forked from pingcap/tidb
-
Notifications
You must be signed in to change notification settings - Fork 0
/
Copy pathexpression.go
1588 lines (1462 loc) · 52.2 KB
/
expression.go
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
649
650
651
652
653
654
655
656
657
658
659
660
661
662
663
664
665
666
667
668
669
670
671
672
673
674
675
676
677
678
679
680
681
682
683
684
685
686
687
688
689
690
691
692
693
694
695
696
697
698
699
700
701
702
703
704
705
706
707
708
709
710
711
712
713
714
715
716
717
718
719
720
721
722
723
724
725
726
727
728
729
730
731
732
733
734
735
736
737
738
739
740
741
742
743
744
745
746
747
748
749
750
751
752
753
754
755
756
757
758
759
760
761
762
763
764
765
766
767
768
769
770
771
772
773
774
775
776
777
778
779
780
781
782
783
784
785
786
787
788
789
790
791
792
793
794
795
796
797
798
799
800
801
802
803
804
805
806
807
808
809
810
811
812
813
814
815
816
817
818
819
820
821
822
823
824
825
826
827
828
829
830
831
832
833
834
835
836
837
838
839
840
841
842
843
844
845
846
847
848
849
850
851
852
853
854
855
856
857
858
859
860
861
862
863
864
865
866
867
868
869
870
871
872
873
874
875
876
877
878
879
880
881
882
883
884
885
886
887
888
889
890
891
892
893
894
895
896
897
898
899
900
901
902
903
904
905
906
907
908
909
910
911
912
913
914
915
916
917
918
919
920
921
922
923
924
925
926
927
928
929
930
931
932
933
934
935
936
937
938
939
940
941
942
943
944
945
946
947
948
949
950
951
952
953
954
955
956
957
958
959
960
961
962
963
964
965
966
967
968
969
970
971
972
973
974
975
976
977
978
979
980
981
982
983
984
985
986
987
988
989
990
991
992
993
994
995
996
997
998
999
1000
// Copyright 2016 PingCAP, Inc.
//
// Licensed under the Apache License, Version 2.0 (the "License");
// you may not use this file except in compliance with the License.
// You may obtain a copy of the License at
//
// http://www.apache.org/licenses/LICENSE-2.0
//
// Unless required by applicable law or agreed to in writing, software
// distributed under the License is distributed on an "AS IS" BASIS,
// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
// See the License for the specific language governing permissions and
// limitations under the License.
package expression
import (
goJSON "encoding/json"
"fmt"
"strconv"
"strings"
"sync"
"sync/atomic"
"github.com/gogo/protobuf/proto"
"github.com/pingcap/errors"
"github.com/pingcap/failpoint"
"github.com/pingcap/tidb/kv"
"github.com/pingcap/tidb/parser/ast"
"github.com/pingcap/tidb/parser/charset"
"github.com/pingcap/tidb/parser/model"
"github.com/pingcap/tidb/parser/mysql"
"github.com/pingcap/tidb/parser/opcode"
"github.com/pingcap/tidb/parser/terror"
"github.com/pingcap/tidb/sessionctx"
"github.com/pingcap/tidb/sessionctx/stmtctx"
"github.com/pingcap/tidb/types"
"github.com/pingcap/tidb/util/chunk"
"github.com/pingcap/tidb/util/generatedexpr"
"github.com/pingcap/tidb/util/logutil"
"github.com/pingcap/tidb/util/size"
"github.com/pingcap/tipb/go-tipb"
"go.uber.org/zap"
)
// These are byte flags used for `HashCode()`.
const (
constantFlag byte = 0
columnFlag byte = 1
scalarFunctionFlag byte = 3
parameterFlag byte = 4
)
// EvalAstExpr evaluates ast expression directly.
// Note: initialized in planner/core
// import expression and planner/core together to use EvalAstExpr
var EvalAstExpr func(sctx sessionctx.Context, expr ast.ExprNode) (types.Datum, error)
// RewriteAstExpr rewrites ast expression directly.
// Note: initialized in planner/core
// import expression and planner/core together to use EvalAstExpr
var RewriteAstExpr func(sctx sessionctx.Context, expr ast.ExprNode, schema *Schema, names types.NameSlice, allowCastArray bool) (Expression, error)
// VecExpr contains all vectorized evaluation methods.
type VecExpr interface {
// Vectorized returns if this expression supports vectorized evaluation.
