opt: don't double count OR selectivity for joins

Previously, an `OR` expression with tight constraints would have its
selecitivy applied in `applyFilters` as expected, without incrementing
`numUnappliedConjuncts`. However, joins call into
`selectivityFromOredEquivalencies`, which would then increment
`numUnappliedConjuncts` for that `OR` if the disjuncts weren't all
conjunctions of equalities. This caused an additional factor of `1/3`
(`memo.unknownFilterSelectivity`) to be applied to the join row count
estimate.

This commit modifies `selectivityFromOredEquivalencies` to avoid
incrementing `numUnappliedConjuncts` for `OR` conditions with tight
constraints. This prevents the double-counting behavior. This commit
also removes a few `FiltersItem` copies from loops.

Fixes cockroachdb#88455

Release note: None

pkg/sql/opt/memo/statistics_builder.go

@@ -4181,7 +4181,11 @@ func (sb *statisticsBuilder) selectivityFromOredEquivalencies(
 	var conjunctSelectivity props.Selectivity
 
 	for f := 0; f < len(h.filters); f++ {
-		disjunction := h.filters[f]
+		disjunction := &h.filters[f]
+		if disjunction.ScalarProps().TightConstraints {
+			// applyFilters will have already handled this filter.
+			continue
+		}
 		var disjuncts []opt.ScalarExpr
 		if orExpr, ok := disjunction.Condition.(*OrExpr); !ok {
 			continue
@@ -4638,7 +4642,7 @@ func (sb *statisticsBuilder) buildStatsFromCheckConstraints(
 	filters := *constraints.(*FiltersExpr)
 	// For each ANDed check constraint...
 	for i := 0; i < len(filters); i++ {
-		filter := filters[i]
+		filter := &filters[i]
 		// This must be some type of comparison operation, or an OR or AND
 		// expression. These operations have at least 2 children.
 		if filter.Condition.ChildCount() < 2 {

pkg/sql/opt/memo/testdata/stats/join

@@ -2040,3 +2040,58 @@ project
  │    └── 82 [type=int]
  └── projections
       └── '1971-10-24' [as=date:19, type=date]
+
+# Regression test for #88455 - don't double-count selectivity for OR expressions
+# with tight constraints in join ON conditions.
+exec-ddl
+CREATE TABLE t0_88455 (c0 INT);
+----
+
+exec-ddl
+CREATE TABLE t1_88455 (c0 INT);
+----
+
+exec-ddl
+ALTER TABLE t0_88455 INJECT STATISTICS '[
+  {
+    "columns": [
+    "c0"
+    ],
+    "created_at": "2022-08-09 09:00:00.00000",
+    "distinct_count": 13,
+    "name": "__auto__",
+    "null_count": 0,
+    "row_count": 13
+  }
+]'
+----
+
+exec-ddl
+ALTER TABLE t1_88455 INJECT STATISTICS '[
+  {
+    "columns": [
+    "c0"
+    ],
+    "created_at": "2022-08-09 09:00:00.00000",
+    "distinct_count": 5,
+    "name": "__auto__",
+    "null_count": 0,
+    "row_count": 5
+  }
+]'
+----
+
+opt format=show-stats
+SELECT * FROM t0_88455 LEFT OUTER JOIN t1_88455 ON t0_88455.c0<1 OR t0_88455.c0>1;
+----
+left-join (cross)
+ ├── columns: c0:1 c0:5
+ ├── stats: [rows=21.66667]
+ ├── scan t0_88455
+ │    ├── columns: t0_88455.c0:1
+ │    └── stats: [rows=13, distinct(1)=13, null(1)=0]
+ ├── scan t1_88455
+ │    ├── columns: t1_88455.c0:5
+ │    └── stats: [rows=5]
+ └── filters
+      └── (t0_88455.c0:1 < 1) OR (t0_88455.c0:1 > 1) [outer=(1), constraints=(/1: (/NULL - /0] [/2 - ]; tight)]

pkg/sql/opt/xform/general_funcs.go

@@ -682,7 +682,7 @@ func (c *CustomFuncs) numAllowedValues(
 	filters := *constraints.(*memo.FiltersExpr)
 	// For each ANDed check constraint...
 	for i := 0; i < len(filters); i++ {
-		filter := filters[i]
+		filter := &filters[i]
 		// This must be some type of comparison operation, or an OR or AND
 		// expression. These operations have at least 2 children.
 		if filter.Condition.ChildCount() < 2 {