planner: allow semi_join_rewrite hint to work for in-subquery (#58830)

close #58829

Author: qw4990

pkg/planner/core/casetest/physicalplantest/testdata/plan_suite_out.json

@@ -3179,7 +3179,7 @@
           "TableReader 10000.00 root  data:TableFullScan",
           "└─TableFullScan 10000.00 cop[tikv] table:t keep order:false, stats:pseudo"
         ],
-        "Warning": "[planner:1815]The SEMI_JOIN_REWRITE hint is not used correctly, maybe it's not in a subquery or the subquery is not EXISTS clause."
+        "Warning": "[planner:1815]The SEMI_JOIN_REWRITE hint is not used correctly, maybe it's not in a subquery or the subquery is not IN/EXISTS clause."
       },
       {
         "SQL": "select * from t where a > (select /*+ SEMI_JOIN_REWRITE() */ min(b) from t t1 where t1.c = t.c)",
@@ -3195,7 +3195,7 @@
           "  └─Selection 9980.01 cop[tikv]  not(isnull(test.t.a)), not(isnull(test.t.c))",
           "    └─TableFullScan 10000.00 cop[tikv] table:t keep order:false, stats:pseudo"
         ],
-        "Warning": "[planner:1815]The SEMI_JOIN_REWRITE hint is not used correctly, maybe it's not in a subquery or the subquery is not EXISTS clause."
+        "Warning": "[planner:1815]The SEMI_JOIN_REWRITE hint is not used correctly, maybe it's not in a subquery or the subquery is not IN/EXISTS clause."
       },
       {
         "SQL": "select * from t where exists (select /*+ SEMI_JOIN_REWRITE() */ 1 from t t1 where t1.a=t.a)",

pkg/planner/core/expression_rewriter.go

@@ -1256,7 +1256,8 @@ func (er *expressionRewriter) handleInSubquery(ctx context.Context, planCtx *exp
 		}
 		planCtx.plan = join
 	} else {
-		planCtx.plan, er.err = planCtx.builder.buildSemiApply(planCtx.plan, np, expression.SplitCNFItems(checkCondition), asScalar, v.Not, false, noDecorrelate)
+		semiRewrite := hintFlags&hint.HintFlagSemiJoinRewrite > 0
+		planCtx.plan, er.err = planCtx.builder.buildSemiApply(planCtx.plan, np, expression.SplitCNFItems(checkCondition), asScalar, v.Not, semiRewrite, noDecorrelate)
 		if er.err != nil {
 			return v, true
 		}

pkg/planner/core/integration_test.go

@@ -2321,3 +2321,15 @@ func TestNestedVirtualGeneratedColumnUpdate(t *testing.T) {
 	tk.MustExec("UPDATE test1 SET col7 = '{\"col10\":\"DDDDD\",\"col9\":[\"abcdefg\"]}';\n")
 	tk.MustExec("DELETE FROM test1 WHERE col1 < 0;\n")
 }
+
+func TestIssue58829(t *testing.T) {
+	store := testkit.CreateMockStore(t)
+	tk := testkit.NewTestKit(t, store)
+	tk.MustExec("use test")
+
+	tk.MustExec(`create table t1 (id varchar(64) not null,  key(id))`)
+	tk.MustExec(`create table t2 (id bigint(20), k int)`)
+
+	// the semi_join_rewrite hint can convert the semi-join to inner-join and finally allow the optimizer to choose the IndexJoin
+	tk.MustHavePlan(`delete from t1 where t1.id in (select /*+ semi_join_rewrite() */ cast(id as char) from t2 where k=1)`, "IndexHashJoin")
+}

pkg/planner/core/logical_plan_builder.go

@@ -3645,6 +3645,7 @@ func (b *PlanBuilder) pushTableHints(hints []*ast.TableOptimizerHint, currentLev
 	warnHandler := b.ctx.GetSessionVars().StmtCtx
 	planHints, subQueryHintFlags, err := h.ParsePlanHints(hints, currentLevel, currentDB,
 		b.hintProcessor, b.ctx.GetSessionVars().StmtCtx.StraightJoinOrder,
+		b.subQueryCtx == handlingInSubquery,
 		b.subQueryCtx == handlingExistsSubquery, b.subQueryCtx == notHandlingSubquery, warnHandler)
 	if err != nil {
 		return

pkg/util/hint/hint.go

@@ -728,7 +728,7 @@ func (*PlanHints) MatchTableName(tables []*HintedTable, hintTables []HintedTable
 func ParsePlanHints(hints []*ast.TableOptimizerHint,
 	currentLevel int, currentDB string,
 	hintProcessor *QBHintHandler, straightJoinOrder bool,
-	handlingExistsSubquery, notHandlingSubquery bool,
+	handlingInSubquery, handlingExistsSubquery, notHandlingSubquery bool,
 	warnHandler hintWarnHandler) (p *PlanHints, subQueryHintFlags uint64, err error) {
 	var (
 		sortMergeTables, inljTables, inlhjTables, inlmjTables, hashJoinTables, bcTables []HintedTable
@@ -872,8 +872,8 @@ func ParsePlanHints(hints []*ast.TableOptimizerHint,
 			}
 			leadingHintCnt++
 		case HintSemiJoinRewrite:
-			if !handlingExistsSubquery {
-				warnHandler.SetHintWarning("The SEMI_JOIN_REWRITE hint is not used correctly, maybe it's not in a subquery or the subquery is not EXISTS clause.")
+			if !handlingExistsSubquery && !handlingInSubquery {
+				warnHandler.SetHintWarning("The SEMI_JOIN_REWRITE hint is not used correctly, maybe it's not in a subquery or the subquery is not IN/EXISTS clause.")
 				continue
 			}
 			subQueryHintFlags |= HintFlagSemiJoinRewrite