forked from simongog/sdsl-lite
-
Notifications
You must be signed in to change notification settings - Fork 1
/
rrr_vector.hpp
717 lines (661 loc) · 29.3 KB
/
rrr_vector.hpp
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
/* sdsl - succinct data structures library
Copyright (C) 2011-2013 Simon Gog
This program is free software: you can redistribute it and/or modify
it under the terms of the GNU General Public License as published by
the Free Software Foundation, either version 3 of the License, or
(at your option) any later version.
This program is distributed in the hope that it will be useful,
but WITHOUT ANY WARRANTY; without even the implied warranty of
MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
GNU General Public License for more details.
You should have received a copy of the GNU General Public License
along with this program. If not, see http://www.gnu.org/licenses/ .
*/
/*! \file rrr_vector.hpp
\brief rrr_vector.hpp contains the sdsl::rrr_vector class, and
classes which support rank and select for rrr_vector.
\author Simon Gog, Matthias Petri
*/
#ifndef INCLUDED_SDSL_RRR_VECTOR
#define INCLUDED_SDSL_RRR_VECTOR
#include "int_vector.hpp"
#include "util.hpp"
#include "rrr_helper.hpp" // for binomial helper class
#include "iterators.hpp"
#include <iostream>
//! Namespace for the succinct data structure library
namespace sdsl
{
// forward declaration needed for friend declaration
template<uint8_t t_b=1, uint16_t t_bs=15, class t_rac=int_vector<>, uint16_t t_k=32>
class rank_support_rrr; // in rrr_vector
// forward declaration needed for friend declaration
template<uint8_t t_b=1, uint16_t t_bs=15, class t_rac=int_vector<>, uint16_t t_k=32>
class select_support_rrr; // in rrr_vector
//! A \f$H_0f$-compressed bitvector representation.
/*!
* \tparam t_bs Size of a basic block.
* \tparam t_rac Random access integer vector. Use to store the block types.
* It is possible to use WTs for t_rac.
* \tparam t_k A rank sample value is stored before every t_k-th basic block.
*
* References:
* - Rasmus Pagh
* Low redundancy in dictionaries with O(1) worst case lookup time
* Technical Report 1998.
* ftp://ftp.cs.au.dk/BRICS/Reports/RS/98/28/BRICS-RS-98-28.pdf, Section 2.
* - Rajeev Raman, V. Raman and S. Srinivasa Rao
* Succinct Indexable Dictionaries with Applications to representations
* of k-ary trees and multi-sets.
* SODA 2002.
* - Francisco Claude, Gonzalo Navarro:
* Practical Rank/Select Queries over Arbitrary Sequences.
* SPIRE 2008: 176-187
* - On the fly-decoding and encoding was discovered in;
* Gonzalo Navarro, Eliana Providel:
* Fast, Small, Simple Rank/Select on Bitmaps.
* SEA 2012
*
* In this version the block size can be adjust by the template parameter t_bs!
* \sa sdsl::rrr_vector for a specialized version for block_size=15
*/
template<uint16_t t_bs=63, class t_rac=int_vector<>, uint16_t t_k=32>
class rrr_vector
{
static_assert(t_bs >= 3 and t_bs <= 256 , "rrr_vector: block size t_bs must be 3 <= t_bs <= 256.");
static_assert(t_k > 1, "rrr_vector: t_k must be > 0.");
public:
typedef bit_vector::size_type size_type;
typedef bit_vector::value_type value_type;
typedef bit_vector::difference_type difference_type;
typedef t_rac rac_type;
typedef random_access_const_iterator<rrr_vector> iterator;
typedef iterator const_iterator;
typedef bv_tag index_category;
typedef rank_support_rrr<1, t_bs, t_rac, t_k> rank_1_type;
typedef rank_support_rrr<0, t_bs, t_rac, t_k> rank_0_type;
typedef select_support_rrr<1, t_bs, t_rac, t_k> select_1_type;
typedef select_support_rrr<0, t_bs, t_rac, t_k> select_0_type;
friend class rank_support_rrr<0, t_bs, t_rac, t_k>;
friend class rank_support_rrr<1, t_bs, t_rac, t_k>;
friend class select_support_rrr<0, t_bs, t_rac, t_k>;
friend class select_support_rrr<1, t_bs, t_rac, t_k>;
typedef rrr_helper<t_bs> rrr_helper_type;
typedef typename rrr_helper_type::number_type number_type;
enum { block_size = t_bs };
private:
size_type m_size = 0; // Size of the original bit_vector.
