*: code preparations for supporting build range for like function on …

…new collation columns (pingcap#48972)

ref pingcap#48181

cmd/explaintest/r/explain_easy.result

@@ -60,7 +60,7 @@ explain format = 'brief' delete from t1 where t1.c2 = 1;
 id	estRows	task	access object	operator info
 Delete	N/A	root		N/A
 └─SelectLock	10.00	root		for update 0
-  └─IndexLookUp	10.00	root		
+  └─IndexLookUp	10.00	root
     ├─IndexRangeScan(Build)	10.00	cop[tikv]	table:t1, index:c2(c2)	range:[1,1], keep order:false, stats:pseudo
     └─TableRowIDScan(Probe)	10.00	cop[tikv]	table:t1	keep order:false, stats:pseudo
 explain format = 'brief' select count(b.c2) from t1 a, t2 b where a.c1 = b.c2 group by a.c1;
@@ -174,11 +174,11 @@ Union	26000.00	root
 │ └─Union	16000.00	root		
 │   ├─StreamAgg	8000.00	root		group by:test.t2.c1, funcs:firstrow(test.t2.c1)->Column#12, funcs:firstrow(test.t2.c1)->Column#10
 │   │ └─IndexReader	8000.00	root		index:StreamAgg
-│   │   └─StreamAgg	8000.00	cop[tikv]		group by:test.t2.c1, 
+│   │   └─StreamAgg	8000.00	cop[tikv]		group by:test.t2.c1,
 │   │     └─IndexFullScan	10000.00	cop[tikv]	table:t2, index:c1(c1)	keep order:true, stats:pseudo
 │   └─StreamAgg	8000.00	root		group by:test.t2.c1, funcs:firstrow(test.t2.c1)->Column#12, funcs:firstrow(test.t2.c1)->Column#10
 │     └─IndexReader	8000.00	root		index:StreamAgg
-│       └─StreamAgg	8000.00	cop[tikv]		group by:test.t2.c1, 
+│       └─StreamAgg	8000.00	cop[tikv]		group by:test.t2.c1,
 │         └─IndexFullScan	10000.00	cop[tikv]	table:t2, index:c1(c1)	keep order:true, stats:pseudo
 └─IndexReader	10000.00	root		index:IndexFullScan
   └─IndexFullScan	10000.00	cop[tikv]	table:t2, index:c1(c1)	keep order:false, stats:pseudo
@@ -188,15 +188,15 @@ HashAgg	24000.00	root		group by:Column#10, funcs:firstrow(Column#11)->Column#10
 └─Union	24000.00	root		
   ├─StreamAgg	8000.00	root		group by:test.t2.c1, funcs:firstrow(test.t2.c1)->Column#11, funcs:firstrow(test.t2.c1)->Column#10
   │ └─IndexReader	8000.00	root		index:StreamAgg
-  │   └─StreamAgg	8000.00	cop[tikv]		group by:test.t2.c1, 
+  │   └─StreamAgg	8000.00	cop[tikv]		group by:test.t2.c1,
   │     └─IndexFullScan	10000.00	cop[tikv]	table:t2, index:c1(c1)	keep order:true, stats:pseudo
   ├─StreamAgg	8000.00	root		group by:test.t2.c1, funcs:firstrow(test.t2.c1)->Column#11, funcs:firstrow(test.t2.c1)->Column#10
   │ └─IndexReader	8000.00	root		index:StreamAgg
-  │   └─StreamAgg	8000.00	cop[tikv]		group by:test.t2.c1, 
+  │   └─StreamAgg	8000.00	cop[tikv]		group by:test.t2.c1,
   │     └─IndexFullScan	10000.00	cop[tikv]	table:t2, index:c1(c1)	keep order:true, stats:pseudo
   └─StreamAgg	8000.00	root		group by:test.t2.c1, funcs:firstrow(test.t2.c1)->Column#11, funcs:firstrow(test.t2.c1)->Column#10
     └─IndexReader	8000.00	root		index:StreamAgg
-      └─StreamAgg	8000.00	cop[tikv]		group by:test.t2.c1, 
+      └─StreamAgg	8000.00	cop[tikv]		group by:test.t2.c1,
         └─IndexFullScan	10000.00	cop[tikv]	table:t2, index:c1(c1)	keep order:true, stats:pseudo
 select * from information_schema.tidb_indexes where table_name='t4';
 TABLE_SCHEMA	TABLE_NAME	NON_UNIQUE	KEY_NAME	SEQ_IN_INDEX	COLUMN_NAME	SUB_PART	INDEX_COMMENT	Expression	INDEX_ID	IS_VISIBLE	CLUSTERED
@@ -797,14 +797,14 @@ create table t(a binary(16) not null, b varchar(2) default null, c varchar(100)
 explain format = 'brief' select * from t where a=x'FA34E1093CB428485734E3917F000000' and b='xb';
 id	estRows	task	access object	operator info
 IndexLookUp	0.10	root		
-├─IndexRangeScan(Build)	0.10	cop[tikv]	table:t, index:a(a, b)	range:[0xFA34E1093CB428485734E3917F000000 "xb",0xFA34E1093CB428485734E3917F000000 "xb"], keep order:false, stats:pseudo
+├─IndexRangeScan(Build)	0.10	cop[tikv]	table:t, index:a(a, b)	range:["\xfa4\xe1\t<\xb4(HW4\xe3\x91\x7f\x00\x00\x00" "xb","\xfa4\xe1\t<\xb4(HW4\xe3\x91\x7f\x00\x00\x00" "xb"], keep order:false, stats:pseudo
 └─TableRowIDScan(Probe)	0.10	cop[tikv]	table:t	keep order:false, stats:pseudo
 explain format = 'brief' update t set c = 'ssss' where a=x'FA34E1093CB428485734E3917F000000' and b='xb';
 id	estRows	task	access object	operator info
 Update	N/A	root		N/A
 └─SelectLock	0.10	root		for update 0
-  └─IndexLookUp	0.10	root		
-    ├─IndexRangeScan(Build)	0.10	cop[tikv]	table:t, index:a(a, b)	range:[0xFA34E1093CB428485734E3917F000000 "xb",0xFA34E1093CB428485734E3917F000000 "xb"], keep order:false, stats:pseudo
+  └─IndexLookUp	0.10	root
+    ├─IndexRangeScan(Build)	0.10	cop[tikv]	table:t, index:a(a, b)	range:["\xfa4\xe1\t<\xb4(HW4\xe3\x91\x7f\x00\x00\x00" "xb","\xfa4\xe1\t<\xb4(HW4\xe3\x91\x7f\x00\x00\x00" "xb"], keep order:false, stats:pseudo
     └─TableRowIDScan(Probe)	0.10	cop[tikv]	table:t	keep order:false, stats:pseudo
 drop table if exists t;
 create table t(a int, b int);

