-
Notifications
You must be signed in to change notification settings - Fork 977
/
assign.c
1336 lines (1283 loc) · 75.7 KB
/
assign.c
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
#include "data.table.h"
static void finalizer(SEXP p)
{
SEXP x;
R_len_t n, l, tl;
if(!R_ExternalPtrAddr(p)) internal_error(__func__, "didn't receive an ExternalPtr"); // # nocov
p = R_ExternalPtrTag(p);
if (!isString(p)) internal_error(__func__, "ExternalPtr doesn't see names in tag"); // # nocov
l = LENGTH(p);
tl = TRUELENGTH(p);
if (l<0 || tl<l) internal_error(__func__, "l=%d, tl=%d", l, tl); // # nocov
n = tl-l;
if (n==0) {
// gc's ReleaseLargeFreeVectors() will have reduced R_LargeVallocSize by the correct amount
// already, so nothing to do (but almost never the case).
return;
}
x = PROTECT(allocVector(INTSXP, 50)); // 50 so it's big enough to be on LargeVector heap. See NodeClassSize in memory.c:allocVector
// INTSXP rather than VECSXP so that GC doesn't inspect contents after LENGTH (thanks to Karl Miller, Jul 2015)
SETLENGTH(x,50+n*2*sizeof(void *)/4); // 1*n for the names, 1*n for the VECSXP itself (both are over allocated).
UNPROTECT(1);
return;
}
void setselfref(SEXP x) {
if(!INHERITS(x, char_datatable)) return; // #5286
SEXP p;
// Store pointer to itself so we can detect if the object has been copied. See
// ?copy for why copies are not just inefficient but cause a problem for over-allocated data.tables.
// Called from C only, not R level, so returns void.
setAttrib(x, SelfRefSymbol, p=R_MakeExternalPtr(
R_NilValue, // for identical() to return TRUE. identical() doesn't look at tag and prot
PROTECT(getAttrib(x, R_NamesSymbol)), // to detect if names has been replaced and its tl lost, e.g. setattr(DT,"names",...)
PROTECT(R_MakeExternalPtr( // to avoid an infinite loop in object.size(), if prot=x here
x, // to know if this data.table has been copied by key<-, attr<-, names<-, etc.
R_NilValue, // this tag and prot currently unused
R_NilValue
))
));
R_RegisterCFinalizerEx(p, finalizer, FALSE);
UNPROTECT(2);
/*
* base::identical doesn't check prot and tag of EXTPTR, just that the ptr itself is the
same in both objects. R_NilValue is always equal to R_NilValue. R_NilValue is a memory
location constant within an R session, but can vary from session to session. So, it
looks like a pointer to a user looking at attributes(DT), but they might wonder how it
works if they realise the selfref of all data.tables all point to the same address (rather
than to the table itself which would be reasonable to expect given the attribute's name).
* p=NULL rather than R_NilValue works too, other than we need something other than NULL
so we can detect tables loaded from disk (which set p to NULL, see 5.13 of R-exts).
* x is wrapped in another EXTPTR because of object.size (called by tables(), and by users).
If the prot (or tag) was x directly it sends object.size into an infinite loop and then
"segfault from C stack overflow" (object.size does count tag and prot, unlike identical,
but doesn't count what's pointed to).
* Could use weak reference possibly, but the fact that they can get be set to R_NilValue
by gc (?) didn't seem appropriate.
* If the .internal.selfref attribute is removed (e.g. by user code), nothing will break, but
an extra copy will just be taken on next :=, with warning, with a new selfref.
* object.size will count size of names twice, but that's ok as only small.
* Thanks to Steve L for suggesting ExtPtr for this, rather than the previous REALSXP
vector which required data.table to do a show/hide dance in a masked identical.
*/
}
/* There are two reasons the finalizer doesn't restore the LENGTH to TRUELENGTH. i) The finalizer
happens too late after GC has already released the memory, and ii) copies by base R (e.g.
[<- in write.table just before test 894) allocate at length LENGTH but copy the TRUELENGTH over.
If the finalizer sets LENGTH to TRUELENGTH, that's a fail as it wasn't really allocated at
TRUELENGTH when R did the copy.
Karl Miller suggested an ENVSXP so that restoring LENGTH in finalizer should work. This is the
closest I got to getting it to pass all tests :
SEXP env = PROTECT(allocSExp(ENVSXP));
defineVar(SelfRefSymbol, x, env);
defineVar(R_NamesSymbol, getAttrib(x, R_NamesSymbol), env);
setAttrib(x, SelfRefSymbol, p = R_MakeExternalPtr(
R_NilValue, // for identical() to return TRUE. identical() doesn't look at tag and prot
R_NilValue, //getAttrib(x, R_NamesSymbol), // to detect if names has been replaced and its tl lost, e.g. setattr(DT,"names",...)
PROTECT( // needed when --enable-strict-barrier it seems, iiuc. TO DO: test under that flag and remove if not needed.
env // wrap x in env to avoid an infinite loop in object.size() if prot=x were here
)
));
R_RegisterCFinalizerEx(p, finalizer, FALSE);
UNPROTECT(2);
Then in finalizer:
SETLENGTH(names, tl)
SETLENGTH(dt, tl)
and that finalizer indeed now happens before the GC releases memory (thanks to the env wrapper).
However, we still have problem (ii) above and it didn't pass tests involving base R copies.
We really need R itself to start setting TRUELENGTH to be the allocated length and then
for GC to release TRUELENGTH not LENGTH. Would really tidy this up.
Moved out of ?setkey Details section in 1.12.2 (Mar 2019). Revisit this w.r.t. to recent versions of R.
The problem (for \code{data.table}) with the copy by \code{key<-} (other than
being slower) is that \R doesn't maintain the over-allocated truelength, but it
looks as though it has. Adding a column by reference using \code{:=} after a
\code{key<-} was therefore a memory overwrite and eventually a segfault; the
over-allocated memory wasn't really there after \code{key<-}'s copy. \code{data.table}s now have an attribute \code{.internal.selfref} to catch and warn about such copies.
This attribute has been implemented in a way that is friendly with
\code{identical()} and \code{object.size()}.
*/
static int _selfrefok(SEXP x, Rboolean checkNames, Rboolean verbose) {
SEXP v, p, tag, prot, names;
v = getAttrib(x, SelfRefSymbol);
if (v==R_NilValue || TYPEOF(v)!=EXTPTRSXP) {
// .internal.selfref missing is expected and normal for i) a pre v1.7.8 data.table loaded
// from disk, and ii) every time a new data.table is over-allocated for the first time.
// Not being an extptr is for when users construct a data.table via structure() using dput, post
// a question, and find the extptr doesn't parse so put quotes around it (for example).
