2016-06-08 05:36:22 +02:00
|
|
|
--
|
|
|
|
-- PARALLEL
|
|
|
|
--
|
|
|
|
|
Clean up duplicate table and function names in regression tests.
Many of the objects we create during the regression tests are put in the
public schema, so that using the same names in different regression tests
creates a hazard of test failures if any two such scripts run concurrently.
This patch cleans up a bunch of latent hazards of that sort, as well as two
live hazards.
The current situation in this regard is far worse than it was a year or two
back, because practically all of the partitioning-related test cases have
reused table names with enthusiasm. I despaired of cleaning up that mess
within the five most-affected tests (create_table, alter_table, insert,
update, inherit); fortunately those don't run concurrently.
Other than partitioning problems, most of the issues boil down to using
names like "foo", "bar", "tmp", etc, without thought for the fact that
other test scripts might use similar names concurrently. I've made an
effort to make all such names more specific.
One of the live hazards was that commit 7421f4b8 caused with.sql to
create a table named "test", conflicting with a similarly-named table
in alter_table.sql; this was exposed in the buildfarm recently.
The other one was that join.sql and transactions.sql both create tables
named "foo" and "bar"; but join.sql's uses of those names date back
only to December or so.
Since commit 7421f4b8 was back-patched to v10, back-patch a minimal
fix for that problem. The rest of this is just future-proofing.
Discussion: https://postgr.es/m/4627.1521070268@sss.pgh.pa.us
2018-03-15 22:08:51 +01:00
|
|
|
create function sp_parallel_restricted(int) returns int as
|
2016-06-17 14:35:47 +02:00
|
|
|
$$begin return $1; end$$ language plpgsql parallel restricted;
|
2016-06-16 18:00:55 +02:00
|
|
|
|
2021-03-15 11:27:08 +01:00
|
|
|
begin;
|
2016-06-08 05:36:22 +02:00
|
|
|
|
2016-06-18 06:28:51 +02:00
|
|
|
-- encourage use of parallel plans
|
2016-06-08 05:36:22 +02:00
|
|
|
set parallel_setup_cost=0;
|
|
|
|
set parallel_tuple_cost=0;
|
Replace min_parallel_relation_size with two new GUCs.
When min_parallel_relation_size was added, the only supported type
of parallel scan was a parallel sequential scan, but there are
pending patches for parallel index scan, parallel index-only scan,
and parallel bitmap heap scan. Those patches introduce two new
types of complications: first, what's relevant is not really the
total size of the relation but the portion of it that we will scan;
and second, index pages and heap pages shouldn't necessarily be
treated in exactly the same way. Typically, the number of index
pages will be quite small, but that doesn't necessarily mean that
a parallel index scan can't pay off.
Therefore, we introduce min_parallel_table_scan_size, which works
out a degree of parallelism for scans based on the number of table
pages that will be scanned (and which is therefore equivalent to
min_parallel_relation_size for parallel sequential scans) and also
min_parallel_index_scan_size which can be used to work out a degree
of parallelism based on the number of index pages that will be
scanned.
Amit Kapila and Robert Haas
Discussion: http://postgr.es/m/CAA4eK1KowGSYYVpd2qPpaPPA5R90r++QwDFbrRECTE9H_HvpOg@mail.gmail.com
Discussion: http://postgr.es/m/CAA4eK1+TnM4pXQbvn7OXqam+k_HZqb0ROZUMxOiL6DWJYCyYow@mail.gmail.com
2017-02-15 19:37:24 +01:00
|
|
|
set min_parallel_table_scan_size=0;
|
2016-06-16 18:00:55 +02:00
|
|
|
set max_parallel_workers_per_gather=4;
|
2016-06-08 05:36:22 +02:00
|
|
|
|
Support Parallel Append plan nodes.
When we create an Append node, we can spread out the workers over the
subplans instead of piling on to each subplan one at a time, which
should typically be a bit more efficient, both because the startup
cost of any plan executed entirely by one worker is paid only once and
also because of reduced contention. We can also construct Append
plans using a mix of partial and non-partial subplans, which may allow
for parallelism in places that otherwise couldn't support it.
Unfortunately, this patch doesn't handle the important case of
parallelizing UNION ALL by running each branch in a separate worker;
the executor infrastructure is added here, but more planner work is
needed.
Amit Khandekar, Robert Haas, Amul Sul, reviewed and tested by
Ashutosh Bapat, Amit Langote, Rafia Sabih, Amit Kapila, and
Rajkumar Raghuwanshi.
Discussion: http://postgr.es/m/CAJ3gD9dy0K_E8r727heqXoBmWZ83HwLFwdcaSSmBQ1+S+vRuUQ@mail.gmail.com
2017-12-05 23:28:39 +01:00
|
|
|
-- Parallel Append with partial-subplans
|
2016-06-08 05:36:22 +02:00
|
|
|
explain (costs off)
|
Support Parallel Append plan nodes.
When we create an Append node, we can spread out the workers over the
subplans instead of piling on to each subplan one at a time, which
should typically be a bit more efficient, both because the startup
cost of any plan executed entirely by one worker is paid only once and
also because of reduced contention. We can also construct Append
plans using a mix of partial and non-partial subplans, which may allow
for parallelism in places that otherwise couldn't support it.
Unfortunately, this patch doesn't handle the important case of
parallelizing UNION ALL by running each branch in a separate worker;
the executor infrastructure is added here, but more planner work is
needed.
Amit Khandekar, Robert Haas, Amul Sul, reviewed and tested by
Ashutosh Bapat, Amit Langote, Rafia Sabih, Amit Kapila, and
Rajkumar Raghuwanshi.
Discussion: http://postgr.es/m/CAJ3gD9dy0K_E8r727heqXoBmWZ83HwLFwdcaSSmBQ1+S+vRuUQ@mail.gmail.com
2017-12-05 23:28:39 +01:00
|
|
|
select round(avg(aa)), sum(aa) from a_star;
|
2017-12-06 04:40:05 +01:00
|
|
|
select round(avg(aa)), sum(aa) from a_star a1;
|
Support Parallel Append plan nodes.
When we create an Append node, we can spread out the workers over the
subplans instead of piling on to each subplan one at a time, which
should typically be a bit more efficient, both because the startup
cost of any plan executed entirely by one worker is paid only once and
also because of reduced contention. We can also construct Append
plans using a mix of partial and non-partial subplans, which may allow
for parallelism in places that otherwise couldn't support it.
Unfortunately, this patch doesn't handle the important case of
parallelizing UNION ALL by running each branch in a separate worker;
the executor infrastructure is added here, but more planner work is
needed.
Amit Khandekar, Robert Haas, Amul Sul, reviewed and tested by
Ashutosh Bapat, Amit Langote, Rafia Sabih, Amit Kapila, and
Rajkumar Raghuwanshi.
Discussion: http://postgr.es/m/CAJ3gD9dy0K_E8r727heqXoBmWZ83HwLFwdcaSSmBQ1+S+vRuUQ@mail.gmail.com
2017-12-05 23:28:39 +01:00
|
|
|
|
|
|
|
-- Parallel Append with both partial and non-partial subplans
|
|
|
|
alter table c_star set (parallel_workers = 0);
|
|
|
|
alter table d_star set (parallel_workers = 0);
|
|
|
|
explain (costs off)
|
|
|
|
select round(avg(aa)), sum(aa) from a_star;
|
2017-12-06 04:40:05 +01:00
|
|
|
select round(avg(aa)), sum(aa) from a_star a2;
|
Support Parallel Append plan nodes.
When we create an Append node, we can spread out the workers over the
subplans instead of piling on to each subplan one at a time, which
should typically be a bit more efficient, both because the startup
cost of any plan executed entirely by one worker is paid only once and
also because of reduced contention. We can also construct Append
plans using a mix of partial and non-partial subplans, which may allow
for parallelism in places that otherwise couldn't support it.
Unfortunately, this patch doesn't handle the important case of
parallelizing UNION ALL by running each branch in a separate worker;
the executor infrastructure is added here, but more planner work is
needed.
Amit Khandekar, Robert Haas, Amul Sul, reviewed and tested by
Ashutosh Bapat, Amit Langote, Rafia Sabih, Amit Kapila, and
Rajkumar Raghuwanshi.
