-
Notifications
You must be signed in to change notification settings - Fork 0
/
sais2.go
1741 lines (1575 loc) · 52.4 KB
/
sais2.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 2019 The Go Authors. All rights reserved.
// Use of this source code is governed by a BSD-style
// license that can be found in the LICENSE file.
// Code generated by go generate; DO NOT EDIT.
package astrobwt
func text_64(text []byte, sa []int64) {
if int(int64(len(text))) != len(text) || len(text) != len(sa) {
panic("suffixarray: misuse of text_64")
}
sais_8_64(text, 256, sa, make([]int64, 2*256))
}
func sais_8_64(text []byte, textMax int, sa, tmp []int64) {
if len(sa) != len(text) || len(tmp) < int(textMax) {
panic("suffixarray: misuse of sais_8_64")
}
// Trivial base cases. Sorting 0 or 1 things is easy.
if len(text) == 0 {
return
}
if len(text) == 1 {
sa[0] = 0
return
}
// Establish slices indexed by text character
// holding character frequency and bucket-sort offsets.
// If there's only enough tmp for one slice,
// we make it the bucket offsets and recompute
// the character frequency each time we need it.
var freq, bucket []int64
if len(tmp) >= 2*textMax {
freq, bucket = tmp[:textMax], tmp[textMax:2*textMax]
freq[0] = -1 // mark as uninitialized
} else {
freq, bucket = nil, tmp[:textMax]
}
// The SAIS algorithm.
// Each of these calls makes one scan through sa.
// See the individual functions for documentation
// about each's role in the algorithm.
numLMS := placeLMS_8_64(text, sa, freq, bucket)
if numLMS <= 1 {
// 0 or 1 items are already sorted. Do nothing.
} else {
induceSubL_8_64(text, sa, freq, bucket)
induceSubS_8_64(text, sa, freq, bucket)
length_8_64(text, sa, numLMS)
maxID := assignID_8_64(text, sa, numLMS)
if maxID < numLMS {
map_64(sa, numLMS)
recurse_64(sa, tmp, numLMS, maxID)
unmap_8_64(text, sa, numLMS)
} else {
// If maxID == numLMS, then each LMS-substring
// is unique, so the relative ordering of two LMS-suffixes
// is determined by just the leading LMS-substring.
// That is, the LMS-suffix sort order matches the
// (simpler) LMS-substring sort order.
// Copy the original LMS-substring order into the
// suffix array destination.
copy(sa, sa[len(sa)-numLMS:])
}
expand_8_64(text, freq, bucket, sa, numLMS)
}
induceL_8_64(text, sa, freq, bucket)
induceS_8_64(text, sa, freq, bucket)
// Mark for caller that we overwrote tmp.
tmp[0] = -1
}
func sais_32(text []int32, textMax int, sa, tmp []int32) {
if len(sa) != len(text) || len(tmp) < int(textMax) {
panic("suffixarray: misuse of sais_32")
}
// Trivial base cases. Sorting 0 or 1 things is easy.
if len(text) == 0 {
return
}
if len(text) == 1 {
sa[0] = 0
return
}
// Establish slices indexed by text character
// holding character frequency and bucket-sort offsets.
// If there's only enough tmp for one slice,
// we make it the bucket offsets and recompute
// the character frequency each time we need it.
var freq, bucket []int32
if len(tmp) >= 2*textMax {
freq, bucket = tmp[:textMax], tmp[textMax:2*textMax]
freq[0] = -1 // mark as uninitialized
} else {
freq, bucket = nil, tmp[:textMax]
}
// The SAIS algorithm.
// Each of these calls makes one scan through sa.
// See the individual functions for documentation
// about each's role in the algorithm.
numLMS := placeLMS_32(text, sa, freq, bucket)
if numLMS <= 1 {
// 0 or 1 items are already sorted. Do nothing.
} else {
induceSubL_32(text, sa, freq, bucket)
induceSubS_32(text, sa, freq, bucket)
length_32(text, sa, numLMS)
maxID := assignID_32(text, sa, numLMS)
if maxID < numLMS {
map_32(sa, numLMS)
recurse_32(sa, tmp, numLMS, maxID)
unmap_32(text, sa, numLMS)
} else {
// If maxID == numLMS, then each LMS-substring
// is unique, so the relative ordering of two LMS-suffixes
// is determined by just the leading LMS-substring.