rac_type m_bt; // Vector for the block types (bt). bt equals the
// number of set bits in the block.
bit_vector m_btnr; // Compressed block type numbers.
int_vector<> m_btnrp; // Sample pointers into m_btnr.
int_vector<> m_rank; // Sample rank values.
public:
//! Default constructor
rrr_vector() {}
//! Constructor
/*!
* \param bv Uncompressed bitvector.
* \param k Store rank samples and pointers each k-th blocks.
*/
rrr_vector(const bit_vector& bv)
{
m_size = bv.size();
int_vector<> bt_array((m_size+t_bs)/((size_type)t_bs), 0, bits::hi(t_bs)+1);
//int_vector<> bt_array;
//bt_array.width(bits::hi(t_bs)+1);
//bt_array.resize((m_size+t_bs)/((size_type)t_bs)); // blocks for the bt_array + a dummy block at the end,
// if m_size%t_bs == 0
// (1) calculate the block types and store them in m_bt
size_type pos = 0, i = 0, x;
size_type btnr_pos = 0;
size_type sum_rank = 0;
while (pos + t_bs <= m_size) { // handle all blocks full blocks
bt_array[ i++ ] = x = rrr_helper_type::get_bt(bv, pos, t_bs);
sum_rank += x;
btnr_pos += rrr_helper_type::space_for_bt(x);
pos += t_bs;
}
if (pos < m_size) { // handle last not full block
bt_array[ i++ ] = x = rrr_helper_type::get_bt(bv, pos, m_size - pos);
sum_rank += x;
btnr_pos += rrr_helper_type::space_for_bt(x);
}
m_btnr = bit_vector(std::max(btnr_pos, (size_type)64), 0); // max necessary for case: t_bs == 1
m_btnrp = int_vector<>((bt_array.size()+t_k-1)/t_k, 0, bits::hi(btnr_pos)+1);
m_rank = int_vector<>((bt_array.size()+t_k-1)/t_k + ((m_size % (t_k*t_bs))>0), 0, bits::hi(sum_rank)+1);
// ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
// only add a finishing block, if the last block of the superblock is not a dummy block
// (2) calculate block type numbers and pointers into btnr and rank samples
pos = 0; i = 0;
btnr_pos= 0, sum_rank = 0;
while (pos + t_bs <= m_size) { // handle all full blocks
if ((i % t_k) == (size_type)0) {
m_rank[ i/t_k ] = sum_rank;
m_btnrp[ i/t_k ] = btnr_pos;
}
uint16_t space_for_bt = rrr_helper_type::space_for_bt(x=bt_array[i++]);
sum_rank += x;
if (space_for_bt) {
number_type bin = rrr_helper_type::decode_btnr(bv, pos, t_bs);
// invert blocks with more than n/2 1-bits
if (x > t_bs/2) {
x = t_bs - x;
bin = (~bin) & rrr_helper_type::binomial::data.L1Mask;
}
number_type nr = rrr_helper_type::bin_to_nr(bin, x);
rrr_helper_type::set_bt(m_btnr, btnr_pos, nr, space_for_bt);
}
btnr_pos += space_for_bt;
pos += t_bs;
}
if (pos < m_size) { // handle last not full block
if ((i % t_k) == (size_type)0) {
m_rank[ i/t_k ] = sum_rank;
m_btnrp[ i/t_k ] = btnr_pos;
}
uint16_t space_for_bt = rrr_helper_type::space_for_bt(x=bt_array[i++]);
// no extra dummy block added to bt_array, therefore this condition should hold
assert(i == bt_array.size());
sum_rank += x;
if (space_for_bt) {
number_type bin = rrr_helper_type::decode_btnr(bv, pos, m_size-pos);
// invert blocks with more than n/2 1-bits
if (x > t_bs/2) {
x = t_bs - x;
bin = (~bin) & rrr_helper_type::binomial::data.L1Mask;
}
number_type nr = rrr_helper_type::bin_to_nr(bin, x);
rrr_helper_type::set_bt(m_btnr, btnr_pos, nr, space_for_bt);
}
btnr_pos += space_for_bt;
}
assert(m_rank.size()-1 == ((i+t_k-1)/t_k));
// for technical reasons we add a last element to m_rank
m_rank[ m_rank.size()-1 ] = sum_rank; // sum_rank contains the total number of set bits in bv
m_bt = bt_array;
}
//! Swap method
void swap(rrr_vector& rrr)
{
if (this != &rrr) {
std::swap(m_size, rrr.m_size);
m_bt.swap(rrr.m_bt);
m_btnr.swap(rrr.m_btnr);
m_btnrp.swap(rrr.m_btnrp);
m_rank.swap(rrr.m_rank);
}
}
//! Accessing the i-th element of the original bit_vector
/*! \param i An index i with \f$ 0 \leq i < size() \f$.