Discussion: http://postgr.es/m/CAJ3gD9dy0K_E8r727heqXoBmWZ83HwLFwdcaSSmBQ1+S+vRuUQ@mail.gmail.com
2017-12-05 23:28:39 +01:00
|
|
|
|
|
|
|
-- Parallel Append with only non-partial subplans
|
|
|
|
alter table a_star set (parallel_workers = 0);
|
|
|
|
alter table b_star set (parallel_workers = 0);
|
|
|
|
alter table e_star set (parallel_workers = 0);
|
|
|
|
alter table f_star set (parallel_workers = 0);
|
|
|
|
explain (costs off)
|
|
|
|
select round(avg(aa)), sum(aa) from a_star;
|
2017-12-06 04:40:05 +01:00
|
|
|
select round(avg(aa)), sum(aa) from a_star a3;
|
Support Parallel Append plan nodes.
When we create an Append node, we can spread out the workers over the
subplans instead of piling on to each subplan one at a time, which
should typically be a bit more efficient, both because the startup
cost of any plan executed entirely by one worker is paid only once and
also because of reduced contention. We can also construct Append
plans using a mix of partial and non-partial subplans, which may allow
for parallelism in places that otherwise couldn't support it.
Unfortunately, this patch doesn't handle the important case of
parallelizing UNION ALL by running each branch in a separate worker;
the executor infrastructure is added here, but more planner work is
needed.
Amit Khandekar, Robert Haas, Amul Sul, reviewed and tested by
Ashutosh Bapat, Amit Langote, Rafia Sabih, Amit Kapila, and
Rajkumar Raghuwanshi.
Discussion: http://postgr.es/m/CAJ3gD9dy0K_E8r727heqXoBmWZ83HwLFwdcaSSmBQ1+S+vRuUQ@mail.gmail.com
2017-12-05 23:28:39 +01:00
|
|
|
|
|
|
|
-- Disable Parallel Append
|
|
|
|
alter table a_star reset (parallel_workers);
|
|
|
|
alter table b_star reset (parallel_workers);
|
|
|
|
alter table c_star reset (parallel_workers);
|
|
|
|
alter table d_star reset (parallel_workers);
|
|
|
|
alter table e_star reset (parallel_workers);
|
|
|
|
alter table f_star reset (parallel_workers);
|
|
|
|
set enable_parallel_append to off;
|
|
|
|
explain (costs off)
|
|
|
|
select round(avg(aa)), sum(aa) from a_star;
|
2017-12-06 04:40:05 +01:00
|
|
|
select round(avg(aa)), sum(aa) from a_star a4;
|
Support Parallel Append plan nodes.
When we create an Append node, we can spread out the workers over the
subplans instead of piling on to each subplan one at a time, which
should typically be a bit more efficient, both because the startup
cost of any plan executed entirely by one worker is paid only once and
also because of reduced contention. We can also construct Append
plans using a mix of partial and non-partial subplans, which may allow
for parallelism in places that otherwise couldn't support it.
Unfortunately, this patch doesn't handle the important case of
parallelizing UNION ALL by running each branch in a separate worker;
the executor infrastructure is added here, but more planner work is
needed.
Amit Khandekar, Robert Haas, Amul Sul, reviewed and tested by
Ashutosh Bapat, Amit Langote, Rafia Sabih, Amit Kapila, and
Rajkumar Raghuwanshi.
Discussion: http://postgr.es/m/CAJ3gD9dy0K_E8r727heqXoBmWZ83HwLFwdcaSSmBQ1+S+vRuUQ@mail.gmail.com
2017-12-05 23:28:39 +01:00
|
|
|
reset enable_parallel_append;
|
2016-06-08 05:36:22 +02:00
|
|
|
|
2018-02-28 16:56:06 +01:00
|
|
|
-- Parallel Append that runs serially
|
Clean up duplicate table and function names in regression tests.
Many of the objects we create during the regression tests are put in the
public schema, so that using the same names in different regression tests
creates a hazard of test failures if any two such scripts run concurrently.
This patch cleans up a bunch of latent hazards of that sort, as well as two
live hazards.
The current situation in this regard is far worse than it was a year or two
back, because practically all of the partitioning-related test cases have
reused table names with enthusiasm. I despaired of cleaning up that mess
within the five most-affected tests (create_table, alter_table, insert,
update, inherit); fortunately those don't run concurrently.
Other than partitioning problems, most of the issues boil down to using
names like "foo", "bar", "tmp", etc, without thought for the fact that
other test scripts might use similar names concurrently. I've made an
effort to make all such names more specific.
One of the live hazards was that commit 7421f4b8 caused with.sql to
create a table named "test", conflicting with a similarly-named table
in alter_table.sql; this was exposed in the buildfarm recently.
The other one was that join.sql and transactions.sql both create tables
named "foo" and "bar"; but join.sql's uses of those names date back
only to December or so.
Since commit 7421f4b8 was back-patched to v10, back-patch a minimal
fix for that problem. The rest of this is just future-proofing.
Discussion: https://postgr.es/m/4627.1521070268@sss.pgh.pa.us
2018-03-15 22:08:51 +01:00
|
|
|
create function sp_test_func() returns setof text as
|
2018-02-28 16:56:06 +01:00
|
|
|
$$ select 'foo'::varchar union all select 'bar'::varchar $$
|
|
|
|
language sql stable;
|
Clean up duplicate table and function names in regression tests.
Many of the objects we create during the regression tests are put in the
public schema, so that using the same names in different regression tests
creates a hazard of test failures if any two such scripts run concurrently.
This patch cleans up a bunch of latent hazards of that sort, as well as two
live hazards.
The current situation in this regard is far worse than it was a year or two
back, because practically all of the partitioning-related test cases have
reused table names with enthusiasm. I despaired of cleaning up that mess
within the five most-affected tests (create_table, alter_table, insert,
update, inherit); fortunately those don't run concurrently.
Other than partitioning problems, most of the issues boil down to using
names like "foo", "bar", "tmp", etc, without thought for the fact that
other test scripts might use similar names concurrently. I've made an
effort to make all such names more specific.
One of the live hazards was that commit 7421f4b8 caused with.sql to
create a table named "test", conflicting with a similarly-named table
in alter_table.sql; this was exposed in the buildfarm recently.
The other one was that join.sql and transactions.sql both create tables
named "foo" and "bar"; but join.sql's uses of those names date back
only to December or so.
Since commit 7421f4b8 was back-patched to v10, back-patch a minimal
fix for that problem. The rest of this is just future-proofing.
Discussion: https://postgr.es/m/4627.1521070268@sss.pgh.pa.us
2018-03-15 22:08:51 +01:00
|
|
|
select sp_test_func() order by 1;
|
2018-02-28 16:56:06 +01:00
|
|
|
|
2018-06-20 04:21:42 +02:00
|
|
|
-- Parallel Append is not to be used when the subpath depends on the outer param
|
|
|
|
create table part_pa_test(a int, b int) partition by range(a);
|
|
|
|
create table part_pa_test_p1 partition of part_pa_test for values from (minvalue) to (0);
|
|
|
|
create table part_pa_test_p2 partition of part_pa_test for values from (0) to (maxvalue);
|
|
|
|
explain (costs off)
|
|
|
|
select (select max((select pa1.b from part_pa_test pa1 where pa1.a = pa2.a)))
|
|
|
|
from part_pa_test pa2;
|
|
|
|
drop table part_pa_test;
|
|
|
|
|
2017-11-15 14:17:29 +01:00
|
|
|
-- test with leader participation disabled
|
|
|
|
set parallel_leader_participation = off;
|
|
|
|
explain (costs off)
|
|
|
|
select count(*) from tenk1 where stringu1 = 'GRAAAA';
|
|
|
|
select count(*) from tenk1 where stringu1 = 'GRAAAA';
|
|
|
|
|
|
|
|
-- test with leader participation disabled, but no workers available (so
|
|
|
|
-- the leader will have to run the plan despite the setting)
|
|
|
|
set max_parallel_workers = 0;
|
|
|
|
explain (costs off)
|
|
|
|
select count(*) from tenk1 where stringu1 = 'GRAAAA';
|
|
|
|
select count(*) from tenk1 where stringu1 = 'GRAAAA';
|
|
|
|
|
|
|
|
reset max_parallel_workers;
|
|
|
|
reset parallel_leader_participation;
|
|
|
|
|
2016-06-16 18:00:55 +02:00
|
|
|
-- test that parallel_restricted function doesn't run in worker
|
|
|
|
alter table tenk1 set (parallel_workers = 4);
|
|
|
|
explain (verbose, costs off)
|
Clean up duplicate table and function names in regression tests.
