seqlog 0.2.0

An embedded append-only sequence log store.
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
#![doc = include_str!("../README.md")]

use crc32fast::hash;
use std::fs::{self, File};
use std::io::{IoSlice, Read, Seek, SeekFrom, Write};
use std::path::{Path, PathBuf};
use std::sync::atomic::{AtomicU64, Ordering};
use std::sync::{Arc, Mutex, RwLock};

mod error;

pub use error::Error;

pub type Result<T> = std::result::Result<T, Error>;

const DATA_SUFFIX: &'static str = ".data";
const INDEX_SUFFIX: &'static str = ".index";
const LOCK_FILE: &'static str = "LOCK";

const INDEX_INTERVAL: u64 = 1024;

const READBUF_SIZE: usize = 128 * 1024; // bigger that ENTRY_MAX_LEN

const LEN_SIZE: usize = std::mem::size_of::<u16>();
const CHSUM_SIZE: usize = 2; // use 2 bytes of CRC32
const HEADER_SIZE: usize = LEN_SIZE + CHSUM_SIZE;

const INDEX_SIZE: usize = std::mem::size_of::<u64>();

const ENTRY_MAX_LEN: usize = u16::MAX as usize;

// The state that is shared amount Writer, Syncer, and multiple Readers.
struct SharedState {
    // read-only
    dir: PathBuf,

    // Writer updates this when rotating.
    // Readers use this to scan files.
    file_seqs: RwLock<Vec<Arc<u64>>>,

    // Writer inserts new file's clone into this when rotating.
    // Syncer takes this.
    current_dup: Mutex<Option<File>>,

    // Syncer updates this after syncing.
    sync_seq: AtomicU64,

    // Writer updates this after appending.
    next_seq: AtomicU64,
}

/// SeqLog instance. It is also the Writer.
pub struct SeqLog {
    _lock: File,

    state: Arc<SharedState>,

    // config
    rotate_size: usize,
    rotate_count: usize,

    // Writer's current file status
    current_data: File,
    current_index: File,
    data_file_size: usize,
    file_seq: u64,
}

impl SeqLog {
    /// Create a SeqLog instance with start sequence number, and open it.
    ///
    /// This requires that the parent directory of the given path already
    /// exists and the path itself does not exist.
    ///
    /// This method creates the directory and some required files. If you
    /// want to create it out of your program, use this script:
    ///
    /// ```bash
    /// mkdir $path
    /// touch $path/LOCK
    /// touch $path/`printf "%020d.data" $start_seq`   # first data file
    /// touch $path/`printf "%020d.index" $start_seq`  # first index file
    /// ```
    pub fn create<P: AsRef<Path>>(path: P, start_seq: u64) -> Result<Self> {
        let dir = path.as_ref();

        fs::create_dir(dir)?;

        // LOCK file
        File::create(dir.join(LOCK_FILE))?;

        // first empty data file and index file
        File::create(data_file_name(dir, start_seq))?;
        File::create(index_file_name(dir, start_seq))?;

        Self::open(dir)
    }

    /// Open an existing SeqLog instance.
    pub fn open<P: AsRef<Path>>(path: P) -> Result<Self> {
        let dir = path.as_ref();

        let lock = File::open(dir.join(LOCK_FILE))?;
        lock.try_lock()?;

        let file_seqs = list_files(dir)?;

        // at least one data file
        let Some(file_seq) = file_seqs.last() else {
            return Err(Error::NoDataFile);
        };

        let file_seq = **file_seq; // deref "&Arc<u64>" to u64

        let fname = data_file_name(dir, file_seq);
        let mut current_data = File::options().append(true).read(true).open(fname)?;

        let fname = index_file_name(dir, file_seq);
        let mut current_index = File::options().append(true).read(true).open(fname)?;

        let (count, file_size) =
            check_current_file(&mut current_data, &mut current_index, file_seq)?;

        let next_seq = file_seq + count;
        let data_file_size = file_size as usize;
        dbg!(next_seq);

        let state = SharedState {
            dir: dir.to_path_buf(),
            file_seqs: RwLock::new(file_seqs),
            current_dup: Mutex::new(None),
            sync_seq: AtomicU64::new(next_seq),
            next_seq: AtomicU64::new(next_seq),
        };

