hermes-core 1.8.34

Core async search engine library with WASM support
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
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
//! Memory-efficient SSTable index structures
//!
//! This module provides two approaches for memory-efficient block indexing:
//!
//! ## Option 1: FST-based Index (native feature)
//! Uses a Finite State Transducer to map keys to block ordinals. The FST can be
//! mmap'd directly without parsing into heap-allocated structures.
//!
//! ## Option 2: Mmap'd Raw Index
//! Keeps the prefix-compressed block index as raw bytes and decodes entries
//! on-demand during binary search. No heap allocation for the index.
//!
//! Both approaches use a compact BlockAddrStore with bitpacked offsets/lengths.

use byteorder::{LittleEndian, ReadBytesExt, WriteBytesExt};
use std::io::{self, Write};
use std::ops::Range;

use crate::directories::OwnedBytes;

/// Block address - offset and length in the data section
#[derive(Debug, Clone, Copy, PartialEq, Eq)]
pub struct BlockAddr {
    pub offset: u64,
    pub length: u32,
}

impl BlockAddr {
    pub fn byte_range(&self) -> Range<u64> {
        self.offset..self.offset + self.length as u64
    }
}

/// Compact storage for block addresses using delta + bitpacking
///
/// Memory layout:
/// - Header: num_blocks (u32) + offset_bits (u8) + length_bits (u8)
/// - Bitpacked data: offsets and lengths interleaved
///
/// Uses delta encoding for offsets (blocks are sequential) and
/// stores lengths directly (typically similar sizes).
#[derive(Debug)]
pub struct BlockAddrStore {
    num_blocks: u32,
    offset_bits: u8,
    length_bits: u8,
    /// Eagerly decoded addresses for O(1) random access
    addrs: Vec<BlockAddr>,
}

impl BlockAddrStore {
    /// Build from a list of block addresses
    pub fn build(addrs: &[BlockAddr]) -> io::Result<Vec<u8>> {
        if addrs.is_empty() {
            let mut buf = Vec::with_capacity(6);
            buf.write_u32::<LittleEndian>(0)?;
            buf.write_u8(0)?;
            buf.write_u8(0)?;
            return Ok(buf);
        }

        // Compute delta offsets and find max values for bit width
        let mut deltas = Vec::with_capacity(addrs.len());
        let mut prev_end: u64 = 0;
        let mut max_delta: u64 = 0;
        let mut max_length: u32 = 0;

        for addr in addrs {
            // Delta from end of previous block (handles gaps)
            let delta = addr.offset.saturating_sub(prev_end);
            deltas.push(delta);
            max_delta = max_delta.max(delta);
            max_length = max_length.max(addr.length);
            prev_end = addr.offset + addr.length as u64;
        }

        // Compute bit widths
        let offset_bits = if max_delta == 0 {
            1
        } else {
            (64 - max_delta.leading_zeros()) as u8
        };
        let length_bits = if max_length == 0 {
            1
        } else {
            (32 - max_length.leading_zeros()) as u8
        };

        // Calculate packed size
        let bits_per_entry = offset_bits as usize + length_bits as usize;
        let total_bits = bits_per_entry * addrs.len();
        let packed_bytes = total_bits.div_ceil(8);

        let mut buf = Vec::with_capacity(6 + packed_bytes);
        buf.write_u32::<LittleEndian>(addrs.len() as u32)?;
        buf.write_u8(offset_bits)?;
        buf.write_u8(length_bits)?;

        // Bitpack the data
        let mut bit_writer = BitWriter::new(&mut buf);
        for (i, addr) in addrs.iter().enumerate() {
            bit_writer.write(deltas[i], offset_bits)?;
            bit_writer.write(addr.length as u64, length_bits)?;
        }
        bit_writer.flush()?;

        Ok(buf)
    }

    /// Load from raw bytes — eagerly decodes all addresses for O(1) access
    pub fn load(data: OwnedBytes) -> io::Result<Self> {
        if data.len() < 6 {
            return Err(io::Error::new(
                io::ErrorKind::InvalidData,
                "BlockAddrStore data too short",
            ));
        }

        let mut reader = data.as_slice();
        let num_blocks = reader.read_u32::<LittleEndian>()?;
        let offset_bits = reader.read_u8()?;
        let length_bits = reader.read_u8()?;