Many of the objects we create during the regression tests are put in the
public schema, so that using the same names in different regression tests
creates a hazard of test failures if any two such scripts run concurrently.
This patch cleans up a bunch of latent hazards of that sort, as well as two
live hazards.
The current situation in this regard is far worse than it was a year or two
back, because practically all of the partitioning-related test cases have
reused table names with enthusiasm. I despaired of cleaning up that mess
within the five most-affected tests (create_table, alter_table, insert,
update, inherit); fortunately those don't run concurrently.
Other than partitioning problems, most of the issues boil down to using
names like "foo", "bar", "tmp", etc, without thought for the fact that
other test scripts might use similar names concurrently. I've made an
effort to make all such names more specific.
One of the live hazards was that commit 7421f4b8 caused with.sql to
create a table named "test", conflicting with a similarly-named table
in alter_table.sql; this was exposed in the buildfarm recently.
The other one was that join.sql and transactions.sql both create tables
named "foo" and "bar"; but join.sql's uses of those names date back
only to December or so.
Since commit 7421f4b8 was back-patched to v10, back-patch a minimal
fix for that problem. The rest of this is just future-proofing.
Discussion: https://postgr.es/m/4627.1521070268@sss.pgh.pa.us
2018-03-15 22:08:51 +01:00
|
|
|
select sp_parallel_restricted(unique1) from tenk1
|
2016-06-16 18:00:55 +02:00
|
|
|
where stringu1 = 'GRAAAA' order by 1;
|
|
|
|
|
2016-06-17 22:25:02 +02:00
|
|
|
-- test parallel plan when group by expression is in target list.
|
|
|
|
explain (costs off)
|
|
|
|
select length(stringu1) from tenk1 group by length(stringu1);
|
|
|
|
select length(stringu1) from tenk1 group by length(stringu1);
|
|
|
|
|
2016-06-18 06:28:51 +02:00
|
|
|
explain (costs off)
|
|
|
|
select stringu1, count(*) from tenk1 group by stringu1 order by stringu1;
|
|
|
|
|
2016-06-17 22:25:02 +02:00
|
|
|
-- test that parallel plan for aggregates is not selected when
|
|
|
|
-- target list contains parallel restricted clause.
|
|
|
|
explain (costs off)
|
Clean up duplicate table and function names in regression tests.
Many of the objects we create during the regression tests are put in the
public schema, so that using the same names in different regression tests
creates a hazard of test failures if any two such scripts run concurrently.
This patch cleans up a bunch of latent hazards of that sort, as well as two
live hazards.
The current situation in this regard is far worse than it was a year or two
back, because practically all of the partitioning-related test cases have
reused table names with enthusiasm. I despaired of cleaning up that mess
within the five most-affected tests (create_table, alter_table, insert,
update, inherit); fortunately those don't run concurrently.
Other than partitioning problems, most of the issues boil down to using
names like "foo", "bar", "tmp", etc, without thought for the fact that
other test scripts might use similar names concurrently. I've made an
effort to make all such names more specific.
One of the live hazards was that commit 7421f4b8 caused with.sql to
create a table named "test", conflicting with a similarly-named table
in alter_table.sql; this was exposed in the buildfarm recently.
The other one was that join.sql and transactions.sql both create tables
named "foo" and "bar"; but join.sql's uses of those names date back
only to December or so.
Since commit 7421f4b8 was back-patched to v10, back-patch a minimal
fix for that problem. The rest of this is just future-proofing.
Discussion: https://postgr.es/m/4627.1521070268@sss.pgh.pa.us
2018-03-15 22:08:51 +01:00
|
|
|
select sum(sp_parallel_restricted(unique1)) from tenk1
|
|
|
|
group by(sp_parallel_restricted(unique1));
|
2016-06-17 22:25:02 +02:00
|
|
|
|
2017-10-27 22:22:39 +02:00
|
|
|
-- test prepared statement
|
|
|
|
prepare tenk1_count(integer) As select count((unique1)) from tenk1 where hundred > $1;
|
|
|
|
explain (costs off) execute tenk1_count(1);
|
|
|
|
execute tenk1_count(1);
|
|
|
|
deallocate tenk1_count;
|
|
|
|
|
2017-02-15 00:09:47 +01:00
|
|
|
-- test parallel plans for queries containing un-correlated subplans.
|
|
|
|
alter table tenk2 set (parallel_workers = 0);
|
|
|
|
explain (costs off)
|
|
|
|
select count(*) from tenk1 where (two, four) not in
|
|
|
|
(select hundred, thousand from tenk2 where thousand > 100);
|
|
|
|
select count(*) from tenk1 where (two, four) not in
|
|
|
|
(select hundred, thousand from tenk2 where thousand > 100);
|
2017-04-18 21:43:56 +02:00
|
|
|
-- this is not parallel-safe due to use of random() within SubLink's testexpr:
|
|
|
|
explain (costs off)
|
|
|
|
select * from tenk1 where (unique1 + random())::integer not in
|
|
|
|
(select ten from tenk2);
|
2017-02-15 00:09:47 +01:00
|
|
|
alter table tenk2 reset (parallel_workers);
|
|
|
|
|
2017-11-16 18:06:14 +01:00
|
|
|
-- test parallel plan for a query containing initplan.
|
|
|
|
set enable_indexscan = off;
|
|
|
|
set enable_indexonlyscan = off;
|
|
|
|
set enable_bitmapscan = off;
|
|
|
|
alter table tenk2 set (parallel_workers = 2);
|
|
|
|
|
|
|
|
explain (costs off)
|
|
|
|
select count(*) from tenk1
|
|
|
|
where tenk1.unique1 = (Select max(tenk2.unique1) from tenk2);
|
|
|
|
select count(*) from tenk1
|
|
|
|
where tenk1.unique1 = (Select max(tenk2.unique1) from tenk2);
|
|
|
|
|
|
|
|
reset enable_indexscan;
|
|
|
|
reset enable_indexonlyscan;
|
|
|
|
reset enable_bitmapscan;
|
|
|
|
alter table tenk2 reset (parallel_workers);
|
|
|
|
|
2017-02-15 19:53:24 +01:00
|
|
|
-- test parallel index scans.
|
|
|
|
set enable_seqscan to off;
|
|
|
|
set enable_bitmapscan to off;
|
Enhance nbtree ScalarArrayOp execution.
Commit 9e8da0f7 taught nbtree to handle ScalarArrayOpExpr quals
natively. This works by pushing down the full context (the array keys)
to the nbtree index AM, enabling it to execute multiple primitive index
scans that the planner treats as one continuous index scan/index path.
This earlier enhancement enabled nbtree ScalarArrayOp index-only scans.
It also allowed scans with ScalarArrayOp quals to return ordered results
(with some notable restrictions, described further down).
Take this general approach a lot further: teach nbtree SAOP index scans
to decide how to execute ScalarArrayOp scans (when and where to start
the next primitive index scan) based on physical index characteristics.
This can be far more efficient. All SAOP scans will now reliably avoid
duplicative leaf page accesses (just like any other nbtree index scan).
SAOP scans whose array keys are naturally clustered together now require
far fewer index descents, since we'll reliably avoid starting a new
primitive scan just to get to a later offset from the same leaf page.
The scan's arrays now advance using binary searches for the array
element that best matches the next tuple's attribute value. Required
scan key arrays (i.e. arrays from scan keys that can terminate the scan)
ratchet forward in lockstep with the index scan. Non-required arrays
(i.e. arrays from scan keys that can only exclude non-matching tuples)
"advance" without the process ever rolling over to a higher-order array.
Naturally, only required SAOP scan keys trigger skipping over leaf pages
(non-required arrays cannot safely end or start primitive index scans).
Consequently, even index scans of a composite index with a high-order
inequality scan key (which we'll mark required) and a low-order SAOP
scan key (which we won't mark required) now avoid repeating leaf page
accesses -- that benefit isn't limited to simpler equality-only cases.
