1
0
mirror of https://github.com/postgres/postgres.git synced 2025-07-08 11:42:09 +03:00

Consider explicit incremental sort for Append and MergeAppend

For an ordered Append or MergeAppend, we need to inject an explicit
sort into any subpath that is not already well enough ordered.
Currently, only explicit full sorts are considered; incremental sorts
are not yet taken into account.

In this patch, for subpaths of an ordered Append or MergeAppend, we
choose to use explicit incremental sort if it is enabled and there are
presorted keys.

The rationale is based on the assumption that incremental sort is
always faster than full sort when there are presorted keys, a premise
that has been applied in various parts of the code.  In addition, the
current cost model tends to favor incremental sort as being cheaper
than full sort in the presence of presorted keys, making it reasonable
not to consider full sort in such cases.

No backpatch as this could result in plan changes.

Author: Richard Guo <guofenglinux@gmail.com>
Reviewed-by: Andrei Lepikhov <lepihov@gmail.com>
Reviewed-by: Robert Haas <robertmhaas@gmail.com>
Discussion: https://postgr.es/m/CAMbWs4_V7a2enTR+T3pOY_YZ-FU8ZsFYym2swOz4jNMqmSgyuw@mail.gmail.com
This commit is contained in:
Richard Guo
2025-07-08 10:21:44 +09:00
parent 7376e60854
commit 55a780e947
7 changed files with 217 additions and 52 deletions

View File

@ -2247,7 +2247,7 @@ append_nonpartial_cost(List *subpaths, int numpaths, int parallel_workers)
* Determines and returns the cost of an Append node.
*/
void
cost_append(AppendPath *apath)
cost_append(AppendPath *apath, PlannerInfo *root)
{
ListCell *l;
@ -2309,26 +2309,52 @@ cost_append(AppendPath *apath)
foreach(l, apath->subpaths)
{
Path *subpath = (Path *) lfirst(l);
Path sort_path; /* dummy for result of cost_sort */
int presorted_keys;
Path sort_path; /* dummy for result of
* cost_sort/cost_incremental_sort */
if (!pathkeys_contained_in(pathkeys, subpath->pathkeys))
if (!pathkeys_count_contained_in(pathkeys, subpath->pathkeys,
&presorted_keys))
{
/*
* We'll need to insert a Sort node, so include costs for
* that. We can use the parent's LIMIT if any, since we
* that. We choose to use incremental sort if it is
* enabled and there are presorted keys; otherwise we use
* full sort.
*
* We can use the parent's LIMIT if any, since we
* certainly won't pull more than that many tuples from
* any child.
*/
cost_sort(&sort_path,
NULL, /* doesn't currently need root */
pathkeys,
subpath->disabled_nodes,
subpath->total_cost,
subpath->rows,
subpath->pathtarget->width,
0.0,
work_mem,
apath->limit_tuples);
if (enable_incremental_sort && presorted_keys > 0)
{
cost_incremental_sort(&sort_path,
root,
pathkeys,
presorted_keys,
subpath->disabled_nodes,
subpath->startup_cost,
subpath->total_cost,
subpath->rows,
subpath->pathtarget->width,
0.0,
work_mem,
apath->limit_tuples);
}
else
{
cost_sort(&sort_path,
root,
pathkeys,
subpath->disabled_nodes,
subpath->total_cost,
subpath->rows,
subpath->pathtarget->width,
0.0,
work_mem,
apath->limit_tuples);
}
subpath = &sort_path;
}