\return The i-th bit of the original bit_vector
*/
value_type operator[](size_type i)const
{
size_type bt_idx = i/t_bs;
uint16_t bt = m_bt[bt_idx];
#ifndef RRR_NO_OPT
if (bt == 0 or bt == t_bs) { // very effective optimization
return bt>0;
}
#endif
size_type sample_pos = bt_idx/t_k;
size_type btnrp = m_btnrp[ sample_pos ];
for (size_type j = sample_pos*t_k; j < bt_idx; ++j) {
btnrp += rrr_helper_type::space_for_bt(m_bt[j]);
}
uint16_t btnrlen = rrr_helper_type::space_for_bt(bt);
number_type btnr = rrr_helper_type::decode_btnr(m_btnr, btnrp, btnrlen);
// blocks with more than n/2 1-bits are inverted
value_type bit = (bt <= t_bs/2) ? rrr_helper_type::decode_bit(bt, btnr, i % t_bs)
: !rrr_helper_type::decode_bit(t_bs - bt, btnr, i % t_bs);
assert(bit == inverse_select(i).first);
return bit;
}
//! Accessing the i-th element of the original bit_vector and rank at
// that position, the umber of 1-bits in the prefix [0..i-1].
/*! \param i An index i with \f$ 0 \leq i < size() \f$.
\return The i-th bit of the original bit_vector and the umber of
1-bits in the prefix [0..i-1] of the original bit_vector.
*/
const std::pair<bool, size_type> inverse_select(size_type i)const
{
assert(i<size());
size_type bt_idx = i/t_bs;
size_type sample_pos = bt_idx/t_k;
size_type rank = m_rank[ sample_pos ];
uint16_t bt = m_bt[ bt_idx ];
#ifndef RRR_NO_OPT
assert(sample_pos+1 < m_rank.size());
if (bt == 0 && m_rank[ sample_pos+1 ] == rank) {
return std::make_pair(false, rank);
} else if (bt == t_bs && m_rank[ sample_pos+1 ] == rank + (size_type)t_bs*t_k) {
return std::make_pair(true, rank + (i - sample_pos*t_k*t_bs));
}
#endif
size_type btnrp = m_btnrp[ sample_pos ];
for (size_type j = sample_pos*t_k; j < bt_idx; ++j) {
rank += m_bt[j];
btnrp += rrr_helper_type::space_for_bt(m_bt[j]);
}
uint16_t off = i % t_bs;
if (bt == 0) {
return std::make_pair(false, rank);
} else if (bt == t_bs) { // very effective optimization
return std::make_pair(true, rank + off);
}
uint16_t btnrlen = rrr_helper_type::space_for_bt(bt);
number_type btnr = rrr_helper_type::decode_btnr(m_btnr, btnrp, btnrlen);
// blocks with more than n/2 1-bits are inverted
std::pair<bool, uint16_t> pair = rrr_helper_type::decode_bit_and_popcount((bt <= t_bs/2) ? bt : t_bs - bt, btnr, off);
return (bt <= t_bs/2) ? std::make_pair(pair.first, rank + pair.second)
: std::make_pair(!pair.first, rank + (off - pair.second));
}
//! Accessing the i-th element of the original bit_vector and if the
// bit is set, query the rank at that position, the umber of 1-bits
// in the prefix [0..i]. If the bit is unset, return zero.