        // Eagerly decode all block addresses once at load time
        let packed_data = &data.as_slice()[6..];
        let mut bit_reader = BitReader::new(packed_data);
        let mut addrs = Vec::with_capacity(num_blocks as usize);
        let mut current_offset: u64 = 0;

        for _ in 0..num_blocks {
            if let (Ok(delta), Ok(length)) =
                (bit_reader.read(offset_bits), bit_reader.read(length_bits))
            {
                current_offset += delta;
                addrs.push(BlockAddr {
                    offset: current_offset,
                    length: length as u32,
                });
                current_offset += length;
            }
        }

        Ok(Self {
            num_blocks,
            offset_bits,
            length_bits,
            addrs,
        })
    }

    /// Number of blocks
    pub fn len(&self) -> usize {
        self.num_blocks as usize
    }

    /// Check if empty
    pub fn is_empty(&self) -> bool {
        self.num_blocks == 0
    }

    /// Get block address by index — O(1) from eagerly decoded array
    #[inline]
    pub fn get(&self, idx: usize) -> Option<BlockAddr> {
        self.addrs.get(idx).copied()
    }

    /// Get all block addresses
    pub fn all(&self) -> Vec<BlockAddr> {
        self.addrs.clone()
    }
}

/// FST-based block index (Option 1)
///
/// Maps keys to block ordinals using an FST. The FST bytes can be mmap'd
/// directly without any parsing or heap allocation.
#[cfg(feature = "fst-index")]
pub struct FstBlockIndex {
    fst: fst::Map<OwnedBytes>,
    block_addrs: BlockAddrStore,
}

#[cfg(feature = "fst-index")]
impl FstBlockIndex {
    /// Build FST index from keys and block addresses
    pub fn build(entries: &[(Vec<u8>, BlockAddr)]) -> io::Result<Vec<u8>> {
        use fst::MapBuilder;

        // Build FST mapping keys to block ordinals
        let mut fst_builder = MapBuilder::memory();
        for (i, (key, _)) in entries.iter().enumerate() {
            fst_builder
                .insert(key, i as u64)
                .map_err(|e| io::Error::new(io::ErrorKind::InvalidData, e))?;
        }
        let fst_bytes = fst_builder
            .into_inner()
            .map_err(|e| io::Error::new(io::ErrorKind::InvalidData, e))?;

        // Build block address store
        let addrs: Vec<BlockAddr> = entries.iter().map(|(_, addr)| *addr).collect();
        let addr_bytes = BlockAddrStore::build(&addrs)?;

        // Combine: fst_len (u32) + fst_bytes + addr_bytes
        let mut result = Vec::with_capacity(4 + fst_bytes.len() + addr_bytes.len());
        result.write_u32::<LittleEndian>(fst_bytes.len() as u32)?;
        result.extend_from_slice(&fst_bytes);
        result.extend_from_slice(&addr_bytes);

        Ok(result)
    }

    /// Load from raw bytes
    pub fn load(data: OwnedBytes) -> io::Result<Self> {
        if data.len() < 4 {
            return Err(io::Error::new(
                io::ErrorKind::InvalidData,
                "FstBlockIndex data too short",
            ));
        }

        let fst_len = u32::from_le_bytes([data[0], data[1], data[2], data[3]]) as usize;

        if data.len() < 4 + fst_len {
            return Err(io::Error::new(
                io::ErrorKind::InvalidData,
                "FstBlockIndex FST data truncated",
            ));
        }

        let fst_data = data.slice(4..4 + fst_len);
        let addr_data = data.slice(4 + fst_len..data.len());

        let fst =
            fst::Map::new(fst_data).map_err(|e| io::Error::new(io::ErrorKind::InvalidData, e))?;
        let block_addrs = BlockAddrStore::load(addr_data)?;

        Ok(Self { fst, block_addrs })
    }

    /// Look up the block index for a key
    /// Returns the block ordinal that could contain this key.
    /// O(key_len) via FST exact lookup + single stream step.
    pub fn locate(&self, key: &[u8]) -> Option<usize> {
        // Fast exact match — O(key_len), no stream allocation
        if let Some(ordinal) = self.fst.get(key) {
            return Some(ordinal as usize);
        }

