zrip-core 0.12.2

Shared types and codecs for zrip (internal crate)
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
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
#![forbid(unsafe_code)]

#[cfg(feature = "alloc")]
use alloc::vec;
#[cfg(feature = "alloc")]
use alloc::vec::Vec;

use crate::bitstream::reader::BitReader;
use crate::bitstream::reader_reverse::ReverseBitReader;
use crate::error::DecompressError;
use crate::fse::decode::FseState;
use crate::fse::table_builder::build_decode_table;

pub fn parse_huffman_weights(data: &[u8]) -> Result<(Vec<u8>, usize), DecompressError> {
    if data.is_empty() {
        return Err(DecompressError::BadHuffmanWeights);
    }

    let header_byte = data[0];

    if header_byte >= 128 {
        parse_direct_weights(data)
    } else {
        parse_fse_compressed_weights(data)
    }
}

#[cfg(feature = "alloc")]
pub fn parse_huffman_weights_into(
    data: &[u8],
    weights: &mut Vec<u8>,
    fse_table: &mut Vec<crate::fse::FseDecodeEntry>,
    fse_symbol_next: &mut Vec<u16>,
    fse_dist: &mut Vec<i16>,
) -> Result<usize, DecompressError> {
    if data.is_empty() {
        return Err(DecompressError::BadHuffmanWeights);
    }

    let header_byte = data[0];

    if header_byte >= 128 {
        parse_direct_weights_into(data, weights)
    } else {
        parse_fse_compressed_weights_into(data, weights, fse_table, fse_symbol_next, fse_dist)
    }
}

fn parse_direct_weights(data: &[u8]) -> Result<(Vec<u8>, usize), DecompressError> {
    let num_symbols = (data[0] as usize) - 127;
    let num_bytes = num_symbols.div_ceil(2);

    if data.len() < 1 + num_bytes {
        return Err(DecompressError::BadHuffmanWeights);
    }

    let mut weights = Vec::with_capacity(num_symbols);
    for i in 0..num_symbols {
        let byte_idx = 1 + i / 2;
        let weight = if i % 2 == 0 {
            data[byte_idx] >> 4
        } else {
            data[byte_idx] & 0x0F
        };
        weights.push(weight);
    }

    Ok((weights, 1 + num_bytes))
}

#[cfg(feature = "alloc")]
fn parse_direct_weights_into(data: &[u8], weights: &mut Vec<u8>) -> Result<usize, DecompressError> {
    let num_symbols = (data[0] as usize) - 127;
    let num_bytes = num_symbols.div_ceil(2);

    if data.len() < 1 + num_bytes {
        return Err(DecompressError::BadHuffmanWeights);
    }

    weights.clear();
    for i in 0..num_symbols {
        let byte_idx = 1 + i / 2;
        let weight = if i % 2 == 0 {
            data[byte_idx] >> 4
        } else {
            data[byte_idx] & 0x0F
        };
        weights.push(weight);
    }

    Ok(1 + num_bytes)
}

fn parse_fse_compressed_weights(data: &[u8]) -> Result<(Vec<u8>, usize), DecompressError> {
    let compressed_size = data[0] as usize;
    if compressed_size == 0 || data.len() < 1 + compressed_size {
        return Err(DecompressError::BadHuffmanWeights);
    }

    let compressed = &data[1..=compressed_size];

    let mut bit_reader = BitReader::new(compressed);
    let (distribution, accuracy_log) =
        crate::fse::table_builder::parse_fse_table_description(&mut bit_reader, 12)?;
    if accuracy_log > 6 {
        return Err(DecompressError::BadHuffmanWeights);
    }

    let table = build_decode_table(&distribution, accuracy_log)
        .map_err(|_| DecompressError::BadHuffmanWeights)?;

    let table_desc_bytes = bit_reader.bytes_consumed();
    let fse_stream = &compressed[table_desc_bytes..];

    if fse_stream.is_empty() {
        return Err(DecompressError::BadHuffmanWeights);
    }

    let mut rev_reader =
        ReverseBitReader::new(fse_stream).map_err(|_| DecompressError::BadHuffmanWeights)?;

    let mut state1 = FseState::new(&table, accuracy_log, &mut rev_reader)
        .map_err(|_| DecompressError::BadHuffmanWeights)?;
    let mut state2 = FseState::new(&table, accuracy_log, &mut rev_reader)
        .map_err(|_| DecompressError::BadHuffmanWeights)?;

    let mut weights = Vec::new();

