timslite 0.1.7

High-performance mmap-backed time-series storage engine with queues and journal.
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
use std::collections::HashMap;
use std::path::PathBuf;
use std::sync::atomic::{AtomicBool, Ordering};
use std::sync::{Arc, Mutex};
use std::time::{Duration, Instant};

use crate::error::{Result, TmslError};
use crate::journal::log::JournalLog;
use crate::queue::{
    now_unix_seconds, ConsumerStateFile, QueueConsumerConfig, QueueNotifier, QueuePollCallback,
    QueuePollCallbackSlot,
};
use crate::util::{is_path_safe_component, PATH_COMPONENT_MAX_LEN};

fn validate_consumer_group_name(group_name: &str) -> Result<()> {
    if is_path_safe_component(group_name) {
        Ok(())
    } else {
        Err(TmslError::InvalidData(format!(
            "journal queue consumer group_name must match ^[0-9A-Za-z_-]+$ and be at most {PATH_COMPONENT_MAX_LEN} bytes"
        )))
    }
}

struct JournalQueueInner {
    consumers: HashMap<String, Arc<Mutex<ConsumerStateFile>>>,
    consumer_configs: HashMap<String, QueueConsumerConfig>,
    closed: Arc<AtomicBool>,
}

impl JournalQueueInner {
    fn new() -> Self {
        Self {
            consumers: HashMap::new(),
            consumer_configs: HashMap::new(),
            closed: Arc::new(AtomicBool::new(false)),
        }
    }
}

#[derive(Clone)]
pub struct JournalQueue {
    log: Arc<Mutex<JournalLog>>,
    queue_dir: PathBuf,
    inner: Arc<Mutex<JournalQueueInner>>,
    notify: Arc<QueueNotifier>,
}

impl JournalQueue {
    pub(crate) fn new(log: Arc<Mutex<JournalLog>>, queue_dir: PathBuf) -> Self {
        Self {
            log,
            queue_dir,
            inner: Arc::new(Mutex::new(JournalQueueInner::new())),
            notify: Arc::new(QueueNotifier::new()),
        }
    }

    pub fn open_consumer(&self, group_name: &str) -> Result<JournalQueueConsumer> {
        self.open_consumer_with_config(group_name, QueueConsumerConfig::default())
    }

    pub fn open_consumer_with_config(
        &self,
        group_name: &str,
        config: QueueConsumerConfig,
    ) -> Result<JournalQueueConsumer> {
        validate_consumer_group_name(group_name)?;
        let mut inner = self
            .inner
            .lock()
            .map_err(|_| TmslError::InvalidData("journal queue mutex poisoned".into()))?;
        if inner.closed.load(Ordering::SeqCst) {
            return Err(TmslError::QueueClosed("journal queue is closed".into()));
        }
        if let Some(existing) = inner.consumer_configs.get(group_name) {
            if *existing != config {
                return Err(TmslError::InvalidData(format!(
                    "journal consumer group {group_name} is already open with a different config"
                )));
            }
        }
        let state_file = if let Some(existing) = inner.consumers.get(group_name) {
            Arc::clone(existing)
        } else {
            let initial_sequence = self
                .log
                .lock()
                .map_err(|_| TmslError::InvalidData("journal log mutex poisoned".into()))?
                .latest_sequence()
                .unwrap_or(0);
            let state_path = self.queue_dir.join(group_name);
            let state = Arc::new(Mutex::new(ConsumerStateFile::open_or_create(
                state_path,
                initial_sequence,
            )?));
            inner
                .consumers
                .insert(group_name.to_string(), Arc::clone(&state));
            inner
                .consumer_configs
                .insert(group_name.to_string(), config);
            state
        };
        let callback_slot = Arc::new(Mutex::new(None));
        self.notify.register_callback_slot(&callback_slot)?;

        Ok(JournalQueueConsumer {
            state_file,
            config,
            log: Arc::clone(&self.log),
            notify: Arc::clone(&self.notify),
            closed: Arc::clone(&inner.closed),
            poll_callback: callback_slot,
        })
    }

    pub fn close(&self) -> Result<()> {
        let mut inner = self
            .inner
            .lock()
            .map_err(|_| TmslError::InvalidData("journal queue mutex poisoned".into()))?;
        if inner.closed.swap(true, Ordering::SeqCst) {
            return Ok(());
        }
        for state in inner.consumers.values() {
            if let Ok(mut guard) = state.lock() {
                let _ = guard.sync();
                let _ = guard.flush();
            }
        }
        inner.consumers.clear();
        inner.consumer_configs.clear();
        drop(inner);
        self.notify_waiters();
        Ok(())
    }