Vectorized() bool
// VecEvalInt evaluates this expression in a vectorized manner.
VecEvalInt(ctx sessionctx.Context, input *chunk.Chunk, result *chunk.Column) error
// VecEvalReal evaluates this expression in a vectorized manner.
VecEvalReal(ctx sessionctx.Context, input *chunk.Chunk, result *chunk.Column) error
// VecEvalString evaluates this expression in a vectorized manner.
VecEvalString(ctx sessionctx.Context, input *chunk.Chunk, result *chunk.Column) error
// VecEvalDecimal evaluates this expression in a vectorized manner.
VecEvalDecimal(ctx sessionctx.Context, input *chunk.Chunk, result *chunk.Column) error
// VecEvalTime evaluates this expression in a vectorized manner.
VecEvalTime(ctx sessionctx.Context, input *chunk.Chunk, result *chunk.Column) error
// VecEvalDuration evaluates this expression in a vectorized manner.
VecEvalDuration(ctx sessionctx.Context, input *chunk.Chunk, result *chunk.Column) error
// VecEvalJSON evaluates this expression in a vectorized manner.
VecEvalJSON(ctx sessionctx.Context, input *chunk.Chunk, result *chunk.Column) error
}
// ReverseExpr contains all resersed evaluation methods.
type ReverseExpr interface {
// SupportReverseEval checks whether the builtinFunc support reverse evaluation.
SupportReverseEval() bool
// ReverseEval evaluates the only one column value with given function result.
ReverseEval(sc *stmtctx.StatementContext, res types.Datum, rType types.RoundingType) (val types.Datum, err error)
}
// Expression represents all scalar expression in SQL.
type Expression interface {
fmt.Stringer
goJSON.Marshaler
VecExpr
ReverseExpr
CollationInfo
// Eval evaluates an expression through a row.
Eval(row chunk.Row) (types.Datum, error)
// EvalInt returns the int64 representation of expression.
EvalInt(ctx sessionctx.Context, row chunk.Row) (val int64, isNull bool, err error)
// EvalReal returns the float64 representation of expression.
EvalReal(ctx sessionctx.Context, row chunk.Row) (val float64, isNull bool, err error)
// EvalString returns the string representation of expression.
EvalString(ctx sessionctx.Context, row chunk.Row) (val string, isNull bool, err error)
// EvalDecimal returns the decimal representation of expression.
EvalDecimal(ctx sessionctx.Context, row chunk.Row) (val *types.MyDecimal, isNull bool, err error)
// EvalTime returns the DATE/DATETIME/TIMESTAMP representation of expression.
EvalTime(ctx sessionctx.Context, row chunk.Row) (val types.Time, isNull bool, err error)
// EvalDuration returns the duration representation of expression.
EvalDuration(ctx sessionctx.Context, row chunk.Row) (val types.Duration, isNull bool, err error)
// EvalJSON returns the JSON representation of expression.
EvalJSON(ctx sessionctx.Context, row chunk.Row) (val types.BinaryJSON, isNull bool, err error)
// GetType gets the type that the expression returns.
GetType() *types.FieldType
// Clone copies an expression totally.
Clone() Expression
// Equal checks whether two expressions are equal.
Equal(ctx sessionctx.Context, e Expression) bool
// IsCorrelated checks if this expression has correlated key.
IsCorrelated() bool
// ConstItem checks if this expression is constant item, regardless of query evaluation state.
// An expression is constant item if it:
// refers no tables.
// refers no correlated column.
// refers no subqueries that refers any tables.
// refers no non-deterministic functions.
// refers no statement parameters.
// refers no param markers when prepare plan cache is enabled.
ConstItem(sc *stmtctx.StatementContext) bool
// Decorrelate try to decorrelate the expression by schema.
Decorrelate(schema *Schema) Expression
// ResolveIndices resolves indices by the given schema. It will copy the original expression and return the copied one.
ResolveIndices(schema *Schema) (Expression, error)
// resolveIndices is called inside the `ResolveIndices` It will perform on the expression itself.
resolveIndices(schema *Schema) error
// ResolveIndicesByVirtualExpr resolves indices by the given schema in terms of virual expression. It will copy the original expression and return the copied one.
