mirror of
https://github.com/timescale/timescaledb.git
synced 2025-05-16 18:43:18 +08:00
406901d838
Files and directories using "hyperstore" as part of the name is moved to the new name using "hypercore".
189 lines
7.4 KiB
SQL
189 lines
7.4 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.
|
|
|
|
create table readings(time timestamptz,
|
|
location text,
|
|
device int,
|
|
temp float,
|
|
humidity float,
|
|
unique (device, location, time)
|
|
);
|
|
|
|
create index on readings(location);
|
|
create index on readings(device);
|
|
|
|
select create_hypertable('readings', 'time');
|
|
|
|
select setseed(1);
|
|
|
|
insert into readings (time, location, device, temp, humidity)
|
|
select t, ceil(random()*3), ceil(random()*30), random()*40, random()*100
|
|
from generate_series('2022-06-01'::timestamptz, '2022-07-01', '5m') t;
|
|
|
|
alter table readings set (
|
|
timescaledb.compress,
|
|
timescaledb.compress_orderby = 'time',
|
|
timescaledb.compress_segmentby = 'device'
|
|
);
|
|
|
|
select format('%I.%I', chunk_schema, chunk_name)::regclass as chunk
|
|
from timescaledb_information.chunks
|
|
where format('%I.%I', hypertable_schema, hypertable_name)::regclass = 'readings'::regclass
|
|
limit 1 \gset
|
|
|
|
alter table :chunk set access method hyperstore;
|
|
|
|
--
|
|
-- Check that TID scan works for both compressed and non-compressed
|
|
-- rows.
|
|
--
|
|
|
|
set timescaledb.enable_transparent_decompression to false;
|
|
|
|
-- Select any row and try to fetch it using CTID. We do not select the
|
|
-- first one just to also try to scan a few rows and make sure the
|
|
-- implementation works.
|
|
select ctid from :chunk limit 1 offset 10 \gset
|
|
|
|
explain (analyze, costs off, timing off, summary off, decompress_cache_stats)
|
|
select * from :chunk where ctid = :'ctid';
|
|
select * from :chunk where ctid = :'ctid';
|
|
|
|
-- Insert a new row, which will then be non-compressed, and fetch it.
|
|
insert into :chunk values ('Wed May 25 17:34:56 2022 PDT', 1, 2, 3.14, 2.14);
|
|
select ctid from :chunk where time = 'Wed May 25 17:34:56 2022 PDT' \gset
|
|
|
|
explain (analyze, costs off, timing off, summary off, decompress_cache_stats)
|
|
select * from :chunk where ctid = :'ctid';
|
|
select * from :chunk where ctid = :'ctid';
|
|
|
|
-- Check that a bad option name generates an error.
|
|
\set ON_ERROR_STOP 0
|
|
explain (analyze, costs off, timing off, summary off, decopress_cache_stats)
|
|
select * from :chunk where device between 5 and 10;
|
|
\set ON_ERROR_STOP 1
|
|
|
|
explain (analyze, costs off, timing off, summary off, decompress_cache_stats)
|
|
select time, temp + humidity from readings where device between 5 and 10 and humidity > 5;
|
|
|
|
-- Check the explain cache information output.
|
|
--
|
|
-- Query 1 and 3 should show the same explain plan, and the plan in
|
|
-- the middle should not include decompress stats:
|
|
explain (analyze, costs off, timing off, summary off, decompress_cache_stats)
|
|
select time, temp + humidity from readings where device between 5 and 10 and humidity > 5;
|
|
|
|
-- Check the explain cache information output. Query 1 and 3 should
|
|
-- show the same explain plan, and the plan in the middle should not
|
|
-- include decompress stats:
|
|
explain (analyze, costs off, timing off, summary off, decompress_cache_stats)
|
|
select * from :chunk where device between 5 and 10;
|
|
explain (analyze, costs off, timing off, summary off, decompress_cache_stats)
|
|
select * from :chunk where device between 5 and 10;
|
|
|
|
-- Queries that will select just a few columns
|
|
set max_parallel_workers_per_gather to 0;
|
|
explain (analyze, costs off, timing off, summary off, decompress_cache_stats)
|
|
select device, humidity from readings where device between 5 and 10;
|
|
|
|
explain (analyze, costs off, timing off, summary off, decompress_cache_stats)
|
|
select device, avg(humidity) from readings where device between 5 and 10
|
|
group by device;
|
|
|
|
-- Test on conflict: insert the same data as before, but throw away
|
|
-- the updates.