In general, all nbtree index scans now output tuples as if they were one
continuous index scan -- even scans that mix a high-order inequality
with lower-order SAOP equalities reliably output tuples in index order.
This allows us to remove a couple of special cases that were applied
when building index paths with SAOP clauses during planning.
Bugfix commit 807a40c5 taught the planner to avoid generating unsafe
path keys: path keys on a multicolumn index path, with a SAOP clause on
any attribute beyond the first/most significant attribute. These cases
are now all safe, so we go back to generating path keys without regard
for the presence of SAOP clauses (just like with any other clause type).
Affected queries can now exploit scan output order in all the usual ways
(e.g., certain "ORDER BY ... LIMIT n" queries can now terminate early).
Also undo changes from follow-up bugfix commit a4523c5a, which taught
the planner to produce alternative index paths, with path keys, but
without low-order SAOP index quals (filter quals were used instead).
We'll no longer generate these alternative paths, since they can no
longer offer any meaningful advantages over standard index qual paths.
Affected queries thereby avoid all of the disadvantages that come from
using filter quals within index scan nodes. They can avoid extra heap
page accesses from using filter quals to exclude non-matching tuples
(index quals will never have that problem). They can also skip over
irrelevant sections of the index in more cases (though only when nbtree
determines that starting another primitive scan actually makes sense).
There is a theoretical risk that removing restrictions on SAOP index
paths from the planner will break compatibility with amcanorder-based
index AMs maintained as extensions. Such an index AM could have the
same limitations around ordered SAOP scans as nbtree had up until now.
Adding a pro forma incompatibility item about the issue to the Postgres
17 release notes seems like a good idea.
Author: Peter Geoghegan <pg@bowt.ie>
Author: Matthias van de Meent <boekewurm+postgres@gmail.com>
Reviewed-By: Heikki Linnakangas <hlinnaka@iki.fi>
Reviewed-By: Matthias van de Meent <boekewurm+postgres@gmail.com>
Reviewed-By: Tomas Vondra <tomas.vondra@enterprisedb.com>
Discussion: https://postgr.es/m/CAH2-Wz=ksvN_sjcnD1+Bt-WtifRA5ok48aDYnq3pkKhxgMQpcw@mail.gmail.com
2024-04-06 17:47:10 +02:00
|
|
|
set random_page_cost = 2;
|
2017-02-15 19:53:24 +01:00
|
|
|
|
|
|
|
explain (costs off)
|
|
|
|
select count((unique1)) from tenk1 where hundred > 1;
|
|
|
|
select count((unique1)) from tenk1 where hundred > 1;
|
|
|
|
|
Enhance nbtree ScalarArrayOp execution.
Commit 9e8da0f7 taught nbtree to handle ScalarArrayOpExpr quals
natively. This works by pushing down the full context (the array keys)
to the nbtree index AM, enabling it to execute multiple primitive index
scans that the planner treats as one continuous index scan/index path.
This earlier enhancement enabled nbtree ScalarArrayOp index-only scans.
It also allowed scans with ScalarArrayOp quals to return ordered results
(with some notable restrictions, described further down).
Take this general approach a lot further: teach nbtree SAOP index scans
to decide how to execute ScalarArrayOp scans (when and where to start
the next primitive index scan) based on physical index characteristics.
This can be far more efficient. All SAOP scans will now reliably avoid
duplicative leaf page accesses (just like any other nbtree index scan).
SAOP scans whose array keys are naturally clustered together now require
far fewer index descents, since we'll reliably avoid starting a new
primitive scan just to get to a later offset from the same leaf page.
The scan's arrays now advance using binary searches for the array
element that best matches the next tuple's attribute value. Required
scan key arrays (i.e. arrays from scan keys that can terminate the scan)
ratchet forward in lockstep with the index scan. Non-required arrays
(i.e. arrays from scan keys that can only exclude non-matching tuples)
"advance" without the process ever rolling over to a higher-order array.
Naturally, only required SAOP scan keys trigger skipping over leaf pages
(non-required arrays cannot safely end or start primitive index scans).
Consequently, even index scans of a composite index with a high-order
inequality scan key (which we'll mark required) and a low-order SAOP
scan key (which we won't mark required) now avoid repeating leaf page
accesses -- that benefit isn't limited to simpler equality-only cases.
In general, all nbtree index scans now output tuples as if they were one
continuous index scan -- even scans that mix a high-order inequality
with lower-order SAOP equalities reliably output tuples in index order.
This allows us to remove a couple of special cases that were applied
when building index paths with SAOP clauses during planning.
Bugfix commit 807a40c5 taught the planner to avoid generating unsafe
path keys: path keys on a multicolumn index path, with a SAOP clause on
any attribute beyond the first/most significant attribute. These cases
are now all safe, so we go back to generating path keys without regard
for the presence of SAOP clauses (just like with any other clause type).
Affected queries can now exploit scan output order in all the usual ways
(e.g., certain "ORDER BY ... LIMIT n" queries can now terminate early).
Also undo changes from follow-up bugfix commit a4523c5a, which taught
the planner to produce alternative index paths, with path keys, but
without low-order SAOP index quals (filter quals were used instead).
We'll no longer generate these alternative paths, since they can no
longer offer any meaningful advantages over standard index qual paths.
Affected queries thereby avoid all of the disadvantages that come from
using filter quals within index scan nodes. They can avoid extra heap
page accesses from using filter quals to exclude non-matching tuples
(index quals will never have that problem). They can also skip over
irrelevant sections of the index in more cases (though only when nbtree
determines that starting another primitive scan actually makes sense).
There is a theoretical risk that removing restrictions on SAOP index
paths from the planner will break compatibility with amcanorder-based
index AMs maintained as extensions. Such an index AM could have the
same limitations around ordered SAOP scans as nbtree had up until now.
Adding a pro forma incompatibility item about the issue to the Postgres
17 release notes seems like a good idea.
Author: Peter Geoghegan <pg@bowt.ie>
Author: Matthias van de Meent <boekewurm+postgres@gmail.com>
Reviewed-By: Heikki Linnakangas <hlinnaka@iki.fi>
Reviewed-By: Matthias van de Meent <boekewurm+postgres@gmail.com>
Reviewed-By: Tomas Vondra <tomas.vondra@enterprisedb.com>
Discussion: https://postgr.es/m/CAH2-Wz=ksvN_sjcnD1+Bt-WtifRA5ok48aDYnq3pkKhxgMQpcw@mail.gmail.com
2024-04-06 17:47:10 +02:00
|
|
|
-- Parallel ScalarArrayOp index scan
|
|
|
|
explain (costs off)
|
|
|
|
select count((unique1)) from tenk1
|
|
|
|
where hundred = any ((select array_agg(i) from generate_series(1, 100, 15) i)::int[]);
|
|
|
|
select count((unique1)) from tenk1
|
|
|
|
where hundred = any ((select array_agg(i) from generate_series(1, 100, 15) i)::int[]);
|
|
|
|
|
2017-02-19 11:23:59 +01:00
|
|
|
-- test parallel index-only scans.
|
|
|
|
explain (costs off)
|
|
|
|
select count(*) from tenk1 where thousand > 95;
|
|
|
|
select count(*) from tenk1 where thousand > 95;
|
|
|
|
|
2017-08-31 19:15:54 +02:00
|
|
|
-- test rescan cases too
|
|
|
|
set enable_material = false;
|
|
|
|
|
|
|
|
explain (costs off)
|
|
|
|
select * from
|
|
|
|
(select count(unique1) from tenk1 where hundred > 10) ss
|
|
|
|
right join (values (1),(2),(3)) v(x) on true;
|
|
|
|
select * from
|
|
|
|
(select count(unique1) from tenk1 where hundred > 10) ss
|
|
|
|
right join (values (1),(2),(3)) v(x) on true;
|
|
|
|
|
|
|
|
explain (costs off)
|
|
|
|
select * from
|
|
|
|
(select count(*) from tenk1 where thousand > 99) ss
|
|
|
|
right join (values (1),(2),(3)) v(x) on true;
|
|
|
|
select * from
|
|
|
|
(select count(*) from tenk1 where thousand > 99) ss
|
|
|
|
right join (values (1),(2),(3)) v(x) on true;
|
|
|
|
|
2019-11-18 09:47:41 +01:00
|
|
|
-- test rescans for a Limit node with a parallel node beneath it.