// That is, the LMS-suffix sort order matches the
// (simpler) LMS-substring sort order.
// Copy the original LMS-substring order into the
// suffix array destination.
copy(sa, sa[len(sa)-numLMS:])
}
expand_32(text, freq, bucket, sa, numLMS)
}
induceL_32(text, sa, freq, bucket)
induceS_32(text, sa, freq, bucket)
// Mark for caller that we overwrote tmp.
tmp[0] = -1
}
func sais_64(text []int64, textMax int, sa, tmp []int64) {
if len(sa) != len(text) || len(tmp) < int(textMax) {
panic("suffixarray: misuse of sais_64")
}
// Trivial base cases. Sorting 0 or 1 things is easy.
if len(text) == 0 {
return
}
if len(text) == 1 {
sa[0] = 0
return
}
// Establish slices indexed by text character
// holding character frequency and bucket-sort offsets.
// If there's only enough tmp for one slice,
// we make it the bucket offsets and recompute
// the character frequency each time we need it.
var freq, bucket []int64
if len(tmp) >= 2*textMax {
freq, bucket = tmp[:textMax], tmp[textMax:2*textMax]
freq[0] = -1 // mark as uninitialized
} else {
freq, bucket = nil, tmp[:textMax]
}
// The SAIS algorithm.
// Each of these calls makes one scan through sa.
// See the individual functions for documentation
// about each's role in the algorithm.
numLMS := placeLMS_64(text, sa, freq, bucket)
if numLMS <= 1 {
// 0 or 1 items are already sorted. Do nothing.
} else {
induceSubL_64(text, sa, freq, bucket)
induceSubS_64(text, sa, freq, bucket)
length_64(text, sa, numLMS)
maxID := assignID_64(text, sa, numLMS)
if maxID < numLMS {
map_64(sa, numLMS)
recurse_64(sa, tmp, numLMS, maxID)
unmap_64(text, sa, numLMS)
} else {
// If maxID == numLMS, then each LMS-substring
// is unique, so the relative ordering of two LMS-suffixes
// is determined by just the leading LMS-substring.
// That is, the LMS-suffix sort order matches the
// (simpler) LMS-substring sort order.
// Copy the original LMS-substring order into the
// suffix array destination.
copy(sa, sa[len(sa)-numLMS:])
}
expand_64(text, freq, bucket, sa, numLMS)
}
induceL_64(text, sa, freq, bucket)
induceS_64(text, sa, freq, bucket)
// Mark for caller that we overwrote tmp.
tmp[0] = -1
}
func freq_8_64(text []byte, freq, bucket []int64) []int64 {
if freq != nil && freq[0] >= 0 {
return freq // already computed
}
if freq == nil {
freq = bucket
}
freq = freq[:256] // eliminate bounds check for freq[c] below
for i := range freq {
freq[i] = 0
}
for _, c := range text {
freq[c]++
}
return freq
}
func freq_32(text []int32, freq, bucket []int32) []int32 {
if freq != nil && freq[0] >= 0 {
return freq // already computed
}
if freq == nil {
freq = bucket
}
for i := range freq {
freq[i] = 0
}
for _, c := range text {
freq[c]++
}
return freq
}
func freq_64(text []int64, freq, bucket []int64) []int64 {
if freq != nil && freq[0] >= 0 {
return freq // already computed
}
if freq == nil {
freq = bucket
}
for i := range freq {
freq[i] = 0
}
for _, c := range text {
freq[c]++
}
return freq
}
func bucketMin_8_64(text []byte, freq, bucket []int64) {
freq = freq_8_64(text, freq, bucket)
freq = freq[:256] // establish len(freq) = 256, so 0 ≤ i < 256 below
bucket = bucket[:256] // eliminate bounds check for bucket[i] below
total := int64(0)
for i, n := range freq {
bucket[i] = total
total += n
}
}
func bucketMin_32(text []int32, freq, bucket []int32) {
freq = freq_32(text, freq, bucket)
total := int32(0)
for i, n := range freq {
bucket[i] = total
total += n
}
}
func bucketMin_64(text []int64, freq, bucket []int64) {
freq = freq_64(text, freq, bucket)
total := int64(0)
for i, n := range freq {
bucket[i] = total
total += n
}
}
func bucketMax_8_64(text []byte, freq, bucket []int64) {
freq = freq_8_64(text, freq, bucket)
freq = freq[:256] // establish len(freq) = 256, so 0 ≤ i < 256 below
bucket = bucket[:256] // eliminate bounds check for bucket[i] below
total := int64(0)
for i, n := range freq {
total += n
bucket[i] = total
}
}
func bucketMax_32(text []int32, freq, bucket []int32) {
freq = freq_32(text, freq, bucket)
total := int32(0)
for i, n := range freq {
total += n
bucket[i] = total
}
}
func bucketMax_64(text []int64, freq, bucket []int64) {
freq = freq_64(text, freq, bucket)
total := int64(0)
for i, n := range freq {
total += n
bucket[i] = total
}
}
func placeLMS_8_64(text []byte, sa, freq, bucket []int64) int {
bucketMax_8_64(text, freq, bucket)
numLMS := 0
lastB := int64(-1)
bucket = bucket[:256] // eliminate bounds check for bucket[c1] below
// The next stanza of code (until the blank line) loop backward
// over text, stopping to execute a code body at each position i
// such that text[i] is an L-character and text[i+1] is an S-character.