// In both cases the selfref is not ok.
return 0;
}
p = R_ExternalPtrAddr(v);
if (p==NULL) {
if (verbose) Rprintf(_(".internal.selfref ptr is NULL. This is expected and normal for a data.table loaded from disk. Please remember to always setDT() immediately after loading to prevent unexpected behavior. If this table was not loaded from disk or you've already run setDT(), please report to data.table issue tracker.\n"));
return -1;
}
if (!isNull(p)) internal_error(__func__, ".internal.selfref ptr is neither NULL nor R_NilValue"); // # nocov
tag = R_ExternalPtrTag(v);
if (!(isNull(tag) || isString(tag))) internal_error(__func__, ".internal.selfref tag is neither NULL nor a character vector"); // # nocov
names = getAttrib(x, R_NamesSymbol);
if (names!=tag && isString(names) && !ALTREP(names)) // !ALTREP for #4734
SET_TRUELENGTH(names, LENGTH(names));
// R copied this vector not data.table; it's not actually over-allocated. It looks over-allocated
// because R copies the original vector's tl over despite allocating length.
prot = R_ExternalPtrProtected(v);
if (TYPEOF(prot) != EXTPTRSXP) // Very rare. Was error(_(".internal.selfref prot is not itself an extptr")).
return 0; // # nocov ; see http://stackoverflow.com/questions/15342227/getting-a-random-internal-selfref-error-in-data-table-for-r
if (x!=R_ExternalPtrAddr(prot) && !ALTREP(x))
SET_TRUELENGTH(x, LENGTH(x)); // R copied this vector not data.table, it's not actually over-allocated
return checkNames ? names==tag : x==R_ExternalPtrAddr(prot);
}
static Rboolean selfrefok(SEXP x, Rboolean verbose) { // for readability
return(_selfrefok(x, FALSE, verbose)==1);
}
static Rboolean selfrefnamesok(SEXP x, Rboolean verbose) {
return(_selfrefok(x, TRUE, verbose)==1);
}
static SEXP shallow(SEXP dt, SEXP cols, R_len_t n)
{
// NEW: cols argument to specify the columns to shallow copy on. If NULL, all columns.
// called from alloccol where n is checked carefully, or from shallow() at R level
// where n is set to truelength (i.e. a shallow copy only with no size change)
int protecti=0;
SEXP newdt = PROTECT(allocVector(VECSXP, n)); protecti++; // to do, use growVector here?
SHALLOW_DUPLICATE_ATTRIB(newdt, dt);
// TO DO: keepattr() would be faster, but can't because shallow isn't merely a shallow copy. It
// also increases truelength. Perhaps make that distinction, then, and split out, but marked
// so that the next change knows to duplicate.
// keepattr() also merely points to the entire attributes list and thus doesn't allow replacing
// some of its elements.
// We copy all attributes that refer to column names so that calling setnames on either
// the original or the shallow copy doesn't break anything.
SEXP index = PROTECT(getAttrib(dt, sym_index)); protecti++;
setAttrib(newdt, sym_index, shallow_duplicate(index));
SEXP sorted = PROTECT(getAttrib(dt, sym_sorted)); protecti++;
setAttrib(newdt, sym_sorted, duplicate(sorted));
SEXP names = PROTECT(getAttrib(dt, R_NamesSymbol)); protecti++;
SEXP newnames = PROTECT(allocVector(STRSXP, n)); protecti++;
const int l = isNull(cols) ? LENGTH(dt) : length(cols);
if (isNull(cols)) {
for (int i=0; i<l; ++i) SET_VECTOR_ELT(newdt, i, VECTOR_ELT(dt,i));
if (length(names)) {
if (length(names) < l) internal_error(__func__, "length(names)>0 but <length(dt)"); // # nocov
for (int i=0; i<l; ++i) SET_STRING_ELT(newnames, i, STRING_ELT(names,i));
}
// else an unnamed data.table is valid e.g. unname(DT) done by ggplot2, and .SD may have its names cleared in dogroups, but shallow will always create names for data.table(NULL) which has 100 slots all empty so you can add to an empty data.table by reference ok.
} else {
for (int i=0; i<l; ++i) SET_VECTOR_ELT(newdt, i, VECTOR_ELT(dt,INTEGER(cols)[i]-1));
if (length(names)) {
// no need to check length(names) < l here. R-level checks if all value
// in 'cols' are valid - in the range of 1:length(names(x))
for (int i=0; i<l; ++i) SET_STRING_ELT( newnames, i, STRING_ELT(names,INTEGER(cols)[i]-1) );
}
}
setAttrib(newdt, R_NamesSymbol, newnames);
// setAttrib appears to change length and truelength, so need to do that first _then_ SET next,
// otherwise (if the SET were were first) the 100 tl is assigned to length.
SETLENGTH(newnames,l);
SET_TRUELENGTH(newnames,n);
SETLENGTH(newdt,l);
SET_TRUELENGTH(newdt,n);
setselfref(newdt);
UNPROTECT(protecti);
return(newdt);
}
// Wrapped in a function so the same message is issued for the data.frame case at the R level
void warn_matrix_column(/* 1-indexed */ int i) {
warning(_("Some columns are a multi-column type (such as a matrix column), for example column %d. setDT will retain these columns as-is but subsequent operations like grouping and joining may fail. Please consider as.data.table() instead which will create a new column for each embedded column."), i);
}
// input validation for setDT() list input; assume is.list(x) was tested in R
SEXP setdt_nrows(SEXP x)
{
int base_length = 0;
bool test_matrix_cols = true;
for (R_len_t i = 0; i < LENGTH(x); ++i) {
SEXP xi = VECTOR_ELT(x, i);
/* allow NULL columns to be created by setDT(list) even though they are not really allowed
* many operations still work in the presence of NULL columns and it might be convenient
* e.g. in package eplusr which calls setDT on a list when parsing JSON. Operations which
* fail for NULL columns will give helpful error at that point, #3480 and #3471 */
if (Rf_isNull(xi)) continue;
if (Rf_inherits(xi, "POSIXlt")) {
error(_("Column %d has class 'POSIXlt'. Please convert it to POSIXct (using as.POSIXct) and run setDT() again. We do not recommend the use of POSIXlt at all because it uses 40 bytes to store one date."), i+1);
}
SEXP dim_xi = getAttrib(xi, R_DimSymbol);
R_len_t len_xi;
// NB: LENGTH() produces an undefined large number here on R 3.3.0.
// There's also a note in NEWS for R 3.1.0 saying length() should always be used by packages,
// but with some overhead for being a function/not macro...
R_len_t n_dim = length(dim_xi);
if (n_dim) {
if (test_matrix_cols && n_dim > 1) {
warn_matrix_column(i+1);
test_matrix_cols = false;
}
len_xi = INTEGER(dim_xi)[0];
} else {
len_xi = LENGTH(xi);
}
if (!base_length) {
base_length = len_xi;
} else if (len_xi != base_length) {
error(_("All elements in argument 'x' to 'setDT' must be of equal length, but input %d has length %d whereas the first non-empty input had length %d"), i+1, len_xi, base_length);
}
}
return ScalarInteger(base_length);
}
SEXP alloccol(SEXP dt, R_len_t n, Rboolean verbose)
{
SEXP names, klass; // klass not class at request of pydatatable because class is reserved word in C++, PR #3129
R_len_t l, tl;
if (isNull(dt)) error(_("alloccol has been passed a NULL dt"));
if (TYPEOF(dt) != VECSXP) error(_("dt passed to alloccol isn't type VECSXP"));
klass = getAttrib(dt, R_ClassSymbol);
if (isNull(klass)) error(_("dt passed to alloccol has no class attribute. Please report result of traceback() to data.table issue tracker."));
l = LENGTH(dt);
names = getAttrib(dt,R_NamesSymbol);
// names may be NULL when null.data.table() passes list() to alloccol for example.
// So, careful to use length() on names, not LENGTH().
if (length(names)!=l) internal_error(__func__, "length of names (%d) is not length of dt (%d)", length(names),l); // # nocov
if (!selfrefok(dt,verbose))
return shallow(dt,R_NilValue,(n>l) ? n : l); // e.g. test 848 and 851 in R > 3.0.2
// added (n>l) ? ... for #970, see test 1481.
// TO DO: test realloc names if selfrefnamesok (users can setattr(x,"name") themselves for example.