ResolveIndicesByVirtualExpr(schema *Schema) (Expression, bool)
// resolveIndicesByVirtualExpr is called inside the `ResolveIndicesByVirtualExpr` It will perform on the expression itself.
resolveIndicesByVirtualExpr(schema *Schema) bool
// RemapColumn remaps columns with provided mapping and returns new expression
RemapColumn(map[int64]*Column) (Expression, error)
// ExplainInfo returns operator information to be explained.
ExplainInfo() string
// ExplainNormalizedInfo returns operator normalized information for generating digest.
ExplainNormalizedInfo() string
// HashCode creates the hashcode for expression which can be used to identify itself from other expression.
// It generated as the following:
// Constant: ConstantFlag+encoded value
// Column: ColumnFlag+encoded value
// ScalarFunction: SFFlag+encoded function name + encoded arg_1 + encoded arg_2 + ...
HashCode(sc *stmtctx.StatementContext) []byte
// MemoryUsage return the memory usage of Expression
MemoryUsage() int64
}
// CNFExprs stands for a CNF expression.
type CNFExprs []Expression
// Clone clones itself.
func (e CNFExprs) Clone() CNFExprs {
cnf := make(CNFExprs, 0, len(e))
for _, expr := range e {
cnf = append(cnf, expr.Clone())
}
return cnf
}
// Shallow makes a shallow copy of itself.
func (e CNFExprs) Shallow() CNFExprs {
cnf := make(CNFExprs, 0, len(e))
cnf = append(cnf, e...)
return cnf
}
func isColumnInOperand(c *Column) bool {
return c.InOperand
}
// IsEQCondFromIn checks if an expression is equal condition converted from `[not] in (subq)`.
func IsEQCondFromIn(expr Expression) bool {
sf, ok := expr.(*ScalarFunction)
if !ok || sf.FuncName.L != ast.EQ {
return false
}
cols := make([]*Column, 0, 1)
cols = ExtractColumnsFromExpressions(cols, sf.GetArgs(), isColumnInOperand)
return len(cols) > 0
}
// ExprNotNull checks if an expression is possible to be null.
func ExprNotNull(expr Expression) bool {
if c, ok := expr.(*Constant); ok {
return !c.Value.IsNull()
}
// For ScalarFunction, the result would not be correct until we support maintaining
// NotNull flag for it.
return mysql.HasNotNullFlag(expr.GetType().GetFlag())
}
// HandleOverflowOnSelection handles Overflow errors when evaluating selection filters.
// We should ignore overflow errors when evaluating selection conditions:
//
// INSERT INTO t VALUES ("999999999999999999");
// SELECT * FROM t WHERE v;
func HandleOverflowOnSelection(sc *stmtctx.StatementContext, val int64, err error) (int64, error) {
if sc.InSelectStmt && err != nil && types.ErrOverflow.Equal(err) {
return -1, nil
}
return val, err
}
// EvalBool evaluates expression list to a boolean value. The first returned value
// indicates bool result of the expression list, the second returned value indicates
// whether the result of the expression list is null, it can only be true when the
// first returned values is false.
func EvalBool(ctx sessionctx.Context, exprList CNFExprs, row chunk.Row) (bool, bool, error) {
hasNull := false
for _, expr := range exprList {
data, err := expr.Eval(row)
if err != nil {
return false, false, err
}
if data.IsNull() {
// For queries like `select a in (select a from s where t.b = s.b) from t`,
// if result of `t.a = s.a` is null, we cannot return immediately until
// we have checked if `t.b = s.b` is null or false, because it means
// subquery is empty, and we should return false as the result of the whole
// exprList in that case, instead of null.
if !IsEQCondFromIn(expr) {
return false, false, nil
}
hasNull = true
continue
}
i, err := data.ToBool(ctx.GetSessionVars().StmtCtx)
if err != nil {
i, err = HandleOverflowOnSelection(ctx.GetSessionVars().StmtCtx, i, err)
if err != nil {
return false, false, err
}
}
if i == 0 {
return false, false, nil
}
}
if hasNull {
return false, true, nil
}
return true, false, nil
}
var (
defaultChunkSize = 1024
selPool = sync.Pool{
New: func() interface{} {
return make([]int, defaultChunkSize)
},
}
zeroPool = sync.Pool{
New: func() interface{} {
return make([]int8, defaultChunkSize)
},
}
)
func allocSelSlice(n int) []int {
if n > defaultChunkSize {
return make([]int, n)
}
return selPool.Get().([]int)
}
func deallocateSelSlice(sel []int) {
if cap(sel) <= defaultChunkSize {
selPool.Put(sel)
}
}
func allocZeroSlice(n int) []int8 {
if n > defaultChunkSize {
return make([]int8, n)
}
return zeroPool.Get().([]int8)
}
func deallocateZeroSlice(isZero []int8) {
if cap(isZero) <= defaultChunkSize {
zeroPool.Put(isZero)
}
}
// VecEvalBool does the same thing as EvalBool but it works in a vectorized manner.