// That is, i+1 is the position of the start of an LMS-substring.
// These could be hoisted out into a function with a callback,
// but at a significant speed cost. Instead, we just write these
// seven lines a few times in this source file. The copies below
// refer back to the pattern established by this original as the
// "LMS-substring iterator".
//
// In every scan through the text, c0, c1 are successive characters of text.
// In this backward scan, c0 == text[i] and c1 == text[i+1].
// By scanning backward, we can keep track of whether the current
// position is type-S or type-L according to the usual definition:
//
// - position len(text) is type S with text[len(text)] == -1 (the sentinel)
// - position i is type S if text[i] < text[i+1], or if text[i] == text[i+1] && i+1 is type S.
// - position i is type L if text[i] > text[i+1], or if text[i] == text[i+1] && i+1 is type L.
//
// The backward scan lets us maintain the current type,
// update it when we see c0 != c1, and otherwise leave it alone.
// We want to identify all S positions with a preceding L.
// Position len(text) is one such position by definition, but we have
// nowhere to write it down, so we eliminate it by untruthfully
// setting isTypeS = false at the start of the loop.
c0, c1, isTypeS := byte(0), byte(0), false
for i := len(text) - 1; i >= 0; i-- {
c0, c1 = text[i], c0
if c0 < c1 {
isTypeS = true
} else if c0 > c1 && isTypeS {
isTypeS = false
// Bucket the index i+1 for the start of an LMS-substring.
b := bucket[c1] - 1
bucket[c1] = b
sa[b] = int64(i + 1)
lastB = b
numLMS++
}
}
// We recorded the LMS-substring starts but really want the ends.
// Luckily, with two differences, the start indexes and the end indexes are the same.
// The first difference is that the rightmost LMS-substring's end index is len(text),
// so the caller must pretend that sa[-1] == len(text), as noted above.
// The second difference is that the first leftmost LMS-substring start index
// does not end an earlier LMS-substring, so as an optimization we can omit
// that leftmost LMS-substring start index (the last one we wrote).
//
// Exception: if numLMS <= 1, the caller is not going to bother with
// the recursion at all and will treat the result as containing LMS-substring starts.
// In that case, we don't remove the final entry.
if numLMS > 1 {
sa[lastB] = 0
}
return numLMS
}
func placeLMS_32(text []int32, sa, freq, bucket []int32) int {
bucketMax_32(text, freq, bucket)
numLMS := 0
lastB := int32(-1)
// The next stanza of code (until the blank line) loop backward
// over text, stopping to execute a code body at each position i
// such that text[i] is an L-character and text[i+1] is an S-character.
// That is, i+1 is the position of the start of an LMS-substring.
// These could be hoisted out into a function with a callback,
// but at a significant speed cost. Instead, we just write these
// seven lines a few times in this source file. The copies below
// refer back to the pattern established by this original as the
// "LMS-substring iterator".
//
// In every scan through the text, c0, c1 are successive characters of text.
// In this backward scan, c0 == text[i] and c1 == text[i+1].
// By scanning backward, we can keep track of whether the current
// position is type-S or type-L according to the usual definition:
//
// - position len(text) is type S with text[len(text)] == -1 (the sentinel)
// - position i is type S if text[i] < text[i+1], or if text[i] == text[i+1] && i+1 is type S.