|
2017-02-15 19:53:24 +01:00
|
|
|
reset enable_seqscan;
|
2019-11-18 09:47:41 +01:00
|
|
|
set enable_indexonlyscan to off;
|
|
|
|
set enable_indexscan to off;
|
|
|
|
alter table tenk1 set (parallel_workers = 0);
|
|
|
|
alter table tenk2 set (parallel_workers = 1);
|
|
|
|
explain (costs off)
|
|
|
|
select count(*) from tenk1
|
|
|
|
left join (select tenk2.unique1 from tenk2 order by 1 limit 1000) ss
|
|
|
|
on tenk1.unique1 < ss.unique1 + 1
|
|
|
|
where tenk1.unique1 < 2;
|
|
|
|
select count(*) from tenk1
|
|
|
|
left join (select tenk2.unique1 from tenk2 order by 1 limit 1000) ss
|
|
|
|
on tenk1.unique1 < ss.unique1 + 1
|
|
|
|
where tenk1.unique1 < 2;
|
|
|
|
--reset the value of workers for each table as it was before this test.
|
|
|
|
alter table tenk1 set (parallel_workers = 4);
|
|
|
|
alter table tenk2 reset (parallel_workers);
|
|
|
|
|
|
|
|
reset enable_material;
|
2017-02-15 19:53:24 +01:00
|
|
|
reset enable_bitmapscan;
|
2019-11-18 09:47:41 +01:00
|
|
|
reset enable_indexonlyscan;
|
|
|
|
reset enable_indexscan;
|
2017-02-15 19:53:24 +01:00
|
|
|
|
Support parallel bitmap heap scans.
The index is scanned by a single process, but then all cooperating
processes can iterate jointly over the resulting set of heap blocks.
In the future, we might also want to support using a parallel bitmap
index scan to set up for a parallel bitmap heap scan, but that's a
job for another day.
Dilip Kumar, with some corrections and cosmetic changes by me. The
larger patch set of which this is a part has been reviewed and tested
by (at least) Andres Freund, Amit Khandekar, Tushar Ahuja, Rafia
Sabih, Haribabu Kommi, Thomas Munro, and me.
Discussion: http://postgr.es/m/CAFiTN-uc4=0WxRGfCzs-xfkMYcSEWUC-Fon6thkJGjkh9i=13A@mail.gmail.com
2017-03-08 18:05:43 +01:00
|
|
|
-- test parallel bitmap heap scan.
|
|
|
|
set enable_seqscan to off;
|
|
|
|
set enable_indexscan to off;
|
2017-04-06 22:36:54 +02:00
|
|
|
set enable_hashjoin to off;
|
|
|
|
set enable_mergejoin to off;
|
|
|
|
set enable_material to off;
|
2017-04-06 23:21:39 +02:00
|
|
|
-- test prefetching, if the platform allows it
|
|
|
|
DO $$
|
|
|
|
BEGIN
|
|
|
|
SET effective_io_concurrency = 50;
|
|
|
|
EXCEPTION WHEN invalid_parameter_value THEN
|
|
|
|
END $$;
|
2017-04-06 22:36:54 +02:00
|
|
|
set work_mem='64kB'; --set small work mem to force lossy pages
|
Support parallel bitmap heap scans.
The index is scanned by a single process, but then all cooperating
processes can iterate jointly over the resulting set of heap blocks.
In the future, we might also want to support using a parallel bitmap
index scan to set up for a parallel bitmap heap scan, but that's a
job for another day.
Dilip Kumar, with some corrections and cosmetic changes by me. The
larger patch set of which this is a part has been reviewed and tested
by (at least) Andres Freund, Amit Khandekar, Tushar Ahuja, Rafia
Sabih, Haribabu Kommi, Thomas Munro, and me.
Discussion: http://postgr.es/m/CAFiTN-uc4=0WxRGfCzs-xfkMYcSEWUC-Fon6thkJGjkh9i=13A@mail.gmail.com
2017-03-08 18:05:43 +01:00
|
|
|
explain (costs off)
|
2017-04-06 22:36:54 +02:00
|
|
|
select count(*) from tenk1, tenk2 where tenk1.hundred > 1 and tenk2.thousand=0;
|
|
|
|
select count(*) from tenk1, tenk2 where tenk1.hundred > 1 and tenk2.thousand=0;
|
|
|
|
|
|
|
|
create table bmscantest (a int, t text);
|
|
|
|
insert into bmscantest select r, 'fooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooo' FROM generate_series(1,100000) r;
|
|
|
|
create index i_bmtest ON bmscantest(a);
|
|
|
|
select count(*) from bmscantest where a>1;
|
Support parallel bitmap heap scans.
The index is scanned by a single process, but then all cooperating
processes can iterate jointly over the resulting set of heap blocks.
In the future, we might also want to support using a parallel bitmap
index scan to set up for a parallel bitmap heap scan, but that's a
job for another day.
Dilip Kumar, with some corrections and cosmetic changes by me. The
larger patch set of which this is a part has been reviewed and tested
by (at least) Andres Freund, Amit Khandekar, Tushar Ahuja, Rafia
Sabih, Haribabu Kommi, Thomas Munro, and me.
Discussion: http://postgr.es/m/CAFiTN-uc4=0WxRGfCzs-xfkMYcSEWUC-Fon6thkJGjkh9i=13A@mail.gmail.com
2017-03-08 18:05:43 +01:00
|
|
|
|
2017-12-19 18:21:56 +01:00
|
|
|
-- test accumulation of stats for parallel nodes
|
Support parallel bitmap heap scans.
The index is scanned by a single process, but then all cooperating
processes can iterate jointly over the resulting set of heap blocks.
In the future, we might also want to support using a parallel bitmap
index scan to set up for a parallel bitmap heap scan, but that's a
job for another day.
Dilip Kumar, with some corrections and cosmetic changes by me. The
larger patch set of which this is a part has been reviewed and tested
by (at least) Andres Freund, Amit Khandekar, Tushar Ahuja, Rafia
Sabih, Haribabu Kommi, Thomas Munro, and me.
Discussion: http://postgr.es/m/CAFiTN-uc4=0WxRGfCzs-xfkMYcSEWUC-Fon6thkJGjkh9i=13A@mail.gmail.com
2017-03-08 18:05:43 +01:00
|
|
|
reset enable_seqscan;
|
2017-12-19 18:21:56 +01:00
|
|
|
alter table tenk2 set (parallel_workers = 0);
|
|
|
|
explain (analyze, timing off, summary off, costs off)
|
|
|
|
select count(*) from tenk1, tenk2 where tenk1.hundred > 1
|
|
|
|
and tenk2.thousand=0;
|
|
|
|
alter table tenk2 reset (parallel_workers);
|
|
|
|
|
|
|
|
reset work_mem;
|
|
|
|
create function explain_parallel_sort_stats() returns setof text
|
|
|
|
language plpgsql as
|
|
|
|
$$
|
|
|
|
declare ln text;
|
|
|
|
begin
|
|
|
|
for ln in
|
|
|
|
explain (analyze, timing off, summary off, costs off)
|
|
|
|
select * from
|
|
|
|
(select ten from tenk1 where ten < 100 order by ten) ss
|
|
|
|
right join (values (1),(2),(3)) v(x) on true
|
|
|
|
loop
|
|
|
|
ln := regexp_replace(ln, 'Memory: \S*', 'Memory: xxx');
|
|
|
|
return next ln;
|
|
|
|
end loop;
|
|
|
|
end;
|
|
|
|
$$;
|
|
|
|
select * from explain_parallel_sort_stats();
|
|
|
|
|
Support parallel bitmap heap scans.
The index is scanned by a single process, but then all cooperating
processes can iterate jointly over the resulting set of heap blocks.
In the future, we might also want to support using a parallel bitmap
index scan to set up for a parallel bitmap heap scan, but that's a
job for another day.
Dilip Kumar, with some corrections and cosmetic changes by me. The
larger patch set of which this is a part has been reviewed and tested
by (at least) Andres Freund, Amit Khandekar, Tushar Ahuja, Rafia
Sabih, Haribabu Kommi, Thomas Munro, and me.