        Ok(Self {
            _lock: lock,
            state: Arc::new(state),

            rotate_size: 1024 * 1024 * 1024, // default value: 1G
            rotate_count: 10,                // default value: 10

            current_data,
            current_index,
            data_file_size,
            file_seq,
        })
    }

    /// Append a batch of entries.
    ///
    /// This issues one `write(2)` syscall. Therefore, appending in batches
    /// reduces the number of system calls.
    ///
    /// This writes data into file (page cache), but does not synchronize
    /// disk. Use [`Self::sync`] or [`SeqLogSyncer`] if needed.
    pub fn append<T>(&mut self, entries: &[T]) -> Result<()>
    where
        T: AsRef<[u8]>,
    {
        // check rotation
        //
        // We must rotate at the beginning of function, but not at the tail.
        // So uses can always sync() the current file after the function.
        if self.data_file_size >= self.rotate_size {
            self.rotate()?;
        }

        let mut headers = Vec::with_capacity(entries.len());
        let mut bufs = Vec::with_capacity(entries.len() * 2); // header + payload

        let mut total_len = 0;
        let mut next_seq = self.next_seq();

        for entry in entries.iter() {
            let entry = entry.as_ref();
            let len = entry.len();

            if len > ENTRY_MAX_LEN {
                return Err(Error::EntryTooLarge(len as usize));
            }

            // encode the header: length + checksum
            let mut header_buf = [0u8; HEADER_SIZE];
            header_buf[..2].copy_from_slice(&(len as u16).to_le_bytes());
            header_buf[2..].copy_from_slice(&(hash(entry) as u16).to_le_bytes());
            headers.push(header_buf);

            bufs.push(IoSlice::new(&[])); // hold the place

            // the payload
            bufs.push(IoSlice::new(entry));

            total_len += HEADER_SIZE + len;
            next_seq += 1;

            // append index
            if (next_seq - self.file_seq) % INDEX_INTERVAL == 0 {
                let offset = self.data_file_size + total_len;
                append_index(&mut self.current_index, offset as u64)?;
            }
        }

        // fix headers in buf
        for (i, header) in headers.iter().enumerate() {
            bufs[i * 2] = IoSlice::new(&header[..]);
        }

        // finally, write into file
        let writen_len = self.current_data.write_vectored(&bufs)?;
        if writen_len == 0 {
            return Err(std::io::Error::from(std::io::ErrorKind::WriteZero).into());
        }
        if writen_len < total_len {
            // truncate the new data
            self.current_data.set_len(self.data_file_size as u64)?;
            return Err(std::io::Error::new(std::io::ErrorKind::WriteZero, "partitial").into());
        }

        // update status
        self.data_file_size += total_len;
        self.state.next_seq.store(next_seq, Ordering::Release);

        Ok(())
    }

    /// Remove files containing only entries before the sequence number.
    ///
    /// If a data file contains entries both before and after the sequence
    /// number, it will not be removed.
    ///
    /// If a data file is in read by any reader, it will not be removed.
    ///
    /// The current file is never removed.
    pub fn purge(&mut self, seq: u64) -> Result<()> {
        let mut file_seqs = self.state.file_seqs.write().unwrap();

        let mut until = file_seqs.len() - 1;
        for (i, file_seq) in file_seqs.iter().enumerate() {
            if **file_seq > seq {
                until = i.saturating_sub(1);
                break;
            }
            if Arc::strong_count(file_seq) > 1 {
                until = i;
                break;
            }
        }

        for file_seq in file_seqs.drain(..until) {
            fs::remove_file(data_file_name(&self.state.dir, *file_seq))?;
            fs::remove_file(index_file_name(&self.state.dir, *file_seq))?;
        }
        Ok(())
    }

    /// Truncate entries from the sequence number, inclusive.
    ///
    /// This does not check if any reader are reading these entries.
    pub fn truncate(&mut self, seq: u64) -> Result<()> {
        if seq >= self.next_seq() {
            return Err(Error::SeqNotReached(seq, self.next_seq()));
        }

        // is the seq in current file?
        if seq < self.file_seq {
            let mut file_seqs = self.state.file_seqs.write().unwrap();

