-
Notifications
You must be signed in to change notification settings - Fork 3.8k
/
mutation_builder.go
1458 lines (1266 loc) · 51.6 KB
/
mutation_builder.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 2018 The Cockroach Authors.
//
// Use of this software is governed by the Business Source License
// included in the file licenses/BSL.txt.
//
// As of the Change Date specified in that file, in accordance with
// the Business Source License, use of this software will be governed
// by the Apache License, Version 2.0, included in the file
// licenses/APL.txt.
package optbuilder
import (
"fmt"
"strings"
"github.com/cockroachdb/cockroach/pkg/server/telemetry"
"github.com/cockroachdb/cockroach/pkg/sql/catalog/colinfo"
"github.com/cockroachdb/cockroach/pkg/sql/catalog/schemaexpr"
"github.com/cockroachdb/cockroach/pkg/sql/opt"
"github.com/cockroachdb/cockroach/pkg/sql/opt/cat"
"github.com/cockroachdb/cockroach/pkg/sql/opt/memo"
"github.com/cockroachdb/cockroach/pkg/sql/opt/props"
"github.com/cockroachdb/cockroach/pkg/sql/parser"
"github.com/cockroachdb/cockroach/pkg/sql/pgwire/pgcode"
"github.com/cockroachdb/cockroach/pkg/sql/pgwire/pgerror"
"github.com/cockroachdb/cockroach/pkg/sql/sem/cast"
"github.com/cockroachdb/cockroach/pkg/sql/sem/tree"
"github.com/cockroachdb/cockroach/pkg/sql/sqlerrors"
"github.com/cockroachdb/cockroach/pkg/sql/sqltelemetry"
"github.com/cockroachdb/cockroach/pkg/sql/types"
"github.com/cockroachdb/cockroach/pkg/util"
"github.com/cockroachdb/errors"
)
// mutationBuilder is a helper struct that supports building Insert, Update,
// Upsert, and Delete operators in stages.
type mutationBuilder struct {
b *Builder
md *opt.Metadata
// opName is the statement's name, used in error messages.
opName string
// tab is the target table.
tab cat.Table
// tabID is the metadata ID of the table.
tabID opt.TableID
// alias is the table alias specified in the mutation statement, or just the
// resolved table name if no alias was specified.
alias tree.TableName
// outScope contains the current set of columns that are in scope, as well as
// the output expression as it is incrementally built. Once the final mutation
// expression is completed, it will be contained in outScope.expr.
outScope *scope
// fetchScope contains the set of columns fetched from the target table.
fetchScope *scope
// insertExpr is the expression that produces the values which will be
// inserted into the target table. It is only populated for INSERT
// expressions. It is currently used to inline constant insert values into
// uniqueness checks.
insertExpr memo.RelExpr
// targetColList is an ordered list of IDs of the table columns into which
// values will be inserted, or which will be updated with new values. It is
// incrementally built as the mutation operator is built.
targetColList opt.ColList
// targetColSet contains the same column IDs as targetColList, but as a set.
targetColSet opt.ColSet
// insertColIDs lists the input column IDs providing values to insert. Its
// length is always equal to the number of columns in the target table,
// including mutation columns. Table columns which will not have values
// inserted are set to 0 (e.g. delete-only mutation columns). insertColIDs
// is empty if this is not an Insert/Upsert operator.
insertColIDs opt.OptionalColList
// fetchColIDs lists the input column IDs storing values which are fetched
// from the target table in order to provide existing values that will form
// lookup and update values. Its length is always equal to the number of
// columns in the target table, including mutation columns. Table columns
// which do not need to be fetched are set to 0. fetchColIDs is empty if
// this is an Insert operator.
fetchColIDs opt.OptionalColList
// updateColIDs lists the input column IDs providing update values. Its
// length is always equal to the number of columns in the target table,
// including mutation columns. Table columns which do not need to be
// updated are set to 0.
updateColIDs opt.OptionalColList
// upsertColIDs lists the input column IDs that choose between an insert or
// update column using a CASE expression:
//
// CASE WHEN canary_col IS NULL THEN ins_col ELSE upd_col END
//
// These columns are used to compute constraints and to return result rows.