// if (TRUELENGTH(getAttrib(dt,R_NamesSymbol))!=tl)
// internal_error(__func__, "tl of dt passes checks, but tl of names (%d) != tl of dt (%d)", tl, TRUELENGTH(getAttrib(dt,R_NamesSymbol))); // # nocov
tl = TRUELENGTH(dt);
// R <= 2.13.2 and we didn't catch uninitialized tl somehow
if (tl<0) internal_error(__func__, "tl of class is marked but tl<0"); // # nocov
if (tl>0 && tl<l) internal_error(__func__, "tl (%d) < l (%d) but tl of class is marked", tl, l); // # nocov
if (tl>l+10000) warning(_("tl (%d) is greater than 10,000 items over-allocated (l = %d). If you didn't set the datatable.alloccol option to be very large, please report to data.table issue tracker including the result of sessionInfo()."),tl,l);
if (n>tl) return(shallow(dt,R_NilValue,n)); // usual case (increasing alloc)
if (n<tl && verbose) Rprintf(_("Attempt to reduce allocation from %d to %d ignored. Can only increase allocation via shallow copy. Please do not use DT[...]<- or DT$someCol<-. Use := inside DT[...] instead."),tl,n);
// otherwise the finalizer can't clear up the Large Vector heap
return(dt);
}
int checkOverAlloc(SEXP x)
{
if (isNull(x))
error(_("Has getOption('datatable.alloccol') somehow become unset? It should be a number, by default 1024."));
if (!isInteger(x) && !isReal(x))
error(_("getOption('datatable.alloccol') should be a number, by default 1024. But its type is '%s'."), type2char(TYPEOF(x)));
if (LENGTH(x) != 1)
error(_("getOption('datatable.alloc') is a numeric vector ok but its length is %d. Its length should be 1."), LENGTH(x));
int ans = asInteger(x);
if (ans<0)
error(_("getOption('datatable.alloc')==%d. It must be >=0 and not NA."), ans);
return ans;
}
SEXP alloccolwrapper(SEXP dt, SEXP overAllocArg, SEXP verbose) {
if (!IS_TRUE_OR_FALSE(verbose))
error(_("%s must be TRUE or FALSE"), "verbose");
int overAlloc = checkOverAlloc(overAllocArg);
SEXP ans = PROTECT(alloccol(dt, length(dt)+overAlloc, LOGICAL(verbose)[0]));
for(R_len_t i = 0; i < LENGTH(ans); i++) {
// clear names; also excluded by copyMostAttrib(). Primarily for data.table and as.data.table, but added here centrally (see #103).
setAttrib(VECTOR_ELT(ans, i), R_NamesSymbol, R_NilValue);
// But don't clear dim and dimnames. Because as from 1.12.4 we keep the matrix column as-is and ask user to use as.data.table to
// unpack matrix columns when they really need to; test 2089.2
// setAttrib(VECTOR_ELT(ans, i), R_DimSymbol, R_NilValue);
// setAttrib(VECTOR_ELT(ans, i), R_DimNamesSymbol, R_NilValue);
}
UNPROTECT(1);
return ans;
}
SEXP shallowwrapper(SEXP dt, SEXP cols) {
// selfref will be FALSE on manually created data.table, e.g., via dput() or structure()
if (!selfrefok(dt, FALSE)) {
int n = isNull(cols) ? length(dt) : length(cols);
return(shallow(dt, cols, n));
} else return(shallow(dt, cols, TRUELENGTH(dt)));
}
SEXP truelength(SEXP x) {
return ScalarInteger(isNull(x) ? 0 : TRUELENGTH(x));
}
SEXP selfrefokwrapper(SEXP x, SEXP verbose) {
return ScalarInteger(_selfrefok(x,FALSE,LOGICAL(verbose)[0]));
}
int *_Last_updated = NULL;
SEXP assign(SEXP dt, SEXP rows, SEXP cols, SEXP newcolnames, SEXP values)
{
// For internal use only by := in [.data.table, and set()
// newcolnames : add these columns (if any)
// cols : column names or numbers corresponding to the values to set
// rows : row numbers to assign
R_len_t numToDo, targetlen, vlen, oldncol, oldtncol, coln, protecti=0, newcolnum, indexLength;
SEXP targetcol, nullint, s, colnam, tmp, key, index, a, assignedNames, indexNames;
bool verbose=GetVerbose();
int ndelete=0; // how many columns are being deleted
const char *c1, *tc1, *tc2;
int *buf, indexNo;
if (isNull(dt)) error(_("assign has been passed a NULL dt"));
if (TYPEOF(dt) != VECSXP) error(_("dt passed to assign isn't type VECSXP"));
if (islocked(dt))
error(_(".SD is locked. Updating .SD by reference using := or set are reserved for future use. Use := in j directly. Or use copy(.SD) as a (slow) last resort, until shallow() is exported."));
// We allow set() on data.frame too; e.g. package Causata uses set() on a data.frame in tests/testTransformationReplay.R
// := is only allowed on a data.table. However, the ":=" = stop(...) message in data.table.R will have already
// detected use on a data.frame before getting to this point.
// For data.frame, can use set() on existing columns but not add new ones because DF are not over-allocated.
bool isDataTable = INHERITS(dt, char_datatable);
if (!isDataTable && !INHERITS(dt, char_dataframe))
internal_error(__func__, "dt is not a data.table or data.frame"); // # nocov
oldncol = LENGTH(dt);
SEXP names = PROTECT(getAttrib(dt, R_NamesSymbol)); protecti++;
if (isNull(names)) error(_("dt passed to assign has no names"));
if (length(names)!=oldncol)
internal_error(__func__, "length of names (%d) is not length of dt (%d)", length(names), oldncol); // # nocov
if (isNull(dt)) {
error(_("data.table is NULL; malformed. A null data.table should be an empty list. typeof() should always return 'list' for data.table.")); // # nocov
// Not possible to test because R won't permit attributes be attached to NULL (which is good and we like); warning from R 3.4.0+ tested by 944.5
}
const int nrow = LENGTH(dt) ? length(VECTOR_ELT(dt,0)) :
(isNewList(values) && length(values) && (length(values)==length(cols)) ? length(VECTOR_ELT(values,0)) : length(values));
// ^ when null data.table the new nrow becomes the fist column added
// ^ distinguish between adding list column or multiple columns at once, #5738
if (isNull(rows)) {
numToDo = nrow;
targetlen = nrow;
if (verbose) Rprintf(_("Assigning to all %d rows\n"), nrow);
// fast way to assign to whole column, without creating 1:nrow(x) vector up in R, or here in C
} else {
if (isReal(rows)) {
rows = PROTECT(coerceVector(rows, INTSXP)); protecti++;
warning(_("Coerced i from numeric to integer. Please pass integer for efficiency; e.g., 2L rather than 2"));
}
if (!isInteger(rows))
error(_("i is type '%s'. Must be integer, or numeric is coerced with warning. If i is a logical subset, simply wrap with which(), and take the which() outside the loop if possible for efficiency."), type2char(TYPEOF(rows)));
targetlen = length(rows);
numToDo = 0;
const int *rowsd = INTEGER(rows);
for (int i=0; i<targetlen; ++i) {
if ((rowsd[i]<0 && rowsd[i]!=NA_INTEGER) || rowsd[i]>nrow)
error(_("i[%d] is %d which is out of range [1,nrow=%d]"), i+1, rowsd[i], nrow); // set() reaches here (test 2005.2); := reaches the same error in subset.c first
if (rowsd[i]>=0) numToDo++;
}
if (verbose) Rprintf(_("Assigning to %d row subset of %d rows\n"), numToDo, nrow);
// TODO: include in message if any rows are assigned several times (e.g. by=.EACHI with dups in i)
if (numToDo==0) {
// isString(cols) is exclusive to calls from set()
if (!length(newcolnames) && !isString(cols)) {
