fastskip 0.1.1

Lock-free arena-backed skip list memtable for LSM-tree storage engines
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

#![cfg(not(miri))] // Thread spawning is extremely slow under Miri

use std::sync::atomic::{AtomicBool, Ordering};
use std::sync::Arc;
use std::thread;

use fastskip::ConcurrentSkipList;

#[test]
fn test_16_writers_5k_each() {
    let sl = Arc::new(ConcurrentSkipList::with_shards(32));
    let mut handles = vec![];

    for tid in 0..16 {
        let sl: Arc<ConcurrentSkipList> = Arc::clone(&sl);
        handles.push(thread::spawn(move || {
            for i in 0..5000 {
                let key = format!("t{:02}_k{:06}", tid, i);
                let val = format!("v{:02}_{:06}", tid, i);
                sl.insert(key.as_bytes(), val.as_bytes());
            }
        }));
    }

    for h in handles {
        h.join().unwrap();
    }

    assert!(sl.len() >= 80_000);

    // Verify all keys retrievable
    for tid in 0..16 {
        for i in 0..5000 {
            let key = format!("t{:02}_k{:06}", tid, i);
            let val = sl.get_live(key.as_bytes()).unwrap();
            let expected = format!("v{:02}_{:06}", tid, i);
            assert_eq!(val, expected.as_bytes());
        }
    }
}

#[test]
fn test_concurrent_insert_delete_read() {
    let sl = Arc::new(ConcurrentSkipList::with_shards(8));

    // Pre-populate
    for i in 0..500 {
        let key = format!("init_{:06}", i);
        sl.insert(key.as_bytes(), b"val");
    }

    let running = Arc::new(AtomicBool::new(true));
    let mut handles = vec![];

    // Writers
    for tid in 0..4 {
        let sl: Arc<ConcurrentSkipList> = Arc::clone(&sl);
        handles.push(thread::spawn(move || {
            for i in 0..2000 {
                let key = format!("w{}_{:06}", tid, i);
                sl.insert(key.as_bytes(), b"val");
            }
        }));
    }

    // Deleters
    for tid in 0..2 {
        let sl: Arc<ConcurrentSkipList> = Arc::clone(&sl);
        handles.push(thread::spawn(move || {
            for i in 0..500 {
                let key = format!("w{}_{:06}", tid, i);
                sl.delete(key.as_bytes());
            }
        }));
    }

    // Readers
    let running_r = Arc::clone(&running);
    let sl_r: Arc<ConcurrentSkipList> = Arc::clone(&sl);
    handles.push(thread::spawn(move || {
        while running_r.load(Ordering::Relaxed) {
            for i in (0..500).step_by(10) {
                let key = format!("init_{:06}", i);
                let _ = sl_r.get(key.as_bytes());
            }
        }
    }));

    for h in handles.into_iter().take(6) {
        h.join().unwrap();
    }
    running.store(false, Ordering::Relaxed);
}

#[test]
fn test_edge_empty_key() {
    let sl = ConcurrentSkipList::new();
    assert!(sl.insert(b"", b"empty_key"));
    match sl.get(b"") {
        Some((v, false)) => assert_eq!(v, b"empty_key"),
        other => panic!(
            "expected value for empty key, got {:?}",
            other.map(|(v, _)| v.len())
        ),
    }
}

#[test]
fn test_edge_empty_value() {
    let sl = ConcurrentSkipList::new();
    assert!(sl.insert(b"key", b""));
    match sl.get(b"key") {
        Some((v, false)) => assert_eq!(v, b""),
        other => panic!("expected empty value, got {:?}", other),
    }
}

#[test]
fn test_edge_large_key_value() {
    let sl = ConcurrentSkipList::new();
    let key = vec![0xABu8; 4096];
    let val = vec![0xCDu8; 8192];
    assert!(sl.insert(&key, &val));
    match sl.get(&key) {
        Some((v, false)) => assert_eq!(v, val.as_slice()),
        _other => panic!("expected large value"),
    }
}

#[test]
fn test_edge_single_element() {
    let sl = ConcurrentSkipList::new();
    assert!(sl.insert(b"only", b"one"));
    assert_eq!(sl.len(), 1);
    assert_eq!(sl.get_live(b"only"), Some(b"one".as_slice()));
    assert_eq!(sl.get_live(b"nope"), None);

    let entries: Vec<_> = sl.iter().collect();
    assert_eq!(entries.len(), 1);
    assert_eq!(entries[0].key, b"only");
}

#[test]
fn test_edge_reverse_order_insert() {
    let sl = ConcurrentSkipList::new();
    for i in (0..100).rev() {
        let key = format!("k{:04}", i);
        let val = format!("v{:04}", i);
        sl.insert(key.as_bytes(), val.as_bytes());
    }

    // Snapshot should yield in sorted order
    let snap = sl.snapshot();
    let entries: Vec<_> = snap.iter().collect();
    assert_eq!(entries.len(), 100);
    for (i, entry) in entries.iter().enumerate() {
        let expected = format!("k{:04}", i);
        assert_eq!(entry.key, expected.as_bytes());
    }
}

#[test]
fn test_concurrent_same_key_insert() {
    let sl = Arc::new(ConcurrentSkipList::with_shards(16));
    let mut handles = vec![];