// The length of upsertColIDs is always equal to the number of columns in
// the target table, including mutation columns. Table columns which do not
// need to be updated are set to 0. upsertColIDs is empty if this is not
// an Upsert operator.
upsertColIDs opt.OptionalColList
// checkColIDs lists the input column IDs storing the boolean results of
// evaluating check constraint expressions defined on the target table. Its
// length is always equal to the number of check constraints on the table
// (see opt.Table.CheckCount).
checkColIDs opt.OptionalColList
// partialIndexPutColIDs lists the input column IDs storing the boolean
// results of evaluating partial index predicate expressions of the target
// table. The predicate expressions are evaluated with their variables
// assigned from newly inserted or updated row values. When these columns
// evaluate to true, it signifies that the inserted or updated row should be
// added to the corresponding partial index. The length of
// partialIndexPutColIDs is always equal to the number of partial indexes on
// the table.
partialIndexPutColIDs opt.OptionalColList
// partialIndexDelColIDs lists the input column IDs storing the boolean
// results of evaluating partial index predicate expressions of the target
// table. The predicate expressions are evaluated with their variables
// assigned from existing row values of deleted or updated rows. When these
// columns evaluate to true, it signifies that the deleted or updated row
// should be removed from the corresponding partial index. The length of
// partialIndexPutColIDs is always equal to the number of partial indexes on
// the table.
partialIndexDelColIDs opt.OptionalColList
// canaryColID is the ID of the column that is used to decide whether to
// insert or update each row. If the canary column's value is null, then it's
// an insert; otherwise it's an update.
canaryColID opt.ColumnID
// arbiters is the set of indexes and unique constraints that are used to
// detect conflicts for UPSERT and INSERT ON CONFLICT statements.
arbiters arbiterSet
// subqueries temporarily stores subqueries that were built during initial
// analysis of SET expressions. They will be used later when the subqueries
// are joined into larger LEFT OUTER JOIN expressions.
subqueries []*scope
// parsedColComputedExprs is a cached set of parsed computed expressions
// from the table schema. These are parsed once and cached for reuse.
parsedColComputedExprs []tree.Expr
// parsedColDefaultExprs is a cached set of parsed default expressions
// from the table schema. These are parsed once and cached for reuse.
parsedColDefaultExprs []tree.Expr
// parsedColOnUpdateExprs is a cached set of parsed ON UPDATE expressions from
// the table schema. These are parsed once and cached for reuse.
parsedColOnUpdateExprs []tree.Expr
// parsedIndexExprs is a cached set of parsed partial index predicate
// expressions from the table schema. These are parsed once and cached for
// reuse.
parsedIndexExprs []tree.Expr
// parsedUniqueConstraintExprs is a cached set of parsed partial unique
// constraint predicate expressions from the table schema. These are parsed
// once and cached for reuse.
parsedUniqueConstraintExprs []tree.Expr
// uniqueChecks contains unique check queries; see buildUnique* methods.
uniqueChecks memo.UniqueChecksExpr
// fkChecks contains foreign key check queries; see buildFK* methods.
fkChecks memo.FKChecksExpr
// cascades contains foreign key check cascades; see buildFK* methods.
cascades memo.FKCascades
// withID is nonzero if we need to buffer the input for FK or uniqueness
// checks.
withID opt.WithID
// extraAccessibleCols stores all the columns that are available to the
// mutation that are not part of the target table. This is useful for
// UPDATE ... FROM queries, as the columns from the FROM tables must be
// made accessible to the RETURNING clause.
extraAccessibleCols []scopeColumn
// fkCheckHelper is used to prevent allocating the helper separately.
fkCheckHelper fkCheckHelper
// uniqueCheckHelper is used to prevent allocating the helper separately.
uniqueCheckHelper uniqueCheckHelper
// arbiterPredicateHelper is used to prevent allocating the helper
// separately.
arbiterPredicateHelper arbiterPredicateHelper
}
func (mb *mutationBuilder) init(b *Builder, opName string, tab cat.Table, alias tree.TableName) {
// This initialization pattern ensures that fields are not unwittingly
// reused. Field reuse must be explicit.
*mb = mutationBuilder{
b: b,
md: b.factory.Metadata(),
opName: opName,
tab: tab,
alias: alias,
}
n := tab.ColumnCount()
mb.targetColList = make(opt.ColList, 0, n)
// Allocate segmented array of column IDs.
numPartialIndexes := partialIndexCount(tab)
colIDs := make(opt.OptionalColList, n*4+tab.CheckCount()+2*numPartialIndexes)
mb.insertColIDs = colIDs[:n]
mb.fetchColIDs = colIDs[n : n*2]
mb.updateColIDs = colIDs[n*2 : n*3]
mb.upsertColIDs = colIDs[n*3 : n*4]
mb.checkColIDs = colIDs[n*4 : n*4+tab.CheckCount()]
mb.partialIndexPutColIDs = colIDs[n*4+tab.CheckCount() : n*4+tab.CheckCount()+numPartialIndexes]
mb.partialIndexDelColIDs = colIDs[n*4+tab.CheckCount()+numPartialIndexes:]
// Add the table and its columns (including mutation columns) to metadata.
mb.tabID = mb.md.AddTable(tab, &mb.alias)
}
// setFetchColIDs sets the list of columns that are fetched in order to provide
// values to the mutation operator. The given columns must come from buildScan.
func (mb *mutationBuilder) setFetchColIDs(cols []scopeColumn) {
for i := range cols {
// Ensure that we don't add system columns to the fetch columns.
if cols[i].kind != cat.System {
mb.fetchColIDs[cols[i].tableOrdinal] = cols[i].id
}
}
}
// buildInputForUpdate constructs a Select expression from the fields in
// the Update operator, similar to this:
//
// SELECT <cols>
// FROM <table>
// WHERE <where>
// ORDER BY <order-by>
// LIMIT <limit>
//
// All columns from the table to update are added to fetchColList.
