Thanks for detaied answer,
On 3/11/2023 00:37, Tomas Vondra wrote:
On 9/25/23 06:30, Andrey Lepikhov wrote:
...
I can't stop thinking about this issue. It is bizarre when Postgres
chooses a non-unique index if a unique index gives us proof of minimum
scan.
That's true, but no one implemented this heuristics. So the "proof of
minimum scan" is merely hypothetical - the optimizer is unaware of it.
See the simple patch in the attachment. There, I have attempted to
resolve situations of uncertainty to avoid making decisions based solely
on the order of indexes in the list.
I don't see a reason to teach the clauselist_selectivity() routine to
estimate UNIQUE indexes. We add some cycles, but it will work with btree
indexes only.
I'm not sure I understand what this is meant to say. Can you elaborate?
We only allow UNIQUE for btree indexes anyway, so what exactly is the
problem here?
Partly, you already answered yourself below: we have unique index
estimation in a few estimation calls, but go through the list of indexes
each time.
Also, for this sake, we would add some input parameters, usually NULL,
because many estimations don't involve indexes at all.
Maybe to change compare_path_costs_fuzzily() and add some heuristic, for
example:
"If selectivity of both paths gives us no more than 1 row, prefer to use
a unique index or an index with least selectivity."
I don't understand how this would work. What do yo mean by "selectivity
of a path"? AFAICS the problem here is that we estimate a path to return
more rows (while we know there can only be 1, thanks to UNIQUE index).
Oops, I meant cardinality. See the patch in the attachment.
So how would you know the path does not give us more than 1 row? Surely
you're not proposing compare_path_costs_fuzzily() to do something
expensive like analyzing the paths, or something.
I solely propose to make optimizer more consecutive in its decisions: if
we have one row for both path candidates, use uniqueness of the index or
value of selectivity as one more parameter.
Also, how would it work in upper levels? If you just change which path
we keep, but leave the inaccurate row estimate in place, that may fix
that level, but it's certainly going to confuse later planning steps.
It is designed for the only scan level.
IMHO the problem here is that we produce wrong estimate, so we better
fix that, instead of adding band-aids to other places.
Agree. I am looking for a solution to help users somehow resolve such
problems. As an alternative solution, I can propose a selectivity hook
or (maybe even better) - use the pathlist approach and add indexes into
the index list with some predefined order - at first positions, place
unique indexes with more columns, etc.
This happens because functional dependencies are very simple type of
statistics - it has some expectations about the input data and also the
queries executed on it. For example it assumes the data is reasonably
homogeneous, so that we can calculate a global "degree".
But the data in the example directly violates that - it has 1000 rows
that are very random (so we'd find no dependencies), and 1000 rows with
perfect dependencies. Hence we end with degree=0.5, which approximates
the dependencies to all data. Not great, true, but that's the price for
simplicity of this statistics kind.
So the simplest solution is to disable dependencies on such data sets.
It's a bit inconvenient/unfortunate we build dependencies by default,
and people may not be aware of there assumptions.
Perhaps we could/should make dependency_degree() more pessimistic when
we find some "violations" of the rule (we intentionally are not strict
about it, because data are imperfect). I don't have a good idea how to
change the formulas, but I'm open to the idea in principle.
Thanks for the explanation!
The other thing we could do is checking for unique indexes in
clauselist_selectivity, and if we find a match we can just skip the
extended statistics altogether. Not sure how expensive this would be,
but for typical cases (with modest number of indexes) perhaps OK. It
wouldn't work without a unique index, but I don't have a better idea.
It looks a bit expensive for me. But I am ready to try, if current
solution doesn't look applicable.
