Merge #120079

120079: opt: fix row count estimates for inverted filters r=mgartner a=mgartner

#### opt: assign table ID to hypothetical inverted index columns

This commit fixes a minor bug in index recommendations where
hypothetical inverted index columns were not assigned table IDs in the
hypothetical table's metadata. In theory this could cause internal
errors, however, there are no known reproductions of errors.

Release note: None

#### opt: decrease row count rounding precision

Row counts are now rounded with less precision so that optimizer test
outputs are the same across different CPU architectures. Follow-up
of #75723.

Release note: None

#### opt: fix row count estimates for inverted filters

Previously, inverted filter row counts were set to their input's row
count. This was often an over-estimation because inverted filters
de-duplicate inverted index tuples from their input by primary key, and
inverted indexes can contain multiple tuples for each logical row.

This commit fixes this over-estimation by setting the row count of
inverted filters to the distinct count of the primary key columns from
the input.

Fixes #120055

Release note (performance improvement): A bug has been fixed that caused
the optimizer to over-estimate the cost of inverted index scans in some
cases. Now, plans with inverted index scans should be selected in more
cases where they are optimal.



Co-authored-by: Marcus Gartner <[email protected]>

Author: craig[bot]

pkg/ccl/logictestccl/testdata/logic_test/explain_redact

@@ -294,7 +294,7 @@ EXPLAIN (OPT, VERBOSE, REDACT) SELECT * FROM q WHERE q > 2
 scan q
  ├── columns: q:1 r:2
  ├── constraint: /1: ‹×›
- ├── stats: [rows=333.3333, distinct(1)=333.333, null(1)=0]
+ ├── stats: [rows=333.333, distinct(1)=333.333, null(1)=0]
  ├── cost: 378.02
  ├── key: (1)
  ├── fd: (1)-->(2)
@@ -307,7 +307,7 @@ EXPLAIN (OPT, TYPES, REDACT) SELECT * FROM q WHERE q > 2
 scan q
  ├── columns: q:1(int!null) r:2(int)
  ├── constraint: /1: ‹×›
- ├── stats: [rows=333.3333, distinct(1)=333.333, null(1)=0]
+ ├── stats: [rows=333.333, distinct(1)=333.333, null(1)=0]
  ├── cost: 378.02
  ├── key: (1)
  ├── fd: (1)-->(2)

pkg/ccl/logictestccl/testdata/logic_test/multi_region_remote_access_error

@@ -1079,7 +1079,7 @@ project
       │    ├── key columns: [26 22] = [13 9]
       │    ├── lookup columns are key
       │    ├── immutable
-      │    ├── stats: [rows=3333.333]
+      │    ├── stats: [rows=3333.33]
       │    ├── cost: 4309.11778
       │    ├── fd: ()-->(13), (9)-->(11)
       │    ├── limit hint: 1.00
@@ -1092,7 +1092,7 @@ project
       │    │    ├── prefix key columns: [21] = [26]
       │    │    ├── inverted-expr
       │    │    │    └── t1.j:24 @> t2.j:3
-      │    │    ├── stats: [rows=3333.333, distinct(21)=1, null(21)=0, distinct(22)=964.524, null(22)=0, distinct(26)=1, null(26)=0]
+      │    │    ├── stats: [rows=3333.33, distinct(21)=1, null(21)=0, distinct(22)=964.524, null(22)=0, distinct(26)=1, null(26)=0]
       │    │    ├── cost: 3837.19889
       │    │    ├── fd: ()-->(26)
       │    │    ├── limit hint: 100.00