// If a FROM clause is defined, we build out each of the table
// expressions required and JOIN them together (LATERAL joins between
// the tables are allowed). We then JOIN the result with the target
// table (the FROM tables can't reference this table) and apply the
// appropriate WHERE conditions.
//
// It is the responsibility of the user to guarantee that the JOIN
// produces a maximum of one row per row of the target table. If multiple
// are found, an arbitrary one is chosen (this row is not readily
// predictable, consistent with the POSTGRES implementation).
// buildInputForUpdate stores the columns of the FROM tables in the
// mutation builder so they can be made accessible to other parts of
// the query (RETURNING clause).
// TODO(andyk): Do needed column analysis to project fewer columns if possible.
func (mb *mutationBuilder) buildInputForUpdate(
inScope *scope,
texpr tree.TableExpr,
from tree.TableExprs,
where *tree.Where,
limit *tree.Limit,
orderBy tree.OrderBy,
) {
var indexFlags *tree.IndexFlags
if source, ok := texpr.(*tree.AliasedTableExpr); ok && source.IndexFlags != nil {
indexFlags = source.IndexFlags
telemetry.Inc(sqltelemetry.IndexHintUseCounter)
telemetry.Inc(sqltelemetry.IndexHintUpdateUseCounter)
}
// Fetch columns from different instance of the table metadata, so that it's
// possible to remap columns, as in this example:
//
// UPDATE abc SET a=b
//
// NOTE: Include mutation columns, but be careful to never use them for any
// reason other than as "fetch columns". See buildScan comment.
mb.fetchScope = mb.b.buildScan(
mb.b.addTable(mb.tab, &mb.alias),
tableOrdinals(mb.tab, columnKinds{
includeMutations: true,
includeSystem: true,
includeInverted: false,
}),
indexFlags,
noRowLocking,
inScope,
false, /* disableNotVisibleIndex */
)
// Set list of columns that will be fetched by the input expression.
mb.setFetchColIDs(mb.fetchScope.cols)
// If there is a FROM clause present, we must join all the tables
// together with the table being updated.
fromClausePresent := len(from) > 0
if fromClausePresent {
fromScope := mb.b.buildFromTables(from, noRowLocking, inScope)
// Check that the same table name is not used multiple times.
mb.b.validateJoinTableNames(mb.fetchScope, fromScope)
// The FROM table columns can be accessed by the RETURNING clause of the
// query and so we have to make them accessible.
mb.extraAccessibleCols = fromScope.cols
// Add the columns in the FROM scope.
// We create a new scope so that fetchScope is not modified. It will be
// used later to build partial index predicate expressions, and we do
// not want ambiguities with column names in the FROM clause.
mb.outScope = mb.fetchScope.replace()
mb.outScope.appendColumnsFromScope(mb.fetchScope)
mb.outScope.appendColumnsFromScope(fromScope)
left := mb.fetchScope.expr
right := fromScope.expr
mb.outScope.expr = mb.b.factory.ConstructInnerJoin(left, right, memo.TrueFilter, memo.EmptyJoinPrivate)
} else {
mb.outScope = mb.fetchScope
}
// WHERE
mb.b.buildWhere(where, mb.outScope)
// SELECT + ORDER BY (which may add projected expressions)
projectionsScope := mb.outScope.replace()
projectionsScope.appendColumnsFromScope(mb.outScope)
orderByScope := mb.b.analyzeOrderBy(orderBy, mb.outScope, projectionsScope, tree.RejectGenerators)
mb.b.buildOrderBy(mb.outScope, projectionsScope, orderByScope)
mb.b.constructProjectForScope(mb.outScope, projectionsScope)
// LIMIT
if limit != nil {
mb.b.buildLimit(limit, inScope, projectionsScope)
}
mb.outScope = projectionsScope
// Build a distinct on to ensure there is at most one row in the joined output
// for every row in the table.
if fromClausePresent {
var pkCols opt.ColSet
// We need to ensure that the join has a maximum of one row for every row
// in the table and we ensure this by constructing a distinct on the primary
// key columns.
primaryIndex := mb.tab.Index(cat.PrimaryIndex)
for i := 0; i < primaryIndex.KeyColumnCount(); i++ {
// If the primary key column is hidden, then we don't need to use it
// for the distinct on.
