Update tdigest_aggregate_support output for PG15+ (#7849)

Regress test tdigest_aggregate_support has been failing since at least
Citus 12.0, when tdigest extension is installed in Postgres. This
appears to be because of an omission by commit 03832f3 and a change in
the implementation of Postgres random() function (pg commit
[d4f109e4a](https://git.postgresql.org/gitweb/?p=postgresql.git;a=commitdiff;h=d4f109e4a)).
To reproduce the test diff:
- Checkout [tdigest ](https://github.com/tvondra/tdigest)and run `make;
make install`
- In citus regress directory run `make check-multi` or
`./citus_tests/run_test.py tdigest_aggregate_support`

There are two parts to this commit:

1. Revert `Output: xxxxx` in EXPLAIN VERBOSE. Citus commit fe4ac51
normalized EXPLAIN VERBOSE output because of a change between pg12 and
pg13. When pg12 support was no longer required, the rule was removed
from normalize.sed and `Output: xxxx` was reverted in the impacted
regress output files (03832f3), but `tdigest_aggregate_support` was
omitted.

2. Adjust the query results; the tdigest_aggregate_support test file has
a comment _verifying results - should be stable due to seed while
inserting the data, if failure due to data these queries could be
removed or check for certain ranges_ but the result values in this
commit are consistent across citus 12.0 (pg 15), citus 12.1 (pg 16) and
citus 13.0 (pg 17), or since the Postgres changed their [implementation
of
random](https://git.postgresql.org/gitweb/?p=postgresql.git;a=commitdiff;h=d4f109e4a),
so proposing to go with these results.
pull/7922/head
Colm 2025-01-20 22:00:33 +00:00 committed by naisila
parent 7e1f22999b
commit a18f8990be
2 changed files with 754 additions and 103 deletions

View File

@ -42,18 +42,18 @@ FROM latencies;
QUERY PLAN QUERY PLAN
--------------------------------------------------------------------- ---------------------------------------------------------------------
Aggregate Aggregate
Output: xxxxxx Output: tdigest(remote_scan.tdigest)
-> Custom Scan (Citus Adaptive) -> Custom Scan (Citus Adaptive)
Output: xxxxxx Output: remote_scan.tdigest
Task Count: 4 Task Count: 4
Tasks Shown: One of 4 Tasks Shown: One of 4
-> Task -> Task
Query: SELECT public.tdigest(latency, 100) AS tdigest FROM tdigest_aggregate_support.latencies_20070000 latencies WHERE true Query: SELECT public.tdigest(latency, 100) AS tdigest FROM tdigest_aggregate_support.latencies_20070000 latencies WHERE true
Node: host=localhost port=xxxxx dbname=regression Node: host=localhost port=xxxxx dbname=regression
-> Aggregate -> Aggregate
Output: xxxxxx Output: tdigest(latency, 100)
-> Seq Scan on tdigest_aggregate_support.latencies_20070000 latencies -> Seq Scan on tdigest_aggregate_support.latencies_20070000 latencies
Output: xxxxxx Output: a, b, latency
(13 rows) (13 rows)
-- explain grouping by distribution column is completely pushed down for tdigest(value, compression) -- explain grouping by distribution column is completely pushed down for tdigest(value, compression)
@ -64,17 +64,17 @@ GROUP BY a;
QUERY PLAN QUERY PLAN
--------------------------------------------------------------------- ---------------------------------------------------------------------
Custom Scan (Citus Adaptive) Custom Scan (Citus Adaptive)
Output: xxxxxx Output: remote_scan.a, remote_scan.tdigest
Task Count: 4 Task Count: 4
Tasks Shown: One of 4 Tasks Shown: One of 4
-> Task -> Task
Query: SELECT a, public.tdigest(latency, 100) AS tdigest FROM tdigest_aggregate_support.latencies_20070000 latencies WHERE true GROUP BY a Query: SELECT a, public.tdigest(latency, 100) AS tdigest FROM tdigest_aggregate_support.latencies_20070000 latencies WHERE true GROUP BY a
Node: host=localhost port=xxxxx dbname=regression Node: host=localhost port=xxxxx dbname=regression
-> HashAggregate -> HashAggregate
Output: xxxxxx Output: a, tdigest(latency, 100)
Group Key: latencies.a Group Key: latencies.a
-> Seq Scan on tdigest_aggregate_support.latencies_20070000 latencies -> Seq Scan on tdigest_aggregate_support.latencies_20070000 latencies
Output: xxxxxx Output: a, b, latency
(12 rows) (12 rows)
-- explain grouping by non-distribution column is partially pushed down for tdigest(value, compression) -- explain grouping by non-distribution column is partially pushed down for tdigest(value, compression)
@ -85,20 +85,20 @@ GROUP BY b;
QUERY PLAN QUERY PLAN
--------------------------------------------------------------------- ---------------------------------------------------------------------
HashAggregate HashAggregate
Output: xxxxxx Output: remote_scan.b, tdigest(remote_scan.tdigest)
Group Key: remote_scan.b Group Key: remote_scan.b
-> Custom Scan (Citus Adaptive) -> Custom Scan (Citus Adaptive)
Output: xxxxxx Output: remote_scan.b, remote_scan.tdigest
Task Count: 4 Task Count: 4
Tasks Shown: One of 4 Tasks Shown: One of 4
-> Task -> Task
Query: SELECT b, public.tdigest(latency, 100) AS tdigest FROM tdigest_aggregate_support.latencies_20070000 latencies WHERE true GROUP BY b Query: SELECT b, public.tdigest(latency, 100) AS tdigest FROM tdigest_aggregate_support.latencies_20070000 latencies WHERE true GROUP BY b
Node: host=localhost port=xxxxx dbname=regression Node: host=localhost port=xxxxx dbname=regression
-> HashAggregate -> HashAggregate
Output: xxxxxx Output: b, tdigest(latency, 100)
Group Key: latencies.b Group Key: latencies.b
-> Seq Scan on tdigest_aggregate_support.latencies_20070000 latencies -> Seq Scan on tdigest_aggregate_support.latencies_20070000 latencies
Output: xxxxxx Output: a, b, latency
(15 rows) (15 rows)
-- explain no grouping to verify partially pushed down for tdigest_precentile(value, compression, quantile) -- explain no grouping to verify partially pushed down for tdigest_precentile(value, compression, quantile)
@ -108,18 +108,18 @@ FROM latencies;
QUERY PLAN QUERY PLAN
--------------------------------------------------------------------- ---------------------------------------------------------------------
Aggregate Aggregate
Output: xxxxxx Output: tdigest_percentile(remote_scan.tdigest_percentile, '0.99'::double precision)
-> Custom Scan (Citus Adaptive) -> Custom Scan (Citus Adaptive)
Output: xxxxxx Output: remote_scan.tdigest_percentile
Task Count: 4 Task Count: 4
Tasks Shown: One of 4 Tasks Shown: One of 4
-> Task -> Task
Query: SELECT public.tdigest(latency, 100) AS tdigest_percentile FROM tdigest_aggregate_support.latencies_20070000 latencies WHERE true Query: SELECT public.tdigest(latency, 100) AS tdigest_percentile FROM tdigest_aggregate_support.latencies_20070000 latencies WHERE true
Node: host=localhost port=xxxxx dbname=regression Node: host=localhost port=xxxxx dbname=regression
-> Aggregate -> Aggregate
Output: xxxxxx Output: tdigest(latency, 100)
-> Seq Scan on tdigest_aggregate_support.latencies_20070000 latencies -> Seq Scan on tdigest_aggregate_support.latencies_20070000 latencies
Output: xxxxxx Output: a, b, latency
(13 rows) (13 rows)
-- explain grouping by distribution column is completely pushed down for tdigest_precentile(value, compression, quantile) -- explain grouping by distribution column is completely pushed down for tdigest_precentile(value, compression, quantile)
@ -130,17 +130,17 @@ GROUP BY a;
QUERY PLAN QUERY PLAN
--------------------------------------------------------------------- ---------------------------------------------------------------------
Custom Scan (Citus Adaptive) Custom Scan (Citus Adaptive)
Output: xxxxxx Output: remote_scan.a, remote_scan.tdigest_percentile
Task Count: 4 Task Count: 4
Tasks Shown: One of 4 Tasks Shown: One of 4
-> Task -> Task
Query: SELECT a, public.tdigest_percentile(latency, 100, '0.99'::double precision) AS tdigest_percentile FROM tdigest_aggregate_support.latencies_20070000 latencies WHERE true GROUP BY a Query: SELECT a, public.tdigest_percentile(latency, 100, '0.99'::double precision) AS tdigest_percentile FROM tdigest_aggregate_support.latencies_20070000 latencies WHERE true GROUP BY a
Node: host=localhost port=xxxxx dbname=regression Node: host=localhost port=xxxxx dbname=regression
-> HashAggregate -> HashAggregate
Output: xxxxxx Output: a, tdigest_percentile(latency, 100, '0.99'::double precision)
Group Key: latencies.a Group Key: latencies.a
-> Seq Scan on tdigest_aggregate_support.latencies_20070000 latencies -> Seq Scan on tdigest_aggregate_support.latencies_20070000 latencies
Output: xxxxxx Output: a, b, latency
(12 rows) (12 rows)
-- explain grouping by non-distribution column is partially pushed down for tdigest_precentile(value, compression, quantile) -- explain grouping by non-distribution column is partially pushed down for tdigest_precentile(value, compression, quantile)
@ -151,20 +151,20 @@ GROUP BY b;
QUERY PLAN QUERY PLAN
--------------------------------------------------------------------- ---------------------------------------------------------------------
HashAggregate HashAggregate
Output: xxxxxx Output: remote_scan.b, tdigest_percentile(remote_scan.tdigest_percentile, '0.99'::double precision)
Group Key: remote_scan.b Group Key: remote_scan.b
-> Custom Scan (Citus Adaptive) -> Custom Scan (Citus Adaptive)
Output: xxxxxx Output: remote_scan.b, remote_scan.tdigest_percentile
Task Count: 4 Task Count: 4