        // Find the first block whose first_key > target (single stream step)
        use fst::{IntoStreamer, Streamer};
        let mut stream = self.fst.range().gt(key).into_stream();
        match stream.next() {
            Some((_, ordinal)) if ordinal > 0 => Some(ordinal as usize - 1),
            Some(_) => None, // key < first block's first key
            None => {
                // No key > target → target is after all keys; use last block
                let len = self.fst.len();
                if len > 0 { Some(len - 1) } else { None }
            }
        }
    }

    /// Get block address by ordinal
    pub fn get_addr(&self, ordinal: usize) -> Option<BlockAddr> {
        self.block_addrs.get(ordinal)
    }

    /// Number of blocks
    pub fn len(&self) -> usize {
        self.block_addrs.len()
    }

    /// Check if empty
    pub fn is_empty(&self) -> bool {
        self.block_addrs.is_empty()
    }

    /// Get all block addresses
    pub fn all_addrs(&self) -> Vec<BlockAddr> {
        self.block_addrs.all()
    }
}

/// Mmap'd raw block index (Option 2)
///
/// Keeps the prefix-compressed block index as raw bytes and decodes
/// entries on-demand. Uses restart points every R entries for O(log N)
/// lookup via binary search instead of O(N) linear scan.
pub struct MmapBlockIndex {
    data: OwnedBytes,
    num_blocks: u32,
    block_addrs: BlockAddrStore,
    /// Offset where the prefix-compressed keys start
    keys_offset: usize,
    /// Offset where the keys section ends (restart array begins)
    keys_end: usize,
    /// Byte offset in data where the restart offsets array starts
    restart_array_offset: usize,
    /// Number of restart points
    restart_count: usize,
    /// Restart interval (R) — a restart point every R entries
    restart_interval: usize,
}

/// Restart interval: store full (uncompressed) key every R entries
const RESTART_INTERVAL: usize = 16;

impl MmapBlockIndex {
    /// Build mmap-friendly index from entries.
    ///
    /// Format: `num_blocks (u32) | BlockAddrStore | prefix-compressed keys
    /// (with restart points) | restart_offsets[..] | restart_count (u32) | restart_interval (u16)`
    pub fn build(entries: &[(Vec<u8>, BlockAddr)]) -> io::Result<Vec<u8>> {
        if entries.is_empty() {
            let mut buf = Vec::with_capacity(16);
            buf.write_u32::<LittleEndian>(0)?; // num_blocks
            buf.extend_from_slice(&BlockAddrStore::build(&[])?);
            // Empty restart array + footer
            buf.write_u32::<LittleEndian>(0)?; // restart_count
            buf.write_u16::<LittleEndian>(RESTART_INTERVAL as u16)?;
            return Ok(buf);
        }

        // Build block address store
        let addrs: Vec<BlockAddr> = entries.iter().map(|(_, addr)| *addr).collect();
        let addr_bytes = BlockAddrStore::build(&addrs)?;

        // Build prefix-compressed keys with restart points
        let mut keys_buf = Vec::new();
        let mut prev_key: Vec<u8> = Vec::new();
        let mut restart_offsets: Vec<u32> = Vec::new();

        for (i, (key, _)) in entries.iter().enumerate() {
            let is_restart = i % RESTART_INTERVAL == 0;

            if is_restart {
                restart_offsets.push(keys_buf.len() as u32);
                // Store full key (no prefix compression)
                write_vint(&mut keys_buf, 0)?;
                write_vint(&mut keys_buf, key.len() as u64)?;
                keys_buf.extend_from_slice(key);
            } else {
                let prefix_len = common_prefix_len(&prev_key, key);
                let suffix = &key[prefix_len..];
                write_vint(&mut keys_buf, prefix_len as u64)?;
                write_vint(&mut keys_buf, suffix.len() as u64)?;
                keys_buf.extend_from_slice(suffix);
            }

            prev_key.clear();
            prev_key.extend_from_slice(key);
        }

        // Combine: num_blocks + addr_bytes + keys + restart_offsets + footer
        let restart_count = restart_offsets.len();
        let mut result =
            Vec::with_capacity(4 + addr_bytes.len() + keys_buf.len() + restart_count * 4 + 6);
        result.write_u32::<LittleEndian>(entries.len() as u32)?;
        result.extend_from_slice(&addr_bytes);
        result.extend_from_slice(&keys_buf);