// TODO(radu): this logic seems fragile, is it assuming that only an
// implicit `rowid` column can be a hidden PK column?
if col := primaryIndex.Column(i); col.Visibility() != cat.Hidden {
pkCols.Add(mb.fetchColIDs[col.Ordinal()])
}
}
if !pkCols.Empty() {
mb.outScope = mb.b.buildDistinctOn(
pkCols, mb.outScope, false /* nullsAreDistinct */, "" /* errorOnDup */)
}
}
}
// buildInputForDelete constructs a Select expression from the fields in
// the Delete operator, similar to this:
//
// SELECT <cols>
// FROM <table>
// WHERE <where>
// ORDER BY <order-by>
// LIMIT <limit>
//
// All columns from the table to update are added to fetchColList.
// TODO(andyk): Do needed column analysis to project fewer columns if possible.
func (mb *mutationBuilder) buildInputForDelete(
inScope *scope, texpr tree.TableExpr, where *tree.Where, limit *tree.Limit, orderBy tree.OrderBy,
) {
var indexFlags *tree.IndexFlags
if source, ok := texpr.(*tree.AliasedTableExpr); ok && source.IndexFlags != nil {
indexFlags = source.IndexFlags
telemetry.Inc(sqltelemetry.IndexHintUseCounter)
telemetry.Inc(sqltelemetry.IndexHintDeleteUseCounter)
}
// Fetch columns from different instance of the table metadata, so that it's
// possible to remap columns, as in this example:
//
// DELETE FROM abc WHERE a=b
//
// NOTE: Include mutation columns, but be careful to never use them for any
// reason other than as "fetch columns". See buildScan comment.
// TODO(andyk): Why does execution engine need mutation columns for Delete?
mb.fetchScope = mb.b.buildScan(
mb.b.addTable(mb.tab, &mb.alias),
tableOrdinals(mb.tab, columnKinds{
includeMutations: true,
includeSystem: true,
includeInverted: false,
}),
indexFlags,
noRowLocking,
inScope,
false, /* disableNotVisibleIndex */
)
mb.outScope = mb.fetchScope
// WHERE
mb.b.buildWhere(where, mb.outScope)
// SELECT + ORDER BY (which may add projected expressions)
projectionsScope := mb.outScope.replace()
projectionsScope.appendColumnsFromScope(mb.outScope)
orderByScope := mb.b.analyzeOrderBy(orderBy, mb.outScope, projectionsScope, tree.RejectGenerators)
mb.b.buildOrderBy(mb.outScope, projectionsScope, orderByScope)
mb.b.constructProjectForScope(mb.outScope, projectionsScope)
// LIMIT
if limit != nil {
mb.b.buildLimit(limit, inScope, projectionsScope)
}
mb.outScope = projectionsScope
// Set list of columns that will be fetched by the input expression.
mb.setFetchColIDs(mb.outScope.cols)
}
// addTargetColsByName adds one target column for each of the names in the given
// list.
func (mb *mutationBuilder) addTargetColsByName(names tree.NameList) {
for _, name := range names {
// Determine the ordinal position of the named column in the table and
// add it as a target column.
if ord := findPublicTableColumnByName(mb.tab, name); ord != -1 {
// System columns are invalid target columns.
if mb.tab.Column(ord).Kind() == cat.System {
panic(pgerror.Newf(pgcode.InvalidColumnReference, "cannot modify system column %q", name))
}
mb.addTargetCol(ord)
continue
}
panic(colinfo.NewUndefinedColumnError(string(name)))
}
}
// addTargetCol adds a target column by its ordinal position in the target
// table. It raises an error if a mutation or computed column is targeted, or if
// the same column is targeted multiple times.
func (mb *mutationBuilder) addTargetCol(ord int) {
tabCol := mb.tab.Column(ord)
// Don't allow targeting of mutation columns.
if tabCol.IsMutation() {
panic(makeBackfillError(tabCol.ColName()))
}
// Computed columns cannot be targeted with input values.
if tabCol.IsComputed() {
panic(schemaexpr.CannotWriteToComputedColError(string(tabCol.ColName())))
}
// Ensure that the name list does not contain duplicates.
colID := mb.tabID.ColumnID(ord)
if mb.targetColSet.Contains(colID) {
panic(pgerror.Newf(pgcode.Syntax,
"multiple assignments to the same column %q", tabCol.ColName()))
}
mb.targetColSet.Add(colID)
mb.targetColList = append(mb.targetColList, colID)
}
// extractValuesInput tests whether the given input is a VALUES clause with no
// WITH, ORDER BY, or LIMIT modifier. If so, it's returned, otherwise nil is
// returned.