Tasks Shown: One of 4 Tasks Shown: One of 4
-> Task -> Task
Query: SELECT b, public.tdigest(latency, 100) AS tdigest_percentile FROM tdigest_aggregate_support.latencies_20070000 latencies WHERE true GROUP BY b Query: SELECT b, public.tdigest(latency, 100) AS tdigest_percentile FROM tdigest_aggregate_support.latencies_20070000 latencies WHERE true GROUP BY b
Node: host=localhost port=xxxxx dbname=regression Node: host=localhost port=xxxxx dbname=regression
-> HashAggregate -> HashAggregate
Output: xxxxxx Output: b, tdigest(latency, 100)
Group Key: latencies.b Group Key: latencies.b
-> Seq Scan on tdigest_aggregate_support.latencies_20070000 latencies -> Seq Scan on tdigest_aggregate_support.latencies_20070000 latencies
Output: xxxxxx Output: a, b, latency
(15 rows) (15 rows)
-- explain no grouping to verify partially pushed down for tdigest_precentile(value, compression, quantiles[]) -- explain no grouping to verify partially pushed down for tdigest_precentile(value, compression, quantiles[])
@ -174,18 +174,18 @@ FROM latencies;
QUERY PLAN QUERY PLAN
--------------------------------------------------------------------- ---------------------------------------------------------------------
Aggregate Aggregate
Output: xxxxxx Output: tdigest_percentile(remote_scan.tdigest_percentile, '{0.99,0.95}'::double precision[])
-> Custom Scan (Citus Adaptive) -> Custom Scan (Citus Adaptive)
Output: xxxxxx Output: remote_scan.tdigest_percentile
Task Count: 4 Task Count: 4
Tasks Shown: One of 4 Tasks Shown: One of 4
-> Task -> Task
Query: SELECT public.tdigest(latency, 100) AS tdigest_percentile FROM tdigest_aggregate_support.latencies_20070000 latencies WHERE true Query: SELECT public.tdigest(latency, 100) AS tdigest_percentile FROM tdigest_aggregate_support.latencies_20070000 latencies WHERE true
Node: host=localhost port=xxxxx dbname=regression Node: host=localhost port=xxxxx dbname=regression
-> Aggregate -> Aggregate
Output: xxxxxx Output: tdigest(latency, 100)
-> Seq Scan on tdigest_aggregate_support.latencies_20070000 latencies -> Seq Scan on tdigest_aggregate_support.latencies_20070000 latencies
Output: xxxxxx Output: a, b, latency
(13 rows) (13 rows)
-- explain grouping by distribution column is completely pushed down for tdigest_precentile(value, compression, quantiles[]) -- explain grouping by distribution column is completely pushed down for tdigest_precentile(value, compression, quantiles[])
@ -196,17 +196,17 @@ GROUP BY a;
QUERY PLAN QUERY PLAN
--------------------------------------------------------------------- ---------------------------------------------------------------------
Custom Scan (Citus Adaptive) Custom Scan (Citus Adaptive)
Output: xxxxxx Output: remote_scan.a, remote_scan.tdigest_percentile
Task Count: 4 Task Count: 4
Tasks Shown: One of 4 Tasks Shown: One of 4
-> Task -> Task
Query: SELECT a, public.tdigest_percentile(latency, 100, '{0.99,0.95}'::double precision[]) AS tdigest_percentile FROM tdigest_aggregate_support.latencies_20070000 latencies WHERE true GROUP BY a Query: SELECT a, public.tdigest_percentile(latency, 100, '{0.99,0.95}'::double precision[]) AS tdigest_percentile FROM tdigest_aggregate_support.latencies_20070000 latencies WHERE true GROUP BY a
Node: host=localhost port=xxxxx dbname=regression Node: host=localhost port=xxxxx dbname=regression
-> HashAggregate -> HashAggregate
Output: xxxxxx Output: a, tdigest_percentile(latency, 100, '{0.99,0.95}'::double precision[])
Group Key: latencies.a Group Key: latencies.a
-> Seq Scan on tdigest_aggregate_support.latencies_20070000 latencies -> Seq Scan on tdigest_aggregate_support.latencies_20070000 latencies
Output: xxxxxx Output: a, b, latency
(12 rows) (12 rows)
-- explain grouping by non-distribution column is partially pushed down for tdigest_precentile(value, compression, quantiles[]) -- explain grouping by non-distribution column is partially pushed down for tdigest_precentile(value, compression, quantiles[])
@ -217,20 +217,20 @@ GROUP BY b;
QUERY PLAN QUERY PLAN
--------------------------------------------------------------------- ---------------------------------------------------------------------
HashAggregate HashAggregate
Output: xxxxxx Output: remote_scan.b, tdigest_percentile(remote_scan.tdigest_percentile, '{0.99,0.95}'::double precision[])
Group Key: remote_scan.b Group Key: remote_scan.b
-> Custom Scan (Citus Adaptive) -> Custom Scan (Citus Adaptive)
Output: xxxxxx Output: remote_scan.b, remote_scan.tdigest_percentile
Task Count: 4 Task Count: 4
Tasks Shown: One of 4 Tasks Shown: One of 4
-> Task -> Task
Query: SELECT b, public.tdigest(latency, 100) AS tdigest_percentile FROM tdigest_aggregate_support.latencies_20070000 latencies WHERE true GROUP BY b Query: SELECT b, public.tdigest(latency, 100) AS tdigest_percentile FROM tdigest_aggregate_support.latencies_20070000 latencies WHERE true GROUP BY b
Node: host=localhost port=xxxxx dbname=regression Node: host=localhost port=xxxxx dbname=regression
-> HashAggregate -> HashAggregate
Output: xxxxxx Output: b, tdigest(latency, 100)
Group Key: latencies.b Group Key: latencies.b
-> Seq Scan on tdigest_aggregate_support.latencies_20070000 latencies -> Seq Scan on tdigest_aggregate_support.latencies_20070000 latencies
Output: xxxxxx Output: a, b, latency
(15 rows) (15 rows)
-- explain no grouping to verify partially pushed down for tdigest_precentile_of(value, compression, hypotetical_value) -- explain no grouping to verify partially pushed down for tdigest_precentile_of(value, compression, hypotetical_value)
@ -240,18 +240,18 @@ FROM latencies;
QUERY PLAN QUERY PLAN
--------------------------------------------------------------------- ---------------------------------------------------------------------
Aggregate Aggregate
Output: xxxxxx Output: tdigest_percentile_of(remote_scan.tdigest_percentile_of, '9000'::double precision)
-> Custom Scan (Citus Adaptive) -> Custom Scan (Citus Adaptive)
Output: xxxxxx Output: remote_scan.tdigest_percentile_of
Task Count: 4 Task Count: 4
Tasks Shown: One of 4 Tasks Shown: One of 4
-> Task -> Task
Query: SELECT public.tdigest(latency, 100) AS tdigest_percentile_of FROM tdigest_aggregate_support.latencies_20070000 latencies WHERE true Query: SELECT public.tdigest(latency, 100) AS tdigest_percentile_of FROM tdigest_aggregate_support.latencies_20070000 latencies WHERE true
Node: host=localhost port=xxxxx dbname=regression Node: host=localhost port=xxxxx dbname=regression
-> Aggregate -> Aggregate
Output: xxxxxx Output: tdigest(latency, 100)
-> Seq Scan on tdigest_aggregate_support.latencies_20070000 latencies -> Seq Scan on tdigest_aggregate_support.latencies_20070000 latencies
Output: xxxxxx Output: a, b, latency
(13 rows) (13 rows)
-- explain grouping by distribution column is completely pushed down for tdigest_precentile_of(value, compression, hypotetical_value) -- explain grouping by distribution column is completely pushed down for tdigest_precentile_of(value, compression, hypotetical_value)
@ -262,17 +262,17 @@ GROUP BY a;
QUERY PLAN QUERY PLAN
--------------------------------------------------------------------- ---------------------------------------------------------------------
Custom Scan (Citus Adaptive) Custom Scan (Citus Adaptive)
Output: xxxxxx Output: remote_scan.a, remote_scan.tdigest_percentile_of
Task Count: 4 Task Count: 4
Tasks Shown: One of 4 Tasks Shown: One of 4
-> Task -> Task
Query: SELECT a, public.tdigest_percentile_of(latency, 100, '9000'::double precision) AS tdigest_percentile_of FROM tdigest_aggregate_support.latencies_20070000 latencies WHERE true GROUP BY a Query: SELECT a, public.tdigest_percentile_of(latency, 100, '9000'::double precision) AS tdigest_percentile_of FROM tdigest_aggregate_support.latencies_20070000 latencies WHERE true GROUP BY a
Node: host=localhost port=xxxxx dbname=regression Node: host=localhost port=xxxxx dbname=regression
-> HashAggregate -> HashAggregate
Output: xxxxxx Output: a, tdigest_percentile_of(latency, 100, '9000'::double precision)
Group Key: latencies.a Group Key: latencies.a
-> Seq Scan on tdigest_aggregate_support.latencies_20070000 latencies -> Seq Scan on tdigest_aggregate_support.latencies_20070000 latencies
Output: xxxxxx Output: a, b, latency
(12 rows) (12 rows)
-- explain grouping by non-distribution column is partially pushed down for tdigest_precentile_of(value, compression, hypotetical_value) -- explain grouping by non-distribution column is partially pushed down for tdigest_precentile_of(value, compression, hypotetical_value)
@ -283,20 +283,20 @@ GROUP BY b;
QUERY PLAN QUERY PLAN
--------------------------------------------------------------------- ---------------------------------------------------------------------
HashAggregate HashAggregate
Output: xxxxxx Output: remote_scan.b, tdigest_percentile_of(remote_scan.tdigest_percentile_of, '9000'::double precision)
Group Key: remote_scan.b Group Key: remote_scan.b
-> Custom Scan (Citus Adaptive) -> Custom Scan (Citus Adaptive)
Output: xxxxxx Output: remote_scan.b, remote_scan.tdigest_percentile_of
Task Count: 4 Task Count: 4
Tasks Shown: One of 4 Tasks Shown: One of 4
-> Task -> Task
