|
|
LIMIT 10) a;
|
|
|
|
-- Ensure that SELECT DISTINCT pushdown happens even with below disabled
|
|
SET timescaledb.enable_per_data_node_queries = false;
|
|
\qecho SELECT DISTINCT works with enable_per_data_node_queries disabled
|
|
:PREFIX
|
|
SELECT DISTINCT device_id
|
|
FROM :TABLE_NAME
|
|
:ORDER_BY_1
|
|
LIMIT 10;
|
|
|
|
SET timescaledb.enable_per_data_node_queries = true;
|
|
SET timescaledb.enable_remote_explain = OFF;
|
|
|
|
-- SELECT DISTINCT should not have duplicate columns
|
|
\qecho SELECT DISTINCT should not have duplicate columns
|
|
:PREFIX
|
|
SELECT DISTINCT device_id, device_id
|
|
FROM :TABLE_NAME
|
|
:ORDER_BY_1;
|
|
|
|
-- SELECT DISTINCT handles whole row correctly
|
|
\qecho SELECT DISTINCT handles whole row correctly
|
|
:PREFIX
|
|
SELECT DISTINCT *
|
|
FROM :TABLE_NAME
|
|
:ORDER_BY_1_2
|
|
LIMIT 10;
|
|
|
|
-- SELECT DISTINCT ON handles whole row correctly
|
|
\qecho SELECT DISTINCT ON (expr) handles whole row correctly
|
|
:PREFIX
|
|
SELECT DISTINCT ON (device_id) *
|
|
FROM :TABLE_NAME
|
|
ORDER BY device_id, time
|
|
LIMIT 10;
|
|
|
|
-- SELECT DISTINCT RECORD works correctly
|
|
\qecho SELECT DISTINCT RECORD works correctly
|
|
:PREFIX
|
|
SELECT DISTINCT :TABLE_NAME r
|
|
FROM :TABLE_NAME
|
|
ORDER BY r
|
|
LIMIT 10;
|
|
|
|
-- SELECT DISTINCT function is not pushed down
|
|
\qecho SELECT DISTINCT FUNCTION_EXPR not pushed down currently
|
|
:PREFIX
|
|
SELECT DISTINCT time_bucket('1h',time) col1
|
|
FROM :TABLE_NAME
|
|
ORDER BY col1
|
|
LIMIT 10;
|
|
|
|
-- SELECT DISTINCT without any var references is handled correctly
|
|
\qecho SELECT DISTINCT without any var references is handled correctly
|
|
:PREFIX
|
|
SELECT DISTINCT 1, 'constx'
|
|
FROM :TABLE_NAME;
|