func (mb *mutationBuilder) extractValuesInput(inputRows *tree.Select) *tree.ValuesClause {
if inputRows == nil {
return nil
}
// Only extract a simple VALUES clause with no modifiers.
if inputRows.With != nil || inputRows.OrderBy != nil || inputRows.Limit != nil {
return nil
}
// Discard parentheses.
if parens, ok := inputRows.Select.(*tree.ParenSelect); ok {
return mb.extractValuesInput(parens.Select)
}
if values, ok := inputRows.Select.(*tree.ValuesClause); ok {
return values
}
return nil
}
// replaceDefaultExprs looks for DEFAULT specifiers in input value expressions
// and replaces them with the corresponding default value expression for the
// corresponding column. This is only possible when the input is a VALUES
// clause. For example:
//
// INSERT INTO t (a, b) (VALUES (1, DEFAULT), (DEFAULT, 2))
//
// Here, the two DEFAULT specifiers are replaced by the default value expression
// for the a and b columns, respectively.
//
// replaceDefaultExprs returns a VALUES expression with replaced DEFAULT values,
// or just the unchanged input expression if there are no DEFAULT values.
func (mb *mutationBuilder) replaceDefaultExprs(inRows *tree.Select) (outRows *tree.Select) {
values := mb.extractValuesInput(inRows)
if values == nil {
return inRows
}
// Ensure that the number of input columns exactly matches the number of
// target columns.
numCols := len(values.Rows[0])
mb.checkNumCols(len(mb.targetColList), numCols)
var newRows []tree.Exprs
for irow, tuple := range values.Rows {
if len(tuple) != numCols {
reportValuesLenError(numCols, len(tuple))
}
// Scan list of tuples in the VALUES row, looking for DEFAULT specifiers.
var newTuple tree.Exprs
for itup, val := range tuple {
if _, ok := val.(tree.DefaultVal); ok {
// Found DEFAULT, so lazily create new rows and tuple lists.
if newRows == nil {
newRows = make([]tree.Exprs, irow, len(values.Rows))
copy(newRows, values.Rows[:irow])
}
if newTuple == nil {
newTuple = make(tree.Exprs, itup, numCols)
copy(newTuple, tuple[:itup])
}
val = mb.parseDefaultExpr(mb.targetColList[itup])
}
if newTuple != nil {
newTuple = append(newTuple, val)
}
}
if newRows != nil {
if newTuple != nil {
newRows = append(newRows, newTuple)
} else {
newRows = append(newRows, tuple)
}
}
}
if newRows != nil {
return &tree.Select{Select: &tree.ValuesClause{Rows: newRows}}
}
return inRows
}
// addSynthesizedDefaultCols is a helper method for addSynthesizedColsForInsert
// and addSynthesizedColsForUpdate that scans the list of Ordinary and WriteOnly
// table columns, looking for any that are not computed and do not yet have
// values provided by the input expression. New columns are synthesized for any
// missing columns.
//
// Values are synthesized for columns based on checking these rules, in order:
// 1. If column has a default value specified for it, use that as its value.
// 2. If column is nullable, use NULL as its value.
// 3. If column is currently being added or dropped (i.e. a mutation column),
// use a default value (0 for INT column, "" for STRING column, etc). Note
// that the existing "fetched" value returned by the scan cannot be used,
// since it may not have been initialized yet by the backfiller.
//
// If includeOrdinary is false, then only WriteOnly columns are considered.
//
// NOTE: colIDs is updated with the column IDs of any synthesized columns which
// are added to mb.outScope.
func (mb *mutationBuilder) addSynthesizedDefaultCols(
colIDs opt.OptionalColList, includeOrdinary bool, applyOnUpdate bool,
) {
// We will construct a new Project operator that will contain the newly
// synthesized column(s).
pb := makeProjectionBuilder(mb.b, mb.outScope)
for i, n := 0, mb.tab.ColumnCount(); i < n; i++ {
tabCol := mb.tab.Column(i)
if kind := tabCol.Kind(); kind == cat.WriteOnly {
// Always include WriteOnly columns.
} else if tabCol.UseOnUpdate(mb.b.evalCtx.SessionData()) && applyOnUpdate {
// Use ON UPDATE columns if specified.
} else if includeOrdinary && kind == cat.Ordinary {
// Include Ordinary columns if indicated.