Query: SELECT b, public.tdigest(latency, 100) AS tdigest_percentile_of FROM tdigest_aggregate_support.latencies_20070000 latencies WHERE true GROUP BY b Query: SELECT b, public.tdigest(latency, 100) AS tdigest_percentile_of FROM tdigest_aggregate_support.latencies_20070000 latencies WHERE true GROUP BY b
Node: host=localhost port=xxxxx dbname=regression Node: host=localhost port=xxxxx dbname=regression
-> HashAggregate -> HashAggregate
Output: xxxxxx Output: b, tdigest(latency, 100)
Group Key: latencies.b Group Key: latencies.b
-> Seq Scan on tdigest_aggregate_support.latencies_20070000 latencies -> Seq Scan on tdigest_aggregate_support.latencies_20070000 latencies
Output: xxxxxx Output: a, b, latency
(15 rows) (15 rows)
-- explain no grouping to verify partially pushed down for tdigest_precentile_of(value, compression, hypotetical_values[]) -- explain no grouping to verify partially pushed down for tdigest_precentile_of(value, compression, hypotetical_values[])
@ -306,18 +306,18 @@ FROM latencies;
QUERY PLAN QUERY PLAN
--------------------------------------------------------------------- ---------------------------------------------------------------------
Aggregate Aggregate
Output: xxxxxx Output: tdigest_percentile_of(remote_scan.tdigest_percentile_of, '{9000,9500}'::double precision[])
-> Custom Scan (Citus Adaptive) -> Custom Scan (Citus Adaptive)
Output: xxxxxx Output: remote_scan.tdigest_percentile_of
Task Count: 4 Task Count: 4
Tasks Shown: One of 4 Tasks Shown: One of 4
-> Task -> Task
Query: SELECT public.tdigest(latency, 100) AS tdigest_percentile_of FROM tdigest_aggregate_support.latencies_20070000 latencies WHERE true Query: SELECT public.tdigest(latency, 100) AS tdigest_percentile_of FROM tdigest_aggregate_support.latencies_20070000 latencies WHERE true
Node: host=localhost port=xxxxx dbname=regression Node: host=localhost port=xxxxx dbname=regression
-> Aggregate -> Aggregate
Output: xxxxxx Output: tdigest(latency, 100)
-> Seq Scan on tdigest_aggregate_support.latencies_20070000 latencies -> Seq Scan on tdigest_aggregate_support.latencies_20070000 latencies
Output: xxxxxx Output: a, b, latency
(13 rows) (13 rows)
-- explain grouping by distribution column is completely pushed down for tdigest_precentile_of(value, compression, hypotetical_values[]) -- explain grouping by distribution column is completely pushed down for tdigest_precentile_of(value, compression, hypotetical_values[])
@ -328,17 +328,17 @@ GROUP BY a;
QUERY PLAN QUERY PLAN
--------------------------------------------------------------------- ---------------------------------------------------------------------
Custom Scan (Citus Adaptive) Custom Scan (Citus Adaptive)
Output: xxxxxx Output: remote_scan.a, remote_scan.tdigest_percentile_of
Task Count: 4 Task Count: 4
Tasks Shown: One of 4 Tasks Shown: One of 4
-> Task -> Task
Query: SELECT a, public.tdigest_percentile_of(latency, 100, '{9000,9500}'::double precision[]) AS tdigest_percentile_of FROM tdigest_aggregate_support.latencies_20070000 latencies WHERE true GROUP BY a Query: SELECT a, public.tdigest_percentile_of(latency, 100, '{9000,9500}'::double precision[]) AS tdigest_percentile_of FROM tdigest_aggregate_support.latencies_20070000 latencies WHERE true GROUP BY a
Node: host=localhost port=xxxxx dbname=regression Node: host=localhost port=xxxxx dbname=regression
-> HashAggregate -> HashAggregate
Output: xxxxxx Output: a, tdigest_percentile_of(latency, 100, '{9000,9500}'::double precision[])
Group Key: latencies.a Group Key: latencies.a
-> Seq Scan on tdigest_aggregate_support.latencies_20070000 latencies -> Seq Scan on tdigest_aggregate_support.latencies_20070000 latencies
Output: xxxxxx Output: a, b, latency
(12 rows) (12 rows)
-- explain grouping by non-distribution column is partially pushed down for tdigest_precentile_of(value, compression, hypotetical_values[]) -- explain grouping by non-distribution column is partially pushed down for tdigest_precentile_of(value, compression, hypotetical_values[])
@ -349,51 +349,51 @@ GROUP BY b;
QUERY PLAN QUERY PLAN
--------------------------------------------------------------------- ---------------------------------------------------------------------
HashAggregate HashAggregate
Output: xxxxxx Output: remote_scan.b, tdigest_percentile_of(remote_scan.tdigest_percentile_of, '{9000,9500}'::double precision[])
Group Key: remote_scan.b Group Key: remote_scan.b
-> Custom Scan (Citus Adaptive) -> Custom Scan (Citus Adaptive)
Output: xxxxxx Output: remote_scan.b, remote_scan.tdigest_percentile_of
Task Count: 4 Task Count: 4
Tasks Shown: One of 4 Tasks Shown: One of 4
-> Task -> Task
Query: SELECT b, public.tdigest(latency, 100) AS tdigest_percentile_of FROM tdigest_aggregate_support.latencies_20070000 latencies WHERE true GROUP BY b Query: SELECT b, public.tdigest(latency, 100) AS tdigest_percentile_of FROM tdigest_aggregate_support.latencies_20070000 latencies WHERE true GROUP BY b
Node: host=localhost port=xxxxx dbname=regression Node: host=localhost port=xxxxx dbname=regression
-> HashAggregate -> HashAggregate
Output: xxxxxx Output: b, tdigest(latency, 100)
Group Key: latencies.b Group Key: latencies.b
-> Seq Scan on tdigest_aggregate_support.latencies_20070000 latencies -> Seq Scan on tdigest_aggregate_support.latencies_20070000 latencies
Output: xxxxxx Output: a, b, latency
(15 rows) (15 rows)
-- verifying results - should be stable due to seed while inserting the data, if failure due to data these queries could be removed or check for certain ranges -- verifying results - should be stable due to seed while inserting the data, if failure due to data these queries could be removed or check for certain ranges
SELECT tdigest(latency, 100) FROM latencies; SELECT tdigest(latency, 100) FROM latencies;
tdigest tdigest
--------------------------------------------------------------------- ---------------------------------------------------------------------
flags 0 count 10000 compression 100 centroids 46 (0.287235, 1) (1.025106, 1) (2.058216, 1) (5.335597, 1) (12.707263, 2) (25.302479, 3) (43.435063, 4) (77.987860, 5) (269.478664, 10) (509.417419, 13) (1227.158879, 22) (3408.256171, 35) (7772.721988, 55) (13840.275516, 65) (32937.127607, 108) (64476.403332, 148) (118260.230644, 199) (239584.293240, 292) (562119.836766, 463) (944722.686313, 547) (1751089.620493, 749) (3751264.745959, 1128) (5877270.108576, 1300) (6224557.402567, 1104) (5804999.258033, 874) (5632316.697114, 755) (4648651.050740, 573) (3460055.227950, 402) (2820271.404686, 314) (2676501.012955, 288) (1649845.166017, 173) (1269335.942008, 131) (813964.853243, 83) (484144.878702, 49) (337179.763016, 34) (198775.241901, 20) (149353.499704, 15) (109688.319223, 11) (79855.926155, 8) (49937.731689, 5) (29971.046175, 3) (19982.538737, 2) (9991.467422, 1) (9992.337047, 1) (9995.578357, 1) (9999.700339, 1) flags 1 count 10000 compression 100 centroids 46 (2.846051, 1) (3.323773, 1) (4.406495, 1) (4.532352, 1) (4.993616, 2) (7.673358, 3) (13.555084, 5) (18.776503, 7) (27.990526, 11) (37.903465, 17) (56.272069, 21) (91.011574, 34) (127.790676, 51) (190.655158, 70) (275.723291, 94) (407.151014, 135) (584.186017, 219) (827.310117, 287) (1121.971646, 345) (1605.113973, 609) (2278.067230, 751) (3126.852770, 1033) (4149.398030, 991) (5374.336553, 1354) (6470.439272, 939) (7319.715295, 777) (8095.598975, 715) (8667.524977, 456) (9077.609863, 374) (9385.068110, 203) (9571.304536, 150) (9702.936696, 118) (9806.254527, 75) (9873.753103, 50) (9918.059273, 33) (9945.081993, 22) (9962.407748, 16) (9974.769012, 9) (9979.796549, 6) (9984.017888, 5) (9985.809833, 3) (9989.863888, 2) (9995.910553, 1) (9995.979459, 1) (9997.355013, 1) (9997.761058, 1)
(1 row) (1 row)
SELECT tdigest_percentile(latency, 100, 0.99) FROM latencies; SELECT tdigest_percentile(latency, 100, 0.99) FROM latencies;
tdigest_percentile tdigest_percentile
--------------------------------------------------------------------- ---------------------------------------------------------------------
9904.28342426494 9900.44356712993
(1 row) (1 row)
SELECT tdigest_percentile(latency, 100, ARRAY[0.99, 0.95]) FROM latencies; SELECT tdigest_percentile(latency, 100, ARRAY[0.99, 0.95]) FROM latencies;
tdigest_percentile tdigest_percentile
--------------------------------------------------------------------- ---------------------------------------------------------------------
{9904.28342426494,9485.49009399385} {9900.44356712993,9484.78109695984}
(1 row) (1 row)
SELECT tdigest_percentile_of(latency, 100, 9000) FROM latencies; SELECT tdigest_percentile_of(latency, 100, 9000) FROM latencies;
tdigest_percentile_of tdigest_percentile_of
--------------------------------------------------------------------- ---------------------------------------------------------------------
0.903462047211138 0.903845993581303
(1 row) (1 row)
SELECT tdigest_percentile_of(latency, 100, ARRAY[9000, 9500]) FROM latencies; SELECT tdigest_percentile_of(latency, 100, ARRAY[9000, 9500]) FROM latencies;
tdigest_percentile_of tdigest_percentile_of
--------------------------------------------------------------------- ---------------------------------------------------------------------
{0.903462047211138,0.95137481812975} {0.903845993581303,0.951492325994142}
(1 row) (1 row)
CREATE TABLE latencies_rollup (a int, tdigest tdigest); CREATE TABLE latencies_rollup (a int, tdigest tdigest);