pkg/ccl/logictestccl/testdata/logic_test/regional_by_row_query_behavior

@@ -3646,7 +3646,7 @@ WHERE
 project
  ├── columns: str:12 id1:2 id2:5 created_at:3 updated_at:4 id1:21 id2:24 created_at:22 updated_at:23
  ├── immutable
- ├── stats: [rows=0.3666667]
+ ├── stats: [rows=0.366667]
  ├── cost: 45.3262257
  ├── fd: ()-->(2,5,21,24), (3)==(22), (22)==(3), (5)==(24), (24)==(5), (4)==(23), (23)==(4), (2)==(21), (21)==(2)
  ├── distribution: ap-southeast-2
@@ -3655,7 +3655,7 @@ project
       ├── columns: a.id:1 a.id1:2 a.created_at:3 a.updated_at:4 a.id2:5 a.crdb_region:6 str:12 abc_id:13 x.id2:14 x.crdb_region:15 b.id:20 b.id1:21 b.created_at:22 b.updated_at:23 b.id2:24 b.crdb_region:25
       ├── key columns: [6 5] = [15 14]
       ├── immutable
-      ├── stats: [rows=0.3666667, distinct(13)=0.366667, null(13)=0, distinct(14)=0.366667, null(14)=0, distinct(15)=0.366667, null(15)=0, distinct(20)=0.366667, null(20)=0, distinct(24)=0.366667, null(24)=0, distinct(25)=0.366667, null(25)=0]
+      ├── stats: [rows=0.366667, distinct(13)=0.366667, null(13)=0, distinct(14)=0.366667, null(14)=0, distinct(15)=0.366667, null(15)=0, distinct(20)=0.366667, null(20)=0, distinct(24)=0.366667, null(24)=0, distinct(25)=0.366667, null(25)=0]
       ├── cost: 45.302559
       ├── fd: ()-->(2,5,6,14,15,21,24,25), (1)-->(3,4), (20)-->(22,23), (1)==(13,20), (13)==(1,20), (20)==(1,13), (5)==(14,24), (14)==(5,24), (24)==(5,14), (6)==(15,25), (15)==(6,25), (25)==(6,15), (2)==(21), (21)==(2), (3)==(22), (22)==(3), (4)==(23), (23)==(4)
       ├── distribution: ap-southeast-2
@@ -3665,7 +3665,7 @@ project
       │    ├── key columns: [25 20] = [25 20]
       │    ├── lookup columns are key
       │    ├── immutable
-      │    ├── stats: [rows=0.1344445, distinct(1)=0.134444, null(1)=0, distinct(2)=0.134444, null(2)=0, distinct(3)=0.134444, null(3)=0, distinct(4)=0.134444, null(4)=0, distinct(5)=0.134444, null(5)=0, distinct(6)=0.134444, null(6)=0, distinct(20)=0.134444, null(20)=0, distinct(21)=0.134444, null(21)=0, distinct(22)=0.134444, null(22)=0, distinct(23)=0.134444, null(23)=0, distinct(24)=0.134444, null(24)=0, distinct(25)=0.134444, null(25)=0]
+      │    ├── stats: [rows=0.134444, distinct(1)=0.134444, null(1)=0, distinct(2)=0.134444, null(2)=0, distinct(3)=0.134444, null(3)=0, distinct(4)=0.134444, null(4)=0, distinct(5)=0.134444, null(5)=0, distinct(6)=0.134444, null(6)=0, distinct(20)=0.134444, null(20)=0, distinct(21)=0.134444, null(21)=0, distinct(22)=0.134444, null(22)=0, distinct(23)=0.134444, null(23)=0, distinct(24)=0.134444, null(24)=0, distinct(25)=0.134444, null(25)=0]
       │    ├── cost: 34.9459129
       │    ├── key: (20)
       │    ├── fd: ()-->(2,5,6,21,24,25), (1)-->(3,4), (20)-->(22,23), (1)==(20), (20)==(1), (2)==(21), (21)==(2), (3)==(22), (22)==(3), (4)==(23), (23)==(4), (5)==(24), (24)==(5), (6)==(25), (25)==(6)
@@ -3677,7 +3677,7 @@ project
       │    │    ├── key columns: [6 2 5 1] = [25 21 24 20]
       │    │    ├── lookup columns are key