func VecEvalBool(ctx sessionctx.Context, exprList CNFExprs, input *chunk.Chunk, selected, nulls []bool) ([]bool, []bool, error) {
// If input.Sel() != nil, we will call input.SetSel(nil) to clear the sel slice in input chunk.
// After the function finished, then we reset the input.Sel().
// The caller will handle the input.Sel() and selected slices.
defer input.SetSel(input.Sel())
input.SetSel(nil)
n := input.NumRows()
selected = selected[:0]
nulls = nulls[:0]
for i := 0; i < n; i++ {
selected = append(selected, false)
nulls = append(nulls, false)
}
sel := allocSelSlice(n)
defer deallocateSelSlice(sel)
sel = sel[:0]
for i := 0; i < n; i++ {
sel = append(sel, i)
}
input.SetSel(sel)
// In isZero slice, -1 means Null, 0 means zero, 1 means not zero
isZero := allocZeroSlice(n)
defer deallocateZeroSlice(isZero)
for _, expr := range exprList {
tp := expr.GetType()
eType := tp.EvalType()
if CanImplicitEvalReal(expr) {
eType = types.ETReal
}
buf, err := globalColumnAllocator.get()
if err != nil {
return nil, nil, err
}
// Take the implicit evalReal path if possible.
if CanImplicitEvalReal(expr) {
if err := implicitEvalReal(ctx, expr, input, buf); err != nil {
return nil, nil, err
}
} else if err := EvalExpr(ctx, expr, eType, input, buf); err != nil {
return nil, nil, err
}
err = toBool(ctx.GetSessionVars().StmtCtx, tp, eType, buf, sel, isZero)
if err != nil {
return nil, nil, err
}
j := 0
isEQCondFromIn := IsEQCondFromIn(expr)
for i := range sel {
if isZero[i] == -1 {
if eType != types.ETInt && !isEQCondFromIn {
continue
}
// In this case, we set this row to null and let it pass this filter.
// The null flag may be set to false later by other expressions in some cases.
nulls[sel[i]] = true
sel[j] = sel[i]
j++
continue
}
if isZero[i] == 0 {
continue
}
sel[j] = sel[i] // this row passes this filter
j++
}
sel = sel[:j]
input.SetSel(sel)
globalColumnAllocator.put(buf)
}
for _, i := range sel {
if !nulls[i] {
selected[i] = true
}
}
return selected, nulls, nil
}
func toBool(sc *stmtctx.StatementContext, tp *types.FieldType, eType types.EvalType, buf *chunk.Column, sel []int, isZero []int8) error {
switch eType {
case types.ETInt:
i64s := buf.Int64s()
for i := range sel {
if buf.IsNull(i) {
isZero[i] = -1
} else {
if i64s[i] == 0 {
isZero[i] = 0
} else {
isZero[i] = 1
}
}
}
case types.ETReal:
f64s := buf.Float64s()
for i := range sel {
if buf.IsNull(i) {
isZero[i] = -1
} else {
if f64s[i] == 0 {
isZero[i] = 0
} else {
isZero[i] = 1
}
}
}
case types.ETDuration:
d64s := buf.GoDurations()
for i := range sel {
if buf.IsNull(i) {
isZero[i] = -1
} else {
if d64s[i] == 0 {
isZero[i] = 0
} else {
isZero[i] = 1
}
}
}
case types.ETDatetime, types.ETTimestamp:
t64s := buf.Times()
for i := range sel {
if buf.IsNull(i) {
isZero[i] = -1
} else {
if t64s[i].IsZero() {
isZero[i] = 0
} else {
isZero[i] = 1
}
}
}
case types.ETString:
for i := range sel {
if buf.IsNull(i) {
isZero[i] = -1
} else {
var fVal float64
var err error
sVal := buf.GetString(i)
if tp.Hybrid() {
switch tp.GetType() {
case mysql.TypeSet, mysql.TypeEnum:
fVal = float64(len(sVal))
if fVal == 0 {
// The elements listed in the column specification are assigned index numbers, beginning
// with 1. The index value of the empty string error value (distinguish from a "normal"
// empty string) is 0. Thus we need to check whether it's an empty string error value when
// `fVal==0`.