--
regards,
Andrei Lepikhov
Postgres Professional
From 21677861e101eed22c829e0b14290a56786a12c2 Mon Sep 17 00:00:00 2001
From: "Andrey V. Lepikhov" <a.lepik...@postgrespro.ru>
Date: Wed, 22 Nov 2023 12:01:39 +0700
Subject: [PATCH] Use an index path with the best selectivity estimation
---
src/backend/optimizer/util/pathnode.c | 40 +++++++++++++++++
.../expected/drop-index-concurrently-1.out | 16 ++++---
src/test/regress/expected/functional_deps.out | 43 +++++++++++++++++++
src/test/regress/expected/join.out | 40 +++++++++--------
src/test/regress/sql/functional_deps.sql | 36 ++++++++++++++++
5 files changed, 149 insertions(+), 26 deletions(-)
diff --git a/src/backend/optimizer/util/pathnode.c
b/src/backend/optimizer/util/pathnode.c
index 0b1d17b9d3..32aca83d67 100644
--- a/src/backend/optimizer/util/pathnode.c
+++ b/src/backend/optimizer/util/pathnode.c
@@ -454,6 +454,46 @@ add_path(RelOptInfo *parent_rel, Path *new_path)
costcmp = compare_path_costs_fuzzily(new_path, old_path,
STD_FUZZ_FACTOR);
+ /*
+ * Apply some heuristics on index paths.
+ */
+ if (IsA(new_path, IndexPath) && IsA(old_path, IndexPath))
+ {
+ IndexPath *inp = (IndexPath *) new_path;
+ IndexPath *iop = (IndexPath *) old_path;
+
+ if (new_path->rows <= 1.0 && old_path->rows <= 1.0)
+ {
+ if (inp->indexinfo->unique &&
!iop->indexinfo->unique)
+ /*
+ * When both paths are predicted to
produce only one tuple,
+ * the optimiser should prefer choosing
a unique index scan
+ * in all cases.
+ */
+ costcmp = COSTS_BETTER1;
+ else if (costcmp != COSTS_DIFFERENT)
+ /*
+ * If the optimiser doesn't have an
obviously stable choice
+ * of unique index, increase the chance
of avoiding mistakes
+ * by choosing an index with smaller
selectivity.
+ * This option makes decision more
conservative and looks
+ * debatable.
+ */
+ costcmp = (inp->indexselectivity <
iop->indexselectivity) ?
+
COSTS_BETTER1 : COSTS_BETTER2;
+ }
+ else if (costcmp == COSTS_EQUAL)
+ /*
+ * The optimizer can't differ the value of two
index paths.
+ * To avoid making a decision that is based on
only an index
+ * order in the list, use some rational
strategy based on
+ * selectivity: prefer touching fewer tuples on
the disk to
+ * filtering them after.
+ */
+ costcmp = (inp->indexselectivity <
iop->indexselectivity) ?
+
COSTS_BETTER1 : COSTS_BETTER2;
+ }
+
/*
* If the two paths compare differently for startup and total
cost,
* then we want to keep both, and we can skip comparing
pathkeys and
diff --git a/src/test/isolation/expected/drop-index-concurrently-1.out
b/src/test/isolation/expected/drop-index-concurrently-1.out
index 1cb2250891..2392cdb033 100644
--- a/src/test/isolation/expected/drop-index-concurrently-1.out
+++ b/src/test/isolation/expected/drop-index-concurrently-1.out
@@ -12,13 +12,15 @@ step preps: PREPARE getrow_seqscan AS SELECT * FROM test_dc
WHERE data = 34 ORDE
step begin: BEGIN;
step disableseq: SET enable_seqscan = false;
step explaini: EXPLAIN (COSTS OFF) EXECUTE getrow_idxscan;
-QUERY PLAN
-----------------------------------------------
-Sort
- Sort Key: id
- -> Index Scan using test_dc_data on test_dc
- Index Cond: (data = 34)
-(4 rows)
+QUERY PLAN
+---------------------------------------------
+Sort
+ Sort Key: id
+ -> Bitmap Heap Scan on test_dc
+ Recheck Cond: (data = 34)
+ -> Bitmap Index Scan on test_dc_data
+ Index Cond: (data = 34)
+(6 rows)
step enableseq: SET enable_seqscan = true;
step explains: EXPLAIN (COSTS OFF) EXECUTE getrow_seqscan;
diff --git a/src/test/regress/expected/functional_deps.out
b/src/test/regress/expected/functional_deps.out
index 32381b8ae7..61af99a041 100644
--- a/src/test/regress/expected/functional_deps.out
+++ b/src/test/regress/expected/functional_deps.out
@@ -230,3 +230,46 @@ EXECUTE foo;
ALTER TABLE articles DROP CONSTRAINT articles_pkey RESTRICT;
EXECUTE foo; -- fail
ERROR: column "articles.keywords" must appear in the GROUP BY clause or be
used in an aggregate function
+/*
+ * Corner cases of PostgreSQL optimizer:
+ *
+ * Correlations between columns aren't found by DBMS.
