opt,sql: use paired-joins with non-covering indexes for left joins

This is done when the left outer/semi/anti join can use a
lookup join. Prior to this, when the non-covering index
could not fully evaluate the filter for left join we could
not generate a lookup join.

With this change:
- Left outer join becomes a pair of two left outer joins.
- Left semi join is a pair of inner join followed by left
  semi join.
- Left anti join is a pair of left outer join followed by
  left anti join.

Informs #55452

Release note (performance improvement): The optimizer can now
generate lookup joins in certain cases for non-covering
indexes, when performing a left outer/semi/anti join.

pkg/sql/distsql_physical_planner.go

@@ -2228,14 +2228,18 @@ func (dsp *DistSQLPlanner) createPlanForLookupJoin(
 	}
 
 	joinReaderSpec := execinfrapb.JoinReaderSpec{
-		Table:                    *n.table.desc.TableDesc(),
-		Type:                     n.joinType,
-		LockingStrength:          n.table.lockingStrength,
-		LockingWaitPolicy:        n.table.lockingWaitPolicy,
-		MaintainOrdering:         len(n.reqOrdering) > 0,
-		HasSystemColumns:         n.table.containsSystemColumns,
-		LeftJoinWithPairedJoiner: n.isSecondJoinInPairedJoiner,
-		LookupBatchBytesLimit:    dsp.distSQLSrv.TestingKnobs.JoinReaderBatchBytesLimit,
+		Table:             *n.table.desc.TableDesc(),
+		Type:              n.joinType,
+		LockingStrength:   n.table.lockingStrength,
+		LockingWaitPolicy: n.table.lockingWaitPolicy,
+		// TODO(sumeer): specifying ordering here using isFirstJoinInPairedJoiner
+		// is late in the sense that the cost of this has not been taken into
+		// account. Make this decision earlier in CustomFuncs.GenerateLookupJoins.
+		MaintainOrdering:                  len(n.reqOrdering) > 0 || n.isFirstJoinInPairedJoiner,
+		HasSystemColumns:                  n.table.containsSystemColumns,
+		LeftJoinWithPairedJoiner:          n.isSecondJoinInPairedJoiner,
+		OutputGroupContinuationForLeftRow: n.isFirstJoinInPairedJoiner,
+		LookupBatchBytesLimit:             dsp.distSQLSrv.TestingKnobs.JoinReaderBatchBytesLimit,
 	}
 	joinReaderSpec.IndexIdx, err = getIndexIdx(n.table.index, n.table.desc)
 	if err != nil {
@@ -2251,7 +2255,7 @@ func (dsp *DistSQLPlanner) createPlanForLookupJoin(
 	joinReaderSpec.LookupColumnsAreKey = n.eqColsAreKey
 
 	numInputNodeCols, planToStreamColMap, post, types :=
-		mappingHelperForLookupJoins(plan, n.input, n.table, false /* addContinuationCol */)
+		mappingHelperForLookupJoins(plan, n.input, n.table, n.isFirstJoinInPairedJoiner)
 
 	// Set the lookup condition.
 	var indexVarMap []int

pkg/sql/distsql_spec_exec_factory.go

@@ -658,6 +658,7 @@ func (e *distSQLSpecExecFactory) ConstructLookupJoin(
 	remoteLookupExpr tree.TypedExpr,
 	lookupCols exec.TableColumnOrdinalSet,
 	onCond tree.TypedExpr,
+	isFirstJoinInPairedJoiner bool,
 	isSecondJoinInPairedJoiner bool,
 	reqOrdering exec.OutputOrdering,
 	locking *tree.LockingItem,

pkg/sql/execinfrapb/flow_diagram.go

@@ -214,6 +214,9 @@ func (jr *JoinReaderSpec) summary() (string, []string) {
 	if jr.LeftJoinWithPairedJoiner {
 		details = append(details, "second join in paired-join")
 	}
+	if jr.OutputGroupContinuationForLeftRow {
+		details = append(details, "first join in paired-join")
+	}
 	return "JoinReader", details
 }
 

pkg/sql/logictest/testdata/logic_test/lookup_join

@@ -378,7 +378,7 @@ SELECT small.c, large.c FROM small LEFT JOIN large ON small.c = large.b AND larg
 27  NULL
 30  NULL
 