            // we have to keep one file at least, so return error if
            // it would remove all files
            let Some(i) = file_seqs.iter().rposition(|f| **f <= seq) else {
                return Err(Error::SeqPurged(seq, *file_seqs[0]));
            };

            // remove some files
            for file_seq in file_seqs.drain(i + 1..) {
                fs::remove_file(data_file_name(&self.state.dir, *file_seq))?;
                fs::remove_file(index_file_name(&self.state.dir, *file_seq))?;
            }

            let file_seq = *file_seqs[i];

            // open new data file and index file
            let fname = data_file_name(&self.state.dir, file_seq);
            self.current_data = File::options().append(true).read(true).open(fname)?;

            let fname = index_file_name(&self.state.dir, file_seq);
            self.current_index = File::options().append(true).read(true).open(fname)?;

            self.data_file_size = 0; // reset later
            self.file_seq = file_seq;

            *self.state.current_dup.lock().unwrap() = Some(self.current_data.try_clone()?);
        }

        // remove entries in current data file
        let mut data_buf = Vec::new();
        let (mut file, data_pos) =
            seek_seq_in_file(&self.state.dir, self.file_seq, seq, &mut data_buf)?;
        self.data_file_size = file.stream_position()? as usize - (data_buf.len() - data_pos);
        self.current_data.set_len(self.data_file_size as u64)?;
        self.current_data.seek(SeekFrom::End(0))?;

        // remove indexes in current index file
        // TODO check index
        let index_file_size = (seq - self.file_seq) / INDEX_INTERVAL * INDEX_SIZE as u64;
        self.current_index.set_len(index_file_size)?;
        self.current_index.seek(SeekFrom::End(0))?;

        // update shared state
        self.state.next_seq.store(seq, Ordering::Relaxed);
        self.state.sync_seq.store(seq, Ordering::Relaxed);

        Ok(())
    }

    /// Return the next sequence number.
    pub fn next_seq(&self) -> u64 {
        self.state.next_seq()
    }

    /// Return the next sequence number to synchronize to disk.
    ///
    /// This is the synchronization version of [`Self::next_seq`].
    pub fn sync_seq(&self) -> u64 {
        self.state.sync_seq()
    }

    /// Synchronizes entries to disk.
    ///
    /// Syncing data to disk is a relatively slow and blocking operation.
    /// If you do not want to block the current thread, create a
    /// [`SeqLogSyncer`] via [`Self::syncer`] and send it to another thread
    /// to perform synchronization there.
    pub fn sync(&self) -> Result<()> {
        let seq = self.next_seq();

        self.current_data.sync_data()?;

        self.state.sync_seq.store(seq, Ordering::Release);
        Ok(())
    }

    /// Create a [`SeqLogSyncer`].
    pub fn syncer(&self) -> Result<SeqLogSyncer> {
        Ok(SeqLogSyncer {
            state: self.state.clone(),
            current: self.current_data.try_clone()?,
        })
    }

    /// Create a new [`SeqLogReader`] with the sequence number.
    ///
    /// If `synced_only` is set to `true`, the reader will only read entries
    /// that have been synchronized to disk via [`Self::sync`] or [`SeqLogSyncer::sync`].
    pub fn reader(&self, next_seq: u64, synced_only: bool) -> Result<SeqLogReader> {
        let mut data_buf = Vec::new();

        let (current, file_seq, data_pos) =
            self.state.seek_seq(next_seq, &mut data_buf, synced_only)?;

        Ok(SeqLogReader {
            state: self.state.clone(),
            data_buf,

            synced_only,

            current,
            file_seq,
            data_pos,
            next_seq,
        })
    }

    /// Configrate rotation.
    ///
    /// `size`: Rotate a data file when it exceeds this size. The default value
    /// is 1G. Setting to 0 means never rotating, and you can call [`Self::rotate`]
    /// to rotate manaully.
    ///
    /// `count`: Number of files to retain. The default value is 10. Setting to
    /// 0 means keeping all files, and you can call [`Self::purge`] to delete
    /// manually.
    pub fn set_rotate(&mut self, size: usize, count: usize) {
        self.rotate_size = if size == 0 { usize::MAX } else { size };
        self.rotate_count = if count == 0 { usize::MAX } else { count };
    }