        // Write restart offsets array
        for &off in &restart_offsets {
            result.write_u32::<LittleEndian>(off)?;
        }

        // Write footer
        result.write_u32::<LittleEndian>(restart_count as u32)?;
        result.write_u16::<LittleEndian>(RESTART_INTERVAL as u16)?;

        Ok(result)
    }

    /// Load from raw bytes
    pub fn load(data: OwnedBytes) -> io::Result<Self> {
        if data.len() < 4 {
            return Err(io::Error::new(
                io::ErrorKind::InvalidData,
                "MmapBlockIndex data too short",
            ));
        }

        let num_blocks = u32::from_le_bytes([data[0], data[1], data[2], data[3]]);

        // Load block addresses
        let addr_data_start = 4;
        let remaining = data.slice(addr_data_start..data.len());
        let block_addrs = BlockAddrStore::load(remaining.clone())?;

        // Calculate where keys start
        let bits_per_entry = block_addrs.offset_bits as usize + block_addrs.length_bits as usize;
        let total_bits = bits_per_entry * num_blocks as usize;
        let addr_packed_size = total_bits.div_ceil(8);
        let keys_offset = addr_data_start + 6 + addr_packed_size; // 6 = header of BlockAddrStore

        // Read footer (last 6 bytes: restart_count u32 + restart_interval u16)
        if data.len() < keys_offset + 6 {
            return Err(io::Error::new(
                io::ErrorKind::InvalidData,
                "MmapBlockIndex missing restart footer",
            ));
        }
        let footer_start = data.len() - 6;
        let restart_count = u32::from_le_bytes([
            data[footer_start],
            data[footer_start + 1],
            data[footer_start + 2],
            data[footer_start + 3],
        ]) as usize;
        let restart_interval =
            u16::from_le_bytes([data[footer_start + 4], data[footer_start + 5]]) as usize;

        // Restart offsets array: restart_count × 4 bytes, just before footer
        let restart_array_offset = footer_start - restart_count * 4;

        // Keys section spans from keys_offset to restart_array_offset
        let keys_end = restart_array_offset;

        Ok(Self {
            data,
            num_blocks,
            block_addrs,
            keys_offset,
            keys_end,
            restart_array_offset,
            restart_count,
            restart_interval,
        })
    }

    /// Read restart offset at given index directly from mmap'd data
    #[inline]
    fn restart_offset(&self, idx: usize) -> u32 {
        let pos = self.restart_array_offset + idx * 4;
        u32::from_le_bytes([
            self.data[pos],
            self.data[pos + 1],
            self.data[pos + 2],
            self.data[pos + 3],
        ])
    }

    /// Decode the full key at a restart point (prefix_len is always 0)
    fn decode_restart_key<'a>(&self, keys_data: &'a [u8], restart_idx: usize) -> &'a [u8] {
        let offset = self.restart_offset(restart_idx) as usize;
        let mut reader = &keys_data[offset..];

        let prefix_len = read_vint(&mut reader).unwrap_or(0) as usize;
        debug_assert_eq!(prefix_len, 0, "restart point should have prefix_len=0");
        let suffix_len = read_vint(&mut reader).unwrap_or(0) as usize;

        // reader now points to the suffix bytes
        &reader[..suffix_len]
    }

    /// O(log(N/R) + R) lookup using binary search on restart points, then
    /// linear scan with prefix decompression within the interval.
    pub fn locate(&self, target: &[u8]) -> Option<usize> {
        if self.num_blocks == 0 {
            return None;
        }

        let keys_data = &self.data.as_slice()[self.keys_offset..self.keys_end];

        // Binary search on restart points to find the interval
        let mut lo = 0usize;
        let mut hi = self.restart_count;

        while lo < hi {
            let mid = lo + (hi - lo) / 2;
            let key = self.decode_restart_key(keys_data, mid);
            match key.cmp(target) {
                std::cmp::Ordering::Equal => {
                    return Some(mid * self.restart_interval);
                }
                std::cmp::Ordering::Less => lo = mid + 1,
                std::cmp::Ordering::Greater => hi = mid,
            }
        }