+ * Selectivities multiplication of many columns increases total selectivity
+ * error. If such non-true selectivity is so small, that rows estimation
+ * give us absolute minimum (1) then the optimizer can't choose between
different
+ * indexes and uses first from the index list (last created).
+ */
+\set scale 100000
+CREATE TABLE t AS (
+ SELECT c1 AS c1, -- selectivity(c1)*selectivity(c2)*nrows <= 1
+ c1 AS c2,
+ -- Create two columns with different selectivity.
+ (c1 % 2) AS c3, -- missing from a good index.
+ (c1 % 4) AS c4 -- missing from a bad index.
+ FROM generate_series(1,:scale) AS c1
+);
+UPDATE t SET c1=1,c2=1,c3=1,c4=2 WHERE c1<:scale/1000;
+CREATE INDEX bad ON t (c1,c2,c3);
+CREATE INDEX good ON t (c1,c2,c4);
+ANALYZE t; -- update stat on the indexes.
+EXPLAIN (COSTS OFF) SELECT * FROM t WHERE c1=1 AND c2=1 AND c3=1 AND c4=1;
+ QUERY PLAN
+----------------------------------------------------
+ Index Scan using good on t
+ Index Cond: ((c1 = 1) AND (c2 = 1) AND (c4 = 1))
+ Filter: (c3 = 1)
+(3 rows)
+
+-- Set the bad index to the first position in the index list.
+DROP INDEX bad;
+CREATE INDEX bad ON t (c1,c2,c3);
+ANALYZE t;
+EXPLAIN (COSTS OFF) SELECT * FROM t WHERE c1=1 AND c2=1 AND c3=1 AND c4=1;
+ QUERY PLAN
+----------------------------------------------------
+ Index Scan using good on t
+ Index Cond: ((c1 = 1) AND (c2 = 1) AND (c4 = 1))
+ Filter: (c3 = 1)
+(3 rows)
+
+DROP TABLE t CASCADE;
diff --git a/src/test/regress/expected/join.out
b/src/test/regress/expected/join.out
index 2c73270143..32b33fabd3 100644
--- a/src/test/regress/expected/join.out
+++ b/src/test/regress/expected/join.out
@@ -8629,14 +8629,15 @@ analyze j2;
explain (costs off) select * from j1
inner join j2 on j1.id1 = j2.id1 and j1.id2 = j2.id2
where j1.id1 % 1000 = 1 and j2.id1 % 1000 = 1;
- QUERY PLAN
------------------------------------------
+ QUERY PLAN
+---------------------------------------------------------
Merge Join
- Merge Cond: (j1.id1 = j2.id1)
- Join Filter: (j2.id2 = j1.id2)
- -> Index Scan using j1_id1_idx on j1
- -> Index Scan using j2_id1_idx on j2
-(5 rows)
+ Merge Cond: ((j1.id1 = j2.id1) AND (j1.id2 = j2.id2))
+ -> Index Only Scan using j1_pkey on j1
+ Filter: ((id1 % 1000) = 1)
+ -> Index Only Scan using j2_pkey on j2
+ Filter: ((id1 % 1000) = 1)
+(6 rows)
select * from j1
inner join j2 on j1.id1 = j2.id1 and j1.id2 = j2.id2
@@ -8651,15 +8652,16 @@ where j1.id1 % 1000 = 1 and j2.id1 % 1000 = 1;
explain (costs off) select * from j1
inner join j2 on j1.id1 = j2.id1 and j1.id2 = j2.id2
where j1.id1 % 1000 = 1 and j2.id1 % 1000 = 1 and j2.id1 = any (array[1]);
- QUERY PLAN
-----------------------------------------------------
+ QUERY PLAN
+---------------------------------------------------------
Merge Join