    /// Rotate the data file manually, e.g. by time or sequence number.
    ///
    /// It does not rotate if the current file is empty.
    ///
    /// You should have called [`Self::set_rotate`] with `size=0` to disable
    /// the automatic rotation.
    pub fn rotate(&mut self) -> Result<()> {
        if self.data_file_size == 0 {
            // do not rotate if current file is empty
            return Ok(());
        }

        let next_seq = self.next_seq();

        // open new file
        self.current_data = File::create_new(data_file_name(&self.state.dir, next_seq))?;
        self.current_index = File::create_new(index_file_name(&self.state.dir, next_seq))?;
        self.data_file_size = 0;
        self.file_seq = next_seq;

        // save new file info
        let mut file_seqs = self.state.file_seqs.write().unwrap();
        file_seqs.push(Arc::new(next_seq));

        // expire
        while file_seqs.len() > self.rotate_count {
            if Arc::strong_count(&file_seqs[0]) > 0 {
                // this file is in used
                break;
            }
            let file_seq = file_seqs.remove(0);
            fs::remove_file(data_file_name(&self.state.dir, *file_seq))?;
            fs::remove_file(index_file_name(&self.state.dir, *file_seq))?;
        }

        // notify the syncer
        *self.state.current_dup.lock().unwrap() = Some(self.current_data.try_clone()?);

        Ok(())
    }
}

/// A sequential reader for scanning entries.
///
/// Compared with key-value stores, SeqLog has two characteristics for reading:
///
/// 1. Seeking to an arbitrary sequence number is not that fast bacause
///    it uses sparse index;
/// 2. Once positioned, sequential reads are very efficient because entries
///    are stored in sequence order.
///
/// This access pattern is similar to a hard disk (HDD).
/// A `SeqLogReader` acts like the disk read head. It is positioned at the
/// specific sequence number when created (by [`SeqLog::reader`]).
/// Subsequent reads (by [`Self::next`]) continue sequentially from that
/// position.
///
/// Each reader maintains its own read position. Multiple readers can
/// coexist and scan the log independently.
///
/// The reader can be sent between threads.
pub struct SeqLogReader {
    state: Arc<SharedState>,

    synced_only: bool,

    current: File,
    file_seq: Arc<u64>,
    data_buf: Vec<u8>,
    data_pos: usize,
    next_seq: u64,
}

impl SeqLogReader {
    /// Return the next entry.
    pub fn next(&mut self) -> Result<Option<&[u8]>> {
        // check if end
        let max_seq = if self.synced_only {
            self.state.sync_seq()
        } else {
            self.state.next_seq()
        };
        if self.next_seq >= max_seq {
            return Ok(None);
        }

        // now there must be an entry

        // if any in the current buffer
        let len = match parse_entry_len(&self.data_buf[self.data_pos..]) {
            Some(len) => len,
            None => {
                // or read more data
                self.read_more_data()?;
                parse_entry_len(&self.data_buf)
                    .ok_or(Error::DataFileTruncated(*self.file_seq, self.next_seq))?
            }
        };

        let entry = &self.data_buf[self.data_pos..];
        let payload = &entry[HEADER_SIZE..HEADER_SIZE + len];

        // check checksum
        let chsum1 = parse_checksum(&entry[LEN_SIZE..]);
        let chsum2 = hash(&payload) as u16;
        if chsum1 != chsum2 {
            return Err(Error::ChecksumMismatch(*self.file_seq, self.next_seq));
        }

        self.next_seq += 1;
        self.data_pos += HEADER_SIZE + len;
        return Ok(Some(payload));
    }

    // We have checked the seq, so there must be one entry at least.
    // Return Error if can not read more data.
    fn read_more_data(&mut self) -> Result<()> {
        let remain_len = self.data_buf.len() - self.data_pos;
        self.data_buf.copy_within(self.data_pos.., 0);

        self.data_buf.resize(READBUF_SIZE, 0);

        let mut len = self.current.read(&mut self.data_buf[remain_len..])?;

        if len == 0 {
            // end of current file, open new file

            if remain_len != 0 {
                return Err(Error::DataFileTruncated(*self.file_seq, self.next_seq));
            }

            let file_seqs = self.state.file_seqs.read().unwrap();
            let Some(file_seq) = file_seqs.iter().rev().find(|&f| **f == self.next_seq) else {
                return Err(Error::DataFileTruncated(self.next_seq, self.next_seq));
            };

            self.file_seq = file_seq.clone();
            self.current = File::open(data_file_name(&self.state.dir, **file_seq))?;

            len = self.current.read(&mut self.data_buf)?;
            if len == 0 {
                return Err(Error::DataFileTruncated(*self.file_seq, self.next_seq));
            }
        }

        self.data_buf.truncate(remain_len + len);
        self.data_pos = 0;
        Ok(())
    }