-## Left join with ON filter on non-covering index
+## Left join with ON filter on non-covering index. Will execute as paired-joins.
 query II rowsort
 SELECT small.c, large.d FROM small LEFT JOIN large ON small.c = large.b AND large.d < 30
 ----
@@ -393,6 +393,26 @@ SELECT small.c, large.d FROM small LEFT JOIN large ON small.c = large.b AND larg
 27  NULL
 30  NULL
 
+## Left semi join with ON filter on non-covering index. Will execute as paired-joins.
+query I rowsort
+SELECT small.c FROM small WHERE EXISTS(SELECT 1 FROM large WHERE small.c = large.b AND large.d < 30)
+----
+6
+12
+
+## Left anti join with ON filter on non-covering index. Will execute as paired-joins.
+query I rowsort
+SELECT small.c FROM small WHERE NOT EXISTS(SELECT 1 FROM large WHERE small.c = large.b AND large.d < 30)
+----
+3
+9
+15
+18
+21
+24
+27
+30
+
 ###########################################################
 #  LOOKUP JOINS ON IMPLICIT INDEX KEY COLUMNS             #
 #  https://github.com/cockroachdb/cockroach/issues/31777  #

pkg/sql/lookup_join.go

@@ -56,13 +56,17 @@ type lookupJoinNode struct {
 	remoteLookupExpr tree.TypedExpr
 
 	// columns are the produced columns, namely the input columns and (unless the
-	// join type is semi or anti join) the columns in the table scanNode.
+	// join type is semi or anti join) the columns in the table scanNode. It
+	// includes an additional continuation column when IsFirstJoinInPairedJoin
+	// is true.
 	columns colinfo.ResultColumns
 
 	// onCond is any ON condition to be used in conjunction with the implicit
 	// equality condition on eqCols or the conditions in lookupExpr.
 	onCond tree.TypedExpr
 
+	// At most one of is{First,Second}JoinInPairedJoiner can be true.
+	isFirstJoinInPairedJoiner  bool
 	isSecondJoinInPairedJoiner bool
 
 	reqOrdering ReqOrdering

pkg/sql/opt/exec/execbuilder/relational.go

@@ -1752,19 +1752,40 @@ func (b *Builder) buildLookupJoin(join *memo.LookupJoinExpr) (execPlan, error) {
 
 	inputCols := join.Input.Relational().OutputCols
 	lookupCols := join.Cols.Difference(inputCols)
+	if join.IsFirstJoinInPairedJoiner {
+		lookupCols.Remove(join.ContinuationCol)
+	}
 
 	lookupOrdinals, lookupColMap := b.getColumns(lookupCols, join.Table)
-	allCols := joinOutputMap(input.outputCols, lookupColMap)
-
+	// allExprCols are the columns used in expressions evaluated by this join.
+	allExprCols := joinOutputMap(input.outputCols, lookupColMap)
+	allCols := allExprCols
+	if join.IsFirstJoinInPairedJoiner {
+		if join.JoinType != opt.InnerJoinOp && join.JoinType != opt.LeftJoinOp {
+			return execPlan{}, errors.AssertionFailedf(
+				"first join in paired joiner must be an inner or left join. found %s",
+				join.JoinType.String(),
+			)
+		}
+		// allCols needs to include the continuation column since it will be
+		// in the result output by this join.
+		allCols = allExprCols.Copy()
+		maxValue, ok := allCols.MaxValue()
+		if !ok {
+			return execPlan{}, errors.AssertionFailedf("allCols should not be empty")
+		}
+		// Assign the continuation column the next unused value in the map.
+		allCols.Set(int(join.ContinuationCol), maxValue+1)
+	}
 	res := execPlan{outputCols: allCols}
 	if join.JoinType == opt.SemiJoinOp || join.JoinType == opt.AntiJoinOp {
 		// For semi and anti join, only the left columns are output.
 		res.outputCols = input.outputCols
 	}
 