       │    │    ├── immutable
-      │    │    ├── stats: [rows=0.003666667, distinct(1)=0.00366667, null(1)=0, distinct(2)=0.00366667, null(2)=0, distinct(5)=0.00366667, null(5)=0, distinct(6)=0.00366667, null(6)=0, distinct(20)=0.00366667, null(20)=0, distinct(21)=0.00366667, null(21)=0, distinct(24)=0.00366667, null(24)=0, distinct(25)=0.00366667, null(25)=0, distinct(21,24,25)=0.00366667, null(21,24,25)=0]
+      │    │    ├── stats: [rows=0.00366667, distinct(1)=0.00366667, null(1)=0, distinct(2)=0.00366667, null(2)=0, distinct(5)=0.00366667, null(5)=0, distinct(6)=0.00366667, null(6)=0, distinct(20)=0.00366667, null(20)=0, distinct(21)=0.00366667, null(21)=0, distinct(24)=0.00366667, null(24)=0, distinct(25)=0.00366667, null(25)=0, distinct(21,24,25)=0.00366667, null(21,24,25)=0]
       │    │    ├── cost: 34.8283629
       │    │    ├── key: (20)
       │    │    ├── fd: ()-->(2,5,6,21,24,25), (1)-->(3,4), (6)==(25), (25)==(6), (2)==(21), (21)==(2), (5)==(24), (24)==(5), (1)==(20), (20)==(1)
@@ -3687,7 +3687,7 @@ project
       │    │    ├── index-join abc
       │    │    │    ├── columns: a.id:1 a.id1:2 a.created_at:3 a.updated_at:4 a.id2:5 a.crdb_region:6
       │    │    │    ├── immutable
-      │    │    │    ├── stats: [rows=0.3666667, distinct(1)=0.366667, null(1)=0, distinct(2)=0.366667, null(2)=0, distinct(3)=0.360675, null(3)=0, distinct(4)=0.360675, null(4)=0, distinct(5)=0.366667, null(5)=0, distinct(6)=0.366667, null(6)=0, distinct(2,5,6)=0.366667, null(2,5,6)=0]
+      │    │    │    ├── stats: [rows=0.366667, distinct(1)=0.366667, null(1)=0, distinct(2)=0.366667, null(2)=0, distinct(3)=0.360675, null(3)=0, distinct(4)=0.360675, null(4)=0, distinct(5)=0.366667, null(5)=0, distinct(6)=0.366667, null(6)=0, distinct(2,5,6)=0.366667, null(2,5,6)=0]
       │    │    │    │   histogram(6)=  0      0.36667
       │    │    │    │                <--- 'ap-southeast-2'
       │    │    │    ├── cost: 33.2056672
@@ -3699,7 +3699,7 @@ project
       │    │    │    └── scan abc@abc_id1_id2_idx [as=a]
       │    │    │         ├── columns: a.id:1 a.id1:2 a.id2:5 a.crdb_region:6
       │    │    │         ├── constraint: /6/2/5/1: [/'ap-southeast-2'/'6da4f356-e526-4b78-b9f9-bbb1a7fc12d6'/'68088706-02c6-47d1-b993-a421cd761f2b' - /'ap-southeast-2'/'6da4f356-e526-4b78-b9f9-bbb1a7fc12d6'/'68088706-02c6-47d1-b993-a421cd761f2b']
-      │    │    │         ├── stats: [rows=0.3666667, distinct(1)=0.366667, null(1)=0, distinct(2)=0.366667, null(2)=0, distinct(5)=0.366667, null(5)=0, distinct(6)=0.366667, null(6)=0, distinct(2,5,6)=0.366667, null(2,5,6)=0]
+      │    │    │         ├── stats: [rows=0.366667, distinct(1)=0.366667, null(1)=0, distinct(2)=0.366667, null(2)=0, distinct(5)=0.366667, null(5)=0, distinct(6)=0.366667, null(6)=0, distinct(2,5,6)=0.366667, null(2,5,6)=0]
       │    │    │         │   histogram(6)=  0      0.36667
       │    │    │         │                <--- 'ap-southeast-2'
       │    │    │         ├── cost: 25.430667