    /// Return the next sequence number.
    pub fn next_seq(&self) -> u64 {
        self.next_seq
    }

    /// Reset the sequence number. Then [`Self::next`] will return
    /// the entry with this sequence number later.
    ///
    /// This is equvalent to [`SeqLog::reader`] to create a new reader.
    /// This is useful only if you can not access the SeqLog instance.
    pub fn reset(&mut self, seq: u64) -> Result<()> {
        let (file, file_seq, data_pos) =
            self.state
                .seek_seq(seq, &mut self.data_buf, self.synced_only)?;

        // update
        self.data_pos = data_pos;
        self.file_seq = file_seq;
        self.current = file;
        self.next_seq = seq;
        Ok(())
    }
}

/// A cross-thread synchronizing handler.
///
/// Unlike [`SeqLog::sync`], a `SeqLogSyncer` can be sent to a different thread
/// and used to perform synchronization there. This allows disk sync operations
/// to be offloaded from the writer thread, avoiding stalls caused by slow and
/// blocking storage operations.
pub struct SeqLogSyncer {
    state: Arc<SharedState>,

    current: File,
}

impl SeqLogSyncer {
    /// Synchronize data to disk.
    pub fn sync(&mut self) -> Result<()> {
        // load next_seq before sync()
        let mut next_seq = self.state.next_seq();

        self.current.sync_data()?;

        // check new file rotated
        if let Some(new_file) = self.state.current_dup.lock().unwrap().take() {
            next_seq = self.state.next_seq();
            self.current = new_file;
            self.current.sync_data()?;
        }

        // store sync_seq after sync()
        self.state.sync_seq.store(next_seq, Ordering::Release);
        Ok(())
    }
}

impl SharedState {
    fn seek_seq(
        &self,
        seq: u64,
        data_buf: &mut Vec<u8>,
        synced_only: bool,
    ) -> Result<(File, Arc<u64>, usize)> {
        let max_seq = if synced_only {
            self.sync_seq()
        } else {
            self.next_seq()
        };
        if seq > max_seq {
            return Err(Error::SeqNotReached(seq, max_seq));
        }

        // locate the file
        let file_seqs = self.file_seqs.read().unwrap();
        let Some(file_seq) = file_seqs.iter().rev().find(|&f| **f <= seq) else {
            return Err(Error::SeqPurged(seq, *file_seqs[0]));
        };

        // lock the file, so the rotate() will not remove this
        let file_seq = file_seq.clone();

        // unlock ASAP
        drop(file_seqs);

        // locate the entry in the file
        let (file, data_pos) = seek_seq_in_file(&self.dir, *file_seq, seq, data_buf)?;

        Ok((file, file_seq, data_pos))
    }

    fn next_seq(&self) -> u64 {
        self.next_seq.load(Ordering::Acquire)
    }

    fn sync_seq(&self) -> u64 {
        self.sync_seq.load(Ordering::Acquire)
    }
}

fn seek_seq_in_file(
    dir: &Path,
    file_seq: u64,
    seq: u64,
    data_buf: &mut Vec<u8>,
) -> Result<(File, usize)> {
    let mut data_file = File::open(data_file_name(dir, file_seq))?;

    // seek by index
    let diff_seq = seq - file_seq;
    if diff_seq >= INDEX_INTERVAL {
        let mut index_file = File::open(index_file_name(dir, file_seq))?;
        index_file.seek(SeekFrom::Start(
            (diff_seq / INDEX_INTERVAL - 1) * INDEX_SIZE as u64,
        ))?;

        let mut index_buf = [0; INDEX_SIZE];
        index_file.read_exact(&mut index_buf)?;

        let offset = parse_index(&index_buf);

        // seek the data file to the offset
        data_file.seek(SeekFrom::Start(offset))?;
    }