/*! \param i An index i with \f$ 0 \leq i < size() \f$.
\return The number of 1-bits in the prefix [0..i], or zero.
*/
size_type conditional_rank(size_type i)const
{
assert(i<size());
size_type bt_idx = i/t_bs;
size_type sample_pos = bt_idx/t_k;
size_type rank = m_rank[ sample_pos ];
uint16_t bt = m_bt[ bt_idx ];
#ifndef RRR_NO_OPT
assert(sample_pos+1 < m_rank.size());
if (bt == 0) {
return 0;
} else if (bt == t_bs && m_rank[ sample_pos+1 ] == rank + (size_type)t_bs*t_k) {
return rank + (i - sample_pos*t_k*t_bs) + 1;
}
#endif
size_type btnrp = m_btnrp[ sample_pos ];
for (size_type j = sample_pos*t_k; j < bt_idx; ++j) {
rank += m_bt[j];
btnrp += rrr_helper_type::space_for_bt(m_bt[j]);
}
uint16_t off = i % t_bs;
#ifdef RRR_NO_OPT
if (bt == 0) {
return 0;
}
#endif
if (bt == t_bs) { // very effective optimization
return rank + off + 1;
}
uint16_t btnrlen = rrr_helper_type::space_for_bt(bt);
number_type btnr = rrr_helper_type::decode_btnr(m_btnr, btnrp, btnrlen);
// blocks with more than n/2 1-bits are inverted
std::pair<bool, uint16_t> pair = rrr_helper_type::decode_bit_and_popcount((bt <= t_bs/2) ? bt : t_bs - bt, btnr, off);
return (bt <= t_bs/2)
? (pair.first ? rank + pair.second + 1 : 0)
: (!pair.first ? rank + (off - pair.second) + 1 : 0);
}
//! Get the integer value of the binary string of length len starting at position idx.
/*! \param idx Starting index of the binary representation of the integer.
* \param len Length of the binary representation of the integer. Default value is 64.
* \returns The integer value of the binary string of length len starting at position idx.
*
* \pre idx+len-1 in [0..size()-1]
* \pre len in [1..64]
*/
uint64_t get_int(size_type idx, uint8_t len=64)const
{
uint64_t res = 0;
size_type bb_idx = idx/t_bs; // begin block index
uint16_t bb_off = idx%t_bs; // begin block offset
uint16_t bt = m_bt[bb_idx];
size_type sample_pos = bb_idx/t_k;
size_type eb_idx = (idx+len-1)/t_bs; // end block index
if (bb_idx == eb_idx) { // extract only in one block
if (bt == 0) { // all bits are zero
res = 0;
} else if (bt == t_bs) { // all bits are ones
res = bits::lo_set[len];
} else {
size_type btnrp = m_btnrp[ sample_pos ];
for (size_type j = sample_pos*t_k; j < bb_idx; ++j) {
btnrp += rrr_helper_type::space_for_bt(m_bt[j]);
}
uint16_t btnrlen = rrr_helper_type::space_for_bt(bt);
number_type btnr = rrr_helper_type::decode_btnr(m_btnr, btnrp, btnrlen);
// blocks with more than n/2 1-bits are inverted
res = (bt <= t_bs/2) ? rrr_helper_type::decode_int(bt, btnr, bb_off, len)
: (~rrr_helper_type::decode_int(t_bs - bt, btnr, bb_off, len)) & bits::lo_set[len];
}
} else { // solve multiple block case by recursion
uint16_t b_len = t_bs-bb_off; // remaining bits in first block
uint16_t b_len_sum = 0;
do {
res |= get_int(idx, b_len) << b_len_sum;
idx += b_len;
b_len_sum += b_len;
len -= b_len;
b_len = std::min((uint16_t)len, t_bs);
} while (len > 0);
}
return res;
}
//! Returns the size of the original bit vector.