} else {
// Wrong kind.
continue
}
if tabCol.IsComputed() {
continue
}
// Skip columns that are already specified.
if colIDs[i] != 0 {
continue
}
// Use ON UPDATE expression if specified, default otherwise
tabColID := mb.tabID.ColumnID(i)
var mutationSuffix string
var expr tree.Expr
if tabCol.UseOnUpdate(mb.b.evalCtx.SessionData()) && applyOnUpdate {
mutationSuffix = "on_update"
expr = mb.parseOnUpdateExpr(tabColID)
} else {
mutationSuffix = "default"
expr = mb.parseDefaultExpr(tabColID)
}
// Add synthesized column. It is important to use the real column
// reference name, as this column may later be referred to by a computed
// column.
colName := scopeColName(tabCol.ColName()).WithMetadataName(
string(tabCol.ColName()) + "_" + mutationSuffix,
)
newCol, _ := pb.Add(colName, expr, tabCol.DatumType())
// Remember id of newly synthesized column.
colIDs[i] = newCol
// Add corresponding target column.
mb.targetColList = append(mb.targetColList, tabColID)
mb.targetColSet.Add(tabColID)
}
mb.outScope = pb.Finish()
}
// addSynthesizedComputedCols is a helper method for addSynthesizedColsForInsert
// and addSynthesizedColsForUpdate that scans the list of table columns, looking
// for any that are computed and do not yet have values provided by the input
// expression. New columns are synthesized for any missing columns using the
// computed column expression.
//
// NOTE: colIDs is updated with the column IDs of any synthesized columns which
// are added to mb.outScope. If restrict is true, only columns that depend on
// columns that were already in the list (plus all write-only columns) are
// updated.
func (mb *mutationBuilder) addSynthesizedComputedCols(colIDs opt.OptionalColList, restrict bool) {
// We will construct a new Project operator that will contain the newly
// synthesized column(s).
pb := makeProjectionBuilder(mb.b, mb.outScope)
var updatedColSet opt.ColSet
if restrict {
updatedColSet = colIDs.ToSet()
}
for i, n := 0, mb.tab.ColumnCount(); i < n; i++ {
tabCol := mb.tab.Column(i)
kind := tabCol.Kind()
if kind != cat.Ordinary && kind != cat.WriteOnly {
// Wrong kind.
continue
}
if !tabCol.IsComputed() {
continue
}
// Skip columns that are already specified (this is possible for upserts).
if colIDs[i] != 0 {
continue
}
// Create a new scope for resolving column references in computed column
// expressions. We cannot use mb.outScope because columns in that scope
// may be ambiguous, by design. We build a scope that contains a single
// column for each column in the target table, representing either an
// existing value (a column from mb.fetchColIDs) or a new value (a
// column from mb.upsertColIDs, mb.updateColIDs, or mb.insertColIDs).
if !pb.HasResolveScope() {
pb.SetResolveScope(mb.computedColumnScope())
}
tabColID := mb.tabID.ColumnID(i)
expr := mb.parseComputedExpr(tabColID)
// Add synthesized column.
colName := scopeColName(tabCol.ColName()).WithMetadataName(
string(tabCol.ColName()) + "_comp",
)
newCol, scalar := pb.Add(colName, expr, tabCol.DatumType())
if restrict && kind != cat.WriteOnly {
// Check if any of the columns referred to in the computed column
// expression are being updated.
var refCols opt.ColSet
if scalar == nil {
// When the expression is a simple column reference, we don't build a
// new scalar; we just use the same column ID.
refCols.Add(newCol)
} else {
var p props.Shared
memo.BuildSharedProps(scalar, &p, mb.b.evalCtx)
refCols = p.OuterCols
}
if !refCols.Intersects(updatedColSet) {
// Normalization rules will clean up the unnecessary projection.
continue
}
}
// Remember id of newly synthesized column.
colIDs[i] = newCol
// Add corresponding target column.
mb.targetColList = append(mb.targetColList, tabColID)
mb.targetColSet.Add(tabColID)
}
mb.outScope = pb.Finish()
}
// addCheckConstraintCols synthesizes a boolean output column for each check
// constraint defined on the target table. The mutation operator will report a
// constraint violation error if the value of the column is false.
//
// Synthesized check columns are not necessary for UPDATE mutations if the
// columns referenced in the check expression are not being mutated. If isUpdate
// is true, check columns that do not reference mutation columns are not added
// to checkColIDs, which allows pruning normalization rules to remove the
// unnecessary projected column.
func (mb *mutationBuilder) addCheckConstraintCols(isUpdate bool) {
if mb.tab.CheckCount() != 0 {
projectionsScope := mb.outScope.replace()
projectionsScope.appendColumnsFromScope(mb.outScope)
mutationCols := mb.mutationColumnIDs()
for i, n := 0, mb.tab.CheckCount(); i < n; i++ {
expr, err := parser.ParseExpr(mb.tab.Check(i).Constraint)
if err != nil {
panic(err)
}
texpr := mb.outScope.resolveAndRequireType(expr, types.Bool)
// Use an anonymous name because the column cannot be referenced
// in other expressions.
