MIME-Version: 1.0
From: Alexandra Wang <alexandra.wang.oss@gmail.com>
Date: Thu, 27 Feb 2025 23:22:21 -0600
Message-ID: <CAK98qZ0MVVdV1tQKhHq=2qVkKN1rWff-YW6BGOOLoO509krv4Q@mail.gmail.com>
Subject: Remove extra Sort node above a btree-compatible Index Scan
To: PostgreSQL Hackers <pgsql-hackers@postgresql.org>
Content-Type: multipart/mixed; boundary="000000000000ebd0a1062f2cffe6"
Archived-At: <https://www.postgresql.org/message-id/CAK98qZ0MVVdV1tQKhHq%3D2qVkKN1rWff-YW6BGOOLoO509krv4Q%40mail.gmail.com>
Precedence: bulk

--000000000000ebd0a1062f2cffe6
Content-Type: multipart/alternative; boundary="000000000000ebd0a0062f2cffe4"

--000000000000ebd0a0062f2cffe4
Content-Type: text/plain; charset="UTF-8"
Content-Transfer-Encoding: quoted-printable

Hello hackers,

While reviewing Mark Dilger=E2=80=99s patch series "Index AM API cleanup" [=
1],
I noticed some unexpected plan diffs when running the xtree and treeb
test modules. Specifically, an additional Sort node appears on top of
an Index Only Scan, even when the index key and sort key are the same.
For example:

--- src/test/modules/xtree/expected/interval.out
+++ src/test/modules/xtree/results/interval.out
@@ -375,10 +375,12 @@
 SET enable_seqscan TO false;
 EXPLAIN (COSTS OFF)
 SELECT f1 FROM INTERVAL_TBL_OF r1 ORDER BY f1;
-                             QUERY PLAN
---------------------------------------------------------------------
- Index Only Scan using interval_tbl_of_f1_idx on interval_tbl_of r1
-(1 row)
+                                QUERY PLAN
+--------------------------------------------------------------------------
+ Sort
+   Sort Key: f1
+   ->  Index Only Scan using interval_tbl_of_f1_idx on interval_tbl_of r1
+(3 rows)


I=E2=80=99ve attached a patch that removes this unnecessary Sort node for
B-tree-compatible indexes. This change should:
- Reduce the number of test failures in the xtree module from 43 to 30
- Reduce the size of regression.diffs from 234K to 123K

Since pathkey comparison is a hot path in query planning and exercised
by many test queries, I plan to gather more performance metrics.
However, in a simple test running make installcheck with and without
the patch, I observed no negative impact on the runtime of the
regression test suite (which doesn=E2=80=99t include other btree-like index=
es)
and a positive impact on the same regression tests for xtree.

Regression tests (same plans):
w/o patch:
make installcheck  1.36s user 2.21s system 12% cpu 28.018 total
w/ patch:
make installcheck  1.32s user 2.12s system 12% cpu 28.089 total

xtree tests:
w/o patch (inferior plan w/ extra sort node):
make installcheck  1.52s user 2.42s system 10% cpu 36.817 total
w/ patch (good plan):
make installcheck  1.52s user 2.48s system 12% cpu 32.201 total

I=E2=80=99ve marked the patch as no-cfbot, as it applies only on top of the
aforementioned patch series [1].

Thoughts?

[1]
https://www.postgresql.org/message-id/a5dfb7cd-7a89-48ab-a913-e5304eee0854%=
40eisentraut.org

Best,
Alex

--000000000000ebd0a0062f2cffe4
Content-Type: text/html; charset="UTF-8"
Content-Transfer-Encoding: quoted-printable

<div dir=3D"ltr">Hello hackers,<br><br>While reviewing Mark Dilger=E2=80=99=
s patch series &quot;Index AM API cleanup&quot; [1],<br>I noticed some unex=
pected plan diffs when running the xtree and treeb<br>test modules. Specifi=
cally, an additional Sort node appears on top of<br>an Index Only Scan, eve=
n when the index key and sort key are the same.<br>For example:<br><br>--- =
src/test/modules/xtree/expected/interval.out<br>+++ src/test/modules/xtree/=
results/interval.out<br>@@ -375,10 +375,12 @@<br>=C2=A0SET enable_seqscan T=
O false;<br>=C2=A0EXPLAIN (COSTS OFF)<br>=C2=A0SELECT f1 FROM INTERVAL_TBL_=
OF r1 ORDER BY f1;<br>- =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =
=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 QUERY PLAN =C2=A0 =C2=A0 =
=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=
=A0 =C2=A0 <br>------------------------------------------------------------=
---------<br>- Index Only Scan using interval_tbl_of_f1_idx on interval_tbl=
_of r1<br>-(1 row)<br>+ =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =
=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0QUERY PLAN =
=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=
=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0<br>+--------------------------------=
------------------------------------------<br>+ Sort<br>+ =C2=A0 Sort Key: =
f1<br>+ =C2=A0 -&gt; =C2=A0Index Only Scan using interval_tbl_of_f1_idx on =
interval_tbl_of r1<br>+(3 rows)<br><br><br>I=E2=80=99ve attached a patch th=
at removes this unnecessary Sort node for<br>B-tree-compatible indexes. Thi=
s change should:<br>- Reduce the number of test failures in the xtree modul=
e from 43 to 30<br>- Reduce the size of regression.diffs from 234K to 123K<=
br><br>Since pathkey comparison is a hot path in query planning and exercis=
ed<br>by many test queries, I plan to gather more performance metrics.<br>H=
owever, in a simple test running make installcheck with and without<br>the =
patch, I observed no negative impact on the runtime of the<br>regression te=
st suite (which doesn=E2=80=99t include other btree-like indexes)<br>and a =
positive impact on the same regression tests for xtree.<br><br>Regression t=
ests (same plans):<br>w/o patch:<br>make installcheck =C2=A01.36s user 2.21=
s system 12% cpu 28.018 total<br>w/ patch:<br>make installcheck =C2=A01.32s=
 user 2.12s system 12% cpu 28.089 total<br><br>xtree tests:<br>w/o patch (i=
nferior plan w/ extra sort node):<br>make installcheck =C2=A01.52s user 2.4=
2s system 10% cpu 36.817 total<br>w/ patch (good plan):<br>make installchec=
k =C2=A01.52s user 2.48s system 12% cpu 32.201 total<br><br>I=E2=80=99ve ma=
rked the patch as no-cfbot, as it applies only on top of the<br>aforementio=
ned patch series [1].<br><br>Thoughts?<br><br>[1] <a href=3D"https://www.po=
stgresql.org/message-id/a5dfb7cd-7a89-48ab-a913-e5304eee0854%40eisentraut.o=
rg">https://www.postgresql.org/message-id/a5dfb7cd-7a89-48ab-a913-e5304eee0=
854%40eisentraut.org</a><br><br>Best,<br>Alex<br><div><br></div></div>

--000000000000ebd0a0062f2cffe4--
--000000000000ebd0a1062f2cffe6
Content-Type: application/octet-stream; 
	name="v1-0001-Remove-unnecessary-Sort-node-above-Index-Scan-of-.patch.no-cfbot"
Content-Disposition: attachment; 
	filename="v1-0001-Remove-unnecessary-Sort-node-above-Index-Scan-of-.patch.no-cfbot"
Content-Transfer-Encoding: base64
Content-ID: <f_m7obq9jx0>
X-Attachment-Id: f_m7obq9jx0
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--000000000000ebd0a1062f2cffe6--