    pub(crate) fn notify_record_appended(&self) {
        let _ = self.notify.notify_data_available();
    }

    fn notify_waiters(&self) {
        let _ = self.notify.notify_waiters();
    }

    pub(crate) fn flush_state_files(&self) -> Result<()> {
        let inner = self
            .inner
            .lock()
            .map_err(|_| TmslError::InvalidData("journal queue mutex poisoned".into()))?;
        for state in inner.consumers.values() {
            let mut guard = state
                .lock()
                .map_err(|_| TmslError::InvalidData("journal state file mutex poisoned".into()))?;
            guard.sync()?;
            guard.flush()?;
        }
        Ok(())
    }
}

#[derive(Clone)]
pub struct JournalQueueConsumer {
    state_file: Arc<Mutex<ConsumerStateFile>>,
    config: QueueConsumerConfig,
    log: Arc<Mutex<JournalLog>>,
    notify: Arc<QueueNotifier>,
    closed: Arc<AtomicBool>,
    poll_callback: QueuePollCallbackSlot,
}

impl JournalQueueConsumer {
    pub fn poll(&self, timeout: Duration) -> Result<Option<(i64, Vec<u8>)>> {
        if self.closed.load(Ordering::SeqCst) {
            return Err(TmslError::QueueClosed("journal queue is closed".into()));
        }
        if let Some(row) = self.try_poll_available()? {
            return Ok(Some(row));
        }
        if timeout.is_zero() {
            return Ok(None);
        }
        let deadline = Instant::now() + timeout;
        let (lock, cvar) = self.notify.wait_parts();
        let mut notified = lock
            .lock()
            .map_err(|_| TmslError::InvalidData("journal notify mutex poisoned".into()))?;
        loop {
            if self.closed.load(Ordering::SeqCst) {
                return Err(TmslError::QueueClosed("journal queue is closed".into()));
            }
            if *notified {
                *notified = false;
                drop(notified);
                if let Some(row) = self.try_poll_available()? {
                    return Ok(Some(row));
                }
                notified = lock
                    .lock()
                    .map_err(|_| TmslError::InvalidData("journal notify mutex poisoned".into()))?;
                continue;
            }
            let Some(remaining) = deadline.checked_duration_since(Instant::now()) else {
                drop(notified);
                return self.try_poll_available();
            };
            let (guard, wait_result) = cvar.wait_timeout(notified, remaining).unwrap();
            notified = guard;
            if wait_result.timed_out() {
                drop(notified);
                return self.try_poll_available();
            }
        }
    }

    pub fn ack(&self, sequence: i64) -> Result<()> {
        if self.closed.load(Ordering::SeqCst) {
            return Err(TmslError::QueueClosed("journal queue is closed".into()));
        }
        let mut state = self
            .state_file
            .lock()
            .map_err(|_| TmslError::InvalidData("journal state file mutex poisoned".into()))?;
        state.ack_pending(sequence)?;
        state.cleanup_acked();
        state.sync()?;
        Ok(())
    }

    pub fn poll_callback(&self, callback: Option<QueuePollCallback>) -> Result<()> {
        let mut slot = self.poll_callback.lock().map_err(|_| {
            TmslError::InvalidData("journal queue callback slot mutex poisoned".into())
        })?;
        if callback.is_some() && slot.is_some() {
            return Err(TmslError::InvalidData(
                "journal queue poll callback is already registered".into(),
            ));
        }
        *slot = callback;
        Ok(())
    }

    fn try_poll_available(&self) -> Result<Option<(i64, Vec<u8>)>> {
        let now = now_unix_seconds();

        // Retry an expired pending delivery. Commit the retry attempt only
        // after the unlocked log read succeeds.
        let retry_scan = {
            let state = self
                .state_file
                .lock()
                .map_err(|_| TmslError::InvalidData("journal state file mutex poisoned".into()))?;
            state.peek_retryable_pending(self.config, now)
        };
        if let Some(sequence) = retry_scan.timestamp {
            match self.read_sequence(sequence) {
                Ok(Some((seq, data))) => {
                    if self.commit_retry_claim(Some(sequence), now)? {
                        return Ok(Some((seq, data)));
                    }
                    // A peer in the same group claimed this retry first.
                }
                Ok(None) => {
                    // The sequence is gone; consume the retry attempt and keep
                    // scanning instead of returning a false `None`.
                    self.commit_retry_claim(Some(sequence), now)?;
                }
                Err(err) => return Err(err),
            }
        } else if retry_scan.has_exhausted {
            self.commit_retry_claim(None, now)?;
        }

        loop {
            let next_sequence = self
                .log
                .lock()
                .map_err(|_| TmslError::InvalidData("journal log mutex poisoned".into()))?
                .next_sequence();

            let candidate = {
                let state = self.state_file.lock().map_err(|_| {
                    TmslError::InvalidData("journal state file mutex poisoned".into())
                })?;
                let mut candidate = state.next_poll_ts();
                while candidate < next_sequence && state.is_in_pending(candidate) {
                    candidate = candidate.saturating_add(1);
                }
                candidate
            };
            if candidate >= next_sequence {
                return Ok(None);
            }