for idx, elem := range tp.GetElems() {
if elem == sVal {
fVal = float64(idx + 1)
break
}
}
}
case mysql.TypeBit:
var bl types.BinaryLiteral = buf.GetBytes(i)
iVal, err := bl.ToInt(sc)
if err != nil {
return err
}
fVal = float64(iVal)
}
} else {
fVal, err = types.StrToFloat(sc, sVal, false)
if err != nil {
return err
}
}
if fVal == 0 {
isZero[i] = 0
} else {
isZero[i] = 1
}
}
}
case types.ETDecimal:
d64s := buf.Decimals()
for i := range sel {
if buf.IsNull(i) {
isZero[i] = -1
} else {
if d64s[i].IsZero() {
isZero[i] = 0
} else {
isZero[i] = 1
}
}
}
case types.ETJson:
for i := range sel {
if buf.IsNull(i) {
isZero[i] = -1
} else {
if buf.GetJSON(i).IsZero() {
isZero[i] = 0
} else {
isZero[i] = 1
}
}
}
}
return nil
}
func implicitEvalReal(ctx sessionctx.Context, expr Expression, input *chunk.Chunk, result *chunk.Column) (err error) {
if expr.Vectorized() && ctx.GetSessionVars().EnableVectorizedExpression {
err = expr.VecEvalReal(ctx, input, result)
} else {
ind, n := 0, input.NumRows()
iter := chunk.NewIterator4Chunk(input)
result.ResizeFloat64(n, false)
f64s := result.Float64s()
for it := iter.Begin(); it != iter.End(); it = iter.Next() {
value, isNull, err := expr.EvalReal(ctx, it)
if err != nil {
return err
}
if isNull {
result.SetNull(ind, isNull)
} else {
f64s[ind] = value
}
ind++
}
}
return
}
// EvalExpr evaluates this expr according to its type.
// And it selects the method for evaluating expression based on
// the environment variables and whether the expression can be vectorized.
// Note: the input argument `evalType` is needed because of that when `expr` is
// of the hybrid type(ENUM/SET/BIT), we need the invoker decide the actual EvalType.
func EvalExpr(ctx sessionctx.Context, expr Expression, evalType types.EvalType, input *chunk.Chunk, result *chunk.Column) (err error) {
if expr.Vectorized() && ctx.GetSessionVars().EnableVectorizedExpression {
switch evalType {
case types.ETInt:
err = expr.VecEvalInt(ctx, input, result)
case types.ETReal:
err = expr.VecEvalReal(ctx, input, result)
case types.ETDuration:
err = expr.VecEvalDuration(ctx, input, result)
case types.ETDatetime, types.ETTimestamp:
err = expr.VecEvalTime(ctx, input, result)
case types.ETString:
err = expr.VecEvalString(ctx, input, result)
case types.ETJson:
err = expr.VecEvalJSON(ctx, input, result)
case types.ETDecimal:
err = expr.VecEvalDecimal(ctx, input, result)
default:
err = fmt.Errorf("invalid eval type %v", expr.GetType().EvalType())
}
} else {
ind, n := 0, input.NumRows()
iter := chunk.NewIterator4Chunk(input)
switch evalType {
case types.ETInt:
result.ResizeInt64(n, false)
i64s := result.Int64s()
for it := iter.Begin(); it != iter.End(); it = iter.Next() {
value, isNull, err := expr.EvalInt(ctx, it)
if err != nil {
return err
}
if isNull {
result.SetNull(ind, isNull)
} else {
i64s[ind] = value
}
ind++
}
case types.ETReal:
result.ResizeFloat64(n, false)
f64s := result.Float64s()
for it := iter.Begin(); it != iter.End(); it = iter.Next() {
value, isNull, err := expr.EvalReal(ctx, it)
if err != nil {
return err
}
if isNull {
result.SetNull(ind, isNull)
} else {
f64s[ind] = value
}
ind++
}
case types.ETDuration:
result.ResizeGoDuration(n, false)
d64s := result.GoDurations()
for it := iter.Begin(); it != iter.End(); it = iter.Next() {
value, isNull, err := expr.EvalDuration(ctx, it)
if err != nil {
return err
}
if isNull {
result.SetNull(ind, isNull)
} else {
d64s[ind] = value.Duration
}
ind++
}
case types.ETDatetime, types.ETTimestamp:
result.ResizeTime(n, false)
t64s := result.Times()
for it := iter.Begin(); it != iter.End(); it = iter.Next() {