size_type size()const
{
return m_size;
}
//! Answers select queries
//! Serializes the data structure into the given ostream
size_type serialize(std::ostream& out, structure_tree_node* v=nullptr, std::string name="")const
{
structure_tree_node* child = structure_tree::add_child(v, name, util::class_name(*this));
size_type written_bytes = 0;
written_bytes += write_member(m_size, out, child, "size");
written_bytes += m_bt.serialize(out, child, "bt");
written_bytes += m_btnr.serialize(out, child, "btnr");
written_bytes += m_btnrp.serialize(out, child, "btnrp");
written_bytes += m_rank.serialize(out, child, "rank_samples");
structure_tree::add_size(child, written_bytes);
return written_bytes;
}
//! Loads the data structure from the given istream.
void load(std::istream& in)
{
read_member(m_size, in);
m_bt.load(in);
m_btnr.load(in);
m_btnrp.load(in);
m_rank.load(in);
}
iterator begin() const
{
return iterator(this, 0);
}
iterator end() const
{
return iterator(this, size());
}
};
template<uint8_t t_bit_pattern>
struct rank_support_rrr_trait {
typedef bit_vector::size_type size_type;
static size_type adjust_rank(size_type r, SDSL_UNUSED size_type n)
{
return r;
}
};
template<>
struct rank_support_rrr_trait<0> {
typedef bit_vector::size_type size_type;
static size_type adjust_rank(size_type r, size_type n)
{
return n - r;
}
};
//! rank_support for the rrr_vector class
/*!
* \tparam t_b The bit pattern of size one. (so `0` or `1`)
* \tparam t_bs The block size of the corresponding rrr_vector
* \tparam t_rac Type used to store the block type in the corresponding rrr_vector.
* TODO: Test if the binary search can be speed up by
* saving the (n/2)-th rank value in T[0], the (n/4)-th in T[1],
* the (3n/4)-th in T[2],... for small number of rank values
* is this called hinted binary search???
* or is this called
*/
template<uint8_t t_b, uint16_t t_bs, class t_rac, uint16_t t_k>
class rank_support_rrr
{
static_assert(t_b == 1u or t_b == 0u , "rank_support_rrr: bit pattern must be `0` or `1`");
public:
typedef rrr_vector<t_bs, t_rac, t_k> bit_vector_type;
typedef typename bit_vector_type::size_type size_type;
typedef typename bit_vector_type::rrr_helper_type rrr_helper_type;
typedef typename rrr_helper_type::number_type number_type;
enum { bit_pat = t_b };
enum { bit_pat_len = (uint8_t)1 };
private:
const bit_vector_type* m_v; //!< Pointer to the rank supported rrr_vector
public:
//! Standard constructor
/*! \param v Pointer to the rrr_vector, which should be supported
*/
explicit rank_support_rrr(const bit_vector_type* v=nullptr)
{
set_vector(v);
}
//! Answers rank queries
/*! \param i Argument for the length of the prefix v[0..i-1], with \f$0\leq i \leq size()\f$.
\returns Number of 1-bits in the prefix [0..i-1] of the original bit_vector.