colName := scopeColName("").WithMetadataName(fmt.Sprintf("check%d", i+1))
scopeCol := projectionsScope.addColumn(colName, texpr)
// TODO(ridwanmsharif): Maybe we can avoid building constraints here
// and instead use the constraints stored in the table metadata.
referencedCols := &opt.ColSet{}
mb.b.buildScalar(texpr, mb.outScope, projectionsScope, scopeCol, referencedCols)
// If the mutation is not an UPDATE, track the synthesized check
// columns in checkColIDS. If the mutation is an UPDATE, only track
// the check columns if the columns referenced in the check
// expression are being mutated.
if !isUpdate || referencedCols.Intersects(mutationCols) {
mb.checkColIDs[i] = scopeCol.id
}
}
mb.b.constructProjectForScope(mb.outScope, projectionsScope)
mb.outScope = projectionsScope
}
}
// mutationColumnIDs returns the set of all column IDs that will be mutated.
func (mb *mutationBuilder) mutationColumnIDs() opt.ColSet {
cols := opt.ColSet{}
for _, col := range mb.insertColIDs {
if col != 0 {
cols.Add(col)
}
}
for _, col := range mb.updateColIDs {
if col != 0 {
cols.Add(col)
}
}
for _, col := range mb.upsertColIDs {
if col != 0 {
cols.Add(col)
}
}
return cols
}
// projectPartialIndexPutCols builds a Project that synthesizes boolean PUT
// columns for each partial index defined on the target table. See
// partialIndexPutColIDs for more info on these columns.
func (mb *mutationBuilder) projectPartialIndexPutCols() {
mb.projectPartialIndexColsImpl(mb.outScope, nil /* delScope */)
}
// projectPartialIndexDelCols builds a Project that synthesizes boolean PUT
// columns for each partial index defined on the target table. See
// partialIndexDelColIDs for more info on these columns.
func (mb *mutationBuilder) projectPartialIndexDelCols() {
mb.projectPartialIndexColsImpl(nil /* putScope */, mb.fetchScope)
}
// projectPartialIndexPutAndDelCols builds a Project that synthesizes boolean
// PUT and DEL columns for each partial index defined on the target table. See
// partialIndexPutColIDs and partialIndexDelColIDs for more info on these
// columns.
func (mb *mutationBuilder) projectPartialIndexPutAndDelCols() {
mb.projectPartialIndexColsImpl(mb.outScope, mb.fetchScope)
}
// projectPartialIndexColsImpl builds a Project that synthesizes boolean PUT and
// DEL columns for each partial index defined on the target table. PUT columns
// are only projected if putScope is non-nil and DEL columns are only projected
// if delScope is non-nil.
//
// NOTE: This function should only be called via projectPartialIndexPutCols,
// projectPartialIndexDelCols, or projectPartialIndexPutAndDelCols.
func (mb *mutationBuilder) projectPartialIndexColsImpl(putScope, delScope *scope) {
if partialIndexCount(mb.tab) > 0 {
projectionScope := mb.outScope.replace()
projectionScope.appendColumnsFromScope(mb.outScope)
ord := 0
for i, n := 0, mb.tab.DeletableIndexCount(); i < n; i++ {
index := mb.tab.Index(i)
// Skip non-partial indexes.
if _, isPartial := index.Predicate(); !isPartial {
continue
}
expr := mb.parsePartialIndexPredicateExpr(i)
// Build synthesized PUT columns.
if putScope != nil {
texpr := putScope.resolveAndRequireType(expr, types.Bool)
// Use an anonymous name because the column cannot be referenced
// in other expressions.
colName := scopeColName("").WithMetadataName(fmt.Sprintf("partial_index_put%d", ord+1))
scopeCol := projectionScope.addColumn(colName, texpr)
mb.b.buildScalar(texpr, putScope, projectionScope, scopeCol, nil)
mb.partialIndexPutColIDs[ord] = scopeCol.id
}
// Build synthesized DEL columns.
if delScope != nil {
texpr := delScope.resolveAndRequireType(expr, types.Bool)
// Use an anonymous name because the column cannot be referenced
// in other expressions.
colName := scopeColName("").WithMetadataName(fmt.Sprintf("partial_index_del%d", ord+1))
scopeCol := projectionScope.addColumn(colName, texpr)
mb.b.buildScalar(texpr, delScope, projectionScope, scopeCol, nil)
mb.partialIndexDelColIDs[ord] = scopeCol.id
}
ord++
}
mb.b.constructProjectForScope(mb.outScope, projectionScope)
mb.outScope = projectionScope
}
}
// computedColumnScope returns a new scope that can be used to build computed
// column expressions. Columns will never be ambiguous because each column in
// the returned scope maps to a single column in the target table.
