add fix dup row

enginetest/queries/script_queries.go

@@ -120,6 +120,26 @@ type ScriptTestAssertion struct {
 // Unlike other engine tests, ScriptTests must be self-contained. No other tables are created outside the definition of
 // the tests.
 var ScriptTests = []ScriptTest{
+{
+// https://github.com/dolthub/dolt/issues/9797
+Name: "EXISTS subquery returns duplicate rows with PRIMARY KEY",
+SetUpScript: []string{
+"CREATE TABLE t(c0 INT, c1 INT, PRIMARY KEY(c0, c1));",
+"INSERT INTO t VALUES (1, 1);",
+"INSERT INTO t VALUES (2, 2);",
+"INSERT INTO t VALUES (2, 3);",
+},
+Assertions: []ScriptTestAssertion{
+{
+Query: "SELECT * FROM t WHERE EXISTS (SELECT 1 FROM t AS x WHERE x.c0 = t.c0);",
+Expected: []sql.Row{
+{1, 1},
+{2, 2},
+{2, 3},
+},
+},
+},
+},
 {
 // https://github.com/dolthub/dolt/issues/9794
 Name: "UPDATE with TRIM function on TEXT column",

sql/analyzer/indexed_joins.go

@@ -472,9 +472,9 @@ func convertSemiToInnerJoin(m *memo.Memo) error {
 }
 
 // join and its commute are a new group
-joinGrp := m.MemoizeInnerJoin(nil, semi.Left, rightGrp, plan.JoinTypeInner, semi.Filter)
+joinGrp := m.MemoizeInnerJoin(nil, semi.Left, rightGrp, plan.JoinTypeSemi, semi.Filter)
 // TODO: can't commute if right SubqueryAlias references outside scope (OuterScopeVisibility/IsLateral)
-m.MemoizeInnerJoin(joinGrp, rightGrp, semi.Left, plan.JoinTypeInner, semi.Filter)
+m.MemoizeInnerJoin(joinGrp, rightGrp, semi.Left, plan.JoinTypeSemi, semi.Filter)
 
 // project belongs to the original group
 leftCols := semi.Left.RelProps.OutputCols()

sql/plan/join.go

@@ -177,7 +177,7 @@ func (i JoinType) IsPartial() bool {
 switch i {
 case JoinTypeSemi, JoinTypeAnti, JoinTypeAntiIncludeNulls, JoinTypeSemiHash,
 JoinTypeAntiHash, JoinTypeAntiHashIncludeNulls, JoinTypeAntiLookup, JoinTypeAntiLookupIncludeNulls,
-JoinTypeSemiLookup:
+JoinTypeSemiLookup, JoinTypeSemiMerge, JoinTypeAntiMerge, JoinTypeAntiMergeIncludeNulls:
 return true
 default:
 return false

sql/plan/subquery.go

@@ -431,6 +431,7 @@ func (s *Subquery) HashMultiple(ctx *sql.Context, row sql.Row) (sql.KeyValueCach
 
 // HasResultRow returns whether the subquery has a result set > 0.
 func (s *Subquery) HasResultRow(ctx *sql.Context, row sql.Row) (bool, error) {
+
 // First check if the query was cached.
 s.cacheMu.Lock()
 cached := s.resultsCached

sql/rowexec/join_iters.go

@@ -367,6 +367,10 @@ func (i *existsIter) Next(ctx *sql.Context) (sql.Row, error) {
 nextState = esIncRight
 }
 case esRet:
+if i.typ.IsSemi() {
+// For semi-joins, after returning a match, move to next left row
+nextState = esIncLeft
+}
 return i.removeParentRow(i.primaryRow.Copy()), nil
 default:
 return nil, fmt.Errorf("invalid exists join state")