    loop {
        weights.push(state1.symbol());
        let nb1 = state1.num_bits();
        if nb1 > 0 && rev_reader.bits_remaining() < nb1 as usize {
            weights.push(state2.symbol());
            break;
        }
        state1
            .update_state(&mut rev_reader)
            .map_err(|_| DecompressError::BadHuffmanWeights)?;

        weights.push(state2.symbol());
        let nb2 = state2.num_bits();
        if nb2 > 0 && rev_reader.bits_remaining() < nb2 as usize {
            weights.push(state1.symbol());
            break;
        }
        state2
            .update_state(&mut rev_reader)
            .map_err(|_| DecompressError::BadHuffmanWeights)?;

        if weights.len() > 255 {
            return Err(DecompressError::BadHuffmanWeights);
        }
    }

    // The final step pushes up to two weights past the in-loop check. At
    // most 255 weights are explicit; the last one is implied.
    if weights.len() > 255 {
        return Err(DecompressError::BadHuffmanWeights);
    }

    Ok((weights, 1 + compressed_size))
}

#[cfg(feature = "alloc")]
fn parse_fse_compressed_weights_into(
    data: &[u8],
    weights: &mut Vec<u8>,
    fse_table: &mut Vec<crate::fse::FseDecodeEntry>,
    fse_symbol_next: &mut Vec<u16>,
    fse_dist: &mut Vec<i16>,
) -> Result<usize, DecompressError> {
    use crate::fse::table_builder::{build_decode_table_into, parse_fse_table_description_into};

    let compressed_size = data[0] as usize;
    if compressed_size == 0 || data.len() < 1 + compressed_size {
        return Err(DecompressError::BadHuffmanWeights);
    }

    let compressed = &data[1..=compressed_size];

    let mut bit_reader = BitReader::new(compressed);
    let accuracy_log = parse_fse_table_description_into(&mut bit_reader, 12, fse_dist)?;
    if accuracy_log > 6 {
        return Err(DecompressError::BadHuffmanWeights);
    }

    build_decode_table_into(fse_dist, accuracy_log, fse_table, fse_symbol_next)
        .map_err(|_| DecompressError::BadHuffmanWeights)?;

    let table_desc_bytes = bit_reader.bytes_consumed();
    let fse_stream = &compressed[table_desc_bytes..];

    if fse_stream.is_empty() {
        return Err(DecompressError::BadHuffmanWeights);
    }

    let mut rev_reader =
        ReverseBitReader::new(fse_stream).map_err(|_| DecompressError::BadHuffmanWeights)?;

    // The weight table has at most 64 entries (accuracy log <= 6). Padding
    // it to 64 lets `& 63` replace bounds checks on every lookup.
    fse_table.resize(
        64,
        crate::fse::FseDecodeEntry {
            base_line: 0,
            num_bits: 0,
            symbol: 0,
        },
    );
    let table: &[crate::fse::FseDecodeEntry; 64] = fse_table[..64].try_into().unwrap();
    let mut state1 = FseState::new(fse_table, accuracy_log, &mut rev_reader)
        .map_err(|_| DecompressError::BadHuffmanWeights)?
        .state() as usize;
    let mut state2 = FseState::new(fse_table, accuracy_log, &mut rev_reader)
        .map_err(|_| DecompressError::BadHuffmanWeights)?
        .state() as usize;

    // Up to 255 weights, plus up to three more from the step that detects
    // the overflow.
    let mut out = [0u8; 260];
    let mut n = 0usize;

    // Fast path: four steps read at most 24 bits (accuracy log <= 6). After a
    // refill, the container holds all remaining bits or at least 57, so with
    // 24 bits left none of the four steps can reach the end of the stream.
    while n <= 251 {
        rev_reader.refill_fast_or_regular();
        if rev_reader.bits_remaining() < 24 {
            break;
        }
        let e1 = table[state1 & 63];
        state1 = e1.base_line as usize + rev_reader.read_bits_branchless(e1.num_bits) as usize;
        let e2 = table[state2 & 63];
        state2 = e2.base_line as usize + rev_reader.read_bits_branchless(e2.num_bits) as usize;
        let e3 = table[state1 & 63];
        state1 = e3.base_line as usize + rev_reader.read_bits_branchless(e3.num_bits) as usize;
        let e4 = table[state2 & 63];
        state2 = e4.base_line as usize + rev_reader.read_bits_branchless(e4.num_bits) as usize;
        out[n..n + 4].copy_from_slice(&[e1.symbol, e2.symbol, e3.symbol, e4.symbol]);
        n += 4;
    }

