sbwt 0.4.2

Indexing sets of DNA k-mers with the spectral Burrow-Wheeler transform.
Documentation
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
//! Miscellaneous utility functions and constants used in the crate.

use std::cmp::min;
use std::io::Cursor;
use std::io::Read;
use std::ops::Range;

use bitvec::prelude::*;
use rayon::iter::ParallelBridge;
use rayon::iter::IntoParallelRefIterator;
use rayon::iter::ParallelIterator;
use simple_sds_sbwt::serialize::Serialize;

type BitVec = bitvec::vec::BitVec<u64, Lsb0>;
type BitSlice = bitvec::slice::BitSlice<u64, Lsb0>;

// Returns the number of bytes written
pub(crate) fn write_bytes<W: std::io::Write>(out: &mut W, bytes: &[u8]) -> std::io::Result<usize>{
    out.write_all(bytes)?;
    Ok(bytes.len() + 8)
}

// Searcher the range [0..n)
// Return the index of the answer, or n if does not exist
pub(crate) fn binary_search_leftmost_that_fulfills_pred<T, Access: Fn(usize) -> T, Pred: Fn(T) -> bool>(access: Access, pred: Pred, n: usize) -> usize {
    let mut ans = n;
    let mut step = n.next_power_of_two();
    while step > 0 {
        while ans as isize - step as isize >= 0 && pred(access(ans-step)) {
            ans -= step;
        }
        step /= 2;
    }
    ans
}

pub const DNA_ALPHABET: [u8; 4] = [b'A', b'C', b'G', b'T'];

// This bit vector of length 256 marks the ascii values of these characters: acgtACGT
const IS_DNA: BitArray<[u32; 8]> = bitarr![const u32, Lsb0; 0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,1,0,1,0,0,0,1,0,0,0,0,0,0,0,0,0,0,0,0,1,0,0,0,0,0,0,0,0,0,0,0,0,1,0,1,0,0,0,1,0,0,0,0,0,0,0,0,0,0,0,0,1,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0];

pub(crate) fn is_dna(c: u8) -> bool {
    IS_DNA[c as usize]
}

// A table mapping mapping ascii A -> 0, C -> 1, G -> 2, T -> 3. Same for lower case.
// All other charcters map to 255. Other code depends on this choice: don't touch it. 
pub const ACGT_TO_0123: [u8; 256] = [255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 0, 255, 1, 255, 255, 255, 2, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 3, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 0, 255, 1, 255, 255, 255, 2, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 3, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255];

#[allow(non_snake_case)] // C-array is an established convention in BWT indexes
fn popcounts_to_C_array(popcounts: &[usize]) -> Vec<usize> {
    let sigma = popcounts.len();
    let mut C: Vec<usize> = vec![0; sigma];
    for c in 0..sigma {
        for d in (c + 1)..sigma {
            C[d] += popcounts[c];
        }
    }

    // Plus one for the ghost dollar
    #[allow(clippy::needless_range_loop)] // Is perfectly clear this way
    for c in 0..sigma {
        C[c] += 1;
    }

    C

}

#[allow(non_snake_case)] // C-array is an established convention in BWT indexes
pub(crate) fn get_C_array(rows: &[bitvec::vec::BitVec::<u64, Lsb0>]) -> Vec<usize> {
    let popcounts: Vec<usize> = rows.iter().map(|row| row.count_ones()).collect(); 
    popcounts_to_C_array(&popcounts)
}

#[allow(non_snake_case, dead_code)] // C-array is an established convention in BWT indexes
pub(crate) fn get_C_array_parallel(rawrows: &[bitvec::vec::BitVec::<u64, Lsb0>], n_threads: usize) -> Vec<usize> {
    let sigma = rawrows.len();
    assert!(sigma > 0);

    let thread_pool = rayon::ThreadPoolBuilder::new().num_threads(n_threads).build().unwrap();
    thread_pool.install(||{
        let popcounts: Vec<usize> = rawrows.par_iter().map(|row| row.count_ones()).collect(); 
        popcounts_to_C_array(&popcounts)
    })
}