// - position i is type L if text[i] > text[i+1], or if text[i] == text[i+1] && i+1 is type L.
//
// The backward scan lets us maintain the current type,
// update it when we see c0 != c1, and otherwise leave it alone.
// We want to identify all S positions with a preceding L.
// Position len(text) is one such position by definition, but we have
// nowhere to write it down, so we eliminate it by untruthfully
// setting isTypeS = false at the start of the loop.
c0, c1, isTypeS := int32(0), int32(0), false
for i := len(text) - 1; i >= 0; i-- {
c0, c1 = text[i], c0
if c0 < c1 {
isTypeS = true
} else if c0 > c1 && isTypeS {
isTypeS = false
// Bucket the index i+1 for the start of an LMS-substring.
b := bucket[c1] - 1
bucket[c1] = b
sa[b] = int32(i + 1)
lastB = b
numLMS++
}
}
// We recorded the LMS-substring starts but really want the ends.
// Luckily, with two differences, the start indexes and the end indexes are the same.
// The first difference is that the rightmost LMS-substring's end index is len(text),
// so the caller must pretend that sa[-1] == len(text), as noted above.
// The second difference is that the first leftmost LMS-substring start index
// does not end an earlier LMS-substring, so as an optimization we can omit
// that leftmost LMS-substring start index (the last one we wrote).
//
// Exception: if numLMS <= 1, the caller is not going to bother with
// the recursion at all and will treat the result as containing LMS-substring starts.
// In that case, we don't remove the final entry.
if numLMS > 1 {
sa[lastB] = 0
}
return numLMS
}
func placeLMS_64(text []int64, sa, freq, bucket []int64) int {
bucketMax_64(text, freq, bucket)
numLMS := 0
lastB := int64(-1)
// The next stanza of code (until the blank line) loop backward
// over text, stopping to execute a code body at each position i
// such that text[i] is an L-character and text[i+1] is an S-character.
// That is, i+1 is the position of the start of an LMS-substring.
// These could be hoisted out into a function with a callback,
// but at a significant speed cost. Instead, we just write these
// seven lines a few times in this source file. The copies below
// refer back to the pattern established by this original as the
// "LMS-substring iterator".
//
// In every scan through the text, c0, c1 are successive characters of text.
// In this backward scan, c0 == text[i] and c1 == text[i+1].
// By scanning backward, we can keep track of whether the current
// position is type-S or type-L according to the usual definition:
//
// - position len(text) is type S with text[len(text)] == -1 (the sentinel)
// - position i is type S if text[i] < text[i+1], or if text[i] == text[i+1] && i+1 is type S.
// - position i is type L if text[i] > text[i+1], or if text[i] == text[i+1] && i+1 is type L.
//
// The backward scan lets us maintain the current type,
// update it when we see c0 != c1, and otherwise leave it alone.
// We want to identify all S positions with a preceding L.
// Position len(text) is one such position by definition, but we have
// nowhere to write it down, so we eliminate it by untruthfully
// setting isTypeS = false at the start of the loop.
c0, c1, isTypeS := int64(0), int64(0), false
for i := len(text) - 1; i >= 0; i-- {
c0, c1 = text[i], c0
if c0 < c1 {
isTypeS = true
} else if c0 > c1 && isTypeS {
isTypeS = false
// Bucket the index i+1 for the start of an LMS-substring.
b := bucket[c1] - 1
bucket[c1] = b
sa[b] = int64(i + 1)
lastB = b
numLMS++
}
}
// We recorded the LMS-substring starts but really want the ends.
// Luckily, with two differences, the start indexes and the end indexes are the same.
// The first difference is that the rightmost LMS-substring's end index is len(text),
// so the caller must pretend that sa[-1] == len(text), as noted above.
// The second difference is that the first leftmost LMS-substring start index
// does not end an earlier LMS-substring, so as an optimization we can omit
// that leftmost LMS-substring start index (the last one we wrote).
//
// Exception: if numLMS <= 1, the caller is not going to bother with
// the recursion at all and will treat the result as containing LMS-substring starts.
// In that case, we don't remove the final entry.
if numLMS > 1 {
sa[lastB] = 0
}
return numLMS
}
func induceSubL_8_64(text []byte, sa, freq, bucket []int64) {
// Initialize positions for left side of character buckets.
bucketMin_8_64(text, freq, bucket)
bucket = bucket[:256] // eliminate bounds check for bucket[cB] below
// As we scan the array left-to-right, each sa[i] = j > 0 is a correctly
// sorted suffix array entry (for text[j:]) for which we know that j-1 is type L.