- Merge Cond: (j1.id1 = j2.id1)
- Join Filter: (j2.id2 = j1.id2)
- -> Index Scan using j1_id1_idx on j1
- -> Index Scan using j2_id1_idx on j2
+ Merge Cond: ((j1.id1 = j2.id1) AND (j1.id2 = j2.id2))
+ -> Index Only Scan using j1_pkey on j1
+ Filter: ((id1 % 1000) = 1)
+ -> Index Only Scan using j2_pkey on j2
Index Cond: (id1 = ANY ('{1}'::integer[]))
-(6 rows)
+ Filter: ((id1 % 1000) = 1)
+(7 rows)
select * from j1
inner join j2 on j1.id1 = j2.id1 and j1.id2 = j2.id2
@@ -8674,12 +8676,12 @@ where j1.id1 % 1000 = 1 and j2.id1 % 1000 = 1 and
j2.id1 = any (array[1]);
explain (costs off) select * from j1
inner join j2 on j1.id1 = j2.id1 and j1.id2 = j2.id2
where j1.id1 % 1000 = 1 and j2.id1 % 1000 = 1 and j2.id1 >= any (array[1,5]);
- QUERY PLAN
--------------------------------------------------------
+ QUERY PLAN
+---------------------------------------------------------
Merge Join
- Merge Cond: (j1.id1 = j2.id1)
- Join Filter: (j2.id2 = j1.id2)
- -> Index Scan using j1_id1_idx on j1
+ Merge Cond: ((j1.id1 = j2.id1) AND (j1.id2 = j2.id2))
+ -> Index Only Scan using j1_pkey on j1
+ Filter: ((id1 % 1000) = 1)
-> Index Only Scan using j2_pkey on j2
Index Cond: (id1 >= ANY ('{1,5}'::integer[]))
Filter: ((id1 % 1000) = 1)
diff --git a/src/test/regress/sql/functional_deps.sql
b/src/test/regress/sql/functional_deps.sql
index 406490b995..f617ee9269 100644
--- a/src/test/regress/sql/functional_deps.sql
+++ b/src/test/regress/sql/functional_deps.sql
@@ -208,3 +208,39 @@ EXECUTE foo;
ALTER TABLE articles DROP CONSTRAINT articles_pkey RESTRICT;
EXECUTE foo; -- fail
+
+/*
+ * Corner cases of PostgreSQL optimizer:
+ *
+ * Correlations between columns aren't found by DBMS.
+ * Selectivities multiplication of many columns increases total selectivity
+ * error. If such non-true selectivity is so small, that rows estimation
+ * give us absolute minimum (1) then the optimizer can't choose between
different
+ * indexes and uses first from the index list (last created).
+ */
+\set scale 100000
+
+CREATE TABLE t AS (
+ SELECT c1 AS c1, -- selectivity(c1)*selectivity(c2)*nrows <= 1
+ c1 AS c2,
+ -- Create two columns with different selectivity.
+ (c1 % 2) AS c3, -- missing from a good index.
+ (c1 % 4) AS c4 -- missing from a bad index.
+ FROM generate_series(1,:scale) AS c1
+);
+UPDATE t SET c1=1,c2=1,c3=1,c4=2 WHERE c1<:scale/1000;
+
+CREATE INDEX bad ON t (c1,c2,c3);
+CREATE INDEX good ON t (c1,c2,c4);
+ANALYZE t; -- update stat on the indexes.
+
+EXPLAIN (COSTS OFF) SELECT * FROM t WHERE c1=1 AND c2=1 AND c3=1 AND c4=1;
+
+-- Set the bad index to the first position in the index list.
+DROP INDEX bad;
+CREATE INDEX bad ON t (c1,c2,c3);
+ANALYZE t;
+
+EXPLAIN (COSTS OFF) SELECT * FROM t WHERE c1=1 AND c2=1 AND c3=1 AND c4=1;
+
+DROP TABLE t CASCADE;
--
2.43.0