View File

@ -1318,6 +1318,7 @@ create_append_plan(PlannerInfo *root, AppendPath *best_path, int flags)
Oid *sortOperators;
Oid *collations;
bool *nullsFirst;
int presorted_keys;
/*
* Compute sort column info, and adjust subplan's tlist as needed.
@ -1353,14 +1354,38 @@ create_append_plan(PlannerInfo *root, AppendPath *best_path, int flags)
numsortkeys * sizeof(bool)) == 0);
/* Now, insert a Sort node if subplan isn't sufficiently ordered */
if (!pathkeys_contained_in(pathkeys, subpath->pathkeys))
if (!pathkeys_count_contained_in(pathkeys, subpath->pathkeys,
&presorted_keys))
{
Sort *sort = make_sort(subplan, numsortkeys,
Plan *sort_plan;
/*
* We choose to use incremental sort if it is enabled and
* there are presorted keys; otherwise we use full sort.
*/
if (enable_incremental_sort && presorted_keys > 0)
{
sort_plan = (Plan *)
make_incrementalsort(subplan, numsortkeys, presorted_keys,
sortColIdx, sortOperators,
collations, nullsFirst);
label_sort_with_costsize(root, sort, best_path->limit_tuples);
subplan = (Plan *) sort;
label_incrementalsort_with_costsize(root,
(IncrementalSort *) sort_plan,
pathkeys,
best_path->limit_tuples);
}
else
{
sort_plan = (Plan *) make_sort(subplan, numsortkeys,
sortColIdx, sortOperators,
collations, nullsFirst);
label_sort_with_costsize(root, (Sort *) sort_plan,
best_path->limit_tuples);
}
subplan = sort_plan;
}
}
@ -1491,6 +1516,7 @@ create_merge_append_plan(PlannerInfo *root, MergeAppendPath *best_path,
Oid *sortOperators;
Oid *collations;
bool *nullsFirst;
int presorted_keys;
/* Build the child plan */
/* Must insist that all children return the same tlist */
@ -1525,14 +1551,38 @@ create_merge_append_plan(PlannerInfo *root, MergeAppendPath *best_path,
numsortkeys * sizeof(bool)) == 0);
/* Now, insert a Sort node if subplan isn't sufficiently ordered */
if (!pathkeys_contained_in(pathkeys, subpath->pathkeys))
if (!pathkeys_count_contained_in(pathkeys, subpath->pathkeys,
&presorted_keys))
{
Sort *sort = make_sort(subplan, numsortkeys,
Plan *sort_plan;
/*
* We choose to use incremental sort if it is enabled and there
* are presorted keys; otherwise we use full sort.
*/
if (enable_incremental_sort && presorted_keys > 0)
{
sort_plan = (Plan *)
make_incrementalsort(subplan, numsortkeys, presorted_keys,
sortColIdx, sortOperators,
collations, nullsFirst);
label_sort_with_costsize(root, sort, best_path->limit_tuples);
subplan = (Plan *) sort;
label_incrementalsort_with_costsize(root,
(IncrementalSort *) sort_plan,
pathkeys,
best_path->limit_tuples);
}
else
{
sort_plan = (Plan *) make_sort(subplan, numsortkeys,
sortColIdx, sortOperators,
collations, nullsFirst);
label_sort_with_costsize(root, (Sort *) sort_plan,
best_path->limit_tuples);
}
subplan = sort_plan;
}
subplans = lappend(subplans, subplan);

View File

@ -1404,12 +1404,12 @@ create_append_path(PlannerInfo *root,
pathnode->path.total_cost = child->total_cost;
}
else
cost_append(pathnode);
cost_append(pathnode, root);
/* Must do this last, else cost_append complains */
pathnode->path.pathkeys = child->pathkeys;
}
else
cost_append(pathnode);
cost_append(pathnode, root);
/* If the caller provided a row estimate, override the computed value. */
if (rows >= 0)
@ -1515,6 +1515,9 @@ create_merge_append_path(PlannerInfo *root,
foreach(l, subpaths)
{
Path *subpath = (Path *) lfirst(l);
int presorted_keys;
Path sort_path; /* dummy for result of
* cost_sort/cost_incremental_sort */
/* All child paths should be unparameterized */
Assert(bms_is_empty(PATH_REQ_OUTER(subpath)));
@ -1523,32 +1526,52 @@ create_merge_append_path(PlannerInfo *root,
pathnode->path.parallel_safe = pathnode->path.parallel_safe &&
subpath->parallel_safe;
if (pathkeys_contained_in(pathkeys, subpath->pathkeys))
if (!pathkeys_count_contained_in(pathkeys, subpath->pathkeys,
&presorted_keys))
{
/* Subpath is adequately ordered, we won't need to sort it */
input_disabled_nodes += subpath->disabled_nodes;
input_startup_cost += subpath->startup_cost;
input_total_cost += subpath->total_cost;
}
else
{
/* We'll need to insert a Sort node, so include cost for that */
Path sort_path; /* dummy for result of cost_sort */
/*
* We'll need to insert a Sort node, so include costs for that. We
* choose to use incremental sort if it is enabled and there are
* presorted keys; otherwise we use full sort.
*
* We can use the parent's LIMIT if any, since we certainly won't
* pull more than that many tuples from any child.
*/
if (enable_incremental_sort && presorted_keys > 0)
{
cost_incremental_sort(&sort_path,
root,
pathkeys,
presorted_keys,
subpath->disabled_nodes,
subpath->startup_cost,
subpath->total_cost,
subpath->rows,
subpath->pathtarget->width,
0.0,
work_mem,
pathnode->limit_tuples);
}
else
{
cost_sort(&sort_path,
root,
pathkeys,
subpath->disabled_nodes,
subpath->total_cost,
subpath->rows,
subpath->pathtarget->width,
0.0,
work_mem,
pathnode->limit_tuples);
}
cost_sort(&sort_path,
root,
pathkeys,
subpath->disabled_nodes,
subpath->total_cost,
subpath->rows,
subpath->pathtarget->width,
0.0,
work_mem,
pathnode->limit_tuples);
input_disabled_nodes += sort_path.disabled_nodes;
input_startup_cost += sort_path.startup_cost;
input_total_cost += sort_path.total_cost;
subpath = &sort_path;
}
input_disabled_nodes += subpath->disabled_nodes;
input_startup_cost += subpath->startup_cost;
input_total_cost += subpath->total_cost;
}
/*