\par Time complexity
\f$ \Order{ sample\_rate of the rrr\_vector} \f$
*/
const size_type rank(size_type i)const
{
assert(m_v != nullptr);
assert(i <= m_v->size());
size_type bt_idx = i/t_bs;
size_type sample_pos = bt_idx/t_k;
size_type rank = m_v->m_rank[ sample_pos ];
uint16_t bt = m_v->m_bt[ bt_idx ];
#ifndef RRR_NO_OPT
if (i < m_v->size()) {
assert(sample_pos+1 < m_v->m_rank.size());
if (bt == 0 && m_v->m_rank[ sample_pos+1 ] == rank) {
return rank_support_rrr_trait<t_b>::adjust_rank(rank, i);
} else if (bt == t_bs && m_v->m_rank[ sample_pos+1 ] == rank + (size_type)t_bs*t_k) {
return rank_support_rrr_trait<t_b>::adjust_rank(
rank + i - sample_pos*t_k*t_bs, i);
}
}
#endif
size_type btnrp = m_v->m_btnrp[ sample_pos ];
for (size_type j = sample_pos*t_k; j < bt_idx; ++j) {
rank += m_v->m_bt[j];
btnrp += rrr_helper_type::space_for_bt(m_v->m_bt[j]);
}
uint16_t off = i % t_bs;
if (!off || bt == 0) {
return rank_support_rrr_trait<t_b>::adjust_rank(rank, i);
} else if (bt == t_bs) { // very effective optimization
return rank_support_rrr_trait<t_b>::adjust_rank(rank + off, i);
}
uint16_t btnrlen = rrr_helper_type::space_for_bt(bt);
number_type btnr = rrr_helper_type::decode_btnr(m_v->m_btnr, btnrp, btnrlen);
// blocks with more than n/2 1-bits are inverted
uint16_t popcnt = (bt <= t_bs/2) ? rrr_helper_type::decode_popcount(bt, btnr, off)
: off - rrr_helper_type::decode_popcount(t_bs - bt, btnr, off);
assert(i >= size() || rank + popcnt == m_v->inverse_select(i).second);
return rank_support_rrr_trait<t_b>::adjust_rank(rank + popcnt, i);
}
//! Short hand for rank(i)
const size_type operator()(size_type i)const
{
return rank(i);
}
//! Returns the size of the original vector
const size_type size()const
{
return m_v->size();
}
//! Set the supported vector.
void set_vector(const bit_vector_type* v=nullptr)
{
m_v = v;
}
void swap(rank_support_rrr&) { }
//! Load the data structure from a stream and set the supported vector.
void load(std::istream&, const bit_vector_type* v=nullptr)
{
set_vector(v);
}
//! Serializes the data structure into a stream.
size_type serialize(std::ostream&, structure_tree_node* v=nullptr, std::string name="")const
{
structure_tree_node* child = structure_tree::add_child(v, name, util::class_name(*this));
structure_tree::add_size(child, 0);
return 0;
}
};
//! Select support for the rrr_vector class.
/*
* \tparam t_b The bit pattern of size one. (so `0` or `1`)
* \tparam t_bs The block size of the corresponding rrr_vector
* \tparam t_rac Type used to store the block type in the corresponding rrr_vector.
*
* Possible TODO: Add heap which contains the 10 first items of
* each binary search could increase performance.
* Experiments on select_support_interleaved showed about
* 25%.
*/
template<uint8_t t_b, uint16_t t_bs, class t_rac, uint16_t t_k>
class select_support_rrr
{
static_assert(t_b == 1u or t_b == 0u , "select_support_rrr: bit pattern must be `0` or `1`");
public:
typedef rrr_vector<t_bs, t_rac, t_k> bit_vector_type;
typedef typename bit_vector_type::size_type size_type;
typedef typename bit_vector_type::rrr_helper_type rrr_helper_type;
typedef typename rrr_helper_type::number_type number_type;
enum { bit_pat = t_b };
enum { bit_pat_len = (uint8_t)1 };
private:
const bit_vector_type* m_v; //!< Pointer to the rank supported rrr_vector
size_type select1(size_type i)const
{
// (1) binary search for the answer in the rank_samples
size_type begin=0, end=m_v->m_rank.size()-1; // min included, max excluded
size_type idx, rank;
// invariant: m_rank[end] >= i