// Because j-1 is type L, inserting it into sa now will sort it correctly.
// But we want to distinguish a j-1 with j-2 of type L from type S.
// We can process the former but want to leave the latter for the caller.
// We record the difference by negating j-1 if it is preceded by type S.
// Either way, the insertion (into the text[j-1] bucket) is guaranteed to
// happen at sa[i´] for some i´ > i, that is, in the portion of sa we have
// yet to scan. A single pass therefore sees indexes j, j-1, j-2, j-3,
// and so on, in sorted but not necessarily adjacent order, until it finds
// one preceded by an index of type S, at which point it must stop.
//
// As we scan through the array, we clear the worked entries (sa[i] > 0) to zero,
// and we flip sa[i] < 0 to -sa[i], so that the loop finishes with sa containing
// only the indexes of the leftmost L-type indexes for each LMS-substring.
//
// The suffix array sa therefore serves simultaneously as input, output,
// and a miraculously well-tailored work queue.
// placeLMS_8_64 left out the implicit entry sa[-1] == len(text),
// corresponding to the identified type-L index len(text)-1.
// Process it before the left-to-right scan of sa proper.
// See body in loop for commentary.
k := len(text) - 1
c0, c1 := text[k-1], text[k]
if c0 < c1 {
k = -k
}
// Cache recently used bucket index:
// we're processing suffixes in sorted order
// and accessing buckets indexed by the
// byte before the sorted order, which still
// has very good locality.
// Invariant: b is cached, possibly dirty copy of bucket[cB].
cB := c1
b := bucket[cB]
sa[b] = int64(k)
b++
for i := 0; i < len(sa); i++ {
j := int(sa[i])
if j == 0 {
// Skip empty entry.
continue
}
if j < 0 {
// Leave discovered type-S index for caller.
sa[i] = int64(-j)
continue
}
sa[i] = 0
// Index j was on work queue, meaning k := j-1 is L-type,
// so we can now place k correctly into sa.
// If k-1 is L-type, queue k for processing later in this loop.
// If k-1 is S-type (text[k-1] < text[k]), queue -k to save for the caller.
k := j - 1
c0, c1 := text[k-1], text[k]
if c0 < c1 {
k = -k
}
if cB != c1 {
bucket[cB] = b
cB = c1
b = bucket[cB]
}
sa[b] = int64(k)
b++
}
}
func induceSubL_32(text []int32, sa, freq, bucket []int32) {
// Initialize positions for left side of character buckets.
bucketMin_32(text, freq, bucket)
// As we scan the array left-to-right, each sa[i] = j > 0 is a correctly
// sorted suffix array entry (for text[j:]) for which we know that j-1 is type L.
// Because j-1 is type L, inserting it into sa now will sort it correctly.
// But we want to distinguish a j-1 with j-2 of type L from type S.
// We can process the former but want to leave the latter for the caller.
// We record the difference by negating j-1 if it is preceded by type S.
// Either way, the insertion (into the text[j-1] bucket) is guaranteed to
// happen at sa[i´] for some i´ > i, that is, in the portion of sa we have
// yet to scan. A single pass therefore sees indexes j, j-1, j-2, j-3,
// and so on, in sorted but not necessarily adjacent order, until it finds
// one preceded by an index of type S, at which point it must stop.
//
// As we scan through the array, we clear the worked entries (sa[i] > 0) to zero,
// and we flip sa[i] < 0 to -sa[i], so that the loop finishes with sa containing
// only the indexes of the leftmost L-type indexes for each LMS-substring.
//
// The suffix array sa therefore serves simultaneously as input, output,
// and a miraculously well-tailored work queue.
// placeLMS_32 left out the implicit entry sa[-1] == len(text),
// corresponding to the identified type-L index len(text)-1.
// Process it before the left-to-right scan of sa proper.
// See body in loop for commentary.
k := len(text) - 1
c0, c1 := text[k-1], text[k]
if c0 < c1 {
k = -k
}
// Cache recently used bucket index:
// we're processing suffixes in sorted order
// and accessing buckets indexed by the
// int32 before the sorted order, which still
// has very good locality.