value, isNull, err := expr.EvalTime(ctx, it)
if err != nil {
return err
}
if isNull {
result.SetNull(ind, isNull)
} else {
t64s[ind] = value
}
ind++
}
case types.ETString:
result.ReserveString(n)
for it := iter.Begin(); it != iter.End(); it = iter.Next() {
value, isNull, err := expr.EvalString(ctx, it)
if err != nil {
return err
}
if isNull {
result.AppendNull()
} else {
result.AppendString(value)
}
}
case types.ETJson:
result.ReserveJSON(n)
for it := iter.Begin(); it != iter.End(); it = iter.Next() {
value, isNull, err := expr.EvalJSON(ctx, it)
if err != nil {
return err
}
if isNull {
result.AppendNull()
} else {
result.AppendJSON(value)
}
}
case types.ETDecimal:
result.ResizeDecimal(n, false)
d64s := result.Decimals()
for it := iter.Begin(); it != iter.End(); it = iter.Next() {
value, isNull, err := expr.EvalDecimal(ctx, it)
if err != nil {
return err
}
if isNull {
result.SetNull(ind, isNull)
} else {
d64s[ind] = *value
}
ind++
}
default:
err = fmt.Errorf("invalid eval type %v", expr.GetType().EvalType())
}
}
return
}
// composeConditionWithBinaryOp composes condition with binary operator into a balance deep tree, which benefits a lot for pb decoder/encoder.
func composeConditionWithBinaryOp(ctx sessionctx.Context, conditions []Expression, funcName string) Expression {
length := len(conditions)
if length == 0 {
return nil
}
if length == 1 {
return conditions[0]
}
expr := NewFunctionInternal(ctx, funcName,
types.NewFieldType(mysql.TypeTiny),
composeConditionWithBinaryOp(ctx, conditions[:length/2], funcName),
composeConditionWithBinaryOp(ctx, conditions[length/2:], funcName))
return expr
}
// ComposeCNFCondition composes CNF items into a balance deep CNF tree, which benefits a lot for pb decoder/encoder.
func ComposeCNFCondition(ctx sessionctx.Context, conditions ...Expression) Expression {
return composeConditionWithBinaryOp(ctx, conditions, ast.LogicAnd)
}
// ComposeDNFCondition composes DNF items into a balance deep DNF tree.
func ComposeDNFCondition(ctx sessionctx.Context, conditions ...Expression) Expression {
return composeConditionWithBinaryOp(ctx, conditions, ast.LogicOr)
}
func extractBinaryOpItems(conditions *ScalarFunction, funcName string) []Expression {
var ret []Expression
for _, arg := range conditions.GetArgs() {
if sf, ok := arg.(*ScalarFunction); ok && sf.FuncName.L == funcName {
ret = append(ret, extractBinaryOpItems(sf, funcName)...)
} else {
ret = append(ret, arg)
}
}
return ret
}
// FlattenDNFConditions extracts DNF expression's leaf item.
// e.g. or(or(a=1, a=2), or(a=3, a=4)), we'll get [a=1, a=2, a=3, a=4].
func FlattenDNFConditions(DNFCondition *ScalarFunction) []Expression {
return extractBinaryOpItems(DNFCondition, ast.LogicOr)
}
// FlattenCNFConditions extracts CNF expression's leaf item.
// e.g. and(and(a>1, a>2), and(a>3, a>4)), we'll get [a>1, a>2, a>3, a>4].
func FlattenCNFConditions(CNFCondition *ScalarFunction) []Expression {
return extractBinaryOpItems(CNFCondition, ast.LogicAnd)
}
// Assignment represents a set assignment in Update, such as
// Update t set c1 = hex(12), c2 = c3 where c2 = 1
type Assignment struct {
Col *Column
// ColName indicates its original column name in table schema. It's used for outputting helping message when executing meets some errors.
ColName model.CIStr
Expr Expression
// LazyErr is used in statement like `INSERT INTO t1 (a) VALUES (1) ON DUPLICATE KEY UPDATE a= (SELECT b FROM source);`, ErrSubqueryMoreThan1Row
// should be evaluated after the duplicate situation is detected in the executing procedure.