 	ctx := buildScalarCtx{
-		ivh:     tree.MakeIndexedVarHelper(nil /* container */, allCols.Len()),
-		ivarMap: allCols,
+		ivh:     tree.MakeIndexedVarHelper(nil /* container */, allExprCols.Len()),
+		ivarMap: allExprCols,
 	}
 	var lookupExpr, remoteLookupExpr tree.TypedExpr
 	if len(join.LookupExpr) > 0 {
@@ -1809,6 +1830,7 @@ func (b *Builder) buildLookupJoin(join *memo.LookupJoinExpr) (execPlan, error) {
 		remoteLookupExpr,
 		lookupOrdinals,
 		onExpr,
+		join.IsFirstJoinInPairedJoiner,
 		join.IsSecondJoinInPairedJoiner,
 		res.reqOrdering(join),
 		locking,

pkg/sql/opt/exec/execbuilder/testdata/lookup_join

@@ -706,8 +706,6 @@ vectorized: true
           spans: FULL SCAN
 
 # Left join with ON filter on non-covering index
-# TODO(radu): this doesn't use lookup join yet, the current rules don't cover
-# left join with ON condition on columns that are not covered by the index.
 query T
 EXPLAIN (VERBOSE) SELECT small.c, large.d FROM small LEFT JOIN large ON small.c = large.b AND large.d < 30
 ----
@@ -718,27 +716,92 @@ vectorized: true
 │ columns: (c, d)
 │ estimated row count: 336
 │
-└── • hash join (right outer)
-    │ columns: (b, d, c)
+└── • project
+    │ columns: (c, b, d)
     │ estimated row count: 336
-    │ equality: (b) = (c)
     │
-    ├── • filter
-    │   │ columns: (b, d)
-    │   │ estimated row count: 3,303
-    │   │ filter: d < 30
-    │   │
-    │   └── • scan
-    │         columns: (b, d)
-    │         estimated row count: 10,000 (100% of the table; stats collected <hidden> ago)
-    │         table: large@large_pkey
-    │         spans: FULL SCAN
+    └── • lookup join (left outer)
+        │ columns: (c, a, b, cont, d)
+        │ table: large@large_pkey
+        │ equality: (a, b) = (a,b)
+        │ equality cols are key
+        │ pred: d < 30
+        │
+        └── • lookup join (left outer)
+            │ columns: (c, a, b, cont)
+            │ estimated row count: 1,000
+            │ table: large@bc
+            │ equality: (c) = (b)
+            │
+            └── • scan
+                  columns: (c)
+                  estimated row count: 100 (100% of the table; stats collected <hidden> ago)
+                  table: small@small_pkey
+                  spans: FULL SCAN
+
+# Left semi-join with ON filter on non-covering index
+query T
+EXPLAIN (VERBOSE) SELECT small.c FROM small WHERE EXISTS(SELECT 1 FROM large WHERE small.c = large.b AND large.d < 30)
+----
+distribution: full
+vectorized: true
+·
+• project
+│ columns: (c)
+│ estimated row count: 100
+│
+└── • lookup join (semi)
+    │ columns: (c, a, b, cont)
+    │ table: large@large_pkey
+    │ equality: (a, b) = (a,b)
+    │ equality cols are key
+    │ pred: d < 30
     │
-    └── • scan
-          columns: (c)
-          estimated row count: 100 (100% of the table; stats collected <hidden> ago)
-          table: small@small_pkey
-          spans: FULL SCAN
+    └── • lookup join (inner)
+        │ columns: (c, a, b, cont)
+        │ estimated row count: 990
+        │ table: large@bc
+        │ equality: (c) = (b)
+        │
+        └── • scan
+              columns: (c)
+              estimated row count: 100 (100% of the table; stats collected <hidden> ago)
+              table: small@small_pkey
+              spans: FULL SCAN
+
+# Left anti-join with ON filter on non-covering index
+query T
+EXPLAIN (VERBOSE) SELECT small.c FROM small WHERE NOT EXISTS(SELECT 1 FROM large WHERE small.c = large.b AND large.d < small.a)
+----
+distribution: full
+vectorized: true
+·
+• project
+│ columns: (c)
+│ estimated row count: 67
+│
+└── • project
+    │ columns: (a, c)
+    │ estimated row count: 67
+    │
+    └── • lookup join (anti)
+        │ columns: (a, c, a, b, cont)
+        │ table: large@large_pkey
+        │ equality: (a, b) = (a,b)
+        │ equality cols are key
+        │ pred: d < a
+        │
+        └── • lookup join (left outer)
+            │ columns: (a, c, a, b, cont)
+            │ estimated row count: 1,000
+            │ table: large@bc
+            │ equality: (c) = (b)
+            │
+            └── • scan
+                  columns: (a, c)
+                  estimated row count: 100 (100% of the table; stats collected <hidden> ago)
+                  table: small@small_pkey
+                  spans: FULL SCAN
 