/// Reverses the given ASCII DNA sequence and replaces each nucleotide with its complement.
pub fn reverse_complement_in_place(seq: &mut [u8]){
    jseqio::reverse_complement_in_place(seq);
}

#[allow(dead_code)]
pub(crate) struct FastXReader{
    inner: jseqio::reader::DynamicFastXReader
}

impl crate::SeqStream for FastXReader{
    fn stream_next(&mut self) -> Option<&[u8]> {
        self.inner.read_next().unwrap().map(|x| x.seq)
    }
}

/// Creates a [crate::SeqStream] out of a slice of ASCII sequences.
pub struct SliceSeqStream<'a>{
    slices: &'a [ &'a [u8]],
    cur_slice_idx: usize,
}

impl<'a> SliceSeqStream<'a> {
    /// Creates a [crate::SeqStream] out of a slice of ASCII sequences.
    pub fn new(slices: &'a [&'a [u8]]) -> Self {
        Self {slices, cur_slice_idx: 0}
    }
}

impl crate::SeqStream for SliceSeqStream<'_> {
    fn stream_next(&mut self) -> Option<&[u8]> {
        if self.cur_slice_idx == self.slices.len() {
            None
        } else {
            let s = self.slices[self.cur_slice_idx];
            self.cur_slice_idx += 1;
            Some(s)
        }
    } 
}

/// Creates a [crate::SeqStream] out of a slice of ascii vectors.
pub struct VecSeqStream<'a> {
    seqs: &'a [Vec<u8>],
    cur_seq_idx: usize,
}

impl<'a> VecSeqStream<'a> {
    /// Creates a [crate::SeqStream] out of a slice of ascii vectors.
    pub fn new(seqs: &'a [Vec<u8>]) -> Self {
        Self {seqs, cur_seq_idx: 0}
    }
}

impl crate::SeqStream for VecSeqStream<'_> {
    fn stream_next(&mut self) -> Option<&[u8]> {
        if self.cur_seq_idx == self.seqs.len() {
            None
        } else {
            let s = &self.seqs[self.cur_seq_idx];
            self.cur_seq_idx += 1;
            Some(s)
        }
    } 
}


// This function runs in parallel, so a rayon thread pool must be initialized.
pub(crate) fn parallel_bitvec_concat(bitvecs: Vec<BitVec>) -> BitVec {
    if bitvecs.len() == 1 {
        return bitvecs.into_iter().next().unwrap(); // Nothing to do
    }
    let total_length = bitvecs.iter().fold(0_usize, |acc, s| acc + s.len());
    let n_words = total_length.div_ceil(64);
    let mut output_data = vec![0_u64; n_words];

    let mut exclusive_input_bitslices = Vec::<&BitSlice>::new(); // One per nonempty input bitvec
    let mut exclusive_output_word_ranges = Vec::<Range<usize>>::new(); // One per nonempty input bitvec

    // Figure out which words will be potentially shared between input slices in the concatenation,
    // and set those bits. Push the exclusive input and output regions to vecs.
    let mut bits_so_far = 0_usize;
    for s in bitvecs.iter() {
        if s.is_empty() {
            continue // Nothing to concatenate
        }

        // Copy the part that falls in the first output word
        let first_word = bits_so_far / 64; // The first word that will be written to
        let first_word_bit_offset = bits_so_far % 64; // Index of the first bit that is written in the first word
        let n_bits_written_to_first_word = min(s.len(), 64 - first_word_bit_offset);
        let first_out = BitSlice::from_element_mut(&mut output_data[first_word]);
        let first_out_slice = &mut first_out[first_word_bit_offset..first_word_bit_offset + n_bits_written_to_first_word];
        first_out_slice.copy_from_bitslice(&s[0..n_bits_written_to_first_word]);