        // lo is the first restart point whose key > target (or restart_count)
        // Search in the interval starting at restart (lo - 1), or 0 if lo == 0
        if lo == 0 {
            // target < first restart key — might be before all keys
            // but we still need to scan from the beginning
        }

        let restart_idx = if lo > 0 { lo - 1 } else { 0 };
        let start_ordinal = restart_idx * self.restart_interval;
        let end_ordinal = if restart_idx + 1 < self.restart_count {
            (restart_idx + 1) * self.restart_interval
        } else {
            self.num_blocks as usize
        };

        // Linear scan from restart point through at most R entries
        let scan_offset = self.restart_offset(restart_idx) as usize;
        let mut reader = &keys_data[scan_offset..];
        let mut current_key = Vec::new();
        let mut last_le_block: Option<usize> = None;

        for i in start_ordinal..end_ordinal {
            let prefix_len = match read_vint(&mut reader) {
                Ok(v) => v as usize,
                Err(_) => break,
            };
            let suffix_len = match read_vint(&mut reader) {
                Ok(v) => v as usize,
                Err(_) => break,
            };

            current_key.truncate(prefix_len);
            if suffix_len > reader.len() {
                break;
            }
            current_key.extend_from_slice(&reader[..suffix_len]);
            reader = &reader[suffix_len..];

            match current_key.as_slice().cmp(target) {
                std::cmp::Ordering::Equal => return Some(i),
                std::cmp::Ordering::Less => last_le_block = Some(i),
                std::cmp::Ordering::Greater => return last_le_block,
            }
        }

        last_le_block
    }

    /// Get block address by ordinal
    pub fn get_addr(&self, ordinal: usize) -> Option<BlockAddr> {
        self.block_addrs.get(ordinal)
    }

    /// Number of blocks
    pub fn len(&self) -> usize {
        self.num_blocks as usize
    }

    /// Check if empty
    pub fn is_empty(&self) -> bool {
        self.num_blocks == 0
    }

    /// Get all block addresses
    pub fn all_addrs(&self) -> Vec<BlockAddr> {
        self.block_addrs.all()
    }

    /// Decode all keys (for debugging/merging)
    pub fn all_keys(&self) -> Vec<Vec<u8>> {
        let mut result = Vec::with_capacity(self.num_blocks as usize);
        let keys_data = &self.data.as_slice()[self.keys_offset..self.keys_end];
        let mut reader = keys_data;
        let mut current_key = Vec::new();

        for _ in 0..self.num_blocks {
            let prefix_len = match read_vint(&mut reader) {
                Ok(v) => v as usize,
                Err(_) => break,
            };
            let suffix_len = match read_vint(&mut reader) {
                Ok(v) => v as usize,
                Err(_) => break,
            };

            current_key.truncate(prefix_len);
            if suffix_len > reader.len() {
                break;
            }
            current_key.extend_from_slice(&reader[..suffix_len]);
            reader = &reader[suffix_len..];

            result.push(current_key.clone());
        }

        result
    }
}

/// Unified block index that can use either FST or mmap'd raw index
pub enum BlockIndex {
    #[cfg(feature = "fst-index")]
    Fst(FstBlockIndex),
    Mmap(MmapBlockIndex),
}

impl BlockIndex {
    /// Locate the block that could contain the key
    pub fn locate(&self, key: &[u8]) -> Option<usize> {
        match self {
            #[cfg(feature = "fst-index")]
            BlockIndex::Fst(idx) => idx.locate(key),
            BlockIndex::Mmap(idx) => idx.locate(key),
        }
    }

    /// Get block address by ordinal
    pub fn get_addr(&self, ordinal: usize) -> Option<BlockAddr> {
        match self {
            #[cfg(feature = "fst-index")]
            BlockIndex::Fst(idx) => idx.get_addr(ordinal),
            BlockIndex::Mmap(idx) => idx.get_addr(ordinal),
        }
    }