*_Last_updated = 0;
UNPROTECT(protecti);
return(dt); // all items of rows either 0 or NA. !length(newcolnames) for #759
}
if (verbose) Rprintf(Pl_(length(newcolnames),
"Added %d new column initialized with all-NA\n",
"Added %d new columns initialized with all-NA\n"),
length(newcolnames));
}
}
if (!length(cols)) {
if (verbose) Rprintf(_("length(LHS)==0; no columns to delete or assign RHS to.")); // test 1295 covers
*_Last_updated = 0;
UNPROTECT(protecti);
return(dt);
}
// FR #2077 - set able to add new cols by reference
if (isString(cols)) {
PROTECT(tmp = chmatch(cols, names, 0)); protecti++;
buf = (int *) R_alloc(length(cols), sizeof(int));
int k=0;
for (int i=0; i<length(cols); ++i) {
if (INTEGER(tmp)[i] == 0) buf[k++] = i;
}
if (k>0) {
if (!isDataTable) error(_("set() on a data.frame is for changing existing columns, not adding new ones. Please use a data.table for that. data.table's are over-allocated and don't shallow copy."));
newcolnames = PROTECT(allocVector(STRSXP, k)); protecti++;
for (int i=0; i<k; ++i) {
SET_STRING_ELT(newcolnames, i, STRING_ELT(cols, buf[i]));
INTEGER(tmp)[buf[i]] = oldncol+i+1;
}
}
cols = tmp;
} else {
if (isReal(cols)) {
cols = PROTECT(coerceVector(cols, INTSXP)); protecti++;
warning(_("Coerced j from numeric to integer. Please pass integer for efficiency; e.g., 2L rather than 2"));
}
if (!isInteger(cols))
error(_("j is type '%s'. Must be integer, character, or numeric is coerced with warning."), type2char(TYPEOF(cols)));
}
if (any_duplicated(cols,FALSE)) error(_("Can't assign to the same column twice in the same query (duplicates detected)."));
if (!isNull(newcolnames) && !isString(newcolnames)) error(_("newcolnames is supplied but isn't a character vector"));
if (Rf_inherits(values, "POSIXlt")) {
warning(_("Values of type POSIXlt detected and converted to POSIXct. We do not recommend the use of POSIXlt at all because it typically takes more than 6 times the storage as an equivalent POSIXct column. Use as.POSIXct() to avoid this warning."));
values = PROTECT(eval(PROTECT(lang2(sym_as_posixct, values)), R_GlobalEnv)); protecti+=2;
}
bool RHS_list_of_columns = TYPEOF(values)==VECSXP && length(cols)>1; // initial value; may be revised below
if (verbose) Rprintf(_("RHS_list_of_columns == %s\n"), RHS_list_of_columns ? "true" : "false");
if (TYPEOF(values)==VECSXP && length(cols)==1 && length(values)==1) {
SEXP item = VECTOR_ELT(values,0);
if (isNull(item) || length(item)==1 || length(item)==targetlen) {
RHS_list_of_columns=true;
if (verbose) Rprintf(_("RHS_list_of_columns revised to true because RHS list has 1 item which is NULL, or whose length %d is either 1 or targetlen (%d). Please unwrap RHS.\n"), length(item), targetlen);
}
}
if (RHS_list_of_columns) {
if (length(values)==0)
error(_("Supplied %d columns to be assigned an empty list (which may be an empty data.table or data.frame since they are lists too). To delete multiple columns use NULL instead. To add multiple empty list columns, use list(list())."), length(cols));
if (length(values)!=length(cols)) {
if (length(values)==1) { // test 351.1; c("colA","colB"):=list(13:15) uses 13:15 for both columns
values = VECTOR_ELT(values,0);
RHS_list_of_columns = false;
if (verbose) Rprintf(_("Recycling single RHS list item across %d columns. Please unwrap RHS.\n"), length(cols));
} else {
error(_("Supplied %d columns to be assigned %d items. Please see NEWS for v1.12.2."), length(cols), length(values));
}
}
}
// Check all inputs :
for (int i=0; i<length(cols); ++i) {
coln = INTEGER(cols)[i];
if (coln<1 || coln>oldncol+length(newcolnames)) {
if (!isDataTable) error(_("Item %d of column numbers in j is %d which is outside range [1,ncol=%d]. set() on a data.frame is for changing existing columns, not adding new ones. Please use a data.table for that."), i+1, coln, oldncol);
else error(_("Item %d of column numbers in j is %d which is outside range [1,ncol=%d]. Use column names instead in j to add new columns."), i+1, coln, oldncol);
}
coln--;
SEXP thisvalue = RHS_list_of_columns ? VECTOR_ELT(values, i) : values;
if (isNull(thisvalue) && !isNull(rows)) {
if (coln >= oldncol)
error(_("Doubly-invalid attempt to delete a non-existent column while also providing i"));
if (TYPEOF(VECTOR_ELT(dt, coln)) == VECSXP)
error(_("Invalid attempt to delete a list column while also providing i; did you intend to add NULL to those rows instead? If so, use list_col := list(list(NULL)).")); // #5526
error(_("When deleting columns, i should not be provided")); // #1082, #3089
}
if (coln+1 <= oldncol) colnam = STRING_ELT(names,coln);
else colnam = STRING_ELT(newcolnames,coln-length(names));
if (coln+1 <= oldncol && isNull(thisvalue)) continue; // delete existing column(s) afterwards, near end of this function
vlen = length(thisvalue);
//if (vlen<1 && nrow>0) {
if (coln+1 <= oldncol && nrow>0 && vlen<1 && numToDo>0) { // numToDo > 0 fixes #2829, see test 1911
error(_("RHS of assignment to existing column '%s' is zero length but not NULL. If you intend to delete the column use NULL. Otherwise, the RHS must have length > 0; e.g., NA_integer_. If you are trying to change the column type to be an empty list column then, as with all column type changes, provide a full length RHS vector such as vector('list',nrow(DT)); i.e., 'plonk' in the new column."), CHAR(STRING_ELT(names,coln)));
}
if (coln+1 > oldncol && TYPEOF(thisvalue)!=VECSXP) { // list() is ok for new columns
newcolnum = coln-length(names);
if (newcolnum<0 || newcolnum>=length(newcolnames))
internal_error(__func__, "length(newcolnames)=%d, length(names)=%d, coln=%d", length(newcolnames), length(names), coln); // # nocov
if (isNull(thisvalue)) {
warning(_("Tried to assign NULL to column '%s', but this column does not exist to remove"), CHAR(STRING_ELT(newcolnames,newcolnum)));
continue;
}
// RHS of assignment to new column is zero length but we'll use its type to create all-NA column of that type
}
{
int j;
if (isMatrix(thisvalue) && (j=INTEGER(getAttrib(thisvalue, R_DimSymbol))[1]) > 1) // matrix passes above (considered atomic vector)
warning(_("%d column matrix RHS of := will be treated as one vector"), j);
}
const SEXP existing = (coln+1)<=oldncol ? VECTOR_ELT(dt,coln) : R_NilValue;
if (isFactor(existing) &&
!isString(thisvalue) && TYPEOF(thisvalue)!=INTSXP && TYPEOF(thisvalue)!=LGLSXP && !isReal(thisvalue) && !isNewList(thisvalue)) { // !=INTSXP includes factor
error(_("Can't assign to column '%s' (type 'factor') a value of type '%s' (not character, factor, integer or numeric)"),
CHAR(STRING_ELT(names,coln)),type2char(TYPEOF(thisvalue)));
}
if (nrow>0 && targetlen>0 && vlen>1 && vlen!=targetlen && !(TYPEOF(existing)==VECSXP && targetlen==1)) {
// We allow assigning objects of arbitrary to single items of list columns for convenience.