        // Copy the part that falls in the last output word (can be the same
        // as the first output word but that's okay).
        bits_so_far += s.len();
        let last_bit = bits_so_far - 1; // >= 0 because s.len() > 0
        let last_word = last_bit / 64;
        let last_word_bit_offset = last_bit % 64; // Index of the last bit that is written in the last word
        let n_bits_written_to_last_word = min(s.len(), last_word_bit_offset + 1);
        let last_out_word = BitSlice::from_element_mut(&mut output_data[last_word]);
        let last_out_slice = &mut last_out_word[(1 + last_word_bit_offset - n_bits_written_to_last_word)..=last_word_bit_offset];
        let last_in_slice = &s[(s.len() - n_bits_written_to_last_word)..];
        last_out_slice.copy_from_bitslice(last_in_slice);

        let first_exclusive_word = (first_word + 1) as isize;
        let last_exclusive_word = last_word as isize - 1;

        if first_exclusive_word <= last_exclusive_word {
            // Nonempty range of exclusive words
            exclusive_input_bitslices.push(&s[n_bits_written_to_first_word..(s.len()-n_bits_written_to_last_word)]);
            exclusive_output_word_ranges.push(first_exclusive_word as usize .. last_exclusive_word as usize +1);
        }

    }

    let exclusive_output_word_ranges = split_to_mut_regions(&mut output_data, exclusive_output_word_ranges);

    // Copy non-overlapping parts in parallel
    assert_eq!(exclusive_input_bitslices.len(), exclusive_output_word_ranges.len());
    exclusive_output_word_ranges.into_iter().enumerate().par_bridge().for_each(|(i, out_word_range)| {
        let out = BitSlice::from_slice_mut(out_word_range);
        out.copy_from_bitslice(exclusive_input_bitslices[i]);
    });

   let mut concat = BitVec::from_vec(output_data); // Reinterpret as BitVec
   concat.truncate(total_length); // Get rid of the tail in the last word
   concat

}

/// Returns a mutable slice for every region in `regions`.
///
/// ## Preconditions  (guaranteed by the caller)
/// * All ranges are inside `0..v.len()`.
/// * The ranges are sorted by `start` and do **not** overlap.
pub(crate) fn split_to_mut_regions(
    v: &mut Vec<u64>,
    regions: Vec<Range<usize>>,
) -> Vec<&mut [u64]> {
    let mut result = Vec::with_capacity(regions.len());
    let mut tail: &mut [u64] = v.as_mut_slice();
    let mut consumed = 0; // absolute index we have reached in `v`

    for r in regions {
        // translate the absolute `start` to an index inside `tail`
        let rel_start = r.start - consumed;
        let len       = r.end   - r.start;

        // First split off everything before the wanted region …
        let (_, after_start)    = tail.split_at_mut(rel_start);
        // ... then split that remainder into the desired region and the rest.
        let (region, after_end) = after_start.split_at_mut(len);

        result.push(region); // keep the region
        tail = after_end; // keep working with the suffix
        consumed = r.end; // advance the absolute cursor
    }

    result
}

pub(crate) fn bitvec_to_simple_sds_raw_bitvec(mut bv: bitvec::vec::BitVec::<u64, Lsb0>) -> simple_sds_sbwt::raw_vector::RawVector {
    // TODO: We really hope that usize equals u64 here, otherwise this this is probably broken.
    // Let's use the deserialization function in simple_sds_sbwt for a raw bitvector.
    // It requires the following header:
    let mut header = [0u64, 0u64]; // bits, words
    header[0] = bv.len() as u64; // Assumes little-endian byte order
    header[1] = bv.len().div_ceil(64) as u64;

    let header_bytes = bytemuck::cast_slice(&header);