@ -413,18 +413,18 @@ FROM latencies_rollup;
QUERY PLAN QUERY PLAN
--------------------------------------------------------------------- ---------------------------------------------------------------------
Aggregate Aggregate
Output: xxxxxx Output: tdigest(remote_scan.tdigest)
-> Custom Scan (Citus Adaptive) -> Custom Scan (Citus Adaptive)
Output: xxxxxx Output: remote_scan.tdigest
Task Count: 4 Task Count: 4
Tasks Shown: One of 4 Tasks Shown: One of 4
-> Task -> Task
Query: SELECT public.tdigest(tdigest) AS tdigest FROM tdigest_aggregate_support.latencies_rollup_20070004 latencies_rollup WHERE true Query: SELECT public.tdigest(tdigest) AS tdigest FROM tdigest_aggregate_support.latencies_rollup_20070004 latencies_rollup WHERE true
Node: host=localhost port=xxxxx dbname=regression Node: host=localhost port=xxxxx dbname=regression
-> Aggregate -> Aggregate
Output: xxxxxx Output: tdigest(tdigest)
-> Seq Scan on tdigest_aggregate_support.latencies_rollup_20070004 latencies_rollup -> Seq Scan on tdigest_aggregate_support.latencies_rollup_20070004 latencies_rollup
Output: xxxxxx Output: a, tdigest
(13 rows) (13 rows)
-- explain grouping by distribution column is completely pushed down for tdigest(tdigest) -- explain grouping by distribution column is completely pushed down for tdigest(tdigest)
@ -435,17 +435,17 @@ GROUP BY a;
QUERY PLAN QUERY PLAN
--------------------------------------------------------------------- ---------------------------------------------------------------------
Custom Scan (Citus Adaptive) Custom Scan (Citus Adaptive)
Output: xxxxxx Output: remote_scan.a, remote_scan.tdigest
Task Count: 4 Task Count: 4
Tasks Shown: One of 4 Tasks Shown: One of 4
-> Task -> Task
Query: SELECT a, public.tdigest(tdigest) AS tdigest FROM tdigest_aggregate_support.latencies_rollup_20070004 latencies_rollup WHERE true GROUP BY a Query: SELECT a, public.tdigest(tdigest) AS tdigest FROM tdigest_aggregate_support.latencies_rollup_20070004 latencies_rollup WHERE true GROUP BY a
Node: host=localhost port=xxxxx dbname=regression Node: host=localhost port=xxxxx dbname=regression
-> HashAggregate -> HashAggregate
Output: xxxxxx Output: a, tdigest(tdigest)
Group Key: latencies_rollup.a Group Key: latencies_rollup.a
-> Seq Scan on tdigest_aggregate_support.latencies_rollup_20070004 latencies_rollup -> Seq Scan on tdigest_aggregate_support.latencies_rollup_20070004 latencies_rollup
Output: xxxxxx Output: a, tdigest
(12 rows) (12 rows)
-- explain no grouping to verify partially pushed down for tdigest_precentile(tdigest, quantile) -- explain no grouping to verify partially pushed down for tdigest_precentile(tdigest, quantile)
@ -455,18 +455,18 @@ FROM latencies_rollup;
QUERY PLAN QUERY PLAN
--------------------------------------------------------------------- ---------------------------------------------------------------------
Aggregate Aggregate
Output: xxxxxx Output: tdigest_percentile(remote_scan.tdigest_percentile, '0.99'::double precision)
-> Custom Scan (Citus Adaptive) -> Custom Scan (Citus Adaptive)
Output: xxxxxx Output: remote_scan.tdigest_percentile
Task Count: 4 Task Count: 4
Tasks Shown: One of 4 Tasks Shown: One of 4
-> Task -> Task
Query: SELECT public.tdigest(tdigest) AS tdigest_percentile FROM tdigest_aggregate_support.latencies_rollup_20070004 latencies_rollup WHERE true Query: SELECT public.tdigest(tdigest) AS tdigest_percentile FROM tdigest_aggregate_support.latencies_rollup_20070004 latencies_rollup WHERE true
Node: host=localhost port=xxxxx dbname=regression Node: host=localhost port=xxxxx dbname=regression
-> Aggregate -> Aggregate
Output: xxxxxx Output: tdigest(tdigest)
-> Seq Scan on tdigest_aggregate_support.latencies_rollup_20070004 latencies_rollup -> Seq Scan on tdigest_aggregate_support.latencies_rollup_20070004 latencies_rollup
Output: xxxxxx Output: a, tdigest
(13 rows) (13 rows)
-- explain grouping by distribution column is completely pushed down for tdigest_precentile(tdigest, quantile) -- explain grouping by distribution column is completely pushed down for tdigest_precentile(tdigest, quantile)
@ -477,17 +477,17 @@ GROUP BY a;
QUERY PLAN QUERY PLAN
--------------------------------------------------------------------- ---------------------------------------------------------------------
Custom Scan (Citus Adaptive) Custom Scan (Citus Adaptive)
Output: xxxxxx Output: remote_scan.a, remote_scan.tdigest_percentile
Task Count: 4 Task Count: 4
Tasks Shown: One of 4 Tasks Shown: One of 4
-> Task -> Task
Query: SELECT a, public.tdigest_percentile(tdigest, '0.99'::double precision) AS tdigest_percentile FROM tdigest_aggregate_support.latencies_rollup_20070004 latencies_rollup WHERE true GROUP BY a Query: SELECT a, public.tdigest_percentile(tdigest, '0.99'::double precision) AS tdigest_percentile FROM tdigest_aggregate_support.latencies_rollup_20070004 latencies_rollup WHERE true GROUP BY a
Node: host=localhost port=xxxxx dbname=regression Node: host=localhost port=xxxxx dbname=regression
-> HashAggregate -> HashAggregate
Output: xxxxxx Output: a, tdigest_percentile(tdigest, '0.99'::double precision)
Group Key: latencies_rollup.a Group Key: latencies_rollup.a
-> Seq Scan on tdigest_aggregate_support.latencies_rollup_20070004 latencies_rollup -> Seq Scan on tdigest_aggregate_support.latencies_rollup_20070004 latencies_rollup
Output: xxxxxx Output: a, tdigest
(12 rows) (12 rows)
-- explain no grouping to verify partially pushed down for tdigest_precentile(value, compression, quantiles[]) -- explain no grouping to verify partially pushed down for tdigest_precentile(value, compression, quantiles[])
@ -497,18 +497,18 @@ FROM latencies_rollup;
QUERY PLAN QUERY PLAN
--------------------------------------------------------------------- ---------------------------------------------------------------------
Aggregate Aggregate
Output: xxxxxx Output: tdigest_percentile(remote_scan.tdigest_percentile, '{0.99,0.95}'::double precision[])
-> Custom Scan (Citus Adaptive) -> Custom Scan (Citus Adaptive)
Output: xxxxxx Output: remote_scan.tdigest_percentile
Task Count: 4 Task Count: 4
Tasks Shown: One of 4 Tasks Shown: One of 4
-> Task -> Task
Query: SELECT public.tdigest(tdigest) AS tdigest_percentile FROM tdigest_aggregate_support.latencies_rollup_20070004 latencies_rollup WHERE true Query: SELECT public.tdigest(tdigest) AS tdigest_percentile FROM tdigest_aggregate_support.latencies_rollup_20070004 latencies_rollup WHERE true
Node: host=localhost port=xxxxx dbname=regression Node: host=localhost port=xxxxx dbname=regression
-> Aggregate -> Aggregate
Output: xxxxxx Output: tdigest(tdigest)
-> Seq Scan on tdigest_aggregate_support.latencies_rollup_20070004 latencies_rollup -> Seq Scan on tdigest_aggregate_support.latencies_rollup_20070004 latencies_rollup
Output: xxxxxx Output: a, tdigest
(13 rows) (13 rows)
-- explain grouping by distribution column is completely pushed down for tdigest_precentile(value, compression, quantiles[]) -- explain grouping by distribution column is completely pushed down for tdigest_precentile(value, compression, quantiles[])
@ -519,17 +519,17 @@ GROUP BY a;
QUERY PLAN QUERY PLAN
--------------------------------------------------------------------- ---------------------------------------------------------------------
Custom Scan (Citus Adaptive) Custom Scan (Citus Adaptive)
Output: xxxxxx Output: remote_scan.a, remote_scan.tdigest_percentile
Task Count: 4 Task Count: 4
Tasks Shown: One of 4 Tasks Shown: One of 4
-> Task -> Task
Query: SELECT a, public.tdigest_percentile(tdigest, '{0.99,0.95}'::double precision[]) AS tdigest_percentile FROM tdigest_aggregate_support.latencies_rollup_20070004 latencies_rollup WHERE true GROUP BY a Query: SELECT a, public.tdigest_percentile(tdigest, '{0.99,0.95}'::double precision[]) AS tdigest_percentile FROM tdigest_aggregate_support.latencies_rollup_20070004 latencies_rollup WHERE true GROUP BY a
Node: host=localhost port=xxxxx dbname=regression Node: host=localhost port=xxxxx dbname=regression
-> HashAggregate -> HashAggregate
Output: xxxxxx Output: a, tdigest_percentile(tdigest, '{0.99,0.95}'::double precision[])
Group Key: latencies_rollup.a Group Key: latencies_rollup.a
-> Seq Scan on tdigest_aggregate_support.latencies_rollup_20070004 latencies_rollup -> Seq Scan on tdigest_aggregate_support.latencies_rollup_20070004 latencies_rollup
Output: xxxxxx Output: a, tdigest
(12 rows) (12 rows)
-- explain no grouping to verify partially pushed down for tdigest_precentile_of(value, compression, hypotetical_value) -- explain no grouping to verify partially pushed down for tdigest_precentile_of(value, compression, hypotetical_value)
@ -539,18 +539,18 @@ FROM latencies_rollup;
QUERY PLAN QUERY PLAN
--------------------------------------------------------------------- ---------------------------------------------------------------------
Aggregate Aggregate
Output: xxxxxx Output: tdigest_percentile_of(remote_scan.tdigest_percentile_of, '9000'::double precision)
-> Custom Scan (Citus Adaptive) -> Custom Scan (Citus Adaptive)
Output: xxxxxx Output: remote_scan.tdigest_percentile_of
Task Count: 4 Task Count: 4
Tasks Shown: One of 4 Tasks Shown: One of 4
-> Task -> Task