 ###########################################################
 #  LOOKUP JOINS ON IMPLICIT INDEX KEY COLUMNS             #

pkg/sql/opt/exec/explain/result_columns.go

@@ -81,7 +81,12 @@ func getResultColumns(
 
 	case lookupJoinOp:
 		a := args.(*lookupJoinArgs)
-		return joinColumns(a.JoinType, inputs[0], tableColumns(a.Table, a.LookupCols)), nil
+		cols := joinColumns(a.JoinType, inputs[0], tableColumns(a.Table, a.LookupCols))
+		// The following matches the behavior of execFactory.ConstructLookupJoin.
+		if a.IsFirstJoinInPairedJoiner {
+			cols = append(cols, colinfo.ResultColumn{Name: "cont", Typ: types.Bool})
+		}
+		return cols, nil
 
 	case ordinalityOp:
 		return appendColumns(inputs[0], colinfo.ResultColumn{

pkg/sql/opt/exec/factory.opt

@@ -280,6 +280,7 @@ define LookupJoin {
     RemoteLookupExpr tree.TypedExpr
     LookupCols exec.TableColumnOrdinalSet
     OnCond tree.TypedExpr
+    IsFirstJoinInPairedJoiner bool
     IsSecondJoinInPairedJoiner bool
     ReqOrdering exec.OutputOrdering
     Locking *tree.LockingItem

pkg/sql/opt/memo/check_expr.go

@@ -235,14 +235,19 @@ func (m *Memo) CheckExpr(e opt.Expr) {
 			panic(errors.AssertionFailedf("lookup join with columns that are not required"))
 		}
 		if t.IsSecondJoinInPairedJoiner {
-			ij, ok := t.Input.(*InvertedJoinExpr)
-			if !ok {
-				panic(errors.AssertionFailedf(
-					"lookup paired-join is paired with %T instead of inverted join", t.Input))
-			}
-			if !ij.IsFirstJoinInPairedJoiner {
-				panic(errors.AssertionFailedf(
-					"lookup paired-join is paired with inverted join that thinks it is unpaired"))
+			switch firstJoin := t.Input.(type) {
+			case *InvertedJoinExpr:
+				if !firstJoin.IsFirstJoinInPairedJoiner {
+					panic(errors.AssertionFailedf(
+						"lookup paired-join is paired with inverted join that thinks it is unpaired"))
+				}
+			case *LookupJoinExpr:
+				if !firstJoin.IsFirstJoinInPairedJoiner {
+					panic(errors.AssertionFailedf(
+						"lookup paired-join is paired with lookup join that thinks it is unpaired"))
+				}
+			default:
+				panic(errors.AssertionFailedf("lookup paired-join is paired with %T", t.Input))
 			}
 		}
 

pkg/sql/opt/ops/relational.opt

@@ -402,10 +402,20 @@ define LookupJoinPrivate {
     # table (and thus each left row matches with at most one table row).
     LookupColsAreTableKey bool
 
+    # At most one of Is{First,Second}JoinInPairedJoiner can be true.
+    #
+    # IsFirstJoinInPairedJoiner is true if this is the first join of a
+    # paired-joiner used for left joins.
+    IsFirstJoinInPairedJoiner bool
+
     # IsSecondJoinInPairedJoiner is true if this is the second join of a
     # paired-joiner used for left joins.
     IsSecondJoinInPairedJoiner bool
 
+    # ContinuationCol is the column ID of the continuation column when
+    # IsFirstJoinInPairedJoiner is true.
+    ContinuationCol ColumnID
+
     # LocalityOptimized is true if this lookup join is part of a locality
     # optimized search strategy. For semi, inner, and left joins, this means
     # that RemoteLookupExpr will be non-nil. See comments above that field for