    // Make sure the leftover bits in the last word are zeros. Simple-sds
    // depends on this, but the bitvec crate does not guarantee this!
    // The undefined padding bytes have broken my code before, so this is
    // crucial.
    let original_len = bv.len();
    bv.resize(original_len.next_multiple_of(64), false);
    bv.resize(original_len, false);

    let raw_data = bytemuck::cast_slice(bv.as_raw_slice());
    let mut data_with_header = Cursor::new(header_bytes).chain(Cursor::new(raw_data));

    simple_sds_sbwt::raw_vector::RawVector::load(&mut data_with_header).unwrap()
}

pub(crate) fn segment_range(range: Range<usize>, n_pieces: usize) -> Vec<Range<usize>> {
    let segment_len = range.len().div_ceil(n_pieces);
    let mut pieces: Vec<Range<usize>> = vec![];
    for t in 0..n_pieces{
        let mut s = range.start + t*segment_len;
        let mut e = range.start + min((t+1)*segment_len, range.len());
        if s >= range.end { // Happens e.g. if range.len() == 1 and n_pieces == 10
            s = range.end;
            e = range.end;
        }
        pieces.push(s..e); // Final segments may be empty. Is ok.
    }
    pieces
}

#[cfg(test)]
pub(crate) fn gen_random_dna_string(len: usize, seed: u64) -> Vec<u8> {
    use rand_chacha::rand_core::{RngCore, SeedableRng};

    let mut rng = rand_chacha::ChaCha20Rng::seed_from_u64(seed);
    (0..len).map(|_| { 
        match rng.next_u64() % 4 {
            0 => b'A',
            1 => b'C',
            2 => b'G',
            3 => b'T',
            _ => panic!("Impossible")
        }
    }).collect()
}


// Calls the callback for each bit in the range. This seems to be
// about 6 times faster than using the iter_ones() function of BitVec
// in the bitvec::crate.
#[inline(always)] // This made a 25% difference in running time
pub fn for_each_one_bit<F>(words: &[u64], range: Range<usize>, mut cb: F)
where
    F: FnMut(usize),
{
    let s = range.start;
    let e = range.end;

    if words.is_empty() || s >= e {
        return;
    }

    let bit_len = words.len() * 64;
    if s >= bit_len {
        return;
    }

    let start_word = s / 64;
    let end_word = (e - 1) / 64; // inclusive

    // Helper: iterate 1-bits in a single word, adding `base_bit` to positions.
    #[inline]
    fn scan_word<F>(mut word: u64, base_bit: usize, cb: &mut F)
    where
        F: FnMut(usize),
    {
        while word != 0 {
            let tz = word.trailing_zeros() as usize;
            cb(base_bit + tz);
            word &= word - 1; // clear lowest set bit
        }
    }

    // Single-word range: mask both ends in the same word.
    if start_word == end_word {
        let start_bit = s % 64;
        let end_bit = e % 64; // exclusive

        let mut word = words[start_word];

        // Mask out bits before s
        word &= !0u64 << start_bit;

        // Mask out bits at/after e
        if end_bit != 0 {
            word &= (1u64 << end_bit) - 1;
        }

        scan_word(word, start_word * 64, &mut cb);
        return;
    }

    // --- First word: mask only the low bits ---
    {
        let start_bit = s % 64;
        let mut word = words[start_word];

        word &= !0u64 << start_bit;
        scan_word(word, start_word * 64, &mut cb);
    }

    // --- Middle words: no masking at all ---
    for (rel_idx, &word) in words[start_word + 1..end_word].iter().enumerate() {
        scan_word(word, (start_word + 1 + rel_idx) * 64, &mut cb);
    }

    // --- Last word: mask only the high bits ---
    {
        let end_bit = e % 64; // exclusive; 0 means "all 64 bits are in range"
        let mut word = words[end_word];

        if end_bit != 0 {
            word &= (1u64 << end_bit) - 1;
        }
        scan_word(word, end_word * 64, &mut cb);
    }
}

#[cfg(test)]
mod tests {

    use rand::{Rng, SeedableRng};

    use super::{for_each_one_bit, *};