Query: SELECT public.tdigest(tdigest) AS tdigest_percentile_of FROM tdigest_aggregate_support.latencies_rollup_20070004 latencies_rollup WHERE true Query: SELECT public.tdigest(tdigest) AS tdigest_percentile_of FROM tdigest_aggregate_support.latencies_rollup_20070004 latencies_rollup WHERE true
Node: host=localhost port=xxxxx dbname=regression Node: host=localhost port=xxxxx dbname=regression
-> Aggregate -> Aggregate
Output: xxxxxx Output: tdigest(tdigest)
-> Seq Scan on tdigest_aggregate_support.latencies_rollup_20070004 latencies_rollup -> Seq Scan on tdigest_aggregate_support.latencies_rollup_20070004 latencies_rollup
Output: xxxxxx Output: a, tdigest
(13 rows) (13 rows)
-- explain grouping by distribution column is completely pushed down for tdigest_precentile_of(value, compression, hypotetical_value) -- explain grouping by distribution column is completely pushed down for tdigest_precentile_of(value, compression, hypotetical_value)
@ -561,17 +561,17 @@ GROUP BY a;
QUERY PLAN QUERY PLAN
--------------------------------------------------------------------- ---------------------------------------------------------------------
Custom Scan (Citus Adaptive) Custom Scan (Citus Adaptive)
Output: xxxxxx Output: remote_scan.a, remote_scan.tdigest_percentile_of
Task Count: 4 Task Count: 4
Tasks Shown: One of 4 Tasks Shown: One of 4
-> Task -> Task
Query: SELECT a, public.tdigest_percentile_of(tdigest, '9000'::double precision) AS tdigest_percentile_of FROM tdigest_aggregate_support.latencies_rollup_20070004 latencies_rollup WHERE true GROUP BY a Query: SELECT a, public.tdigest_percentile_of(tdigest, '9000'::double precision) AS tdigest_percentile_of FROM tdigest_aggregate_support.latencies_rollup_20070004 latencies_rollup WHERE true GROUP BY a
Node: host=localhost port=xxxxx dbname=regression Node: host=localhost port=xxxxx dbname=regression
-> HashAggregate -> HashAggregate
Output: xxxxxx Output: a, tdigest_percentile_of(tdigest, '9000'::double precision)
Group Key: latencies_rollup.a Group Key: latencies_rollup.a
-> Seq Scan on tdigest_aggregate_support.latencies_rollup_20070004 latencies_rollup -> Seq Scan on tdigest_aggregate_support.latencies_rollup_20070004 latencies_rollup
Output: xxxxxx Output: a, tdigest
(12 rows) (12 rows)
-- explain no grouping to verify partially pushed down for tdigest_precentile_of(value, compression, hypotetical_values[]) -- explain no grouping to verify partially pushed down for tdigest_precentile_of(value, compression, hypotetical_values[])
@ -581,18 +581,18 @@ FROM latencies_rollup;
QUERY PLAN QUERY PLAN
--------------------------------------------------------------------- ---------------------------------------------------------------------
Aggregate Aggregate
Output: xxxxxx Output: tdigest_percentile_of(remote_scan.tdigest_percentile_of, '{9000,9500}'::double precision[])
-> Custom Scan (Citus Adaptive) -> Custom Scan (Citus Adaptive)
Output: xxxxxx Output: remote_scan.tdigest_percentile_of
Task Count: 4 Task Count: 4
Tasks Shown: One of 4 Tasks Shown: One of 4
-> Task -> Task
Query: SELECT public.tdigest(tdigest) AS tdigest_percentile_of FROM tdigest_aggregate_support.latencies_rollup_20070004 latencies_rollup WHERE true Query: SELECT public.tdigest(tdigest) AS tdigest_percentile_of FROM tdigest_aggregate_support.latencies_rollup_20070004 latencies_rollup WHERE true
Node: host=localhost port=xxxxx dbname=regression Node: host=localhost port=xxxxx dbname=regression
-> Aggregate -> Aggregate
Output: xxxxxx Output: tdigest(tdigest)
-> Seq Scan on tdigest_aggregate_support.latencies_rollup_20070004 latencies_rollup -> Seq Scan on tdigest_aggregate_support.latencies_rollup_20070004 latencies_rollup
Output: xxxxxx Output: a, tdigest
(13 rows) (13 rows)
-- explain grouping by distribution column is completely pushed down for tdigest_precentile_of(value, compression, hypotetical_values[]) -- explain grouping by distribution column is completely pushed down for tdigest_precentile_of(value, compression, hypotetical_values[])
@ -603,48 +603,48 @@ GROUP BY a;
QUERY PLAN QUERY PLAN
--------------------------------------------------------------------- ---------------------------------------------------------------------
Custom Scan (Citus Adaptive) Custom Scan (Citus Adaptive)
Output: xxxxxx Output: remote_scan.a, remote_scan.tdigest_percentile_of
Task Count: 4 Task Count: 4
Tasks Shown: One of 4 Tasks Shown: One of 4
-> Task -> Task
Query: SELECT a, public.tdigest_percentile_of(tdigest, '{9000,9500}'::double precision[]) AS tdigest_percentile_of FROM tdigest_aggregate_support.latencies_rollup_20070004 latencies_rollup WHERE true GROUP BY a Query: SELECT a, public.tdigest_percentile_of(tdigest, '{9000,9500}'::double precision[]) AS tdigest_percentile_of FROM tdigest_aggregate_support.latencies_rollup_20070004 latencies_rollup WHERE true GROUP BY a
Node: host=localhost port=xxxxx dbname=regression Node: host=localhost port=xxxxx dbname=regression
-> HashAggregate -> HashAggregate
Output: xxxxxx Output: a, tdigest_percentile_of(tdigest, '{9000,9500}'::double precision[])
Group Key: latencies_rollup.a Group Key: latencies_rollup.a
-> Seq Scan on tdigest_aggregate_support.latencies_rollup_20070004 latencies_rollup -> Seq Scan on tdigest_aggregate_support.latencies_rollup_20070004 latencies_rollup
Output: xxxxxx Output: a, tdigest
(12 rows) (12 rows)
-- verifying results - should be stable due to seed while inserting the data, if failure due to data these queries could be removed or check for certain ranges -- verifying results - should be stable due to seed while inserting the data, if failure due to data these queries could be removed or check for certain ranges
SELECT tdigest(tdigest) FROM latencies_rollup; SELECT tdigest(tdigest) FROM latencies_rollup;
tdigest tdigest
--------------------------------------------------------------------- ---------------------------------------------------------------------
flags 0 count 10000 compression 100 centroids 47 (0.287235, 1) (1.025106, 1) (2.058216, 1) (5.335597, 1) (12.707263, 2) (25.302479, 3) (43.435063, 4) (77.987860, 5) (241.681030, 9) (402.696604, 11) (999.675875, 20) (2310.848640, 27) (4374.387978, 37) (9722.896547, 56) (21713.805492, 87) (39735.065966, 112) (87335.860853, 177) (182744.906162, 262) (336766.886786, 338) (661263.339724, 464) (1228663.222377, 623) (2146097.038498, 805) (2854487.701653, 827) (5292830.156590, 1195) (6168185.834602, 1104) (6399734.303813, 966) (5778088.854724, 773) (5213381.984997, 637) (3763042.148296, 431) (3036786.646485, 333) (1948238.134602, 207) (1456568.605821, 152) (999888.715345, 103) (715935.892988, 73) (543464.906535, 55) (327339.982973, 33) (198853.838033, 20) (159362.743852, 16) (79807.827301, 8) (69877.414438, 7) (49937.731689, 5) (29971.046175, 3) (19982.538737, 2) (9991.467422, 1) (9992.337047, 1) (9995.578357, 1) (9999.700339, 1) flags 1 count 10000 compression 100 centroids 47 (2.846051, 1) (3.323773, 1) (4.406495, 1) (4.532352, 1) (4.993616, 2) (6.530962, 2) (11.814064, 4) (16.758727, 6) (24.422807, 7) (29.634013, 9) (38.230799, 17) (63.206944, 26) (93.000466, 33) (133.739050, 52) (203.937122, 73) (296.554766, 104) (433.471165, 156) (621.440087, 231) (848.382844, 249) (1169.617895, 397) (1644.617827, 549) (2300.974970, 785) (3167.406918, 921) (4133.429028, 1070) (5250.107199, 1221) (6441.973764, 1145) (7456.776221, 841) (8182.469509, 600) (8686.514890, 423) (9061.061505, 310) (9331.300632, 234) (9538.562658, 162) (9684.285952, 114) (9777.125865, 75) (9847.735791, 61) (9903.652590, 39) (9936.297173, 28) (9959.245853, 17) (9971.565967, 11) (9979.422819, 8) (9984.033648, 5) (9985.809833, 3) (9989.863888, 2) (9995.910553, 1) (9995.979459, 1) (9997.355013, 1) (9997.761058, 1)
(1 row) (1 row)
SELECT tdigest_percentile(tdigest, 0.99) FROM latencies_rollup; SELECT tdigest_percentile(tdigest, 0.99) FROM latencies_rollup;
tdigest_percentile tdigest_percentile
--------------------------------------------------------------------- ---------------------------------------------------------------------
9903.76070790358 9900.8567500684
(1 row) (1 row)
SELECT tdigest_percentile(tdigest, ARRAY[0.99, 0.95]) FROM latencies_rollup; SELECT tdigest_percentile(tdigest, ARRAY[0.99, 0.95]) FROM latencies_rollup;
tdigest_percentile tdigest_percentile
--------------------------------------------------------------------- ---------------------------------------------------------------------
{9903.76070790358,9492.7106302226} {9900.8567500684,9484.13020701958}
(1 row) (1 row)
SELECT tdigest_percentile_of(tdigest, 9000) FROM latencies_rollup; SELECT tdigest_percentile_of(tdigest, 9000) FROM latencies_rollup;
tdigest_percentile_of tdigest_percentile_of
--------------------------------------------------------------------- ---------------------------------------------------------------------
0.902852659582396 0.902275031655923
(1 row) (1 row)
SELECT tdigest_percentile_of(tdigest, ARRAY[9000, 9500]) FROM latencies_rollup; SELECT tdigest_percentile_of(tdigest, ARRAY[9000, 9500]) FROM latencies_rollup;
tdigest_percentile_of tdigest_percentile_of
--------------------------------------------------------------------- ---------------------------------------------------------------------