|
|
explain (analyze, costs off, timing off, summary off, decompress_cache_stats)
|
|
insert into readings (time, location, device, temp, humidity)
|
|
select t, ceil(random()*10), ceil(random()*30), random()*40, random()*100
|
|
from generate_series('2022-06-01'::timestamptz, '2022-07-01', '5m') t
|
|
on conflict (location, device, time) do nothing;
|
|
|
|
-- This should show values for all columns
|
|
explain (analyze, costs off, timing off, summary off, decompress_cache_stats)
|
|
select time, temp + humidity from readings where device between 5 and 10 and humidity > 5 limit 5;
|
|
select time, temp + humidity from readings where device between 5 and 10 and humidity > 5 limit 5;
|
|
|
|
-- Get the compressed chunk
|
|
select format('%I.%I', c2.schema_name, c2.table_name)::regclass as cchunk
|
|
from _timescaledb_catalog.chunk c1
|
|
join _timescaledb_catalog.chunk c2
|
|
on (c1.compressed_chunk_id = c2.id)
|
|
where format('%I.%I', c1.schema_name, c1.table_name)::regclass = :'chunk'::regclass \gset
|
|
|
|
-- Show that location is using dictionary encoding
|
|
select (_timescaledb_functions.compressed_data_info(location)).* from :cchunk limit 1;
|
|
|
|
-- Test that vectorized filtering on text column works
|
|
set enable_indexscan=off;
|
|
explain (analyze, costs off, timing off, summary off, decompress_cache_stats)
|
|
select time, location, temp from :chunk
|
|
where location = 1::text
|
|
order by time desc;
|
|
|
|
-- Save the data for comparison with seqscan
|
|
create temp table chunk_saved as
|
|
select time, location, temp from :chunk
|
|
where location = 1::text
|
|
order by time desc;
|
|
|
|
-- Show same query with seqscan and compare output
|
|
set timescaledb.enable_columnarscan=off;
|
|
explain (analyze, costs off, timing off, summary off)
|
|
select time, location, temp from :chunk
|
|
where location = 1::text
|
|
order by time desc;
|
|
|
|
-- If output is the same, this query should return nothing
|
|
(select time, location, temp from :chunk
|
|
where location = 1::text
|
|
order by time desc)
|
|
except
|
|
select * from chunk_saved;
|
|
|
|
-- Insert some non-compressed values to see that vectorized filtering
|
|
-- works on those non-compressed text columns.
|
|
insert into :chunk values ('2022-06-01 15:30'::timestamptz, 1, 2, 3.14, 2.14), ('2022-06-01 15:30'::timestamptz, 2, 2, 3.14, 2.14);
|
|
|
|
-- Query should only return the one non-compressed row that has location=1
|
|
(select time, location, temp from :chunk
|
|
where location = 1::text
|
|
order by time desc)
|
|
except
|
|
select * from chunk_saved;
|
|
|
|
-- Test that a ColumnarScan doesn't decompress anything if there are
|
|
-- no referenced columns, or the referenced column is a segmentby
|
|
-- column
|
|
set timescaledb.enable_columnarscan=true;
|
|
explain (analyze, costs off, timing off, summary off, decompress_cache_stats)
|
|
select count(*) from :chunk where device = 1;
|
|
|
|
explain (analyze, costs off, timing off, summary off, decompress_cache_stats)
|
|
select device from :chunk where device = 1;
|
|
|
|
-- Using a non-segmentby column will decompress that column
|
|
explain (analyze, costs off, timing off, summary off, decompress_cache_stats)
|
|
select count(*) from :chunk where location = 1::text;
|
|
|
|
-- Testing same thing with SeqScan. It still decompresses in the
|
|
-- count(*) case, although it shouldn't have to. So, probably an
|
|
-- opportunity to optimize.
|
|
set timescaledb.enable_columnarscan=false;
|
|
explain (analyze, costs off, timing off, summary off, decompress_cache_stats)
|
|
select count(*) from :chunk where device = 1;
|
|
|
|
explain (analyze, costs off, timing off, summary off, decompress_cache_stats)
|
|
select device from :chunk where device = 1;
|
|
|
|
explain (analyze, costs off, timing off, summary off, decompress_cache_stats)
|
|
select count(*) from :chunk where location = 1::text;
|
|
|
|
-- ColumnarScan declares itself as projection capable. This query
|
|
-- would add a Result node on top if ColumnarScan couldn't project.
|
|
set timescaledb.enable_columnarscan=true;
|
|
explain
|
|
select time, device+device as device_x2 from :chunk limit 1;
|
|
select time, device+device as device_x2 from :chunk limit 1;
|
|
|