    #[test]
    fn parallel_concatenation() {
        // Generate 1001 pseudorandom bit vectors with lengths between 0 and 256 bits
        let seed = 42;
        let mut rng = rand::rngs::StdRng::seed_from_u64(seed);
        let mut bitvecs: Vec<BitVec> = Vec::new();
        for i in 0..1000 {
            let len = rng.gen_range(0, 256);
            if i == 500 {
                // Add an empty bit vector
                bitvecs.push(BitVec::new());
            }
            let mut bits = BitVec::with_capacity(len);
            for _ in 0..len {
                bits.push(rng.gen_bool(0.5));
            }
            bitvecs.push(bits);
        }

        let true_concat = bitvecs.iter().fold(BitVec::new(), |mut acc, v| {
            acc.extend_from_bitslice(v);
            acc
        });

        let thread_pool = rayon::ThreadPoolBuilder::new().num_threads(3).build().unwrap();
        let our_concat = thread_pool.install(|| {
            parallel_bitvec_concat(bitvecs)
        });

        assert_eq!(true_concat.len(), our_concat.len());
        assert_eq!(true_concat, our_concat);
    }

    fn check_for_each_one_bit(words: &[u64], range: Range<usize>) {
        let bv: bitvec::vec::BitVec::<u64, Lsb0> = bitvec::vec::BitVec::from_vec(words.to_vec());
        let true_bits: Vec<usize> = bv.iter_ones().filter(|b| range.contains(b)).collect();

        let mut our_bits = Vec::<usize>::new();
        for_each_one_bit(words, range, |i| our_bits.push(i));

        assert_eq!(true_bits, our_bits);
    }

    #[test]
    fn for_each_one_bit_zero_word() {
        let words = [0_u64];
        check_for_each_one_bit(&words, 0..64);
    }

    #[test]
    fn for_each_one_bit_empty() {
        let words = [u64::MAX, u64::MAX, u64::MAX];

        for s in 0..=64*3 {
            check_for_each_one_bit(&words, s..s);
        }
        check_for_each_one_bit(&words, 64*3..1000); // Out of bounds
    }

    #[test]
    fn for_each_one_bit_one_word() {
        let mut rng = rand::rngs::StdRng::seed_from_u64(1234);
        let mut bits = bitvec![u64, Lsb0;];
        let range = 10..50;
        for i in 0..64 {
            if range.contains(&i) {
                bits.push(rng.gen_bool(0.5));
            } else {
                bits.push(true); // These bits should be masked out
            }
        }
        eprintln!("{}", bits);
        let words = bits.into_vec();
        check_for_each_one_bit(words.as_slice(), 0..64);
        check_for_each_one_bit(words.as_slice(), range);
        check_for_each_one_bit(words.as_slice(), 20..40);
    }

    #[test]
    fn for_each_one_bit_multiple_words() {
        let mut rng = rand::rngs::StdRng::seed_from_u64(4123);
        let mut bits = bitvec![u64, Lsb0;];
        let n_bits = 512;
        for s in (0..n_bits).step_by(32) {
            for e in (s..n_bits).step_by(128) {
                let range = s..e;
                for i in 0..n_bits {
                    if range.contains(&i) {
                        bits.push(rng.gen_bool(0.5));
                    } else {
                        bits.push(true); // These bits should be masked out
                    }
                }
                //eprintln!("{}", bits);
                let words = bits.clone().into_vec();
                check_for_each_one_bit(words.as_slice(), s..e);
            }
        }
    }

    #[test]
    fn for_each_one_bit_hand_crafted_edge_cases() {
        let words: &[u64] = &[0,0,0,u64::MAX,u64::MAX,0,u64::MAX,u64::MAX];
        check_for_each_one_bit(words, 0..words.len()*64);
        check_for_each_one_bit(words, 10..words.len()*64-10);
    }
}

pub fn for_each_run_with_key<T: Eq, KeyType: Eq, F1: Fn(&T) -> KeyType, F2: FnMut(Range<usize>)>(items: &[T], key_fn: F1, mut callback: F2) {
    if items.is_empty() { return }

    let mut run_start = 0;
    let n = items.len();
    for i in 1..n {
        if key_fn(&items[i]) != key_fn(&items[i-1]) {
            callback(run_start..i);
            run_start = i;
        }
    }
    // Final run
    callback(run_start..n);
}