    // Each step reads at most `accuracy_log` bits. After the remaining-bits
    // check, a refill leaves at least that many bits in the container, so the
    // unchecked read is exact.
    loop {
        let e1 = table[state1 & 63];
        out[n] = e1.symbol;
        n += 1;
        if e1.num_bits > 0 && rev_reader.bits_remaining() < e1.num_bits as usize {
            out[n] = table[state2 & 63].symbol;
            n += 1;
            break;
        }
        rev_reader.refill();
        state1 = e1.base_line as usize + rev_reader.read_bits_branchless(e1.num_bits) as usize;

        let e2 = table[state2 & 63];
        out[n] = e2.symbol;
        n += 1;
        if e2.num_bits > 0 && rev_reader.bits_remaining() < e2.num_bits as usize {
            out[n] = table[state1 & 63].symbol;
            n += 1;
            break;
        }
        rev_reader.refill();
        state2 = e2.base_line as usize + rev_reader.read_bits_branchless(e2.num_bits) as usize;

        if n > 255 {
            return Err(DecompressError::BadHuffmanWeights);
        }
    }

    // The final step pushes up to two weights past the in-loop check. At
    // most 255 weights are explicit; the last one is implied.
    if n > 255 {
        return Err(DecompressError::BadHuffmanWeights);
    }

    weights.clear();
    weights.extend_from_slice(&out[..n]);
    Ok(1 + compressed_size)
}

#[cfg(feature = "alloc")]
pub fn build_huffman_decode_table(
    weights: &[u8],
) -> Result<(Vec<crate::huffman::HuffmanDecodeEntry>, u8), DecompressError> {
    use crate::huffman::{HuffmanDecodeEntry, MAX_BITS};

    if weights.is_empty() || weights.len() > 255 {
        return Err(DecompressError::BadHuffmanWeights);
    }

    let mut weight_sum: u32 = 0;
    let mut max_weight = 0u8;
    for &w in weights.iter() {
        max_weight = max_weight.max(w);
        if w > 0 {
            weight_sum += 1u32 << (w - 1);
        }
    }

    if max_weight > MAX_BITS + 1 {
        return Err(DecompressError::BadHuffmanWeights);
    }

    if weight_sum == 0 {
        return Err(DecompressError::BadHuffmanWeights);
    }

    let table_log = high_bit(weight_sum) + 1;
    if table_log > MAX_BITS as u32 {
        return Err(DecompressError::BadHuffmanWeights);
    }

    let total_capacity = 1u32 << table_log;
    let last_weight_value = total_capacity - weight_sum;
    if last_weight_value == 0 || !last_weight_value.is_power_of_two() {
        return Err(DecompressError::BadHuffmanWeights);
    }
    let last_weight = high_bit(last_weight_value) as u8 + 1;

    let mut all_weights: Vec<u8> = weights.to_vec();
    all_weights.push(last_weight);

    let mut table = vec![HuffmanDecodeEntry::default(); super::DECODE_TABLE_SIZE];

    let max_w = table_log as u8 + 1;

    let mut rank_count = vec![0u32; max_w as usize + 1];
    for &w in all_weights.iter() {
        if w > 0 && w <= max_w {
            rank_count[w as usize] += 1;
        }
    }

    let mut rank_start = vec![0u32; max_w as usize + 1];
    {
        let mut cumul = 0u32;
        for w in 1..=max_w {
            rank_start[w as usize] = cumul;
            cumul += rank_count[w as usize] * (1u32 << (w - 1));
        }
    }

    for (symbol, &w) in all_weights.iter().enumerate() {
        if w == 0 {
            continue;
        }
        let num_bits = (table_log as u8 + 1) - w;
        let entries = 1usize << (w - 1);
        let start = rank_start[w as usize] as usize;
        rank_start[w as usize] += entries as u32;
        let entry = HuffmanDecodeEntry {
            symbol: symbol as u8,
            num_bits,
        };
        table[start..start + entries].fill(entry);
    }