    /// Number of blocks
    pub fn len(&self) -> usize {
        match self {
            #[cfg(feature = "fst-index")]
            BlockIndex::Fst(idx) => idx.len(),
            BlockIndex::Mmap(idx) => idx.len(),
        }
    }

    /// Check if empty
    pub fn is_empty(&self) -> bool {
        self.len() == 0
    }

    /// Get all block addresses
    pub fn all_addrs(&self) -> Vec<BlockAddr> {
        match self {
            #[cfg(feature = "fst-index")]
            BlockIndex::Fst(idx) => idx.all_addrs(),
            BlockIndex::Mmap(idx) => idx.all_addrs(),
        }
    }
}

// ============================================================================
// Helper functions
// ============================================================================

fn common_prefix_len(a: &[u8], b: &[u8]) -> usize {
    a.iter().zip(b.iter()).take_while(|(x, y)| x == y).count()
}

fn write_vint<W: Write>(writer: &mut W, mut value: u64) -> io::Result<()> {
    loop {
        let byte = (value & 0x7F) as u8;
        value >>= 7;
        if value == 0 {
            writer.write_all(&[byte])?;
            return Ok(());
        } else {
            writer.write_all(&[byte | 0x80])?;
        }
    }
}

fn read_vint(reader: &mut &[u8]) -> io::Result<u64> {
    let mut result = 0u64;
    let mut shift = 0;

    loop {
        if reader.is_empty() {
            return Err(io::Error::new(
                io::ErrorKind::UnexpectedEof,
                "Unexpected end of varint",
            ));
        }
        let byte = reader[0];
        *reader = &reader[1..];
        result |= ((byte & 0x7F) as u64) << shift;
        if byte & 0x80 == 0 {
            return Ok(result);
        }
        shift += 7;
        if shift >= 64 {
            return Err(io::Error::new(
                io::ErrorKind::InvalidData,
                "Varint too long",
            ));
        }
    }
}

/// Simple bit writer for packing
struct BitWriter<'a> {
    output: &'a mut Vec<u8>,
    buffer: u64,
    bits_in_buffer: u8,
}

impl<'a> BitWriter<'a> {
    fn new(output: &'a mut Vec<u8>) -> Self {
        Self {
            output,
            buffer: 0,
            bits_in_buffer: 0,
        }
    }

    fn write(&mut self, value: u64, num_bits: u8) -> io::Result<()> {
        debug_assert!(num_bits <= 64);

        self.buffer |= value << self.bits_in_buffer;
        self.bits_in_buffer += num_bits;

        while self.bits_in_buffer >= 8 {
            self.output.push(self.buffer as u8);
            self.buffer >>= 8;
            self.bits_in_buffer -= 8;
        }

        Ok(())
    }

    fn flush(&mut self) -> io::Result<()> {
        if self.bits_in_buffer > 0 {
            self.output.push(self.buffer as u8);
            self.buffer = 0;
            self.bits_in_buffer = 0;
        }
        Ok(())
    }
}

/// Simple bit reader for unpacking
struct BitReader<'a> {
    data: &'a [u8],
    byte_pos: usize,
    bit_pos: u8,
}

impl<'a> BitReader<'a> {
    fn new(data: &'a [u8]) -> Self {
        Self {
            data,
            byte_pos: 0,
            bit_pos: 0,
        }
    }

    fn read(&mut self, num_bits: u8) -> io::Result<u64> {
        if num_bits == 0 {
            return Ok(0);
        }

        let mut result: u64 = 0;
        let mut bits_read: u8 = 0;

        while bits_read < num_bits {
            if self.byte_pos >= self.data.len() {
                return Err(io::Error::new(
                    io::ErrorKind::UnexpectedEof,
                    "Not enough bits",
                ));
            }

            let bits_available = 8 - self.bit_pos;
            let bits_to_read = (num_bits - bits_read).min(bits_available);
            // Handle edge case where bits_to_read == 8 to avoid overflow
            let mask = if bits_to_read >= 8 {
                0xFF
            } else {
                (1u8 << bits_to_read) - 1
            };
            let bits = (self.data[self.byte_pos] >> self.bit_pos) & mask;

            result |= (bits as u64) << bits_read;
            bits_read += bits_to_read;
            self.bit_pos += bits_to_read;

            if self.bit_pos >= 8 {
                self.byte_pos += 1;
                self.bit_pos = 0;
            }
        }