{0.902852659582396,0.950865574659141} {0.902275031655923,0.951516061128103}
(1 row) (1 row)
SET client_min_messages TO WARNING; -- suppress cascade messages SET client_min_messages TO WARNING; -- suppress cascade messages

View File

@ -0,0 +1,651 @@
--
-- TDIGEST_AGGREGATE_SUPPORT
-- test the integration of github.com/tvondra/tdigest aggregates into the citus planner
-- for push down parts of the aggregate to use parallelized execution and reduced data
-- transfer sizes for aggregates not grouped by the distribution column
--
SET citus.next_shard_id TO 20070000;
CREATE SCHEMA tdigest_aggregate_support;
SET search_path TO tdigest_aggregate_support, public;
-- create the tdigest extension when installed
SELECT CASE WHEN COUNT(*) > 0
THEN 'CREATE EXTENSION tdigest WITH SCHEMA public'
ELSE 'SELECT false AS tdigest_present' END
AS create_cmd FROM pg_available_extensions()
WHERE name = 'tdigest'
\gset
:create_cmd;
SET citus.shard_count TO 4;
SET citus.coordinator_aggregation_strategy TO 'disabled'; -- prevent aggregate execution when the aggregate can't be pushed down
CREATE TABLE latencies (a int, b int, latency double precision);
SELECT create_distributed_table('latencies', 'a');
create_distributed_table
---------------------------------------------------------------------
(1 row)
SELECT setseed(0.42); -- make the random data inserted deterministic
setseed
---------------------------------------------------------------------
(1 row)
INSERT INTO latencies
SELECT (random()*20)::int AS a,
(random()*20)::int AS b,
random()*10000.0 AS latency
FROM generate_series(1, 10000);
-- explain no grouping to verify partially pushed down for tdigest(value, compression)
EXPLAIN (COSTS OFF, VERBOSE)
SELECT tdigest(latency, 100)
FROM latencies;
QUERY PLAN
---------------------------------------------------------------------
Aggregate
Output: tdigest(remote_scan.tdigest)
-> Custom Scan (Citus Adaptive)
Output: remote_scan.tdigest
Task Count: 4
Tasks Shown: One of 4
-> Task
Query: SELECT public.tdigest(latency, 100) AS tdigest FROM tdigest_aggregate_support.latencies_20070000 latencies WHERE true
Node: host=localhost port=xxxxx dbname=regression
-> Aggregate
Output: tdigest(latency, 100)
-> Seq Scan on tdigest_aggregate_support.latencies_20070000 latencies
Output: a, b, latency
(13 rows)
-- explain grouping by distribution column is completely pushed down for tdigest(value, compression)
EXPLAIN (COSTS OFF, VERBOSE)
SELECT a, tdigest(latency, 100)
FROM latencies
GROUP BY a;
QUERY PLAN
---------------------------------------------------------------------
Custom Scan (Citus Adaptive)
Output: remote_scan.a, remote_scan.tdigest
Task Count: 4
Tasks Shown: One of 4
-> Task
Query: SELECT a, public.tdigest(latency, 100) AS tdigest FROM tdigest_aggregate_support.latencies_20070000 latencies WHERE true GROUP BY a
Node: host=localhost port=xxxxx dbname=regression
-> HashAggregate
Output: a, tdigest(latency, 100)
Group Key: latencies.a
-> Seq Scan on tdigest_aggregate_support.latencies_20070000 latencies
Output: a, b, latency
(12 rows)
-- explain grouping by non-distribution column is partially pushed down for tdigest(value, compression)
EXPLAIN (COSTS OFF, VERBOSE)
SELECT b, tdigest(latency, 100)
FROM latencies
GROUP BY b;
QUERY PLAN
---------------------------------------------------------------------
HashAggregate
Output: remote_scan.b, tdigest(remote_scan.tdigest)
Group Key: remote_scan.b
-> Custom Scan (Citus Adaptive)
Output: remote_scan.b, remote_scan.tdigest
Task Count: 4
Tasks Shown: One of 4
-> Task
Query: SELECT b, public.tdigest(latency, 100) AS tdigest FROM tdigest_aggregate_support.latencies_20070000 latencies WHERE true GROUP BY b
Node: host=localhost port=xxxxx dbname=regression
-> HashAggregate
Output: b, tdigest(latency, 100)
Group Key: latencies.b
-> Seq Scan on tdigest_aggregate_support.latencies_20070000 latencies
Output: a, b, latency
(15 rows)
-- explain no grouping to verify partially pushed down for tdigest_precentile(value, compression, quantile)
EXPLAIN (COSTS OFF, VERBOSE)
SELECT tdigest_percentile(latency, 100, 0.99)
FROM latencies;
QUERY PLAN
---------------------------------------------------------------------
Aggregate
Output: tdigest_percentile(remote_scan.tdigest_percentile, '0.99'::double precision)
-> Custom Scan (Citus Adaptive)
Output: remote_scan.tdigest_percentile
Task Count: 4
Tasks Shown: One of 4
-> Task
Query: SELECT public.tdigest(latency, 100) AS tdigest_percentile FROM tdigest_aggregate_support.latencies_20070000 latencies WHERE true
Node: host=localhost port=xxxxx dbname=regression
-> Aggregate
Output: tdigest(latency, 100)
-> Seq Scan on tdigest_aggregate_support.latencies_20070000 latencies
Output: a, b, latency
(13 rows)
-- explain grouping by distribution column is completely pushed down for tdigest_precentile(value, compression, quantile)
EXPLAIN (COSTS OFF, VERBOSE)
SELECT a, tdigest_percentile(latency, 100, 0.99)
FROM latencies
GROUP BY a;
QUERY PLAN
---------------------------------------------------------------------
Custom Scan (Citus Adaptive)
Output: remote_scan.a, remote_scan.tdigest_percentile
Task Count: 4
Tasks Shown: One of 4
-> Task
Query: SELECT a, public.tdigest_percentile(latency, 100, '0.99'::double precision) AS tdigest_percentile FROM tdigest_aggregate_support.latencies_20070000 latencies WHERE true GROUP BY a
Node: host=localhost port=xxxxx dbname=regression
-> HashAggregate
Output: a, tdigest_percentile(latency, 100, '0.99'::double precision)
Group Key: latencies.a
-> Seq Scan on tdigest_aggregate_support.latencies_20070000 latencies
Output: a, b, latency
(12 rows)
-- explain grouping by non-distribution column is partially pushed down for tdigest_precentile(value, compression, quantile)
EXPLAIN (COSTS OFF, VERBOSE)
SELECT b, tdigest_percentile(latency, 100, 0.99)
FROM latencies
GROUP BY b;
QUERY PLAN
---------------------------------------------------------------------
HashAggregate
Output: remote_scan.b, tdigest_percentile(remote_scan.tdigest_percentile, '0.99'::double precision)
Group Key: remote_scan.b
-> Custom Scan (Citus Adaptive)
Output: remote_scan.b, remote_scan.tdigest_percentile
Task Count: 4
Tasks Shown: One of 4
-> Task
Query: SELECT b, public.tdigest(latency, 100) AS tdigest_percentile FROM tdigest_aggregate_support.latencies_20070000 latencies WHERE true GROUP BY b
Node: host=localhost port=xxxxx dbname=regression
-> HashAggregate
Output: b, tdigest(latency, 100)
Group Key: latencies.b
-> Seq Scan on tdigest_aggregate_support.latencies_20070000 latencies
Output: a, b, latency
(15 rows)
-- explain no grouping to verify partially pushed down for tdigest_precentile(value, compression, quantiles[])
EXPLAIN (COSTS OFF, VERBOSE)
SELECT tdigest_percentile(latency, 100, ARRAY[0.99, 0.95])
FROM latencies;
QUERY PLAN
---------------------------------------------------------------------
Aggregate
Output: tdigest_percentile(remote_scan.tdigest_percentile, '{0.99,0.95}'::double precision[])
-> Custom Scan (Citus Adaptive)
Output: remote_scan.tdigest_percentile
Task Count: 4
Tasks Shown: One of 4
-> Task
Query: SELECT public.tdigest(latency, 100) AS tdigest_percentile FROM tdigest_aggregate_support.latencies_20070000 latencies WHERE true
Node: host=localhost port=xxxxx dbname=regression
-> Aggregate
Output: tdigest(latency, 100)
-> Seq Scan on tdigest_aggregate_support.latencies_20070000 latencies
Output: a, b, latency
(13 rows)
-- explain grouping by distribution column is completely pushed down for tdigest_precentile(value, compression, quantiles[])
EXPLAIN (COSTS OFF, VERBOSE)
SELECT a, tdigest_percentile(latency, 100, ARRAY[0.99, 0.95])
FROM latencies
GROUP BY a;
QUERY PLAN
---------------------------------------------------------------------
Custom Scan (Citus Adaptive)
Output: remote_scan.a, remote_scan.tdigest_percentile
Task Count: 4
Tasks Shown: One of 4
-> Task
Query: SELECT a, public.tdigest_percentile(latency, 100, '{0.99,0.95}'::double precision[]) AS tdigest_percentile FROM tdigest_aggregate_support.latencies_20070000 latencies WHERE true GROUP BY a
Node: host=localhost port=xxxxx dbname=regression
-> HashAggregate
Output: a, tdigest_percentile(latency, 100, '{0.99,0.95}'::double precision[])
Group Key: latencies.a
-> Seq Scan on tdigest_aggregate_support.latencies_20070000 latencies
Output: a, b, latency
(12 rows)
-- explain grouping by non-distribution column is partially pushed down for tdigest_precentile(value, compression, quantiles[])
EXPLAIN (COSTS OFF, VERBOSE)
SELECT b, tdigest_percentile(latency, 100, ARRAY[0.99, 0.95])
FROM latencies
GROUP BY b;
QUERY PLAN
---------------------------------------------------------------------
HashAggregate
Output: remote_scan.b, tdigest_percentile(remote_scan.tdigest_percentile, '{0.99,0.95}'::double precision[])
Group Key: remote_scan.b
-> Custom Scan (Citus Adaptive)
Output: remote_scan.b, remote_scan.tdigest_percentile
Task Count: 4
Tasks Shown: One of 4
-> Task
Query: SELECT b, public.tdigest(latency, 100) AS tdigest_percentile FROM tdigest_aggregate_support.latencies_20070000 latencies WHERE true GROUP BY b
Node: host=localhost port=xxxxx dbname=regression
-> HashAggregate
Output: b, tdigest(latency, 100)
Group Key: latencies.b
-> Seq Scan on tdigest_aggregate_support.latencies_20070000 latencies