    Ok((table, table_log as u8))
}

/// Builds the decode table for `weights` into `table` and returns the table
/// log.
#[cfg(feature = "alloc")]
pub fn build_huffman_decode_table_into(
    weights: &[u8],
    table: &mut Vec<crate::huffman::HuffmanDecodeEntry>,
    all_weights: &mut Vec<u8>,
) -> Result<u8, DecompressError> {
    use crate::huffman::{HuffmanDecodeEntry, MAX_BITS};

    if weights.is_empty() || weights.len() > 255 {
        return Err(DecompressError::BadHuffmanWeights);
    }

    let mut weight_sum: u32 = 0;
    let mut max_weight = 0u8;
    for &w in weights.iter() {
        max_weight = max_weight.max(w);
        if w > 0 {
            weight_sum += 1u32 << (w - 1);
        }
    }

    if max_weight > MAX_BITS + 1 {
        return Err(DecompressError::BadHuffmanWeights);
    }

    if weight_sum == 0 {
        return Err(DecompressError::BadHuffmanWeights);
    }

    let table_log = high_bit(weight_sum) + 1;
    if table_log > MAX_BITS as u32 {
        return Err(DecompressError::BadHuffmanWeights);
    }

    let total_capacity = 1u32 << table_log;
    let last_weight_value = total_capacity - weight_sum;
    if last_weight_value == 0 || !last_weight_value.is_power_of_two() {
        return Err(DecompressError::BadHuffmanWeights);
    }
    let last_weight = high_bit(last_weight_value) as u8 + 1;

    all_weights.clear();
    all_weights.extend_from_slice(weights);
    all_weights.push(last_weight);

    if table.len() != super::DECODE_TABLE_SIZE {
        table.clear();
        table.resize(super::DECODE_TABLE_SIZE, HuffmanDecodeEntry::default());
    }

    let max_w = table_log as usize + 1;

    // Every weight is at most `table_log`, because `weight_sum` includes its
    // `1 << (w - 1)`. So `w & 15` is `w` and only drops the bounds check.
    // Zero weights go to bucket 0 instead of being skipped: zero and nonzero
    // weights interleave unpredictably, so a branch per symbol mispredicts.
    //
    // The histogram and the counting sort below run on four contiguous
    // partitions of the symbols, each with its own counters. With one shared
    // counter array, consecutive symbols of the same weight waited on each
    // other's counter store. Contiguous partitions keep symbols of equal
    // weight in increasing order, as the canonical code needs. Padding
    // symbols get weight 0 and land in bucket 0, which the fill skips.
    let q = all_weights.len().div_ceil(4);
    all_weights.resize(4 * q, 0);
    let aw = &all_weights[..];
    let mut hist = [[0u32; 16]; 4];
    for i in 0..q {
        for (k, h) in hist.iter_mut().enumerate() {
            h[(aw[k * q + i] & 15) as usize] += 1;
        }
    }
    let mut rank_count = [0u32; 16];
    for (w, count) in rank_count.iter_mut().enumerate() {
        *count = hist[0][w] + hist[1][w] + hist[2][w] + hist[3][w];
    }

    // Group symbols by weight (counting sort), weight 0 first, and within a
    // weight by partition. Each nonzero weight then fills one contiguous range
    // with a fixed number of entries per symbol, which compiles to fixed-width
    // stores instead of a variable-length fill per symbol. There are at most
    // 256 symbols, so `& 255` only drops the bounds check.
    let mut next = [[0u32; 16]; 4];
    let mut acc = 0u32;
    for w in 0..16 {
        for k in 0..4 {
            next[k][w] = acc;
            acc += hist[k][w];
        }
    }
    let mut sorted = [0u8; 256];
    for i in 0..q {
        for (k, slots) in next.iter_mut().enumerate() {
            let symbol = k * q + i;
            let slot = &mut slots[(aw[symbol] & 15) as usize];
            sorted[*slot as usize & 255] = symbol as u8;
            *slot += 1;
        }
    }
    let mut first = rank_count[0] as usize;
    let mut start = 0usize;
    for (w, &count) in rank_count.iter().enumerate().take(max_w + 1).skip(1) {
        let count = count as usize;
        let symbols = &sorted[first..first + count];
        first += count;
        let num_bits = (max_w - w) as u8;
        match w {
            1 => fill_runs::<1>(table, start, symbols, num_bits),
            2 => fill_runs::<2>(table, start, symbols, num_bits),
            3 => fill_runs::<4>(table, start, symbols, num_bits),
            4 => fill_runs::<8>(table, start, symbols, num_bits),
            5 => fill_runs::<16>(table, start, symbols, num_bits),
            _ => {
                let entries = 1usize << (w - 1);
                for (i, &symbol) in symbols.iter().enumerate() {
                    let pos = start + i * entries;
                    table[pos..pos + entries].fill(HuffmanDecodeEntry { symbol, num_bits });
                }
            }
        }
        start += count << (w - 1);
    }