// Note that isNewList(R_NilValue) is true so it needs to be !(TYPEOF(existing)==VECSXP) above
error(_("Supplied %d items to be assigned to %d items of column '%s'. If you wish to 'recycle' the RHS please use rep() to make this intent clear to readers of your code."), vlen, targetlen, CHAR(colnam));
}
}
// having now checked the inputs, from this point there should be no errors so we can now proceed to
// modify DT by reference. Other than if new columns are being added and the allocVec() fails with
// out-of-memory. In that case the user will receive hard halt and know to rerun.
if (length(newcolnames)) {
oldtncol = TRUELENGTH(dt); // TO DO: oldtncol can be just called tl now, as we won't realloc here any more.
if (oldtncol<oldncol) {
if (oldtncol==0) error(_("This data.table has either been loaded from disk (e.g. using readRDS()/load()) or constructed manually (e.g. using structure()). Please run setDT() or setalloccol() on it first (to pre-allocate space for new columns) before assigning by reference to it.")); // #2996
internal_error(__func__, "oldtncol(%d) < oldncol(%d)", oldtncol, oldncol); // # nocov
}
if (oldtncol>oldncol+10000L) warning(_("truelength (%d) is greater than 10,000 items over-allocated (length = %d). See ?truelength. If you didn't set the datatable.alloccol option very large, please report to data.table issue tracker including the result of sessionInfo()."),oldtncol, oldncol);
if (oldtncol < oldncol+LENGTH(newcolnames))
internal_error(__func__, "input dt has not been allocated enough column slots. l=%d, tl=%d, adding %d", oldncol, oldtncol, LENGTH(newcolnames)); // # nocov
if (!selfrefnamesok(dt,verbose))
error(_("It appears that at some earlier point, names of this data.table have been reassigned. Please ensure to use setnames() rather than names<- or colnames<-. Otherwise, please report to data.table issue tracker.")); // # nocov
// Can growVector at this point easily enough, but it shouldn't happen in first place so leave it as
// strong error message for now.
else if (TRUELENGTH(names) != oldtncol)
// Use (long long) to cast R_xlen_t to a fixed type to robustly avoid -Wformat compiler warnings, see #5768, PRId64 didn't work
internal_error(__func__, "selfrefnames is ok but tl names [%lld] != tl [%d]", (long long)TRUELENGTH(names), oldtncol); // # nocov
SETLENGTH(dt, oldncol+LENGTH(newcolnames));
SETLENGTH(names, oldncol+LENGTH(newcolnames));
for (int i=0; i<LENGTH(newcolnames); ++i)
SET_STRING_ELT(names,oldncol+i,STRING_ELT(newcolnames,i));
// truelengths of both already set by alloccol
if (oldncol==0) {
// adding columns to null data.table needs row.names set, #4597
SEXP rn;
PROTECT(rn = allocVector(INTSXP, 2)); protecti++;
INTEGER(rn)[0] = NA_INTEGER;
INTEGER(rn)[1] = -nrow;
setAttrib(dt, R_RowNamesSymbol, rn);
}
}
for (int i=0; i<length(cols); ++i) {
coln = INTEGER(cols)[i]-1;
SEXP thisvalue = RHS_list_of_columns ? VECTOR_ELT(values, i) : values;
if (TYPEOF(thisvalue)==NILSXP) {
if (!isNull(rows)) internal_error(__func__, "earlier error 'When deleting columns, i should not be provided' did not happen"); // # nocov
ndelete++;
continue; // delete column(s) afterwards, below this loop
}
vlen = length(thisvalue);
if (length(rows)==0 && targetlen==vlen && (vlen>0 || nrow==0)) {
if ( MAYBE_SHARED(thisvalue) || // set() protects the NAMED of atomic vectors from .Call setting arguments to 2 by wrapping with list
(TYPEOF(values)==VECSXP && i>LENGTH(values)-1) || // recycled RHS would have columns pointing to others, #185.
(TYPEOF(values)!=VECSXP && i>0) // assigning the same values to a second column. Have to ensure a copy #2540
) {
if (verbose) {
Rprintf(_("RHS for item %d has been duplicated because MAYBE_REFERENCED==%d MAYBE_SHARED==%d, but then is being plonked. length(values)==%d; length(cols)==%d)\n"),
i+1, MAYBE_REFERENCED(thisvalue), MAYBE_SHARED(thisvalue), length(values), length(cols));
}
thisvalue = copyAsPlain(thisvalue); // PROTECT not needed as assigned as element to protected list below.
} else {
if (verbose) Rprintf(_("Direct plonk of unnamed RHS, no copy. MAYBE_REFERENCED==%d, MAYBE_SHARED==%d\n"), MAYBE_REFERENCED(thisvalue), MAYBE_SHARED(thisvalue)); // e.g. DT[,a:=as.character(a)] as tested by 754.5
}
SET_VECTOR_ELT(dt, coln, thisvalue); // plonk new column in as it's already the correct length
setAttrib(thisvalue, R_NamesSymbol, R_NilValue); // clear names such as DT[,a:=mapvector[a]]
setAttrib(thisvalue, R_DimSymbol, R_NilValue); // so that matrix is treated as vector
setAttrib(thisvalue, R_DimNamesSymbol, R_NilValue); // the 3rd of the 3 attribs not copied by copyMostAttrib, for consistency.
continue;
}
if (coln+1 > oldncol) { // new column
SET_VECTOR_ELT(dt, coln, targetcol=allocNAVectorLike(thisvalue, nrow));
// initialize with NAs for when 'rows' is a subset and it doesn't touch
// do not try to save the time to NA fill (contiguous branch free assign anyway) since being
// sure all items will be written to (isNull(rows), length(rows), vlen<1, targetlen) is not worth the risk.
if (isVectorAtomic(thisvalue)) copyMostAttrib(thisvalue,targetcol); // class etc but not names
// else for lists (such as data.frame and data.table) treat them as raw lists and drop attribs
if (vlen<1) continue; // e.g. DT[,newcol:=integer()] (adding new empty column)
} else { // existing column
targetcol = VECTOR_ELT(dt,coln);
}
const char *ret = memrecycle(targetcol, rows, 0, targetlen, thisvalue, 0, -1, coln+1, CHAR(STRING_ELT(names, coln)));
if (ret) warning("%s", ret);
}
*_Last_updated = numToDo; // the updates have taken place with no error, so update .Last.updated now
assignedNames = PROTECT(allocVector(STRSXP, LENGTH(cols))); protecti++;
for (int i=0; i<LENGTH(cols); ++i) SET_STRING_ELT(assignedNames,i,STRING_ELT(names,INTEGER(cols)[i]-1));
key = getAttrib(dt, sym_sorted);
if (length(key)) {
// if assigning to at least one key column, the key is truncated to one position before the first changed column.
//any() and subsetVector() don't seem to be exposed by R API at C level, so this is done here long hand.