            // Read outside the state lock, then revalidate + claim.
            let Some((sequence, data)) = self.read_sequence(candidate)? else {
                // Journal sequences are contiguous; do not skip ahead.
                return Ok(None);
            };
            let claimed = {
                let mut state = self.state_file.lock().map_err(|_| {
                    TmslError::InvalidData("journal state file mutex poisoned".into())
                })?;
                // Recheck close under the state-file lock: a close that raced
                // the unlocked log read must not be followed by a claim.
                let claimed = self.try_claim_pending_checked(&mut state, candidate, now)?;
                if claimed {
                    state.sync()?;
                }
                claimed
            };
            if claimed {
                return Ok(Some((sequence, data)));
            }
            // A peer claimed the candidate; recompute the next candidate.
        }
    }

    fn try_claim_pending_checked(
        &self,
        state: &mut ConsumerStateFile,
        sequence: i64,
        now: i64,
    ) -> Result<bool> {
        if self.closed.load(Ordering::SeqCst) {
            return Err(TmslError::QueueClosed("journal queue is closed".into()));
        }
        state.try_claim_pending(sequence, now)
    }

    fn commit_retry_claim(&self, sequence: Option<i64>, now: i64) -> Result<bool> {
        let mut state = self
            .state_file
            .lock()
            .map_err(|_| TmslError::InvalidData("journal state file mutex poisoned".into()))?;
        if self.closed.load(Ordering::SeqCst) {
            return Err(TmslError::QueueClosed("journal queue is closed".into()));
        }
        let commit = state.commit_retryable_pending(self.config, now, sequence)?;
        if commit.changed {
            state.sync()?;
        }
        Ok(commit.claimed)
    }

    fn read_sequence(&self, sequence: i64) -> Result<Option<(i64, Vec<u8>)>> {
        self.log
            .lock()
            .map_err(|_| TmslError::InvalidData("journal log mutex poisoned".into()))?
            .read(sequence)
    }
}

#[cfg(test)]
mod tests {
    use super::*;
    use crate::config::StoreConfig;
    use crate::journal::log::JournalLog;
    use std::sync::{Arc, Mutex};
    use std::time::Duration;

    fn temp_dir(name: &str) -> std::path::PathBuf {
        let dir = std::env::temp_dir().join(format!(
            "timslite_journal_queue_{name}_{:?}",
            std::time::SystemTime::now()
                .duration_since(std::time::UNIX_EPOCH)
                .unwrap()
                .as_nanos()
        ));
        std::fs::create_dir_all(&dir).unwrap();
        dir
    }

    fn make_log(name: &str) -> (Arc<Mutex<JournalLog>>, std::path::PathBuf) {
        let dir = temp_dir(name);
        let config = StoreConfig {
            data_segment_size: 4096,
            initial_data_segment_size: 4096,
            ..StoreConfig::default()
        };
        let log = JournalLog::open_or_create(dir.clone(), &config).unwrap();
        (Arc::new(Mutex::new(log)), dir)
    }

    #[test]
    fn journal_queue_new_consumer_starts_after_current_latest() {
        let (log, dir) = make_log("starts_after_latest");
        log.lock().unwrap().append(b"existing").unwrap();
        let queue = JournalQueue::new(log.clone(), dir.join("queue"));
        let consumer = queue.open_consumer("group1").unwrap();

        assert!(consumer.poll(Duration::from_millis(1)).unwrap().is_none());

        let seq = log.lock().unwrap().append(b"next").unwrap();
        queue.notify_record_appended();
        let row = consumer.poll(Duration::from_millis(100)).unwrap().unwrap();
        assert_eq!(row.0, seq);
        assert_eq!(row.1, b"next");
    }

    #[test]
    fn journal_queue_polls_realtime_sequence_one_based() {
        let (log, dir) = make_log("one_based");
        let queue = JournalQueue::new(log.clone(), dir.join("queue"));
        let consumer = queue.open_consumer("group1").unwrap();

        assert_eq!(log.lock().unwrap().append(b"first").unwrap(), 1);
        queue.notify_record_appended();

        let row = consumer.poll(Duration::from_millis(100)).unwrap().unwrap();
        assert_eq!(row.0, 1);
        assert_eq!(row.1, b"first");
    }