    Ok(table_log as u8)
}

/// Writes `L` copies of each symbol's entry, one symbol after another,
/// starting at `start`.
#[inline(always)]
fn fill_runs<const L: usize>(
    table: &mut [crate::huffman::HuffmanDecodeEntry],
    start: usize,
    symbols: &[u8],
    num_bits: u8,
) {
    let run = &mut table[start..start + symbols.len() * L];
    for (chunk, &symbol) in run.as_chunks_mut::<L>().0.iter_mut().zip(symbols) {
        *chunk = [crate::huffman::HuffmanDecodeEntry { symbol, num_bits }; L];
    }
}

fn high_bit(val: u32) -> u32 {
    debug_assert!(val > 0);
    31 - val.leading_zeros()
}

#[cfg(test)]
mod tests {
    use super::*;

    #[test]
    fn parse_direct_2_symbols() {
        // header = 127 + 2 = 129
        let data = [129, 0x42];
        let (weights, consumed) = parse_huffman_weights(&data).unwrap();
        assert_eq!(weights.len(), 2);
        assert_eq!(weights[0], 4);
        assert_eq!(weights[1], 2);
        assert_eq!(consumed, 2);
    }

    #[test]
    fn parse_direct_3_symbols() {
        // header = 127 + 3 = 130
        let data = [130, 0x31, 0x20];
        let (weights, consumed) = parse_huffman_weights(&data).unwrap();
        assert_eq!(weights.len(), 3);
        assert_eq!(weights[0], 3);
        assert_eq!(weights[1], 1);
        assert_eq!(weights[2], 2);
        assert_eq!(consumed, 3);
    }

    #[test]
    fn build_table_two_symbols() {
        // weight 2 means 2^1=2 slots. Sum=2. Implied symbol needs 2 slots too.
        // Total = 4 = 2^2, so table_log = 2.
        let weights = vec![2];
        let (table, table_log) = build_huffman_decode_table(&weights).unwrap();
        assert_eq!(table_log, 2);
        assert_eq!(table.len(), crate::huffman::DECODE_TABLE_SIZE);
        let sym0_count = table[..4].iter().filter(|e| e.symbol == 0).count();
        let sym1_count = table[..4].iter().filter(|e| e.symbol == 1).count();
        assert_eq!(sym0_count, 2);
        assert_eq!(sym1_count, 2);
    }

    /// FSE-compressed weight header whose stream decodes to `count` weights.
    /// With two symbols at 16/32, every state reads one bit: two 5-bit
    /// initial states, then one bit per weight except the last two.
    fn fse_weights_header(count: usize) -> Vec<u8> {
        use crate::bitstream::writer::BitWriter;
        use crate::fse::table_builder::serialize_fse_table_description;

        let mut compressed = serialize_fse_table_description(&[16, 16], 5);
        let mut stream = BitWriter::new();
        for _ in 0..10 + count - 2 {
            stream.write_bits(0, 1);
        }
        stream.close_reverse_stream();
        compressed.extend_from_slice(&stream.into_bytes());
        let mut data = vec![compressed.len() as u8];
        data.extend_from_slice(&compressed);
        data
    }

    #[test]
    fn parse_fse_weights_limit() {
        let data = fse_weights_header(255);
        let (weights, _) = parse_huffman_weights(&data).unwrap();
        assert_eq!(weights.len(), 255);
        let mut out = Vec::new();
        parse_huffman_weights_into(
            &data,
            &mut out,
            &mut Vec::new(),
            &mut Vec::new(),
            &mut Vec::new(),
        )
        .unwrap();
        assert_eq!(out.len(), 255);

        for count in [256, 257] {
            let data = fse_weights_header(count);
            assert!(parse_huffman_weights(&data).is_err(), "{count} weights");
            let result = parse_huffman_weights_into(
                &data,
                &mut out,
                &mut Vec::new(),
                &mut Vec::new(),
                &mut Vec::new(),
            );
            assert_eq!(
                result,
                Err(DecompressError::BadHuffmanWeights),
                "{count} weights"
            );
        }
    }