//
// The columns included in the scope depend on the state of mb.upsertColIDs,
// mb.updateColIDs, mb.fetchColIDs, and mb.insertColIDs, using the same order of
// preference as disambiguateColumns (see mapToReturnColID). Therefore, this
// function will return different scopes at different stages of building a
// mutation statement. For example, when building the scan portion of an UPDATE,
// the scope will include columns from mb.fetchColIDs, while it will include
// columns from mb.updateColIDs or mb.fetchColIDs when building the SET portion
// of an UPDATE.
func (mb *mutationBuilder) computedColumnScope() *scope {
s := mb.b.allocScope()
for i, n := 0, mb.tab.ColumnCount(); i < n; i++ {
colID := mb.mapToReturnColID(i)
if colID == 0 {
continue
}
col := mb.outScope.getColumn(colID)
if col == nil {
panic(errors.AssertionFailedf("expected to find column %d in scope", colID))
}
targetCol := mb.tab.Column(i)
s.cols = append(s.cols, scopeColumn{
name: scopeColName(targetCol.ColName()),
typ: col.typ,
id: col.id,
})
}
return s
}
// disambiguateColumns ranges over the scope and ensures that at most one column
// has each table column name, and that name refers to the column with the final
// value that the mutation applies.
func (mb *mutationBuilder) disambiguateColumns() {
// Determine the set of input columns that will have their names preserved.
var preserve opt.ColSet
for i, n := 0, mb.tab.ColumnCount(); i < n; i++ {
if colID := mb.mapToReturnColID(i); colID != 0 {
preserve.Add(colID)
}
}
// Clear names of all non-preserved columns.
for i := range mb.outScope.cols {
if !preserve.Contains(mb.outScope.cols[i].id) {
mb.outScope.cols[i].clearName()
}
}
}
// makeMutationPrivate builds a MutationPrivate struct containing the table and
// column metadata needed for the mutation operator.
func (mb *mutationBuilder) makeMutationPrivate(needResults bool) *memo.MutationPrivate {
// Helper function that returns nil if there are no non-zero column IDs in a
// given list. A zero column ID indicates that column does not participate
// in this mutation operation.
checkEmptyList := func(colIDs opt.OptionalColList) opt.OptionalColList {
if colIDs.IsEmpty() {
return nil
}
return colIDs
}
private := &memo.MutationPrivate{
Table: mb.tabID,
InsertCols: checkEmptyList(mb.insertColIDs),
FetchCols: checkEmptyList(mb.fetchColIDs),
UpdateCols: checkEmptyList(mb.updateColIDs),
CanaryCol: mb.canaryColID,
ArbiterIndexes: mb.arbiters.IndexOrdinals(),
ArbiterConstraints: mb.arbiters.UniqueConstraintOrdinals(),
CheckCols: checkEmptyList(mb.checkColIDs),
PartialIndexPutCols: checkEmptyList(mb.partialIndexPutColIDs),
PartialIndexDelCols: checkEmptyList(mb.partialIndexDelColIDs),
FKCascades: mb.cascades,
}
// If we didn't actually plan any checks or cascades, don't buffer the input.
if len(mb.uniqueChecks) > 0 || len(mb.fkChecks) > 0 || len(mb.cascades) > 0 {
private.WithID = mb.withID
}
if needResults {
private.ReturnCols = make(opt.OptionalColList, mb.tab.ColumnCount())
for i, n := 0, mb.tab.ColumnCount(); i < n; i++ {
if kind := mb.tab.Column(i).Kind(); kind != cat.Ordinary {
// Only non-mutation and non-system columns are output columns.
continue
}
retColID := mb.mapToReturnColID(i)
if retColID == 0 {
panic(errors.AssertionFailedf("column %d is not available in the mutation input", i))
}
private.ReturnCols[i] = retColID
}
}
return private
}
// mapToReturnColID returns the ID of the input column that provides the final
// value for the column at the given ordinal position in the table. This value
// might mutate the column, or it might be returned by the mutation statement,
// or it might not be used at all. Columns take priority in this order:
//
// upsert, update, fetch, insert
//
// If an upsert column is available, then it already combines an update/fetch
// value with an insert value, so it takes priority. If an update column is
// available, then it overrides any fetch value. Finally, the relative priority
// of fetch and insert columns doesn't matter, since they're only used together
// in the upsert case where an upsert column would be available.
//
// If the column is never referenced by the statement, then mapToReturnColID
// returns 0. This would be the case for delete-only columns in an Insert
// statement, because they're neither fetched nor mutated.