View File

@ -118,7 +118,7 @@ extern void cost_incremental_sort(Path *path,
Cost input_startup_cost, Cost input_total_cost,
double input_tuples, int width, Cost comparison_cost, int sort_mem,
double limit_tuples);
extern void cost_append(AppendPath *apath);
extern void cost_append(AppendPath *apath, PlannerInfo *root);
extern void cost_merge_append(Path *path, PlannerInfo *root,
List *pathkeys, int n_streams,
int input_disabled_nodes,

View File

@ -1722,3 +1722,43 @@ order by t1.four, t1.two limit 1;
-> Seq Scan on tenk1 t2
(12 rows)
--
-- Test incremental sort for Append/MergeAppend
--
create table prt_tbl (a int, b int) partition by range (a);
create table prt_tbl_1 partition of prt_tbl for values from (0) to (100);
create table prt_tbl_2 partition of prt_tbl for values from (100) to (200);
insert into prt_tbl select i%200, i from generate_series(1,1000)i;
create index on prt_tbl_1(a);
create index on prt_tbl_2(a, b);
analyze prt_tbl;
set enable_seqscan to off;
set enable_bitmapscan to off;
-- Ensure we get an incremental sort for the subpath of Append
explain (costs off) select * from prt_tbl order by a, b;
QUERY PLAN
------------------------------------------------------------
Append
-> Incremental Sort
Sort Key: prt_tbl_1.a, prt_tbl_1.b
Presorted Key: prt_tbl_1.a
-> Index Scan using prt_tbl_1_a_idx on prt_tbl_1
-> Index Only Scan using prt_tbl_2_a_b_idx on prt_tbl_2
(6 rows)
-- Ensure we get an incremental sort for the subpath of MergeAppend
explain (costs off) select * from prt_tbl_1 union all select * from prt_tbl_2 order by a, b;
QUERY PLAN
------------------------------------------------------------
Merge Append
Sort Key: prt_tbl_1.a, prt_tbl_1.b
-> Incremental Sort
Sort Key: prt_tbl_1.a, prt_tbl_1.b
Presorted Key: prt_tbl_1.a
-> Index Scan using prt_tbl_1_a_idx on prt_tbl_1
-> Index Only Scan using prt_tbl_2_a_b_idx on prt_tbl_2
(7 rows)
reset enable_bitmapscan;
reset enable_seqscan;
drop table prt_tbl;

View File

@ -1898,10 +1898,11 @@ ORDER BY thousand, tenthous;
Merge Append
Sort Key: tenk1.thousand, tenk1.tenthous
-> Index Only Scan using tenk1_thous_tenthous on tenk1
-> Sort
-> Incremental Sort
Sort Key: tenk1_1.thousand, tenk1_1.thousand
Presorted Key: tenk1_1.thousand
-> Index Only Scan using tenk1_thous_tenthous on tenk1 tenk1_1
(6 rows)
(7 rows)
explain (costs off)
SELECT thousand, tenthous, thousand+tenthous AS x FROM tenk1
@ -1982,10 +1983,11 @@ ORDER BY x, y;
Merge Append
Sort Key: a.thousand, a.tenthous
-> Index Only Scan using tenk1_thous_tenthous on tenk1 a
-> Sort
-> Incremental Sort
Sort Key: b.unique2, b.unique2
Presorted Key: b.unique2
-> Index Only Scan using tenk1_unique2 on tenk1 b
(6 rows)
(7 rows)
-- exercise rescan code path via a repeatedly-evaluated subquery
explain (costs off)

View File

@ -298,3 +298,27 @@ explain (costs off)
select * from
(select * from tenk1 order by four) t1 join tenk1 t2 on t1.four = t2.four and t1.two = t2.two
order by t1.four, t1.two limit 1;
--
-- Test incremental sort for Append/MergeAppend
--
create table prt_tbl (a int, b int) partition by range (a);
create table prt_tbl_1 partition of prt_tbl for values from (0) to (100);
create table prt_tbl_2 partition of prt_tbl for values from (100) to (200);
insert into prt_tbl select i%200, i from generate_series(1,1000)i;
create index on prt_tbl_1(a);
create index on prt_tbl_2(a, b);
analyze prt_tbl;
set enable_seqscan to off;
set enable_bitmapscan to off;
-- Ensure we get an incremental sort for the subpath of Append
explain (costs off) select * from prt_tbl order by a, b;
-- Ensure we get an incremental sort for the subpath of MergeAppend
explain (costs off) select * from prt_tbl_1 union all select * from prt_tbl_2 order by a, b;
reset enable_bitmapscan;
reset enable_seqscan;
drop table prt_tbl;