executor/point_get.go

@@ -540,11 +540,13 @@ func EncodeUniqueIndexValuesForKey(ctx sessionctx.Context, tblInfo *model.TableI
 		colInfo := tblInfo.Columns[idxInfo.Columns[i].Offset]
 		// table.CastValue will append 0x0 if the string value's length is smaller than the BINARY column's length.
 		// So we don't use CastValue for string value for now.
-		// TODO: merge two if branch.
+		// TODO: The first if branch should have been removed, because the functionality of set the collation of the datum
+		// have been moved to util/ranger (normal path) and getNameValuePairs/getPointGetValue (fast path). But this change
+		// will be cherry-picked to a hotfix, so we choose to be a bit conservative and keep this for now.
 		if colInfo.GetType() == mysql.TypeString || colInfo.GetType() == mysql.TypeVarString || colInfo.GetType() == mysql.TypeVarchar {
 			var str string
 			str, err = idxVals[i].ToString()
-			idxVals[i].SetString(str, colInfo.FieldType.GetCollate())
+			idxVals[i].SetString(str, idxVals[i].Collation())
 		} else if colInfo.GetType() == mysql.TypeEnum && (idxVals[i].Kind() == types.KindString || idxVals[i].Kind() == types.KindBytes || idxVals[i].Kind() == types.KindBinaryLiteral) {
 			var str string
 			var e types.Enum

planner/core/point_get_plan.go

@@ -1434,7 +1434,15 @@ func getNameValuePairs(ctx sessionctx.Context, tbl *model.TableInfo, tblName mod
 		col := model.FindColumnInfo(tbl.Cols(), colName.Name.Name.L)
 		if col == nil { // Handling the case when the column is _tidb_rowid.
 			return append(nvPairs, nameValuePair{colName: colName.Name.Name.L, colFieldType: types.NewFieldType(mysql.TypeLonglong), value: d, con: con}), false
-		} else if col.GetType() == mysql.TypeString && col.GetCollate() == charset.CollationBin { // This type we needn't to pad `\0` in here.
+		}
+
+		// As in buildFromBinOp in util/ranger, when we build key from the expression to do range scan or point get on
+		// a string column, we should set the collation of the string datum to collation of the column.
+		if col.FieldType.EvalType() == types.ETString && (d.Kind() == types.KindString || d.Kind() == types.KindBinaryLiteral) {
+			d.SetString(d.GetString(), col.FieldType.GetCollate())
+		}
+
+		if col.GetType() == mysql.TypeString && col.GetCollate() == charset.CollationBin { // This type we needn't to pad `\0` in here.
 			return append(nvPairs, nameValuePair{colName: colName.Name.Name.L, colFieldType: &col.FieldType, value: d, con: con}), false
 		}
 		if !checkCanConvertInPointGet(col, d) {
@@ -1464,6 +1472,11 @@ func getPointGetValue(stmtCtx *stmtctx.StatementContext, col *model.ColumnInfo,
 	if !checkCanConvertInPointGet(col, *d) {
 		return nil
 	}
+	// As in buildFromBinOp in util/ranger, when we build key from the expression to do range scan or point get on
+	// a string column, we should set the collation of the string datum to collation of the column.
+	if col.FieldType.EvalType() == types.ETString && (d.Kind() == types.KindString || d.Kind() == types.KindBinaryLiteral) {
+		d.SetString(d.GetString(), col.FieldType.GetCollate())
+	}
 	dVal, err := d.ConvertTo(stmtCtx, &col.FieldType)
 	if err != nil {
 		return nil