Output: a, b, latency
(15 rows)
-- explain no grouping to verify partially pushed down for tdigest_precentile_of(value, compression, hypotetical_value)
EXPLAIN (COSTS OFF, VERBOSE)
SELECT tdigest_percentile_of(latency, 100, 9000)
FROM latencies;
QUERY PLAN
---------------------------------------------------------------------
Aggregate
Output: tdigest_percentile_of(remote_scan.tdigest_percentile_of, '9000'::double precision)
-> Custom Scan (Citus Adaptive)
Output: remote_scan.tdigest_percentile_of
Task Count: 4
Tasks Shown: One of 4
-> Task
Query: SELECT public.tdigest(latency, 100) AS tdigest_percentile_of FROM tdigest_aggregate_support.latencies_20070000 latencies WHERE true
Node: host=localhost port=xxxxx dbname=regression
-> Aggregate
Output: tdigest(latency, 100)
-> Seq Scan on tdigest_aggregate_support.latencies_20070000 latencies
Output: a, b, latency
(13 rows)
-- explain grouping by distribution column is completely pushed down for tdigest_precentile_of(value, compression, hypotetical_value)
EXPLAIN (COSTS OFF, VERBOSE)
SELECT a, tdigest_percentile_of(latency, 100, 9000)
FROM latencies
GROUP BY a;
QUERY PLAN
---------------------------------------------------------------------
Custom Scan (Citus Adaptive)
Output: remote_scan.a, remote_scan.tdigest_percentile_of
Task Count: 4
Tasks Shown: One of 4
-> Task
Query: SELECT a, public.tdigest_percentile_of(latency, 100, '9000'::double precision) AS tdigest_percentile_of FROM tdigest_aggregate_support.latencies_20070000 latencies WHERE true GROUP BY a
Node: host=localhost port=xxxxx dbname=regression
-> HashAggregate
Output: a, tdigest_percentile_of(latency, 100, '9000'::double precision)
Group Key: latencies.a
-> Seq Scan on tdigest_aggregate_support.latencies_20070000 latencies
Output: a, b, latency
(12 rows)
-- explain grouping by non-distribution column is partially pushed down for tdigest_precentile_of(value, compression, hypotetical_value)
EXPLAIN (COSTS OFF, VERBOSE)
SELECT b, tdigest_percentile_of(latency, 100, 9000)
FROM latencies
GROUP BY b;
QUERY PLAN
---------------------------------------------------------------------
HashAggregate
Output: remote_scan.b, tdigest_percentile_of(remote_scan.tdigest_percentile_of, '9000'::double precision)
Group Key: remote_scan.b
-> Custom Scan (Citus Adaptive)
Output: remote_scan.b, remote_scan.tdigest_percentile_of
Task Count: 4
Tasks Shown: One of 4
-> Task
Query: SELECT b, public.tdigest(latency, 100) AS tdigest_percentile_of FROM tdigest_aggregate_support.latencies_20070000 latencies WHERE true GROUP BY b
Node: host=localhost port=xxxxx dbname=regression
-> HashAggregate
Output: b, tdigest(latency, 100)
Group Key: latencies.b
-> Seq Scan on tdigest_aggregate_support.latencies_20070000 latencies
Output: a, b, latency
(15 rows)
-- explain no grouping to verify partially pushed down for tdigest_precentile_of(value, compression, hypotetical_values[])
EXPLAIN (COSTS OFF, VERBOSE)
SELECT tdigest_percentile_of(latency, 100, ARRAY[9000, 9500])
FROM latencies;
QUERY PLAN
---------------------------------------------------------------------
Aggregate
Output: tdigest_percentile_of(remote_scan.tdigest_percentile_of, '{9000,9500}'::double precision[])
-> Custom Scan (Citus Adaptive)
Output: remote_scan.tdigest_percentile_of
Task Count: 4
Tasks Shown: One of 4
-> Task
Query: SELECT public.tdigest(latency, 100) AS tdigest_percentile_of FROM tdigest_aggregate_support.latencies_20070000 latencies WHERE true
Node: host=localhost port=xxxxx dbname=regression
-> Aggregate
Output: tdigest(latency, 100)
-> Seq Scan on tdigest_aggregate_support.latencies_20070000 latencies
Output: a, b, latency
(13 rows)
-- explain grouping by distribution column is completely pushed down for tdigest_precentile_of(value, compression, hypotetical_values[])
EXPLAIN (COSTS OFF, VERBOSE)
SELECT a, tdigest_percentile_of(latency, 100, ARRAY[9000, 9500])
FROM latencies
GROUP BY a;
QUERY PLAN
---------------------------------------------------------------------
Custom Scan (Citus Adaptive)
Output: remote_scan.a, remote_scan.tdigest_percentile_of
Task Count: 4
Tasks Shown: One of 4
-> Task
Query: SELECT a, public.tdigest_percentile_of(latency, 100, '{9000,9500}'::double precision[]) AS tdigest_percentile_of FROM tdigest_aggregate_support.latencies_20070000 latencies WHERE true GROUP BY a
Node: host=localhost port=xxxxx dbname=regression
-> HashAggregate
Output: a, tdigest_percentile_of(latency, 100, '{9000,9500}'::double precision[])
Group Key: latencies.a
-> Seq Scan on tdigest_aggregate_support.latencies_20070000 latencies
Output: a, b, latency
(12 rows)
-- explain grouping by non-distribution column is partially pushed down for tdigest_precentile_of(value, compression, hypotetical_values[])
EXPLAIN (COSTS OFF, VERBOSE)
SELECT b, tdigest_percentile_of(latency, 100, ARRAY[9000, 9500])
FROM latencies
GROUP BY b;
QUERY PLAN
---------------------------------------------------------------------
HashAggregate
Output: remote_scan.b, tdigest_percentile_of(remote_scan.tdigest_percentile_of, '{9000,9500}'::double precision[])
Group Key: remote_scan.b
-> Custom Scan (Citus Adaptive)
Output: remote_scan.b, remote_scan.tdigest_percentile_of
Task Count: 4
Tasks Shown: One of 4
-> Task
Query: SELECT b, public.tdigest(latency, 100) AS tdigest_percentile_of FROM tdigest_aggregate_support.latencies_20070000 latencies WHERE true GROUP BY b
Node: host=localhost port=xxxxx dbname=regression
-> HashAggregate
Output: b, tdigest(latency, 100)
Group Key: latencies.b
-> Seq Scan on tdigest_aggregate_support.latencies_20070000 latencies
Output: a, b, latency
(15 rows)
-- verifying results - should be stable due to seed while inserting the data, if failure due to data these queries could be removed or check for certain ranges
SELECT tdigest(latency, 100) FROM latencies;
tdigest
---------------------------------------------------------------------
flags 1 count 10000 compression 100 centroids 46 (0.287235, 1) (1.025106, 1) (2.058216, 1) (5.335597, 1) (6.353631, 2) (8.434160, 3) (10.858766, 4) (15.597572, 5) (26.947867, 10) (39.185955, 13) (55.779949, 22) (97.378748, 35) (141.322218, 55) (212.927316, 65) (304.973404, 108) (435.651374, 148) (594.272516, 199) (820.494155, 292) (1214.081721, 463) (1727.098147, 547) (2337.903365, 749) (3325.589314, 1128) (4520.977007, 1300) (5638.186053, 1104) (6641.875581, 874) (7460.022115, 755) (8112.829059, 573) (8607.102557, 402) (8981.756066, 314) (9293.406295, 288) (9536.677260, 173) (9689.587344, 131) (9806.805461, 83) (9880.507729, 49) (9917.051853, 34) (9938.762095, 20) (9956.899980, 15) (9971.665384, 11) (9981.990769, 8) (9987.546338, 5) (9990.348725, 3) (9991.269368, 2) (9991.467422, 1) (9992.337047, 1) (9995.578357, 1) (9999.700339, 1)
(1 row)
SELECT tdigest_percentile(latency, 100, 0.99) FROM latencies;
tdigest_percentile
---------------------------------------------------------------------
9903.84313359954
(1 row)
SELECT tdigest_percentile(latency, 100, ARRAY[0.99, 0.95]) FROM latencies;
tdigest_percentile
---------------------------------------------------------------------
{9903.84313359954,9485.4900939991}
(1 row)
SELECT tdigest_percentile_of(latency, 100, 9000) FROM latencies;
tdigest_percentile_of
---------------------------------------------------------------------
0.903462047211093
(1 row)
SELECT tdigest_percentile_of(latency, 100, ARRAY[9000, 9500]) FROM latencies;
tdigest_percentile_of
---------------------------------------------------------------------
{0.903462047211093,0.951374818129208}
(1 row)
CREATE TABLE latencies_rollup (a int, tdigest tdigest);
SELECT create_distributed_table('latencies_rollup', 'a', colocate_with => 'latencies');
create_distributed_table
---------------------------------------------------------------------
(1 row)
INSERT INTO latencies_rollup
SELECT a, tdigest(latency, 100)
FROM latencies
GROUP BY a;
EXPLAIN (COSTS OFF, VERBOSE)
SELECT tdigest(tdigest)
FROM latencies_rollup;
QUERY PLAN
---------------------------------------------------------------------
Aggregate
Output: tdigest(remote_scan.tdigest)
-> Custom Scan (Citus Adaptive)
Output: remote_scan.tdigest
Task Count: 4
Tasks Shown: One of 4
-> Task
Query: SELECT public.tdigest(tdigest) AS tdigest FROM tdigest_aggregate_support.latencies_rollup_20070004 latencies_rollup WHERE true
Node: host=localhost port=xxxxx dbname=regression
-> Aggregate
Output: tdigest(tdigest)
-> Seq Scan on tdigest_aggregate_support.latencies_rollup_20070004 latencies_rollup
Output: a, tdigest
(13 rows)
-- explain grouping by distribution column is completely pushed down for tdigest(tdigest)
EXPLAIN (COSTS OFF, VERBOSE)
SELECT a, tdigest(tdigest)
FROM latencies_rollup
GROUP BY a;
QUERY PLAN
---------------------------------------------------------------------
Custom Scan (Citus Adaptive)
Output: remote_scan.a, remote_scan.tdigest
Task Count: 4
Tasks Shown: One of 4
-> Task
Query: SELECT a, public.tdigest(tdigest) AS tdigest FROM tdigest_aggregate_support.latencies_rollup_20070004 latencies_rollup WHERE true GROUP BY a
Node: host=localhost port=xxxxx dbname=regression
-> HashAggregate
Output: a, tdigest(tdigest)
Group Key: latencies_rollup.a
-> Seq Scan on tdigest_aggregate_support.latencies_rollup_20070004 latencies_rollup
Output: a, tdigest