pkg/sql/logictest/testdata/logic_test/prepare

@@ -1182,7 +1182,7 @@ EXECUTE e
 select
  ├── columns: k:1 str:2
  ├── immutable
- ├── stats: [rows=333.3333]
+ ├── stats: [rows=333.333]
  ├── cost: 1118.85
  ├── key: (1)
  ├── fd: (1)-->(2)

pkg/sql/opt/exec/execbuilder/testdata/distsql_agg

@@ -1026,7 +1026,7 @@ group-by (streaming)
  ├── columns: b:2 count:7
  ├── grouping columns: b:2
  ├── internal-ordering: +2 opt(1)
- ├── stats: [rows=9.561793, distinct(2)=9.56179, null(2)=0]
+ ├── stats: [rows=9.56179, distinct(2)=9.56179, null(2)=0]
  ├── cost: 29.145618
  ├── key: (2)
  ├── fd: (2)-->(7)

pkg/sql/opt/exec/execbuilder/testdata/enums

@@ -180,14 +180,14 @@ union
  ├── interesting orderings: (+15)
  ├── project
  │    ├── columns: nulls.x:1
- │    ├── stats: [rows=333.3333, distinct(1)=5, null(1)=3.33333]
+ │    ├── stats: [rows=333.333, distinct(1)=5, null(1)=3.33333]
  │    ├── cost: 1142.40333
  │    ├── ordering: +1
  │    ├── distribution: test
  │    ├── interesting orderings: (+1)
  │    └── select
  │         ├── columns: nulls.x:1 y:2
- │         ├── stats: [rows=333.3333, distinct(1)=5, null(1)=3.33333, distinct(2)=33.3333, null(2)=0]
+ │         ├── stats: [rows=333.333, distinct(1)=5, null(1)=3.33333, distinct(2)=33.3333, null(2)=0]
  │         ├── cost: 1139.05
  │         ├── ordering: +1
  │         ├── distribution: test
@@ -204,14 +204,14 @@ union
  │              └── y:2 < 0 [outer=(2), constraints=(/2: (/NULL - /-1]; tight)]
  └── project
       ├── columns: nulls.x:8
-      ├── stats: [rows=333.3333, distinct(8)=5, null(8)=3.33333]
+      ├── stats: [rows=333.333, distinct(8)=5, null(8)=3.33333]
       ├── cost: 1142.40333
       ├── ordering: +8
       ├── distribution: test
       ├── interesting orderings: (+8)
       └── select
            ├── columns: nulls.x:8 y:9
-           ├── stats: [rows=333.3333, distinct(8)=5, null(8)=3.33333, distinct(9)=33.3333, null(9)=0]
+           ├── stats: [rows=333.333, distinct(8)=5, null(8)=3.33333, distinct(9)=33.3333, null(9)=0]
            ├── cost: 1139.05
            ├── ordering: +8
            ├── distribution: test