LazyErr error
}
// MemoryUsage return the memory usage of Assignment
func (a *Assignment) MemoryUsage() (sum int64) {
if a == nil {
return
}
sum = size.SizeOfPointer + a.ColName.MemoryUsage() + size.SizeOfInterface*2
if a.Expr != nil {
sum += a.Expr.MemoryUsage()
}
return
}
// VarAssignment represents a variable assignment in Set, such as set global a = 1.
type VarAssignment struct {
Name string
Expr Expression
IsDefault bool
IsGlobal bool
IsSystem bool
ExtendValue *Constant
}
// splitNormalFormItems split CNF(conjunctive normal form) like "a and b and c", or DNF(disjunctive normal form) like "a or b or c"
func splitNormalFormItems(onExpr Expression, funcName string) []Expression {
//nolint: revive
switch v := onExpr.(type) {
case *ScalarFunction:
if v.FuncName.L == funcName {
var ret []Expression
for _, arg := range v.GetArgs() {
ret = append(ret, splitNormalFormItems(arg, funcName)...)
}
return ret
}
}
return []Expression{onExpr}
}
// SplitCNFItems splits CNF items.
// CNF means conjunctive normal form, e.g. "a and b and c".
func SplitCNFItems(onExpr Expression) []Expression {
return splitNormalFormItems(onExpr, ast.LogicAnd)
}
// SplitDNFItems splits DNF items.
// DNF means disjunctive normal form, e.g. "a or b or c".
func SplitDNFItems(onExpr Expression) []Expression {
return splitNormalFormItems(onExpr, ast.LogicOr)
}
// EvaluateExprWithNull sets columns in schema as null and calculate the final result of the scalar function.
// If the Expression is a non-constant value, it means the result is unknown.
func EvaluateExprWithNull(ctx sessionctx.Context, schema *Schema, expr Expression) Expression {
if MaybeOverOptimized4PlanCache(ctx, []Expression{expr}) {
ctx.GetSessionVars().StmtCtx.SetSkipPlanCache(errors.New("skip plan-cache: %v affects null check"))
}
if ctx.GetSessionVars().StmtCtx.InNullRejectCheck {
expr, _ = evaluateExprWithNullInNullRejectCheck(ctx, schema, expr)
return expr
}
return evaluateExprWithNull(ctx, schema, expr)
}
func evaluateExprWithNull(ctx sessionctx.Context, schema *Schema, expr Expression) Expression {
switch x := expr.(type) {
case *ScalarFunction:
args := make([]Expression, len(x.GetArgs()))
for i, arg := range x.GetArgs() {
args[i] = evaluateExprWithNull(ctx, schema, arg)
}
return NewFunctionInternal(ctx, x.FuncName.L, x.RetType.Clone(), args...)
case *Column:
if !schema.Contains(x) {
return x
}
return &Constant{Value: types.Datum{}, RetType: types.NewFieldType(mysql.TypeNull)}
case *Constant:
if x.DeferredExpr != nil {
return FoldConstant(x)
}
}
return expr
}
// evaluateExprWithNullInNullRejectCheck sets columns in schema as null and calculate the final result of the scalar function.
// If the Expression is a non-constant value, it means the result is unknown.
// The returned bool values indicates whether the value is influenced by the Null Constant transformed from schema column
// when the value is Null Constant.
func evaluateExprWithNullInNullRejectCheck(ctx sessionctx.Context, schema *Schema, expr Expression) (Expression, bool) {
switch x := expr.(type) {
case *ScalarFunction:
args := make([]Expression, len(x.GetArgs()))
nullFromSets := make([]bool, len(x.GetArgs()))
for i, arg := range x.GetArgs() {
args[i], nullFromSets[i] = evaluateExprWithNullInNullRejectCheck(ctx, schema, arg)
}
allArgsNullFromSet := true
for i := range args {
if cons, ok := args[i].(*Constant); ok && cons.Value.IsNull() && !nullFromSets[i] {
allArgsNullFromSet = false
break
}
}
// If one of the args of `AND` and `OR` are Null Constant from the column schema, and the another argument is Constant, then we should keep it.