// Invariant: b is cached, possibly dirty copy of bucket[cB].
cB := c1
b := bucket[cB]
sa[b] = int32(k)
b++
for i := 0; i < len(sa); i++ {
j := int(sa[i])
if j == 0 {
// Skip empty entry.
continue
}
if j < 0 {
// Leave discovered type-S index for caller.
sa[i] = int32(-j)
continue
}
sa[i] = 0
// Index j was on work queue, meaning k := j-1 is L-type,
// so we can now place k correctly into sa.
// If k-1 is L-type, queue k for processing later in this loop.
// If k-1 is S-type (text[k-1] < text[k]), queue -k to save for the caller.
k := j - 1
c0, c1 := text[k-1], text[k]
if c0 < c1 {
k = -k
}
if cB != c1 {
bucket[cB] = b
cB = c1
b = bucket[cB]
}
sa[b] = int32(k)
b++
}
}
func induceSubL_64(text []int64, sa, freq, bucket []int64) {
// Initialize positions for left side of character buckets.
bucketMin_64(text, freq, bucket)
// As we scan the array left-to-right, each sa[i] = j > 0 is a correctly
// sorted suffix array entry (for text[j:]) for which we know that j-1 is type L.
// Because j-1 is type L, inserting it into sa now will sort it correctly.
// But we want to distinguish a j-1 with j-2 of type L from type S.
// We can process the former but want to leave the latter for the caller.
// We record the difference by negating j-1 if it is preceded by type S.
// Either way, the insertion (into the text[j-1] bucket) is guaranteed to
// happen at sa[i´] for some i´ > i, that is, in the portion of sa we have
// yet to scan. A single pass therefore sees indexes j, j-1, j-2, j-3,
// and so on, in sorted but not necessarily adjacent order, until it finds
// one preceded by an index of type S, at which point it must stop.
//
// As we scan through the array, we clear the worked entries (sa[i] > 0) to zero,
// and we flip sa[i] < 0 to -sa[i], so that the loop finishes with sa containing
// only the indexes of the leftmost L-type indexes for each LMS-substring.
//
// The suffix array sa therefore serves simultaneously as input, output,
// and a miraculously well-tailored work queue.
// placeLMS_64 left out the implicit entry sa[-1] == len(text),
// corresponding to the identified type-L index len(text)-1.
// Process it before the left-to-right scan of sa proper.
// See body in loop for commentary.
k := len(text) - 1
c0, c1 := text[k-1], text[k]
if c0 < c1 {
k = -k
}
// Cache recently used bucket index:
// we're processing suffixes in sorted order
// and accessing buckets indexed by the
// int64 before the sorted order, which still
// has very good locality.
// Invariant: b is cached, possibly dirty copy of bucket[cB].
cB := c1
b := bucket[cB]
sa[b] = int64(k)
b++
for i := 0; i < len(sa); i++ {
j := int(sa[i])
if j == 0 {
// Skip empty entry.
continue
}
if j < 0 {
// Leave discovered type-S index for caller.
sa[i] = int64(-j)
continue
}
sa[i] = 0
// Index j was on work queue, meaning k := j-1 is L-type,
// so we can now place k correctly into sa.
// If k-1 is L-type, queue k for processing later in this loop.
// If k-1 is S-type (text[k-1] < text[k]), queue -k to save for the caller.
k := j - 1
c0, c1 := text[k-1], text[k]
if c0 < c1 {
k = -k
}
if cB != c1 {
bucket[cB] = b
cB = c1
b = bucket[cB]
}
sa[b] = int64(k)
b++
}
}
func induceSubS_8_64(text []byte, sa, freq, bucket []int64) {
// Initialize positions for right side of character buckets.
bucketMax_8_64(text, freq, bucket)
bucket = bucket[:256] // eliminate bounds check for bucket[cB] below
// Analogous to induceSubL_8_64 above,
// as we scan the array right-to-left, each sa[i] = j > 0 is a correctly
// sorted suffix array entry (for text[j:]) for which we know that j-1 is type S.
// Because j-1 is type S, inserting it into sa now will sort it correctly.
// But we want to distinguish a j-1 with j-2 of type S from type L.
// We can process the former but want to leave the latter for the caller.
// We record the difference by negating j-1 if it is preceded by type L.