    #[test]
    fn journal_claim_recheck_rejects_closed_queue() {
        let (log, dir) = make_log("closed_claim");
        let queue = JournalQueue::new(log.clone(), dir.join("queue"));
        let consumer = queue.open_consumer("group").unwrap();
        log.lock().unwrap().append(b"payload").unwrap();

        // Close only after data exists so the poll's unlocked log read can
        // complete before the locked claim recheck observes the closed flag.
        queue.close().unwrap();
        let err = consumer.try_poll_available().unwrap_err();
        assert!(matches!(err, TmslError::QueueClosed(_)));

        let state = consumer.state_file.lock().unwrap();
        assert!(!state.is_in_pending(1));
        drop(state);

        // The retry commit path must reject close under the same lock too.
        let err = consumer
            .commit_retry_claim(Some(1), now_unix_seconds())
            .unwrap_err();
        assert!(matches!(err, TmslError::QueueClosed(_)));
        let _ = std::fs::remove_dir_all(&dir);
    }

    #[test]
    fn journal_queue_same_group_concurrent_claims_are_unique() {
        use std::collections::HashSet;
        use std::sync::mpsc;
        use std::sync::{Arc, Barrier};
        use std::thread;

        const CONSUMERS: usize = 4;
        const RECORDS: i64 = 200;

        let (log, dir) = make_log("concurrent_unique");
        let queue = JournalQueue::new(log.clone(), dir.join("queue"));
        // Consumers must be opened before the records exist: a fresh group
        // starts at the current latest sequence.
        let consumers: Vec<_> = (0..CONSUMERS)
            .map(|_| queue.open_consumer("same_group").unwrap())
            .collect();
        for i in 0..RECORDS {
            log.lock()
                .unwrap()
                .append(format!("record-{i}").as_bytes())
                .unwrap();
        }

        let barrier = Arc::new(Barrier::new(CONSUMERS));
        let (done_tx, done_rx) = mpsc::channel();
        let mut handles = Vec::new();
        for consumer in consumers {
            let barrier = Arc::clone(&barrier);
            let done_tx = done_tx.clone();
            handles.push(thread::spawn(move || {
                barrier.wait();
                let mut seen = Vec::new();
                loop {
                    match consumer.poll(Duration::ZERO) {
                        Ok(Some((sequence, _))) => {
                            seen.push(sequence);
                            // Acking is best-effort here: a duplicate claim may
                            // already have been cleaned up by the winning peer.
                            let _ = consumer.ack(sequence);
                        }
                        Ok(None) => break,
                        Err(err) => panic!("journal poll failed: {err:?}"),
                    }
                }
                let _ = done_tx.send(seen);
            }));
        }
        drop(done_tx);

        let mut claimed = Vec::new();
        for _ in 0..CONSUMERS {
            match done_rx.recv_timeout(Duration::from_secs(10)) {
                Ok(seen) => claimed.extend(seen),
                Err(_) => panic!("journal concurrent claim test timed out"),
            }
        }
        for handle in handles {
            handle.join().expect("journal claim worker panicked");
        }

        let unique: HashSet<i64> = claimed.iter().copied().collect();
        assert_eq!(
            claimed.len(),
            RECORDS as usize,
            "each journal record must be claimed exactly once"
        );
        assert_eq!(unique.len(), RECORDS as usize, "duplicate claims detected");
        for expected in 1..=RECORDS {
            assert!(unique.contains(&expected), "missing sequence {expected}");
        }
        let _ = std::fs::remove_dir_all(&dir);
    }

    #[test]
    fn journal_queue_ack_persists_processed_sequence() {
        let (log, dir) = make_log("ack_persists");
        let queue_dir = dir.join("queue");
        let queue = JournalQueue::new(log.clone(), queue_dir.clone());
        let consumer = queue.open_consumer("group1").unwrap();
        log.lock().unwrap().append(b"first").unwrap();
        queue.notify_record_appended();
        let row = consumer.poll(Duration::from_millis(100)).unwrap().unwrap();
        consumer.ack(row.0).unwrap();
        queue.flush_state_files().unwrap();
        drop(consumer);
        drop(queue);

        let queue = JournalQueue::new(log, queue_dir);
        let consumer = queue.open_consumer("group1").unwrap();

        assert!(consumer.poll(Duration::from_millis(1)).unwrap().is_none());
    }
}