PROTECT(tmp = chin(key, assignedNames)); protecti++;
int newKeyLength = length(key);
for (int i=0; i<LENGTH(tmp); ++i) if (LOGICAL(tmp)[i]) {
// If a key column is being assigned to, set newKeyLength to the key element before since everything after that may have changed in order.
newKeyLength = i;
break;
}
if(newKeyLength == 0){
// no valid key columns remain, remove the key
setAttrib(dt, sym_sorted, R_NilValue);
} else if (newKeyLength < length(key)){
// new key is shorter than original one. Reassign
PROTECT(tmp = allocVector(STRSXP, newKeyLength)); protecti++;
for (int i=0; i<newKeyLength; ++i) SET_STRING_ELT(tmp, i, STRING_ELT(key, i));
setAttrib(dt, sym_sorted, tmp);
}
//else: no key column changed, nothing to be done
}
index = getAttrib(dt, install("index"));
if (index != R_NilValue) {
s = ATTRIB(index);
indexNo = 0;
// get a vector with all index names
PROTECT(indexNames = allocVector(STRSXP, xlength(s))); protecti++;
while(s != R_NilValue){
SET_STRING_ELT(indexNames, indexNo, PRINTNAME(TAG(s)));
indexNo++;
s = CDR(s);
}
s = ATTRIB(index); // reset to first element
indexNo = 0;
while(s != R_NilValue) {
a = TAG(s);
indexLength = xlength(CAR(s));
tc1 = c1 = CHAR(PRINTNAME(a)); // the index name; e.g. "__col1__col2"
if (*tc1!='_' || *(tc1+1)!='_') {
// fix for #1396
if (verbose) {
Rprintf(_("Dropping index '%s' as it doesn't have '__' at the beginning of its name. It was very likely created by v1.9.4 of data.table.\n"), tc1);
}
setAttrib(index, a, R_NilValue);
indexNo++;
s = CDR(s);
continue; // with next index
}
tc1 += 2; // tc1 always marks the start of a key column
if (!*tc1) internal_error(__func__, "index name ends with trailing __"); // # nocov
// check the position of the first appearance of an assigned column in the index.
// the new index will be truncated to this position.
char *s4 = (char*) malloc(strlen(c1) + 3);
if(s4 == NULL){
internal_error(__func__, "Couldn't allocate memory for s4"); // # nocov
}
memcpy(s4, c1, strlen(c1));
memset(s4 + strlen(c1), '\0', 1);
strcat(s4, "__"); // add trailing '__' to newKey so we can search for pattern '__colName__' also at the end of the index.
int newKeyLength = strlen(c1);
for(int i = 0; i < xlength(assignedNames); i++){
tc2 = CHAR(STRING_ELT(assignedNames, i));
char *s5 = (char*) malloc(strlen(tc2) + 5); //4 * '_' + \0
if(s5 == NULL){
free(s4); // # nocov
internal_error(__func__, "Couldn't allocate memory for s5"); // # nocov
}
memset(s5, '_', 2);
memset(s5 + 2, '\0', 1);
strcat(s5, tc2);
strcat(s5, "__");
tc2 = strstr(s4, s5);
if(tc2 == NULL){ // column is not part of key
free(s5);
continue;
}
if(tc2 - s4 < newKeyLength){ // new column match is before last match
newKeyLength = tc2 - s4;
}
free(s5);
}
memset(s4 + newKeyLength, '\0', 1); // truncate the new key to the new length
if(newKeyLength == 0){ // no valid key column remains. Drop the key
setAttrib(index, a, R_NilValue);
SET_STRING_ELT(indexNames, indexNo, NA_STRING);
if (verbose) {
Rprintf(_("Dropping index '%s' due to an update on a key column\n"), c1+2);
}
} else if(newKeyLength < strlen(c1)) {
SEXP s4Str = PROTECT(mkString(s4));
if(indexLength == 0 && // shortened index can be kept since it is just information on the order (see #2372)
LOGICAL(chin(s4Str, indexNames))[0] == 0) {// index with shortened name not present yet
SET_TAG(s, install(s4));
SET_STRING_ELT(indexNames, indexNo, mkChar(s4));
if (verbose)
Rprintf(_("Shortening index '%s' to '%s' due to an update on a key column\n"), c1+2, s4 + 2);
} else { // indexLength > 0 || shortened name present already
// indexLength > 0 indicates reordering. Drop it to avoid spurious reordering in non-indexed columns (#2372)
// shortened name already present indicates that index needs to be dropped to avoid duplicate indices.
setAttrib(index, a, R_NilValue);
SET_STRING_ELT(indexNames, indexNo, NA_STRING);
if (verbose)
Rprintf(_("Dropping index '%s' due to an update on a key column\n"), c1+2);
}
UNPROTECT(1); // s4Str
} //else: index is not affected by assign: nothing to be done
free(s4);
indexNo ++;
s = CDR(s);
}
}
if (ndelete) {
// delete any columns assigned NULL (there was a 'continue' earlier in loop above)
int *tt = (int *)R_alloc(ndelete, sizeof(int));
const int *colsd=INTEGER(cols), ncols=length(cols), ndt=length(dt);
for (int i=0, k=0; i<ncols; ++i) { // find which ones to delete and put them in tt
// Aside: a new column being assigned NULL (something odd to do) would have been warned above, added above, and now deleted. Just
// easier to code it this way; e.g. so that other columns may be added or removed ok by the same query.
coln = colsd[i]-1;
SEXP thisvalue = RHS_list_of_columns ? VECTOR_ELT(values, i) : values;
if (isNull(thisvalue)) tt[k++] = coln;
}
R_isort(tt, ndelete); // sort the column-numbers-to-delete into ascending order
for (int i=0; i<ndelete-1; ++i) {
if (tt[i]>=tt[i+1])
internal_error(__func__, "%d column numbers to delete not now in strictly increasing order. No-dups were checked earlier.", i); // # nocov
}
for (int i=tt[0], j=1, k=tt[0]+1; i<ndt-ndelete; ++i, ++k) { // i moves up from the first non-deleted column and is the target of write
while (j<ndelete && k==tt[j]) { j++; k++; } // move k up to the next non-deleted column; j is the next position in tt
SET_VECTOR_ELT(dt, i, VECTOR_ELT(dt, k));
SET_STRING_ELT(names, i, STRING_ELT(names, k));
}
for (int i=ndt-ndelete; i<ndt; ++i) { // blank out the ndelete slots at the end
SET_VECTOR_ELT(dt, i, R_NilValue);
SET_STRING_ELT(names, i, NA_STRING); // release reference to the CHARSXP
}
SETLENGTH(dt, ndt-ndelete);
SETLENGTH(names, ndt-ndelete);
if (LENGTH(names)==0) {
// That was last column deleted, leaving NULL data.table, so we need to reset .row_names, so that it really is the NULL data.table.
PROTECT(nullint=allocVector(INTSXP, 0)); protecti++;
setAttrib(dt, R_RowNamesSymbol, nullint); // i.e. .set_row_names(0)
}
}
UNPROTECT(protecti);
return(dt); // needed for `*tmp*` mechanism (when := isn't used), and to return the new object after a := for compound syntax.
}
#define MSGSIZE 1000
static char memrecycle_message[MSGSIZE+1]; // returned to rbindlist so it can prefix with which one of the list of data.table-like objects
const char *memrecycle(const SEXP target, const SEXP where, const int start, const int len, SEXP source, const int sourceStart, const int sourceLen, const int colnum, const char *colname)
// like memcpy but recycles single-item source
// 'where' a 1-based INTEGER vector subset of target to assign to, or NULL or integer()
// assigns to target[start:start+len-1] or target[where[start:start+len-1]] where start is 0-based
// if sourceLen==-1 then all of source is used (if it is 1 item then it is recycled, or its length must match) for convenience to avoid
// having to use length(source) (repeating source expression) in each call
// sourceLen==1 is used in dogroups to recycle the group values into ans to match the nrow of each group's result; sourceStart is set to each group value row.