Discussion: http://postgr.es/m/CAFiTN-uc4=0WxRGfCzs-xfkMYcSEWUC-Fon6thkJGjkh9i=13A@mail.gmail.com
2017-03-08 18:05:43 +01:00
|
|
|
reset enable_indexscan;
|
2017-04-06 22:36:54 +02:00
|
|
|
reset enable_hashjoin;
|
|
|
|
reset enable_mergejoin;
|
|
|
|
reset enable_material;
|
|
|
|
reset effective_io_concurrency;
|
|
|
|
drop table bmscantest;
|
2017-12-19 18:21:56 +01:00
|
|
|
drop function explain_parallel_sort_stats();
|
Support parallel bitmap heap scans.
The index is scanned by a single process, but then all cooperating
processes can iterate jointly over the resulting set of heap blocks.
In the future, we might also want to support using a parallel bitmap
index scan to set up for a parallel bitmap heap scan, but that's a
job for another day.
Dilip Kumar, with some corrections and cosmetic changes by me. The
larger patch set of which this is a part has been reviewed and tested
by (at least) Andres Freund, Amit Khandekar, Tushar Ahuja, Rafia
Sabih, Haribabu Kommi, Thomas Munro, and me.
Discussion: http://postgr.es/m/CAFiTN-uc4=0WxRGfCzs-xfkMYcSEWUC-Fon6thkJGjkh9i=13A@mail.gmail.com
2017-03-08 18:05:43 +01:00
|
|
|
|
2017-03-07 17:49:49 +01:00
|
|
|
-- test parallel merge join path.
|
|
|
|
set enable_hashjoin to off;
|
|
|
|
set enable_nestloop to off;
|
|
|
|
|
|
|
|
explain (costs off)
|
|
|
|
select count(*) from tenk1, tenk2 where tenk1.unique1 = tenk2.unique1;
|
|
|
|
select count(*) from tenk1, tenk2 where tenk1.unique1 = tenk2.unique1;
|
|
|
|
|
|
|
|
reset enable_hashjoin;
|
|
|
|
reset enable_nestloop;
|
2017-03-09 13:40:36 +01:00
|
|
|
|
2017-08-15 00:21:26 +02:00
|
|
|
-- test gather merge
|
|
|
|
set enable_hashagg = false;
|
2017-03-09 13:40:36 +01:00
|
|
|
|
|
|
|
explain (costs off)
|
2017-08-15 00:21:26 +02:00
|
|
|
select count(*) from tenk1 group by twenty;
|
2017-03-09 13:40:36 +01:00
|
|
|
|
2017-08-15 00:21:26 +02:00
|
|
|
select count(*) from tenk1 group by twenty;
|
2017-03-09 13:40:36 +01:00
|
|
|
|
2017-11-13 22:33:56 +01:00
|
|
|
--test expressions in targetlist are pushed down for gather merge
|
Clean up duplicate table and function names in regression tests.
Many of the objects we create during the regression tests are put in the
public schema, so that using the same names in different regression tests
creates a hazard of test failures if any two such scripts run concurrently.
This patch cleans up a bunch of latent hazards of that sort, as well as two
live hazards.
The current situation in this regard is far worse than it was a year or two
back, because practically all of the partitioning-related test cases have
reused table names with enthusiasm. I despaired of cleaning up that mess
within the five most-affected tests (create_table, alter_table, insert,
update, inherit); fortunately those don't run concurrently.
Other than partitioning problems, most of the issues boil down to using
names like "foo", "bar", "tmp", etc, without thought for the fact that
other test scripts might use similar names concurrently. I've made an
effort to make all such names more specific.
One of the live hazards was that commit 7421f4b8 caused with.sql to
create a table named "test", conflicting with a similarly-named table
in alter_table.sql; this was exposed in the buildfarm recently.
The other one was that join.sql and transactions.sql both create tables
named "foo" and "bar"; but join.sql's uses of those names date back
only to December or so.
Since commit 7421f4b8 was back-patched to v10, back-patch a minimal
fix for that problem. The rest of this is just future-proofing.
Discussion: https://postgr.es/m/4627.1521070268@sss.pgh.pa.us
2018-03-15 22:08:51 +01:00
|
|
|
create function sp_simple_func(var1 integer) returns integer
|
2017-11-13 22:33:56 +01:00
|
|
|
as $$
|
|
|
|
begin
|
|
|
|
return var1 + 10;
|
|
|
|
end;
|
|
|
|
$$ language plpgsql PARALLEL SAFE;
|
|
|
|
|
|
|
|
explain (costs off, verbose)
|
Clean up duplicate table and function names in regression tests.
Many of the objects we create during the regression tests are put in the
public schema, so that using the same names in different regression tests
creates a hazard of test failures if any two such scripts run concurrently.
This patch cleans up a bunch of latent hazards of that sort, as well as two
live hazards.
The current situation in this regard is far worse than it was a year or two
back, because practically all of the partitioning-related test cases have
reused table names with enthusiasm. I despaired of cleaning up that mess
within the five most-affected tests (create_table, alter_table, insert,
update, inherit); fortunately those don't run concurrently.
Other than partitioning problems, most of the issues boil down to using
names like "foo", "bar", "tmp", etc, without thought for the fact that
other test scripts might use similar names concurrently. I've made an
effort to make all such names more specific.
One of the live hazards was that commit 7421f4b8 caused with.sql to
create a table named "test", conflicting with a similarly-named table
in alter_table.sql; this was exposed in the buildfarm recently.
The other one was that join.sql and transactions.sql both create tables
named "foo" and "bar"; but join.sql's uses of those names date back
only to December or so.
Since commit 7421f4b8 was back-patched to v10, back-patch a minimal
fix for that problem. The rest of this is just future-proofing.
Discussion: https://postgr.es/m/4627.1521070268@sss.pgh.pa.us
2018-03-15 22:08:51 +01:00
|
|
|
select ten, sp_simple_func(ten) from tenk1 where ten < 100 order by ten;
|
2017-11-13 22:33:56 +01:00
|
|
|
|
Clean up duplicate table and function names in regression tests.
Many of the objects we create during the regression tests are put in the
public schema, so that using the same names in different regression tests
creates a hazard of test failures if any two such scripts run concurrently.
This patch cleans up a bunch of latent hazards of that sort, as well as two
live hazards.
The current situation in this regard is far worse than it was a year or two
back, because practically all of the partitioning-related test cases have
reused table names with enthusiasm. I despaired of cleaning up that mess
within the five most-affected tests (create_table, alter_table, insert,
update, inherit); fortunately those don't run concurrently.
Other than partitioning problems, most of the issues boil down to using
names like "foo", "bar", "tmp", etc, without thought for the fact that
other test scripts might use similar names concurrently. I've made an
effort to make all such names more specific.
One of the live hazards was that commit 7421f4b8 caused with.sql to
create a table named "test", conflicting with a similarly-named table
in alter_table.sql; this was exposed in the buildfarm recently.
The other one was that join.sql and transactions.sql both create tables
named "foo" and "bar"; but join.sql's uses of those names date back
only to December or so.
Since commit 7421f4b8 was back-patched to v10, back-patch a minimal
fix for that problem. The rest of this is just future-proofing.