        Ok(result)
    }
}

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

    #[test]
    fn test_block_addr_store_roundtrip() {
        let addrs = vec![
            BlockAddr {
                offset: 0,
                length: 1000,
            },
            BlockAddr {
                offset: 1000,
                length: 1500,
            },
            BlockAddr {
                offset: 2500,
                length: 800,
            },
            BlockAddr {
                offset: 3300,
                length: 2000,
            },
        ];

        let bytes = BlockAddrStore::build(&addrs).unwrap();
        let store = BlockAddrStore::load(OwnedBytes::new(bytes)).unwrap();

        assert_eq!(store.len(), 4);
        for (i, expected) in addrs.iter().enumerate() {
            let actual = store.get(i).unwrap();
            assert_eq!(actual.offset, expected.offset, "offset mismatch at {}", i);
            assert_eq!(actual.length, expected.length, "length mismatch at {}", i);
        }
    }

    #[test]
    fn test_block_addr_store_empty() {
        let bytes = BlockAddrStore::build(&[]).unwrap();
        let store = BlockAddrStore::load(OwnedBytes::new(bytes)).unwrap();
        assert_eq!(store.len(), 0);
        assert!(store.get(0).is_none());
    }

    #[test]
    fn test_mmap_block_index_roundtrip() {
        let entries = vec![
            (
                b"aaa".to_vec(),
                BlockAddr {
                    offset: 0,
                    length: 100,
                },
            ),
            (
                b"bbb".to_vec(),
                BlockAddr {
                    offset: 100,
                    length: 150,
                },
            ),
            (
                b"ccc".to_vec(),
                BlockAddr {
                    offset: 250,
                    length: 200,
                },
            ),
        ];

        let bytes = MmapBlockIndex::build(&entries).unwrap();
        let index = MmapBlockIndex::load(OwnedBytes::new(bytes)).unwrap();

        assert_eq!(index.len(), 3);

        // Test locate
        assert_eq!(index.locate(b"aaa"), Some(0));
        assert_eq!(index.locate(b"bbb"), Some(1));
        assert_eq!(index.locate(b"ccc"), Some(2));
        assert_eq!(index.locate(b"aab"), Some(0)); // Between aaa and bbb
        assert_eq!(index.locate(b"ddd"), Some(2)); // After all keys
        assert_eq!(index.locate(b"000"), None); // Before all keys
    }

    #[cfg(feature = "fst-index")]
    #[test]
    fn test_fst_block_index_roundtrip() {
        let entries = vec![
            (
                b"aaa".to_vec(),
                BlockAddr {
                    offset: 0,
                    length: 100,
                },
            ),
            (
                b"bbb".to_vec(),
                BlockAddr {
                    offset: 100,
                    length: 150,
                },
            ),
            (
                b"ccc".to_vec(),
                BlockAddr {
                    offset: 250,
                    length: 200,
                },
            ),
        ];

        let bytes = FstBlockIndex::build(&entries).unwrap();
        let index = FstBlockIndex::load(OwnedBytes::new(bytes)).unwrap();

        assert_eq!(index.len(), 3);

        // Test locate
        assert_eq!(index.locate(b"aaa"), Some(0));
        assert_eq!(index.locate(b"bbb"), Some(1));
        assert_eq!(index.locate(b"ccc"), Some(2));
        assert_eq!(index.locate(b"aab"), Some(0)); // Between aaa and bbb
        assert_eq!(index.locate(b"ddd"), Some(2)); // After all keys
    }

    #[test]
    fn test_bit_writer_reader() {
        let mut buf = Vec::new();
        let mut writer = BitWriter::new(&mut buf);

        writer.write(5, 3).unwrap(); // 101
        writer.write(3, 2).unwrap(); // 11
        writer.write(15, 4).unwrap(); // 1111
        writer.flush().unwrap();

        let mut reader = BitReader::new(&buf);
        assert_eq!(reader.read(3).unwrap(), 5);
        assert_eq!(reader.read(2).unwrap(), 3);
        assert_eq!(reader.read(4).unwrap(), 15);
    }
}