{
if (len<1) return NULL;
int slen = sourceLen>=0 ? sourceLen : length(source); // since source may get reassigned to a scalar, we should not mark it as const
if (slen==0) return NULL;
if (sourceStart<0 || sourceStart+slen>length(source))
internal_error(__func__, "sourceStart=%d sourceLen=%d length(source)=%d", sourceStart, sourceLen, length(source)); // # nocov
if (!length(where) && start+len>length(target))
internal_error(__func__, "start=%d len=%d length(target)=%d", start, len, length(target)); // # nocov
const int soff = sourceStart;
if (slen>1 && slen!=len && (!isNewList(target) || isNewList(source)))
internal_error(__func__, "recycle length error not caught earlier. slen=%d len=%d", slen, len); // # nocov
// Internal error because the column has already been added to the DT, so length mismatch should have been caught before adding the column.
// for 5647 this used to limit slen to len, but no longer
if (colname==NULL)
internal_error(__func__, "received NULL colname"); // # nocov
*memrecycle_message = '\0';
int protecti=0;
const bool sourceIsFactor=isFactor(source), targetIsFactor=isFactor(target);
const bool sourceIsI64=isReal(source) && INHERITS(source, char_integer64);
const bool targetIsI64=isReal(target) && INHERITS(target, char_integer64);
if (sourceIsFactor || targetIsFactor) {
if (!targetIsFactor) {
if (!isString(target) && !isNewList(target))
error(_("Cannot assign 'factor' to '%s'. Factors can only be assigned to factor, character or list columns."), type2char(TYPEOF(target)));
// else assigning factor to character is left to later below, avoiding wasteful asCharacterFactor
} else if (!sourceIsFactor && !isString(source)) {
// target is factor
if (allNA(source, false)) { // return false for list and other types that allNA does not support
source = ScalarLogical(NA_LOGICAL); slen = 1; // a global constant in R and won't allocate; fall through to regular zero-copy coerce
} else if (isInteger(source) || isReal(source)) {
// allow assigning level numbers to factor columns; test 425, 426, 429 and 1945
const int nlevel = length(getAttrib(target, R_LevelsSymbol));
if (isInteger(source)) {
const int *sd = INTEGER(source);
for (int i=0; i<slen; ++i) {
const int val = sd[i+soff];
if ((val<1 && val!=NA_INTEGER) || val>nlevel) {
if (colnum == 0)
error(_("Assigning factor numbers to target vector. But %d is outside the level range [1,%d]"), val, nlevel);
else
error(_("Assigning factor numbers to column %d named '%s'. But %d is outside the level range [1,%d]"), colnum, colname, val, nlevel);
}
}
} else {
const double *sd = REAL(source);
for (int i=0; i<slen; ++i) {
const double val = sd[i+soff];
// Since nlevel is an int, val < 1 || val > nlevel will deflect UB guarded against in PR #5832
if (!ISNAN(val) && (val < 1 || val > nlevel || val != (int)val)) {
if (colnum == 0)
error(_("Assigning factor numbers to target vector. But %f is outside the level range [1,%d], or is not a whole number."), val, nlevel);
else
error(_("Assigning factor numbers to column %d named '%s'. But %f is outside the level range [1,%d], or is not a whole number."), colnum, colname, val, nlevel);
}
}
}
// Now just let the valid level numbers fall through to regular assign by BODY below
} else {
error(_("Cannot assign '%s' to 'factor'. Factor columns can be assigned factor, character, NA in any type, or level numbers."), type2char(TYPEOF(source)));
}
} else {
// either factor or character being assigned to factor column
SEXP targetLevels = PROTECT(getAttrib(target, R_LevelsSymbol)); protecti++;
SEXP sourceLevels = source; // character source
if (sourceIsFactor) { sourceLevels=PROTECT(getAttrib(source, R_LevelsSymbol)); protecti++; }
if (!sourceIsFactor || !R_compute_identical(sourceLevels, targetLevels, 0)) { // !sourceIsFactor for test 2115.6
const int nTargetLevels=length(targetLevels), nSourceLevels=length(sourceLevels);
const SEXP *targetLevelsD=STRING_PTR_RO(targetLevels), *sourceLevelsD=STRING_PTR_RO(sourceLevels);
SEXP newSource = PROTECT(allocVector(INTSXP, length(source))); protecti++;
savetl_init();
for (int k=0; k<nTargetLevels; ++k) {
const SEXP s = targetLevelsD[k];
const int tl = TRUELENGTH(s);
if (tl>0) {
savetl(s);
} else if (tl<0) {
// # nocov start
for (int j=0; j<k; ++j) SET_TRUELENGTH(s, 0); // wipe our negative usage and restore 0
savetl_end(); // then restore R's own usage (if any)
internal_error(__func__, "levels of target are either not unique or have truelength<0"); // # nocov
// # nocov end
}
SET_TRUELENGTH(s, -k-1);
}
int nAdd = 0;
for (int k=0; k<nSourceLevels; ++k) {
const SEXP s = sourceLevelsD[k];
const int tl = TRUELENGTH(s);
if (tl>=0) {
if (!sourceIsFactor && s==NA_STRING) continue; // don't create NA factor level when assigning character to factor; test 2117
if (tl>0) savetl(s);
SET_TRUELENGTH(s, -nTargetLevels-(++nAdd));
} // else, when sourceIsString, it's normal for there to be duplicates here
}
const int nSource = length(source);
int *newSourceD = INTEGER(newSource);
if (sourceIsFactor) {
const int *sourceD = INTEGER(source);
for (int i=0; i<nSource; ++i) { // convert source integers to refer to target levels
const int val = sourceD[i];
newSourceD[i] = val==NA_INTEGER ? NA_INTEGER : -TRUELENGTH(sourceLevelsD[val-1]); // retains NA factor levels here via TL(NA_STRING); e.g. ordered factor
}
} else {
const SEXP *sourceD = STRING_PTR_RO(source);
for (int i=0; i<nSource; ++i) { // convert source integers to refer to target levels
const SEXP val = sourceD[i];
newSourceD[i] = val==NA_STRING ? NA_INTEGER : -TRUELENGTH(val);
}
}
source = newSource;
for (int k=0; k<nTargetLevels; ++k) SET_TRUELENGTH(targetLevelsD[k], 0); // don't need those anymore
if (nAdd) {
// cannot grow the levels yet as that would be R call which could fail to alloc and we have no hook to clear up
SEXP *temp = (SEXP *)malloc(nAdd * sizeof(SEXP *));
if (!temp) {
// # nocov start
for (int k=0; k<nSourceLevels; ++k) SET_TRUELENGTH(sourceLevelsD[k], 0);
savetl_end();
error(_("Unable to allocate working memory of %zu bytes to combine factor levels"), nAdd*sizeof(SEXP *));
// # nocov end
}
for (int k=0, thisAdd=0; thisAdd<nAdd; ++k) { // thisAdd<nAdd to stop early when the added ones are all reached
SEXP s = sourceLevelsD[k];
int tl = TRUELENGTH(s);
if (tl) { // tl negative here