// m_rank[begin] < i
while (end-begin > 1) {
idx = (begin+end) >> 1; // idx in [0..m_rank.size()-1]
rank = m_v->m_rank[idx];
if (rank >= i)
end = idx;
else { // rank < i
begin = idx;
}
}
// (2) linear search between the samples
rank = m_v->m_rank[begin]; // now i>rank
idx = begin * t_k; // initialize idx for select result
if (m_v->m_rank[end] < i) {
return size();
}
#ifndef RRR_NO_OPT
if (m_v->m_rank[end] == rank + (size_type)t_bs*t_k) {// optimisation for select<1>
return idx*t_bs + i-rank -1;
}
#endif
size_type btnrp = m_v->m_btnrp[ begin ];
uint16_t bt = 0, btnrlen = 0; // temp variables for block_type and space for block type
while (i > rank) {
bt = m_v->m_bt[idx++];
rank += bt;
btnrp += (btnrlen=rrr_helper_type::space_for_bt(bt));
}
rank -= bt;
#ifndef RRR_NO_OPT
assert(bt != 0);
if (bt == t_bs) { // very effective optimization
return (idx-1) * t_bs + i-rank-1;
}
#endif
number_type btnr = rrr_helper_type::decode_btnr(m_v->m_btnr, btnrp-btnrlen, btnrlen);
// blocks with more than n/2 1-bits are inverted
return (idx-1) * t_bs + ((bt <= t_bs/2)
? rrr_helper_type::decode_select(bt, btnr, i-rank)
: rrr_helper_type::decode_select0(t_bs - bt, btnr, i-rank));
}
size_type select0(size_type i)const
{
// (1) binary search for the answer in the rank_samples
size_type begin=0, end=m_v->m_rank.size()-1; // min included, max excluded
size_type idx, rank;
// invariant: m_rank[end] >= i
// m_rank[begin] < i
while (end-begin > 1) {
idx = (begin+end) >> 1; // idx in [0..m_rank.size()-1]
rank = idx*t_bs*t_k - m_v->m_rank[idx];
if (rank >= i)
end = idx;
else { // rank < i
begin = idx;
}
}
// (2) linear search between the samples
rank = begin*t_bs*t_k - m_v->m_rank[begin]; // now i>rank
idx = begin * t_k; // initialize idx for select result
if (size() - m_v->m_rank[end] < i) {
return size();
}
#ifndef RRR_NO_OPT
if (m_v->m_rank[end] == m_v->m_rank[begin]) { // only for select<0>
return idx*t_bs + i-rank -1;
}
#endif
size_type btnrp = m_v->m_btnrp[ begin ];
uint16_t bt = 0, btnrlen = 0; // temp variables for block_type and space for block type
while (i > rank) {
bt = m_v->m_bt[idx++];
rank += (t_bs-bt);
btnrp += (btnrlen=rrr_helper_type::space_for_bt(bt));
}
rank -= (t_bs-bt);
#ifndef RRR_NO_OPT
assert(bt != t_bs);
if (bt == 0) { // very effective optimization
return (idx-1) * t_bs + i-rank-1;
}
#endif
number_type btnr = rrr_helper_type::decode_btnr(m_v->m_btnr, btnrp-btnrlen, btnrlen);
// blocks with more than n/2 1-bits are inverted
return (idx-1) * t_bs + ((bt <= t_bs/2)
? rrr_helper_type::decode_select0(bt, btnr, i-rank)
: rrr_helper_type::decode_select(t_bs - bt, btnr, i-rank));
}
public:
explicit select_support_rrr(const bit_vector_type* v=nullptr)
{
set_vector(v);
}
//! Answers select queries
size_type select(size_type i)const
{
return t_b ? select1(i) : select0(i);
}
const size_type operator()(size_type i)const
{
return select(i);
}
const size_type size()const
{
return m_v->size();
}
void set_vector(const bit_vector_type* v=nullptr)
{
m_v = v;
}
select_support_rrr& operator=(const select_support_rrr& rs)
{
if (this != &rs) {
set_vector(rs.m_v);
}
return *this;
}
void swap(select_support_rrr&) { }
void load(std::istream&, const bit_vector_type* v=nullptr)
{
set_vector(v);
}
size_type serialize(std::ostream&, structure_tree_node* v=nullptr, std::string name="")const
{
structure_tree_node* child = structure_tree::add_child(v, name, util::class_name(*this));
structure_tree::add_size(child, 0);
return 0;
}
};
}// end namespace sdsl
#include "rrr_vector_15.hpp" // include specialization
#endif