    #[test]
    fn parse_fse_weights_into_matches_reference() {
        use crate::fse::table_builder::{normalize_counts, serialize_fse_table_description};

        let mut seed = 0x9e37_79b9_7f4a_7c15u64;
        let mut rand = move || {
            seed ^= seed << 13;
            seed ^= seed >> 7;
            seed ^= seed << 17;
            seed
        };
        let (mut out, mut table, mut next, mut dist) =
            (Vec::new(), Vec::new(), Vec::new(), Vec::new());
        let (mut ok, mut too_many) = (0, 0);
        for _ in 0..20_000 {
            let accuracy_log = 5 + (rand() % 2) as u8;
            let symbols = 2 + (rand() % 12) as usize;
            // A dominant symbol gives zero- and one-bit states, so some
            // streams decode more than 255 weights.
            let freqs: Vec<u32> = (0..symbols)
                .map(|_| {
                    if rand() % 4 == 0 {
                        1000
                    } else {
                        (rand() % 50) as u32 + 1
                    }
                })
                .collect();
            let mut compressed = serialize_fse_table_description(
                &normalize_counts(&freqs, accuracy_log),
                accuracy_log,
            );
            let stream_len = 1 + (rand() % 100) as usize;
            compressed.extend((0..stream_len).map(|_| rand() as u8));
            *compressed.last_mut().unwrap() |= 1 << (rand() % 8);
            if compressed.len() > 127 {
                continue;
            }
            let mut data = vec![compressed.len() as u8];
            data.extend_from_slice(&compressed);

            let reference = parse_huffman_weights(&data);
            let result =
                parse_huffman_weights_into(&data, &mut out, &mut table, &mut next, &mut dist);
            match (reference, result) {
                (Ok((weights, consumed)), Ok(result_consumed)) => {
                    assert_eq!(out, weights);
                    assert_eq!(result_consumed, consumed);
                    ok += 1;
                }
                (Err(e), Err(result_e)) => {
                    assert_eq!(result_e, e);
                    too_many += usize::from(e == DecompressError::BadHuffmanWeights);
                }
                (reference, result) => panic!("{reference:?} vs {result:?} for {data:?}"),
            }
        }
        assert!(ok > 1000 && too_many > 100, "ok {ok}, too many {too_many}");
    }

    #[test]
    fn build_table_into_matches_reference() {
        let mut seed = 0x7a3d_5c19_e0b4_8f21u64;
        let mut rand = move || {
            seed ^= seed << 13;
            seed ^= seed >> 7;
            seed ^= seed << 17;
            seed
        };
        let (mut table, mut all) = (Vec::new(), Vec::new());
        for _ in 0..3000 {
            // A random complete prefix code with 2 to 256 leaves, at most 11
            // bits deep, over random symbols. The highest symbol's weight is
            // implied.
            let leaves = 2 + (rand() as usize) % 255;
            let mut depths = vec![1u8, 1];
            while depths.len() < leaves {
                let i = (rand() as usize) % depths.len();
                if depths[i] < 11 {
                    depths[i] += 1;
                    depths.push(depths[i]);
                }
            }
            let table_log = *depths.iter().max().unwrap();
            let last_symbol = leaves - 1 + (rand() as usize) % (256 - leaves + 1);
            let mut symbols: Vec<usize> = (0..last_symbol).collect();
            for i in (1..symbols.len()).rev() {
                symbols.swap(i, (rand() as usize) % (i + 1));
            }
            let mut weights = vec![0u8; last_symbol];
            for (&symbol, &depth) in symbols[..leaves - 1].iter().zip(&depths) {
                weights[symbol] = table_log + 1 - depth;
            }

            let (expected, expected_log) = build_huffman_decode_table(&weights).unwrap();
            let log = build_huffman_decode_table_into(&weights, &mut table, &mut all).unwrap();
            assert_eq!(log, expected_log);
            let n = 1 << log;
            let fields = |t: &[crate::huffman::HuffmanDecodeEntry]| -> Vec<(u8, u8)> {
                t[..n].iter().map(|e| (e.symbol, e.num_bits)).collect()
            };
            assert_eq!(fields(&table), fields(&expected), "{weights:?}");
        }
    }

    #[test]
    fn build_table_rejects_more_than_255_weights() {
        // 256 weight-1 symbols and one weight-8 symbol sum to 384. The
        // implied last symbol (weight 8) completes a 512-entry table, but
        // the alphabet would have 258 symbols.
        let mut weights = vec![1u8; 256];
        weights.push(8);
        let mut table = Vec::new();
        let result = build_huffman_decode_table_into(&weights, &mut table, &mut Vec::new());
        assert_eq!(result, Err(DecompressError::BadHuffmanWeights));
    }
}