if (tl != -nTargetLevels-thisAdd-1) internal_error(__func__, "extra level check sum failed"); // # nocov
temp[thisAdd++] = s;
SET_TRUELENGTH(s,0);
}
}
savetl_end();
setAttrib(target, R_LevelsSymbol, targetLevels=growVector(targetLevels, nTargetLevels + nAdd));
for (int k=0; k<nAdd; ++k) {
SET_STRING_ELT(targetLevels, nTargetLevels+k, temp[k]);
}
free(temp);
} else {
// all source levels were already in target levels, but not with the same integers; we're done
savetl_end();
}
// now continue, but with the mapped integers in the (new) source
}
}
} else if (isString(source) && !isString(target) && !isNewList(target)) {
if (colnum == 0)
warning(_("Coercing 'character' RHS to '%s' to match the type of target vector."), targetIsI64?"integer64":type2char(TYPEOF(target)));
else
warning(_("Coercing 'character' RHS to '%s' to match the type of column %d named '%s'."), targetIsI64?"integer64":type2char(TYPEOF(target)), colnum, colname);
// this "Coercing ..." warning first to give context in case coerceVector warns 'NAs introduced by coercion'
// and also because 'character' to integer/double coercion is often a user mistake (e.g. wrong target column, or wrong
// variable on RHS) which they are more likely to appreciate than find inconvenient
source = PROTECT(coerceVector(source, TYPEOF(target))); protecti++;
} else if (isNewList(source) && !isNewList(target)) {
if (targetIsI64) {
if (colnum == 0)
error(_("Cannot coerce 'list' RHS to 'integer64' to match the type of target vector."));
else
error(_("Cannot coerce 'list' RHS to 'integer64' to match the type of column %d named '%s'."), colnum, colname);
// because R's coerceVector doesn't know about integer64
}
// as in base R; e.g. let as.double(list(1,2,3)) work but not as.double(list(1,c(2,4),3))
// relied on by NNS, simstudy and table.express; tests 1294.*
if (colnum == 0)
warning(_("Coercing 'list' RHS to '%s' to match the type of target vector."), type2char(TYPEOF(target)));
else
warning(_("Coercing 'list' RHS to '%s' to match the type of column %d named '%s'."), type2char(TYPEOF(target)), colnum, colname);
source = PROTECT(coerceVector(source, TYPEOF(target))); protecti++;
} else if ((TYPEOF(target)!=TYPEOF(source) || targetIsI64!=sourceIsI64) && !isNewList(target)) {
if (GetVerbose()>=3) {
// only take the (small) cost of GetVerbose() (search of options() list) when types don't match
const char *sourceType = sourceIsI64 ? "integer64" : type2char(TYPEOF(source));
const char *targetType = targetIsI64 ? "integer64" : type2char(TYPEOF(target));
if (colnum == 0)
Rprintf(_("Zero-copy coerce when assigning '%s' to '%s' target vector.\n"), sourceType, targetType);
else
Rprintf(_("Zero-copy coerce when assigning '%s' to column %d named '%s' which is '%s'.\n"), sourceType, colnum, colname, targetType);
}
// The following checks are up front here, otherwise we'd need them twice in the two branches
// inside BODY that cater for 'where' or not. Maybe there's a way to merge the two macros in future.
// The idea is to do these range checks without calling coerceVector() (which allocates)
#define CHECK_RANGE(STYPE, RFUN, COND, FMT, TO, FMTVAL) {{ \
const STYPE *sd = (const STYPE *)RFUN(source); \
for (int i=0; i<slen; ++i) { \
const STYPE val = sd[i+soff]; \
if (COND) { \
const char *sType = sourceIsI64 ? "integer64" : type2char(TYPEOF(source)); \
const char *tType = targetIsI64 ? "integer64" : type2char(TYPEOF(target)); \
snprintf(memrecycle_message, MSGSIZE, \
colnum == 0 \
? _("%"FMT" (type '%s') at RHS position %d "TO" when assigning to type '%s' (target vector)") \
: _("%"FMT" (type '%s') at RHS position %d "TO" when assigning to type '%s' (column %d named '%s')"), \
FMTVAL, sType, i+1, tType, colnum, colname); \
/* string returned so that rbindlist/dogroups can prefix it with which item of its list this refers to */ \
break; \
} \
} \
} break; }
switch(TYPEOF(target)) {
case LGLSXP:
switch (TYPEOF(source)) {
case RAWSXP: CHECK_RANGE(Rbyte, RAW, val!=0 && val!=1, "d", "taken as TRUE", val)
case INTSXP: CHECK_RANGE(int, INTEGER, val!=0 && val!=1 && val!=NA_INTEGER, "d", "taken as TRUE", val)
case REALSXP: if (sourceIsI64)
CHECK_RANGE(int64_t, REAL, val!=0 && val!=1 && val!=NA_INTEGER64, PRId64, "taken as TRUE", val)
else CHECK_RANGE(double, REAL, !ISNAN(val) && val!=0.0 && val!=1.0, "f", "taken as TRUE", val)
} break;
case RAWSXP:
switch (TYPEOF(source)) {
case INTSXP: CHECK_RANGE(int, INTEGER, val<0 || val>255, "d", "taken as 0", val)
case REALSXP: if (sourceIsI64)
CHECK_RANGE(int64_t, REAL, val<0 || val>255, PRId64, "taken as 0", val)
else CHECK_RANGE(double, REAL, !R_FINITE(val) || val<0.0 || val>256.0 || (int)val!=val, "f", "either truncated (precision lost) or taken as 0", val)
} break;
case INTSXP:
switch (TYPEOF(source)) {
case REALSXP: if (sourceIsI64)
CHECK_RANGE(int64_t, REAL, val!=NA_INTEGER64 && (val<=NA_INTEGER || val>INT_MAX), PRId64, "out-of-range (NA)", val)
else CHECK_RANGE(double, REAL, !ISNAN(val) && (!within_int32_repres(val) || (int)val!=val), "f", "out-of-range(NA) or truncated (precision lost)", val)
case CPLXSXP: CHECK_RANGE(Rcomplex, COMPLEX, !((ISNAN(val.i) || (R_FINITE(val.i) && val.i==0.0)) &&
(ISNAN(val.r) || (within_int32_repres(val.r) && (int)val.r==val.r))), "f", "either imaginary part discarded or real part truncated (precision lost)", val.r)
} break;
case REALSXP:
switch (TYPEOF(source)) {
case REALSXP: if (targetIsI64 && !sourceIsI64)
CHECK_RANGE(double, REAL, !ISNAN(val) && (!within_int64_repres(val) || (int64_t)val!=val), "f", "out-of-range(NA) or truncated (precision lost)", val)
break;
case CPLXSXP: if (targetIsI64)
CHECK_RANGE(Rcomplex, COMPLEX, !((ISNAN(val.i) || (R_FINITE(val.i) && val.i==0.0)) &&
(ISNAN(val.r) || (R_FINITE(val.r) && (int64_t)val.r==val.r))), "f", "either imaginary part discarded or real part truncated (precision lost)", val.r)
else CHECK_RANGE(Rcomplex, COMPLEX, !(ISNAN(val.i) || (R_FINITE(val.i) && val.i==0.0)), "f", "imaginary part discarded", val.i)
}
}
}
#undef BODY
#define BODY(STYPE, RFUN, CTYPE, CAST, ASSIGN) {{ \
const STYPE *sd = (const STYPE *)RFUN(source); \
if (length(where)) { \
if (slen==1) { \
const STYPE val = sd[soff]; \
const CTYPE cval = CAST; \
for (int wi=0; wi<len; ++wi) { \
const int w = wd[wi]; \
if (w<1) continue; /*0 or NA*/ \
const int i = w-1; \
ASSIGN; \
} \