// Otherwise, we shouldn't let Null Constant which affected by the column schema participate in computing in `And` and `OR`
// due to the result of `AND` and `OR` are uncertain if one of the arguments is NULL.
if x.FuncName.L == ast.LogicAnd || x.FuncName.L == ast.LogicOr {
hasNonConstantArg := false
for _, arg := range args {
if _, ok := arg.(*Constant); !ok {
hasNonConstantArg = true
break
}
}
if hasNonConstantArg {
for i := range args {
if cons, ok := args[i].(*Constant); ok && cons.Value.IsNull() && nullFromSets[i] {
if x.FuncName.L == ast.LogicAnd {
args[i] = NewOne()
break
}
if x.FuncName.L == ast.LogicOr {
args[i] = NewZero()
break
}
}
}
}
}
c := NewFunctionInternal(ctx, x.FuncName.L, x.RetType.Clone(), args...)
cons, ok := c.(*Constant)
// If the return expr is Null Constant, and all the Null Constant arguments are affected by column schema,
// then we think the result Null Constant is also affected by the column schema
return c, ok && cons.Value.IsNull() && allArgsNullFromSet
case *Column:
if !schema.Contains(x) {
return x, false
}
return &Constant{Value: types.Datum{}, RetType: types.NewFieldType(mysql.TypeNull)}, true
case *Constant:
if x.DeferredExpr != nil {
return FoldConstant(x), false
}
}
return expr, false
}
// TableInfo2SchemaAndNames converts the TableInfo to the schema and name slice.
func TableInfo2SchemaAndNames(ctx sessionctx.Context, dbName model.CIStr, tbl *model.TableInfo) (*Schema, []*types.FieldName, error) {
cols, names, err := ColumnInfos2ColumnsAndNames(ctx, dbName, tbl.Name, tbl.Cols(), tbl)
if err != nil {
return nil, nil, err
}
keys := make([]KeyInfo, 0, len(tbl.Indices)+1)
for _, idx := range tbl.Indices {
if !idx.Unique || idx.State != model.StatePublic {
continue
}
ok := true
newKey := make([]*Column, 0, len(idx.Columns))
for _, idxCol := range idx.Columns {
find := false
for i, col := range tbl.Columns {
if idxCol.Name.L == col.Name.L {
if !mysql.HasNotNullFlag(col.GetFlag()) {
break
}
newKey = append(newKey, cols[i])
find = true
break
}
}
if !find {
ok = false
break
}
}
if ok {
keys = append(keys, newKey)
}
}
if tbl.PKIsHandle {
for i, col := range tbl.Columns {
if mysql.HasPriKeyFlag(col.GetFlag()) {
keys = append(keys, KeyInfo{cols[i]})
break
}
}
}
schema := NewSchema(cols...)
schema.SetUniqueKeys(keys)
return schema, names, nil
}
// ColumnInfos2ColumnsAndNames converts the ColumnInfo to the *Column and NameSlice.
func ColumnInfos2ColumnsAndNames(ctx sessionctx.Context, dbName, tblName model.CIStr, colInfos []*model.ColumnInfo, tblInfo *model.TableInfo) ([]*Column, types.NameSlice, error) {
columns := make([]*Column, 0, len(colInfos))
names := make([]*types.FieldName, 0, len(colInfos))
for i, col := range colInfos {
names = append(names, &types.FieldName{
OrigTblName: tblName,
OrigColName: col.Name,
DBName: dbName,
TblName: tblName,
ColName: col.Name,
})
newCol := &Column{
RetType: col.FieldType.Clone(),
ID: col.ID,
UniqueID: ctx.GetSessionVars().AllocPlanColumnID(),
Index: col.Offset,
OrigName: names[i].String(),
IsHidden: col.Hidden,
}
columns = append(columns, newCol)
}
// Resolve virtual generated column.
mockSchema := NewSchema(columns...)
// Ignore redundant warning here.
save := ctx.GetSessionVars().StmtCtx.IgnoreTruncate
defer func() {
ctx.GetSessionVars().StmtCtx.IgnoreTruncate = save
}()
ctx.GetSessionVars().StmtCtx.IgnoreTruncate = true
for i, col := range colInfos {
if col.IsGenerated() && !col.GeneratedStored {
expr, err := generatedexpr.ParseExpression(col.GeneratedExprString)
if err != nil {
return nil, nil, errors.Trace(err)
}
expr, err = generatedexpr.SimpleResolveName(expr, tblInfo)
if err != nil {
return nil, nil, errors.Trace(err)
}