    // read data file
    data_buf.resize(READBUF_SIZE, 0);
    let len = data_file.read(data_buf)?;
    data_buf.truncate(len);

    let mut count = diff_seq % INDEX_INTERVAL;
    if count == 0 {
        return Ok((data_file, 0));
    }

    // walk the entries
    loop {
        let mut data_pos = 0;
        while let Some(len) = parse_entry_len(&data_buf[data_pos..]) {
            data_pos += HEADER_SIZE + len;
            count -= 1;
            if count == 0 {
                return Ok((data_file, data_pos));
            }
        }

        if data_buf.len() < READBUF_SIZE {
            return Err(Error::DataFileTruncated(file_seq, seq));
        }

        data_buf.copy_within(data_pos.., 0);

        let len = data_file.read(&mut data_buf[READBUF_SIZE - data_pos..])?;
        if len == 0 {
            return Err(Error::DataFileTruncated(file_seq, seq));
        }
        data_buf.truncate(READBUF_SIZE - data_pos + len);
    }
}

// list all files and return the seqs
fn list_files(dir: &Path) -> Result<Vec<Arc<u64>>> {
    let mut files = Vec::new();

    for entry in fs::read_dir(dir)? {
        let path = entry?.path().to_path_buf();
        let fname = path
            .file_name()
            .expect("invalid filename")
            .to_str()
            .expect("fail to parse filename");
        if !fname.ends_with(DATA_SUFFIX) {
            continue;
        }
        let Ok(seq) = fname[..fname.len() - DATA_SUFFIX.len()].parse() else {
            return Err(Error::InvalidDataFilename(fname.to_string()));
        };

        files.push(Arc::new(seq));
    }

    files.sort();

    Ok(files)
}

fn check_current_file(
    data_file: &mut File,
    index_file: &mut File,
    file_seq: u64,
) -> Result<(u64, u64)> {
    let index_file_len = index_file.metadata()?.len() as usize;
    if index_file_len % INDEX_SIZE != 0 {
        return Err(Error::InvalidIndexFile(file_seq));
    }

    // read the last index, if any
    let mut count = 0;
    let mut offset = 0;
    if index_file_len != 0 {
        // read the index, get the offset
        let mut index_buf = [0; INDEX_SIZE];
        index_file.seek(SeekFrom::End(-(INDEX_SIZE as i64)))?;
        index_file.read_exact(&mut index_buf)?;
        offset = parse_index(&index_buf);

        // seek the data file to the offset
        data_file.seek(SeekFrom::Start(offset))?;

        count = (index_file_len / INDEX_SIZE) as u64 * INDEX_INTERVAL;
    }

    // read data
    let mut data_buf = Vec::new();
    data_file.read_to_end(&mut data_buf)?;

    // walk entries to the end
    let mut entry = &data_buf[..];
    while let Some(len) = parse_entry_len(entry) {
        entry = &entry[HEADER_SIZE + len..];

        count += 1;

        offset += (HEADER_SIZE + len) as u64;
        if count % INDEX_INTERVAL == 0 {
            append_index(index_file, offset)?;
        }
    }

    Ok((count, offset))
}

// append index to index file
fn append_index(file: &mut File, offset: u64) -> Result<()> {
    Ok(file.write_all(&offset.to_le_bytes())?)
}

// build data file name
fn data_file_name(dir: &Path, seq: u64) -> PathBuf {
    dir.join(format!("{:020}{}", seq, DATA_SUFFIX))
}

// build index file name
fn index_file_name(dir: &Path, seq: u64) -> PathBuf {
    dir.join(format!("{:020}{}", seq, INDEX_SUFFIX))
}

// return None if not complete
fn parse_entry_len(buf: &[u8]) -> Option<usize> {
    if buf.len() < HEADER_SIZE {
        return None;
    }
    let len = u16::from_le_bytes(buf[0..LEN_SIZE].try_into().unwrap()) as usize;
    if buf.len() < HEADER_SIZE + len {
        return None;
    }
    Some(len)
}

fn parse_checksum(buf: &[u8]) -> u16 {
    u16::from_le_bytes(buf[0..CHSUM_SIZE].try_into().unwrap())
}

fn parse_index(buf: &[u8]) -> u64 {
    u64::from_le_bytes(buf[..8].try_into().unwrap())
}