// Either way, the insertion (into the text[j-1] bucket) is guaranteed to
// happen at sa[i´] for some i´ < i, that is, in the portion of sa we have
// yet to scan. A single pass therefore sees indexes j, j-1, j-2, j-3,
// and so on, in sorted but not necessarily adjacent order, until it finds
// one preceded by an index of type L, at which point it must stop.
// That index (preceded by one of type L) is an LMS-substring start.
//
// As we scan through the array, we clear the worked entries (sa[i] > 0) to zero,
// and we flip sa[i] < 0 to -sa[i] and compact into the top of sa,
// so that the loop finishes with the top of sa containing exactly
// the LMS-substring start indexes, sorted by LMS-substring.
// Cache recently used bucket index:
cB := byte(0)
b := bucket[cB]
top := len(sa)
for i := len(sa) - 1; i >= 0; i-- {
j := int(sa[i])
if j == 0 {
// Skip empty entry.
continue
}
sa[i] = 0
if j < 0 {
// Leave discovered LMS-substring start index for caller.
top--
sa[top] = int64(-j)
continue
}
// Index j was on work queue, meaning k := j-1 is S-type,
// so we can now place k correctly into sa.
// If k-1 is S-type, queue k for processing later in this loop.
// If k-1 is L-type (text[k-1] > text[k]), queue -k to save for the caller.
k := j - 1
c1 := text[k]
c0 := text[k-1]
if c0 > c1 {
k = -k
}
if cB != c1 {
bucket[cB] = b
cB = c1
b = bucket[cB]
}
b--
sa[b] = int64(k)
}
}
func induceSubS_32(text []int32, sa, freq, bucket []int32) {
// Initialize positions for right side of character buckets.
bucketMax_32(text, freq, bucket)
// Analogous to induceSubL_32 above,
// as we scan the array right-to-left, each sa[i] = j > 0 is a correctly
// sorted suffix array entry (for text[j:]) for which we know that j-1 is type S.
// Because j-1 is type S, inserting it into sa now will sort it correctly.
// But we want to distinguish a j-1 with j-2 of type S from type L.
// We can process the former but want to leave the latter for the caller.
// We record the difference by negating j-1 if it is preceded by type L.
// Either way, the insertion (into the text[j-1] bucket) is guaranteed to
// happen at sa[i´] for some i´ < i, that is, in the portion of sa we have
// yet to scan. A single pass therefore sees indexes j, j-1, j-2, j-3,
// and so on, in sorted but not necessarily adjacent order, until it finds
// one preceded by an index of type L, at which point it must stop.
// That index (preceded by one of type L) is an LMS-substring start.
//
// As we scan through the array, we clear the worked entries (sa[i] > 0) to zero,
// and we flip sa[i] < 0 to -sa[i] and compact into the top of sa,
// so that the loop finishes with the top of sa containing exactly
// the LMS-substring start indexes, sorted by LMS-substring.
// Cache recently used bucket index:
cB := int32(0)
b := bucket[cB]
top := len(sa)
for i := len(sa) - 1; i >= 0; i-- {
j := int(sa[i])
if j == 0 {
// Skip empty entry.
continue
}
sa[i] = 0
if j < 0 {
// Leave discovered LMS-substring start index for caller.
top--
sa[top] = int32(-j)
continue
}
// Index j was on work queue, meaning k := j-1 is S-type,
// so we can now place k correctly into sa.
// If k-1 is S-type, queue k for processing later in this loop.
// If k-1 is L-type (text[k-1] > text[k]), queue -k to save for the caller.
k := j - 1
c1 := text[k]
c0 := text[k-1]
if c0 > c1 {
k = -k
}
if cB != c1 {
bucket[cB] = b
cB = c1
b = bucket[cB]
}
b--
sa[b] = int32(k)
}
}
func induceSubS_64(text []int64, sa, freq, bucket []int64) {
// Initialize positions for right side of character buckets.
bucketMax_64(text, freq, bucket)
// Analogous to induceSubL_64 above,
// as we scan the array right-to-left, each sa[i] = j > 0 is a correctly
// sorted suffix array entry (for text[j:]) for which we know that j-1 is type S.
// Because j-1 is type S, inserting it into sa now will sort it correctly.
// But we want to distinguish a j-1 with j-2 of type S from type L.
// We can process the former but want to leave the latter for the caller.
// We record the difference by negating j-1 if it is preceded by type L.