Discussion: https://postgr.es/m/4627.1521070268@sss.pgh.pa.us
2018-03-15 22:08:51 +01:00
|
|
|
drop function sp_simple_func(integer);
|
2017-11-13 22:33:56 +01:00
|
|
|
|
2018-06-21 16:58:42 +02:00
|
|
|
-- test handling of SRFs in targetlist (bug in 10.0)
|
|
|
|
|
|
|
|
explain (costs off)
|
|
|
|
select count(*), generate_series(1,2) from tenk1 group by twenty;
|
|
|
|
|
|
|
|
select count(*), generate_series(1,2) from tenk1 group by twenty;
|
|
|
|
|
2017-11-15 14:17:29 +01:00
|
|
|
-- test gather merge with parallel leader participation disabled
|
|
|
|
set parallel_leader_participation = off;
|
|
|
|
|
|
|
|
explain (costs off)
|
|
|
|
select count(*) from tenk1 group by twenty;
|
|
|
|
|
|
|
|
select count(*) from tenk1 group by twenty;
|
|
|
|
|
|
|
|
reset parallel_leader_participation;
|
|
|
|
|
2017-08-30 15:59:23 +02:00
|
|
|
--test rescan behavior of gather merge
|
|
|
|
set enable_material = false;
|
|
|
|
|
|
|
|
explain (costs off)
|
|
|
|
select * from
|
|
|
|
(select string4, count(unique2)
|
|
|
|
from tenk1 group by string4 order by string4) ss
|
|
|
|
right join (values (1),(2),(3)) v(x) on true;
|
|
|
|
|
|
|
|
select * from
|
|
|
|
(select string4, count(unique2)
|
|
|
|
from tenk1 group by string4 order by string4) ss
|
|
|
|
right join (values (1),(2),(3)) v(x) on true;
|
|
|
|
|
|
|
|
reset enable_material;
|
|
|
|
|
2017-08-29 19:12:23 +02:00
|
|
|
reset enable_hashagg;
|
|
|
|
|
2017-11-14 21:03:55 +01:00
|
|
|
-- check parallelized int8 aggregate (bug #14897)
|
|
|
|
explain (costs off)
|
|
|
|
select avg(unique1::int8) from tenk1;
|
|
|
|
|
|
|
|
select avg(unique1::int8) from tenk1;
|
|
|
|
|
2017-08-29 19:12:23 +02:00
|
|
|
-- gather merge test with a LIMIT
|
|
|
|
explain (costs off)
|
|
|
|
select fivethous from tenk1 order by fivethous limit 4;
|
|
|
|
|
|
|
|
select fivethous from tenk1 order by fivethous limit 4;
|
|
|
|
|
2017-08-15 00:21:26 +02:00
|
|
|
-- gather merge test with 0 worker
|
|
|
|
set max_parallel_workers = 0;
|
|
|
|
explain (costs off)
|
|
|
|
select string4 from tenk1 order by string4 limit 5;
|
|
|
|
select string4 from tenk1 order by string4 limit 5;
|
2017-11-15 14:17:29 +01:00
|
|
|
|
|
|
|
-- gather merge test with 0 workers, with parallel leader
|
|
|
|
-- participation disabled (the leader will have to run the plan
|
|
|
|
-- despite the setting)
|
|
|
|
set parallel_leader_participation = off;
|
|
|
|
explain (costs off)
|
|
|
|
select string4 from tenk1 order by string4 limit 5;
|
|
|
|
select string4 from tenk1 order by string4 limit 5;
|
|
|
|
|
|
|
|
reset parallel_leader_participation;
|
2017-08-15 00:21:26 +02:00
|
|
|
reset max_parallel_workers;
|
2017-03-09 13:40:36 +01:00
|
|
|
|
2024-01-30 22:10:59 +01:00
|
|
|
create function parallel_safe_volatile(a int) returns int as
|
|
|
|
$$ begin return a; end; $$ parallel safe volatile language plpgsql;
|
|
|
|
|
|
|
|
-- Test gather merge atop of a sort of a partial path
|
|
|
|
explain (costs off)
|
|
|
|
select * from tenk1 where four = 2
|
|
|
|
order by four, hundred, parallel_safe_volatile(thousand);
|
|
|
|
|
|
|
|
-- Test gather merge atop of an incremental sort a of partial path
|
|
|
|
set min_parallel_index_scan_size = 0;
|
|
|
|
set enable_seqscan = off;
|
|
|
|
|
|
|
|
explain (costs off)
|
|
|
|
select * from tenk1 where four = 2
|
|
|
|
order by four, hundred, parallel_safe_volatile(thousand);
|
|
|
|
|
|
|
|
reset min_parallel_index_scan_size;
|
|
|
|
reset enable_seqscan;
|
|
|
|
|
|
|
|
-- Test GROUP BY with a gather merge path atop of a sort of a partial path
|
|
|
|
explain (costs off)
|
|
|
|
select count(*) from tenk1
|
|
|
|
group by twenty, parallel_safe_volatile(two);
|
|
|
|
|
|
|
|
drop function parallel_safe_volatile(int);
|
|
|
|
|
2017-08-25 02:42:49 +02:00
|
|
|
SAVEPOINT settings;
|
2023-02-15 09:21:59 +01:00
|
|
|
SET LOCAL debug_parallel_query = 1;
|
2016-06-08 05:36:22 +02:00
|
|
|
explain (costs off)
|
|
|
|
select stringu1::int2 from tenk1 where unique1 = 1;
|
2017-08-25 02:42:49 +02:00
|
|
|
ROLLBACK TO SAVEPOINT settings;
|
|
|
|
|
|
|
|
-- exercise record typmod remapping between backends
|
Clean up duplicate table and function names in regression tests.
Many of the objects we create during the regression tests are put in the
public schema, so that using the same names in different regression tests
creates a hazard of test failures if any two such scripts run concurrently.
This patch cleans up a bunch of latent hazards of that sort, as well as two
live hazards.
The current situation in this regard is far worse than it was a year or two
back, because practically all of the partitioning-related test cases have
reused table names with enthusiasm. I despaired of cleaning up that mess
within the five most-affected tests (create_table, alter_table, insert,
update, inherit); fortunately those don't run concurrently.
Other than partitioning problems, most of the issues boil down to using
names like "foo", "bar", "tmp", etc, without thought for the fact that
other test scripts might use similar names concurrently. I've made an
effort to make all such names more specific.
One of the live hazards was that commit 7421f4b8 caused with.sql to
create a table named "test", conflicting with a similarly-named table
in alter_table.sql; this was exposed in the buildfarm recently.
The other one was that join.sql and transactions.sql both create tables
named "foo" and "bar"; but join.sql's uses of those names date back
only to December or so.
Since commit 7421f4b8 was back-patched to v10, back-patch a minimal
fix for that problem. The rest of this is just future-proofing.
Discussion: https://postgr.es/m/4627.1521070268@sss.pgh.pa.us
2018-03-15 22:08:51 +01:00
|
|
|
CREATE FUNCTION make_record(n int)
|
2017-08-25 02:42:49 +02:00
|
|
|
RETURNS RECORD LANGUAGE plpgsql PARALLEL SAFE AS
|
|
|
|
$$
|
|
|
|
BEGIN
|
|
|
|
RETURN CASE n
|
|
|
|
WHEN 1 THEN ROW(1)
|
|
|
|
WHEN 2 THEN ROW(1, 2)
|
|
|
|
WHEN 3 THEN ROW(1, 2, 3)
|
|
|
|
WHEN 4 THEN ROW(1, 2, 3, 4)
|
|
|
|
ELSE ROW(1, 2, 3, 4, 5)
|
|
|
|
END;
|
|
|
|
END;
|
|
|
|
$$;
|
|
|
|
SAVEPOINT settings;
|
2023-02-15 09:21:59 +01:00
|
|
|
SET LOCAL debug_parallel_query = 1;
|
2017-08-25 02:42:49 +02:00
|
|
|
SELECT make_record(x) FROM (SELECT generate_series(1, 5) x) ss ORDER BY x;
|
|
|
|
ROLLBACK TO SAVEPOINT settings;
|
|
|
|
DROP function make_record(n int);
|
2016-06-08 05:36:22 +02:00
|
|
|
|
2017-10-29 08:28:40 +01:00
|
|
|
-- test the sanity of parallel query after the active role is dropped.
|
|
|
|
drop role if exists regress_parallel_worker;
|
|
|
|
create role regress_parallel_worker;
|
|
|
|
set role regress_parallel_worker;
|
|
|
|
reset session authorization;
|
|
|
|
drop role regress_parallel_worker;
|
2023-02-15 09:21:59 +01:00
|
|
|
set debug_parallel_query = 1;
|
2017-10-29 08:28:40 +01:00
|
|
|
select count(*) from tenk1;
|
2023-02-15 09:21:59 +01:00
|
|
|
reset debug_parallel_query;
|
2017-10-29 08:28:40 +01:00
|
|
|
reset role;
|
|
|
|
|
2018-09-04 06:36:09 +02:00
|
|
|
-- Window function calculation can't be pushed to workers.
|
|
|
|
explain (costs off, verbose)
|
|
|
|
select count(*) from tenk1 a where (unique1, two) in
|
|
|
|
(select unique1, row_number() over() from tenk1 b);
|
|
|
|
|
|
|
|
|
2018-09-14 06:06:30 +02:00
|
|
|
-- LIMIT/OFFSET within sub-selects can't be pushed to workers.