pkg/sql/opt/exec/execbuilder/testdata/explain

@@ -1769,7 +1769,7 @@ EXPLAIN (OPT, VERBOSE) SELECT * FROM tc WHERE a + 2 * b > 1 ORDER BY a*b
 distribute
  ├── columns: a:1 b:2  [hidden: column8:8]
  ├── immutable
- ├── stats: [rows=333.3333]
+ ├── stats: [rows=333.333]
  ├── cost: 1418.32951
  ├── fd: (1,2)-->(8)
  ├── ordering: +8
@@ -1780,7 +1780,7 @@ distribute
  └── sort
       ├── columns: a:1 b:2 column8:8
       ├── immutable
-      ├── stats: [rows=333.3333]
+      ├── stats: [rows=333.333]
       ├── cost: 1218.30951
       ├── fd: (1,2)-->(8)
       ├── ordering: +8
@@ -1789,15 +1789,15 @@ distribute
       └── project
            ├── columns: column8:8 a:1 b:2
            ├── immutable
-           ├── stats: [rows=333.3333]
+           ├── stats: [rows=333.333]
            ├── cost: 1145.73667
            ├── fd: (1,2)-->(8)
            ├── prune: (1,2,8)
            ├── interesting orderings: (+1)
            ├── select
            │    ├── columns: a:1 b:2
            │    ├── immutable
-           │    ├── stats: [rows=333.3333]
+           │    ├── stats: [rows=333.333]
            │    ├── cost: 1139.05
            │    ├── interesting orderings: (+1)
            │    ├── scan tc
@@ -1817,7 +1817,7 @@ EXPLAIN (OPT, VERBOSE, REDACT) SELECT * FROM tc WHERE a + 2 * b > 1 ORDER BY a*b
 distribute
  ├── columns: a:1 b:2  [hidden: column8:8]
  ├── immutable
- ├── stats: [rows=333.3333]
+ ├── stats: [rows=333.333]
  ├── cost: 1418.32951
  ├── fd: (1,2)-->(8)
  ├── ordering: +8
@@ -1828,7 +1828,7 @@ distribute
  └── sort
       ├── columns: a:1 b:2 column8:8
       ├── immutable
-      ├── stats: [rows=333.3333]
+      ├── stats: [rows=333.333]
       ├── cost: 1218.30951
       ├── fd: (1,2)-->(8)
       ├── ordering: +8
@@ -1837,15 +1837,15 @@ distribute
       └── project
            ├── columns: column8:8 a:1 b:2
            ├── immutable
-           ├── stats: [rows=333.3333]
+           ├── stats: [rows=333.333]
            ├── cost: 1145.73667
            ├── fd: (1,2)-->(8)
            ├── prune: (1,2,8)
            ├── interesting orderings: (+1)
            ├── select
            │    ├── columns: a:1 b:2
            │    ├── immutable
-           │    ├── stats: [rows=333.3333]
+           │    ├── stats: [rows=333.333]
            │    ├── cost: 1139.05
            │    ├── interesting orderings: (+1)
            │    ├── scan tc
@@ -1865,7 +1865,7 @@ EXPLAIN (OPT, TYPES) SELECT * FROM tc WHERE a + 2 * b > 1 ORDER BY a*b
 distribute
  ├── columns: a:1(int) b:2(int)  [hidden: column8:8(int)]
  ├── immutable
- ├── stats: [rows=333.3333]
+ ├── stats: [rows=333.333]
  ├── cost: 1418.32951
  ├── fd: (1,2)-->(8)
  ├── ordering: +8
@@ -1876,7 +1876,7 @@ distribute
  └── sort
       ├── columns: a:1(int) b:2(int) column8:8(int)
       ├── immutable
-      ├── stats: [rows=333.3333]
+      ├── stats: [rows=333.333]
       ├── cost: 1218.30951
       ├── fd: (1,2)-->(8)
       ├── ordering: +8
@@ -1885,15 +1885,15 @@ distribute
       └── project
            ├── columns: column8:8(int) a:1(int) b:2(int)
            ├── immutable
-           ├── stats: [rows=333.3333]
+           ├── stats: [rows=333.333]
            ├── cost: 1145.73667
            ├── fd: (1,2)-->(8)
            ├── prune: (1,2,8)
            ├── interesting orderings: (+1)
            ├── select
            │    ├── columns: a:1(int) b:2(int)
            │    ├── immutable
-           │    ├── stats: [rows=333.3333]
+           │    ├── stats: [rows=333.333]
            │    ├── cost: 1139.05
            │    ├── interesting orderings: (+1)
            │    ├── scan tc
@@ -1921,7 +1921,7 @@ EXPLAIN (OPT, TYPES, REDACT) SELECT * FROM tc WHERE a + 2 * b > 1 ORDER BY a*b
 distribute
  ├── columns: a:1(int) b:2(int)  [hidden: column8:8(int)]
  ├── immutable
- ├── stats: [rows=333.3333]
+ ├── stats: [rows=333.333]
  ├── cost: 1418.32951
  ├── fd: (1,2)-->(8)
  ├── ordering: +8
@@ -1932,7 +1932,7 @@ distribute
  └── sort
       ├── columns: a:1(int) b:2(int) column8:8(int)
       ├── immutable
-      ├── stats: [rows=333.3333]
+      ├── stats: [rows=333.333]
       ├── cost: 1218.30951
       ├── fd: (1,2)-->(8)
       ├── ordering: +8
@@ -1941,15 +1941,15 @@ distribute
       └── project
            ├── columns: column8:8(int) a:1(int) b:2(int)
            ├── immutable
-           ├── stats: [rows=333.3333]
+           ├── stats: [rows=333.333]
            ├── cost: 1145.73667
            ├── fd: (1,2)-->(8)
            ├── prune: (1,2,8)
            ├── interesting orderings: (+1)
            ├── select
            │    ├── columns: a:1(int) b:2(int)
            │    ├── immutable
-           │    ├── stats: [rows=333.3333]
+           │    ├── stats: [rows=333.333]
            │    ├── cost: 1139.05
            │    ├── interesting orderings: (+1)
            │    ├── scan tc