(12 rows)
-- explain no grouping to verify partially pushed down for tdigest_precentile(tdigest, quantile)
EXPLAIN (COSTS OFF, VERBOSE)
SELECT tdigest_percentile(tdigest, 0.99)
FROM latencies_rollup;
QUERY PLAN
---------------------------------------------------------------------
Aggregate
Output: tdigest_percentile(remote_scan.tdigest_percentile, '0.99'::double precision)
-> Custom Scan (Citus Adaptive)
Output: remote_scan.tdigest_percentile
Task Count: 4
Tasks Shown: One of 4
-> Task
Query: SELECT public.tdigest(tdigest) AS tdigest_percentile FROM tdigest_aggregate_support.latencies_rollup_20070004 latencies_rollup WHERE true
Node: host=localhost port=xxxxx dbname=regression
-> Aggregate
Output: tdigest(tdigest)
-> Seq Scan on tdigest_aggregate_support.latencies_rollup_20070004 latencies_rollup
Output: a, tdigest
(13 rows)
-- explain grouping by distribution column is completely pushed down for tdigest_precentile(tdigest, quantile)
EXPLAIN (COSTS OFF, VERBOSE)
SELECT a, tdigest_percentile(tdigest, 0.99)
FROM latencies_rollup
GROUP BY a;
QUERY PLAN
---------------------------------------------------------------------
Custom Scan (Citus Adaptive)
Output: remote_scan.a, remote_scan.tdigest_percentile
Task Count: 4
Tasks Shown: One of 4
-> Task
Query: SELECT a, public.tdigest_percentile(tdigest, '0.99'::double precision) AS tdigest_percentile FROM tdigest_aggregate_support.latencies_rollup_20070004 latencies_rollup WHERE true GROUP BY a
Node: host=localhost port=xxxxx dbname=regression
-> HashAggregate
Output: a, tdigest_percentile(tdigest, '0.99'::double precision)
Group Key: latencies_rollup.a
-> Seq Scan on tdigest_aggregate_support.latencies_rollup_20070004 latencies_rollup
Output: a, tdigest
(12 rows)
-- explain no grouping to verify partially pushed down for tdigest_precentile(value, compression, quantiles[])
EXPLAIN (COSTS OFF, VERBOSE)
SELECT tdigest_percentile(tdigest, ARRAY[0.99, 0.95])
FROM latencies_rollup;
QUERY PLAN
---------------------------------------------------------------------
Aggregate
Output: tdigest_percentile(remote_scan.tdigest_percentile, '{0.99,0.95}'::double precision[])
-> Custom Scan (Citus Adaptive)
Output: remote_scan.tdigest_percentile
Task Count: 4
Tasks Shown: One of 4
-> Task
Query: SELECT public.tdigest(tdigest) AS tdigest_percentile FROM tdigest_aggregate_support.latencies_rollup_20070004 latencies_rollup WHERE true
Node: host=localhost port=xxxxx dbname=regression
-> Aggregate
Output: tdigest(tdigest)
-> Seq Scan on tdigest_aggregate_support.latencies_rollup_20070004 latencies_rollup
Output: a, tdigest
(13 rows)
-- explain grouping by distribution column is completely pushed down for tdigest_precentile(value, compression, quantiles[])
EXPLAIN (COSTS OFF, VERBOSE)
SELECT a, tdigest_percentile(tdigest, ARRAY[0.99, 0.95])
FROM latencies_rollup
GROUP BY a;
QUERY PLAN
---------------------------------------------------------------------
Custom Scan (Citus Adaptive)
Output: remote_scan.a, remote_scan.tdigest_percentile
Task Count: 4
Tasks Shown: One of 4
-> Task
Query: SELECT a, public.tdigest_percentile(tdigest, '{0.99,0.95}'::double precision[]) AS tdigest_percentile FROM tdigest_aggregate_support.latencies_rollup_20070004 latencies_rollup WHERE true GROUP BY a
Node: host=localhost port=xxxxx dbname=regression
-> HashAggregate
Output: a, tdigest_percentile(tdigest, '{0.99,0.95}'::double precision[])
Group Key: latencies_rollup.a
-> Seq Scan on tdigest_aggregate_support.latencies_rollup_20070004 latencies_rollup
Output: a, tdigest
(12 rows)
-- explain no grouping to verify partially pushed down for tdigest_precentile_of(value, compression, hypotetical_value)
EXPLAIN (COSTS OFF, VERBOSE)
SELECT tdigest_percentile_of(tdigest, 9000)
FROM latencies_rollup;
QUERY PLAN
---------------------------------------------------------------------
Aggregate
Output: tdigest_percentile_of(remote_scan.tdigest_percentile_of, '9000'::double precision)
-> Custom Scan (Citus Adaptive)
Output: remote_scan.tdigest_percentile_of
Task Count: 4
Tasks Shown: One of 4
-> Task
Query: SELECT public.tdigest(tdigest) AS tdigest_percentile_of FROM tdigest_aggregate_support.latencies_rollup_20070004 latencies_rollup WHERE true
Node: host=localhost port=xxxxx dbname=regression
-> Aggregate
Output: tdigest(tdigest)
-> Seq Scan on tdigest_aggregate_support.latencies_rollup_20070004 latencies_rollup
Output: a, tdigest
(13 rows)
-- explain grouping by distribution column is completely pushed down for tdigest_precentile_of(value, compression, hypotetical_value)
EXPLAIN (COSTS OFF, VERBOSE)
SELECT a, tdigest_percentile_of(tdigest, 9000)
FROM latencies_rollup
GROUP BY a;
QUERY PLAN
---------------------------------------------------------------------
Custom Scan (Citus Adaptive)
Output: remote_scan.a, remote_scan.tdigest_percentile_of
Task Count: 4
Tasks Shown: One of 4
-> Task
Query: SELECT a, public.tdigest_percentile_of(tdigest, '9000'::double precision) AS tdigest_percentile_of FROM tdigest_aggregate_support.latencies_rollup_20070004 latencies_rollup WHERE true GROUP BY a
Node: host=localhost port=xxxxx dbname=regression
-> HashAggregate
Output: a, tdigest_percentile_of(tdigest, '9000'::double precision)
Group Key: latencies_rollup.a
-> Seq Scan on tdigest_aggregate_support.latencies_rollup_20070004 latencies_rollup
Output: a, tdigest
(12 rows)
-- explain no grouping to verify partially pushed down for tdigest_precentile_of(value, compression, hypotetical_values[])
EXPLAIN (COSTS OFF, VERBOSE)
SELECT tdigest_percentile_of(tdigest, ARRAY[9000, 9500])
FROM latencies_rollup;
QUERY PLAN
---------------------------------------------------------------------
Aggregate
Output: tdigest_percentile_of(remote_scan.tdigest_percentile_of, '{9000,9500}'::double precision[])
-> Custom Scan (Citus Adaptive)
Output: remote_scan.tdigest_percentile_of
Task Count: 4
Tasks Shown: One of 4
-> Task
Query: SELECT public.tdigest(tdigest) AS tdigest_percentile_of FROM tdigest_aggregate_support.latencies_rollup_20070004 latencies_rollup WHERE true
Node: host=localhost port=xxxxx dbname=regression
-> Aggregate
Output: tdigest(tdigest)
-> Seq Scan on tdigest_aggregate_support.latencies_rollup_20070004 latencies_rollup
Output: a, tdigest
(13 rows)
-- explain grouping by distribution column is completely pushed down for tdigest_precentile_of(value, compression, hypotetical_values[])
EXPLAIN (COSTS OFF, VERBOSE)
SELECT a, tdigest_percentile_of(tdigest, ARRAY[9000, 9500])
FROM latencies_rollup
GROUP BY a;
QUERY PLAN
---------------------------------------------------------------------
Custom Scan (Citus Adaptive)
Output: remote_scan.a, remote_scan.tdigest_percentile_of
Task Count: 4
Tasks Shown: One of 4
-> Task
Query: SELECT a, public.tdigest_percentile_of(tdigest, '{9000,9500}'::double precision[]) AS tdigest_percentile_of FROM tdigest_aggregate_support.latencies_rollup_20070004 latencies_rollup WHERE true GROUP BY a
Node: host=localhost port=xxxxx dbname=regression
-> HashAggregate
Output: a, tdigest_percentile_of(tdigest, '{9000,9500}'::double precision[])
Group Key: latencies_rollup.a
-> Seq Scan on tdigest_aggregate_support.latencies_rollup_20070004 latencies_rollup
Output: a, tdigest
(12 rows)
-- verifying results - should be stable due to seed while inserting the data, if failure due to data these queries could be removed or check for certain ranges
SELECT tdigest(tdigest) FROM latencies_rollup;
tdigest
---------------------------------------------------------------------
flags 1 count 10000 compression 100 centroids 47 (0.287235, 1) (1.025106, 1) (2.058216, 1) (5.335597, 1) (6.353631, 2) (8.434160, 3) (10.858766, 4) (15.597572, 5) (26.853448, 9) (36.608782, 11) (49.983794, 20) (85.586987, 27) (118.226702, 37) (173.623153, 56) (249.583971, 87) (354.777375, 112) (493.422943, 177) (697.499642, 262) (996.351736, 338) (1425.136508, 464) (1972.172107, 623) (2665.959054, 805) (3451.617535, 827) (4429.146575, 1195) (5587.124850, 1104) (6624.983751, 966) (7474.888557, 773) (8184.273132, 637) (8730.956261, 431) (9119.479419, 333) (9411.778428, 207) (9582.688196, 152) (9707.657431, 103) (9807.341000, 73) (9881.180119, 55) (9919.393423, 33) (9942.691902, 20) (9960.171491, 16) (9975.978413, 8) (9982.487777, 7) (9987.546338, 5) (9990.348725, 3) (9991.269368, 2) (9991.467422, 1) (9992.337047, 1) (9995.578357, 1) (9999.700339, 1)
(1 row)
SELECT tdigest_percentile(tdigest, 0.99) FROM latencies_rollup;
tdigest_percentile
---------------------------------------------------------------------
9903.32646582201
(1 row)
SELECT tdigest_percentile(tdigest, ARRAY[0.99, 0.95]) FROM latencies_rollup;
tdigest_percentile
---------------------------------------------------------------------
{9903.32646582201,9492.23455844906}
(1 row)
SELECT tdigest_percentile_of(tdigest, 9000) FROM latencies_rollup;
tdigest_percentile_of
---------------------------------------------------------------------
0.902852659582613
(1 row)
SELECT tdigest_percentile_of(tdigest, ARRAY[9000, 9500]) FROM latencies_rollup;
tdigest_percentile_of
---------------------------------------------------------------------
{0.902852659582613,0.950865574658712}
(1 row)
SET client_min_messages TO WARNING; -- suppress cascade messages
DROP SCHEMA tdigest_aggregate_support CASCADE;