|
|
|
|
explain (costs off)
|
|
|
|
select * from tenk1 a where two in
|
|
|
|
(select two from tenk1 b where stringu1 like '%AAAA' limit 3);
|
|
|
|
|
2017-04-06 22:44:48 +02:00
|
|
|
-- to increase the parallel query test coverage
|
2017-08-25 02:42:49 +02:00
|
|
|
SAVEPOINT settings;
|
2023-02-15 09:21:59 +01:00
|
|
|
SET LOCAL debug_parallel_query = 1;
|
2017-04-06 22:44:48 +02:00
|
|
|
EXPLAIN (analyze, timing off, summary off, costs off) SELECT * FROM tenk1;
|
2017-08-25 02:42:49 +02:00
|
|
|
ROLLBACK TO SAVEPOINT settings;
|
2017-04-06 22:44:48 +02:00
|
|
|
|
2016-08-22 18:00:00 +02:00
|
|
|
-- provoke error in worker
|
2020-09-03 22:52:09 +02:00
|
|
|
-- (make the error message long enough to require multiple bufferloads)
|
2017-08-25 02:42:49 +02:00
|
|
|
SAVEPOINT settings;
|
2023-02-15 09:21:59 +01:00
|
|
|
SET LOCAL debug_parallel_query = 1;
|
2020-09-03 22:52:09 +02:00
|
|
|
select (stringu1 || repeat('abcd', 5000))::int2 from tenk1 where unique1 = 1;
|
2017-08-25 02:42:49 +02:00
|
|
|
ROLLBACK TO SAVEPOINT settings;
|
2016-06-08 05:36:22 +02:00
|
|
|
|
Let Parallel Append over simple UNION ALL have partial subpaths.
A simple UNION ALL gets flattened into an appendrel of subquery
RTEs, but up until now it's been impossible for the appendrel to use
the partial paths for the subqueries, so we can implement the
appendrel as a Parallel Append but only one with non-partial paths
as children.
There are three separate obstacles to removing that limitation.
First, when planning a subquery, propagate any partial paths to the
final_rel so that they are potentially visible to outer query levels
(but not if they have initPlans attached, because that wouldn't be
safe). Second, after planning a subquery, propagate any partial paths
for the final_rel to the subquery RTE in the outer query level in the
same way we do for non-partial paths. Third, teach finalize_plan() to
account for the possibility that the fake parameter we use for rescan
signalling when the plan contains a Gather (Merge) node may be
propagated from an outer query level.
Patch by me, reviewed and tested by Amit Khandekar, Rajkumar
Raghuwanshi, and Ashutosh Bapat. Test cases based on examples by
Rajkumar Raghuwanshi.
Discussion: http://postgr.es/m/CA+Tgmoa6L9A1nNCk3aTDVZLZ4KkHDn1+tm7mFyFvP+uQPS7bAg@mail.gmail.com
2018-03-13 21:34:08 +01:00
|
|
|
-- test interaction with set-returning functions
|
|
|
|
SAVEPOINT settings;
|
|
|
|
|
|
|
|
-- multiple subqueries under a single Gather node
|
|
|
|
-- must set parallel_setup_cost > 0 to discourage multiple Gather nodes
|
|
|
|
SET LOCAL parallel_setup_cost = 10;
|
|
|
|
EXPLAIN (COSTS OFF)
|
|
|
|
SELECT unique1 FROM tenk1 WHERE fivethous = tenthous + 1
|
|
|
|
UNION ALL
|
|
|
|
SELECT unique1 FROM tenk1 WHERE fivethous = tenthous + 1;
|
|
|
|
ROLLBACK TO SAVEPOINT settings;
|
|
|
|
|
|
|
|
-- can't use multiple subqueries under a single Gather node due to initPlans
|
|
|
|
EXPLAIN (COSTS OFF)
|
|
|
|
SELECT unique1 FROM tenk1 WHERE fivethous =
|
|
|
|
(SELECT unique1 FROM tenk1 WHERE fivethous = 1 LIMIT 1)
|
|
|
|
UNION ALL
|
|
|
|
SELECT unique1 FROM tenk1 WHERE fivethous =
|
|
|
|
(SELECT unique2 FROM tenk1 WHERE fivethous = 1 LIMIT 1)
|
|
|
|
ORDER BY 1;
|
|
|
|
|
|
|
|
-- test interaction with SRFs
|
|
|
|
SELECT * FROM information_schema.foreign_data_wrapper_options
|
|
|
|
ORDER BY 1, 2, 3;
|
|
|
|
|
Fix mis-planning of repeated application of a projection.
create_projection_plan contains a hidden assumption (here made
explicit by an Assert) that a projection-capable Path will yield a
projection-capable Plan. Unfortunately, that assumption is violated
only a few lines away, by create_projection_plan itself. This means
that two stacked ProjectionPaths can yield an outcome where we try to
jam the upper path's tlist into a non-projection-capable child node,
resulting in an invalid plan.
There isn't any good reason to have stacked ProjectionPaths; indeed the
whole concept is faulty, since the set of Vars/Aggs/etc needed by the
upper one wouldn't necessarily be available in the output of the lower
one, nor could the lower one create such values if they weren't
available from its input. Hence, we can fix this by adjusting
create_projection_path to strip any top-level ProjectionPath from the
subpath it's given. (This amounts to saying "oh, we changed our
minds about what we need to project here".)
The test case added here only fails in v13 and HEAD; before that, we
don't attempt to shove the Sort into the parallel part of the plan,
for reasons that aren't entirely clear to me. However, all the
directly-related code looks generally the same as far back as v11,
where the hazard was introduced (by d7c19e62a). So I've got no faith
that the same type of bug doesn't exist in v11 and v12, given the
right test case. Hence, back-patch the code changes, but not the
irrelevant test case, into those branches.
Per report from Bas Poot.
Discussion: https://postgr.es/m/534fca83789c4a378c7de379e9067d4f@politie.nl
2021-05-31 18:03:00 +02:00
|
|
|
EXPLAIN (VERBOSE, COSTS OFF)
|
|
|
|
SELECT generate_series(1, two), array(select generate_series(1, two))
|
|
|
|
FROM tenk1 ORDER BY tenthous;
|
|
|
|
|
2022-08-04 17:11:22 +02:00
|
|
|
-- must disallow pushing sort below gather when pathkey contains an SRF
|
|
|
|
EXPLAIN (VERBOSE, COSTS OFF)
|
|
|
|
SELECT unnest(ARRAY[]::integer[]) + 1 AS pathkey
|
|
|
|
FROM tenk1 t1 JOIN tenk1 t2 ON TRUE
|
|
|
|
ORDER BY pathkey;
|
|
|
|
|
2018-10-02 07:31:33 +02:00
|
|
|
-- test passing expanded-value representations to workers
|
|
|
|
CREATE FUNCTION make_some_array(int,int) returns int[] as
|
|
|
|
$$declare x int[];
|
|
|
|
begin
|
|
|
|
x[1] := $1;
|
|
|
|
x[2] := $2;
|
|
|
|
return x;
|
|
|
|
end$$ language plpgsql parallel safe;
|
|
|
|
CREATE TABLE fooarr(f1 text, f2 int[], f3 text);
|
|
|
|
INSERT INTO fooarr VALUES('1', ARRAY[1,2], 'one');
|
|
|
|
|
|
|
|
PREPARE pstmt(text, int[]) AS SELECT * FROM fooarr WHERE f1 = $1 AND f2 = $2;
|
|
|
|
EXPLAIN (COSTS OFF) EXECUTE pstmt('1', make_some_array(1,2));
|
|
|
|
EXECUTE pstmt('1', make_some_array(1,2));
|
|
|
|
DEALLOCATE pstmt;
|
|
|
|
|
|
|
|
-- test interaction between subquery and partial_paths
|
2018-04-25 21:14:14 +02:00
|
|
|
CREATE VIEW tenk1_vw_sec WITH (security_barrier) AS SELECT * FROM tenk1;
|
|
|
|
EXPLAIN (COSTS OFF)
|
2018-10-14 20:02:59 +02:00
|
|
|
SELECT 1 FROM tenk1_vw_sec
|
|
|
|
WHERE (SELECT sum(f1) FROM int4_tbl WHERE f1 < unique1) < 100;
|
2018-04-25 21:14:14 +02:00
|
|
|
|
2016-06-08 05:36:22 +02:00
|
|
|
rollback;
|