// Either way, the insertion (into the text[j-1] bucket) is guaranteed to
// happen at sa[i´] for some i´ < i, that is, in the portion of sa we have
// yet to scan. A single pass therefore sees indexes j, j-1, j-2, j-3,
// and so on, in sorted but not necessarily adjacent order, until it finds
// one preceded by an index of type L, at which point it must stop.
// That index (preceded by one of type L) is an LMS-substring start.
//
// As we scan through the array, we clear the worked entries (sa[i] > 0) to zero,
// and we flip sa[i] < 0 to -sa[i] and compact into the top of sa,
// so that the loop finishes with the top of sa containing exactly
// the LMS-substring start indexes, sorted by LMS-substring.
// Cache recently used bucket index:
cB := int64(0)
b := bucket[cB]
top := len(sa)
for i := len(sa) - 1; i >= 0; i-- {
j := int(sa[i])
if j == 0 {
// Skip empty entry.
continue
}
sa[i] = 0
if j < 0 {
// Leave discovered LMS-substring start index for caller.
top--
sa[top] = int64(-j)
continue
}
// Index j was on work queue, meaning k := j-1 is S-type,
// so we can now place k correctly into sa.
// If k-1 is S-type, queue k for processing later in this loop.
// If k-1 is L-type (text[k-1] > text[k]), queue -k to save for the caller.
k := j - 1
c1 := text[k]
c0 := text[k-1]
if c0 > c1 {
k = -k
}
if cB != c1 {
bucket[cB] = b
cB = c1
b = bucket[cB]
}
b--
sa[b] = int64(k)
}
}
func length_8_64(text []byte, sa []int64, numLMS int) {
end := 0 // index of current LMS-substring end (0 indicates final LMS-substring)
// The encoding of N text bytes into a “length” word
// adds 1 to each byte, packs them into the bottom
// N*8 bits of a word, and then bitwise inverts the result.
// That is, the text sequence A B C (hex 41 42 43)
// encodes as ^uint64(0x42_43_44).
// LMS-substrings can never start or end with 0xFF.
// Adding 1 ensures the encoded byte sequence never
// starts or ends with 0x00, so that present bytes can be
// distinguished from zero-padding in the top bits,
// so the length need not be separately encoded.
// Inverting the bytes increases the chance that a
// 4-byte encoding will still be ≥ len(text).
// In particular, if the first byte is ASCII (<= 0x7E, so +1 <= 0x7F)
// then the high bit of the inversion will be set,
// making it clearly not a valid length (it would be a negative one).
//
// cx holds the pre-inverted encoding (the packed incremented bytes).
cx := uint64(0) // byte-only
// This stanza (until the blank line) is the "LMS-substring iterator",
// described in placeLMS_8_64 above, with one line added to maintain cx.
c0, c1, isTypeS := byte(0), byte(0), false
for i := len(text) - 1; i >= 0; i-- {
c0, c1 = text[i], c0
cx = cx<<8 | uint64(c1+1) // byte-only
if c0 < c1 {
isTypeS = true
} else if c0 > c1 && isTypeS {
isTypeS = false
// Index j = i+1 is the start of an LMS-substring.
// Compute length or encoded text to store in sa[j/2].
j := i + 1
var code int64
if end == 0 {
code = 0
} else {
code = int64(end - j)
if code <= 64/8 && ^cx >= uint64(len(text)) { // byte-only
code = int64(^cx) // byte-only
} // byte-only
}
sa[j>>1] = code
end = j + 1
cx = uint64(c1 + 1) // byte-only
}
}
}
func length_32(text []int32, sa []int32, numLMS int) {
end := 0 // index of current LMS-substring end (0 indicates final LMS-substring)
// The encoding of N text int32s into a “length” word
// adds 1 to each int32, packs them into the bottom
// N*8 bits of a word, and then bitwise inverts the result.
// That is, the text sequence A B C (hex 41 42 43)
// encodes as ^uint32(0x42_43_44).
// LMS-substrings can never start or end with 0xFF.
// Adding 1 ensures the encoded int32 sequence never
// starts or ends with 0x00, so that present int32s can be
// distinguished from zero-padding in the top bits,
// so the length need not be separately encoded.
// Inverting the int32s increases the chance that a
// 4-int32 encoding will still be ≥ len(text).
// In particular, if the first int32 is ASCII (<= 0x7E, so +1 <= 0x7F)
// then the high bit of the inversion will be set,
// making it clearly not a valid length (it would be a negative one).
//