    // All threads try to insert the same key
    for _ in 0..8 {
        let sl: Arc<ConcurrentSkipList> = Arc::clone(&sl);
        handles.push(thread::spawn(move || {
            sl.insert(b"same_key", b"value");
        }));
    }

    // Only one should succeed. Verify the key exists and wasn't corrupted.
    for h in handles {
        h.join().unwrap();
    }

    match sl.get(b"same_key") {
        Some((v, false)) => assert_eq!(v, b"value"),
        other => panic!("expected value, got {:?}", other),
    }
}

#[test]
fn test_delete_then_reinsert_different_thread() {
    let sl = Arc::new(ConcurrentSkipList::new());
    sl.insert(b"key", b"v1");

    let sl2: Arc<ConcurrentSkipList> = Arc::clone(&sl);
    let h = thread::spawn(move || {
        sl2.delete(b"key");
    });
    h.join().unwrap();

    // Should be tombstoned
    match sl.get(b"key") {
        Some((_, true)) => {}
        _other => panic!("expected tombstone"),
    }
    assert_eq!(sl.get_live(b"key"), None);
}

#[test]
fn test_seal_full_lifecycle_with_concurrent_writers() {
    // Simulate a realistic LSM memtable lifecycle:
    // 1. Concurrent writers fill the memtable
    // 2. Seal it
    // 3. Concurrent readers iterate the frozen memtable
    // 4. Fresh memtable gets new writes
    let sl = Arc::new(ConcurrentSkipList::with_shards(8));

    // Phase 1: Concurrent writes
    let mut handles = vec![];
    for tid in 0..4 {
        let sl = Arc::clone(&sl);
        handles.push(thread::spawn(move || {
            for i in 0..1000 {
                let key = format!("t{}_k{:04}", tid, i);
                let val = format!("v{}_{}", tid, i);
                sl.insert(key.as_bytes(), val.as_bytes());
            }
        }));
    }
    for h in handles {
        h.join().unwrap();
    }

    assert_eq!(sl.len(), 4000);

    // Phase 2: Seal (Arc::try_unwrap ensures exclusive access)
    let sl = Arc::try_unwrap(sl).unwrap();
    let (frozen, fresh) = sl.seal().unwrap();

    assert_eq!(frozen.len(), 4000);
    assert!(fresh.is_empty());

    // Phase 3: Concurrent readers on frozen memtable
    let frozen = Arc::new(frozen);
    let mut readers = vec![];
    for _ in 0..4 {
        let frozen = Arc::clone(&frozen);
        readers.push(thread::spawn(move || {
            let mut count = 0;
            for entry in frozen.iter() {
                if !entry.is_tombstone {
                    count += 1;
                }
            }
            count
        }));
    }

    let mut total = 0;
    for r in readers {
        total += r.join().unwrap();
    }
    assert_eq!(total, 4000 * 4); // 4 readers, each sees 4000

    // Phase 4: Fresh memtable accepts writes
    let fresh = Arc::new(fresh);
    let mut writers = vec![];
    for tid in 0..4 {
        let fresh = Arc::clone(&fresh);
        writers.push(thread::spawn(move || {
            for i in 0..500 {
                let key = format!("new_t{}_k{:04}", tid, i);
                fresh.insert(key.as_bytes(), b"val");
            }
        }));
    }
    for w in writers {
        w.join().unwrap();
    }
    assert_eq!(fresh.len(), 2000);

    // Phase 5: Drop frozen, verify memory reclaimed
    let mem_before = frozen.memory_usage();
    assert!(mem_before > 0);
    std::mem::drop(frozen);
    // fresh memtable has its own arena
    assert!(fresh.memory_usage() > 0);
}

#[test]
fn test_cursor_concurrent_reads() {
    let sl = Arc::new(ConcurrentSkipList::new());
    for i in 0..100 {
        let key = format!("k{:04}", i);
        let val = format!("v{:04}", i);
        sl.insert(key.as_bytes(), val.as_bytes());
    }

    // Multiple threads using cursors simultaneously
    let mut handles = vec![];
    for _ in 0..4 {
        let sl = Arc::clone(&sl);
        handles.push(thread::spawn(move || {
            let cursor = sl.cursor_at(b"k0050").unwrap();
            let keys: Vec<_> = cursor.map(|e| e.key.to_vec()).collect();
            assert_eq!(keys.len(), 50); // k0050..k0099
            assert_eq!(keys[0], b"k0050");
            assert_eq!(keys[49], b"k0099");
        }));
    }

    for h in handles {
        h.join().unwrap();
    }
}

#[test]
fn test_snapshot_under_concurrent_writes() {
    let sl = Arc::new(ConcurrentSkipList::with_shards(8));

    // Pre-populate
    for i in 0..100 {
        let key = format!("key_{:04}", i);
        sl.insert(key.as_bytes(), b"val");
    }

    // Take snapshot
    let snap = sl.snapshot();

    // Concurrent writers insert more entries
    let mut handles = vec![];
    for tid in 0..4 {
        let sl = Arc::clone(&sl);
        handles.push(thread::spawn(move || {
            for i in 0..500 {
                let key = format!("new_t{}_{:04}", tid, i);
                sl.insert(key.as_bytes(), b"val");
            }
        }));
    }

    // Iterate snapshot while writers are active
    let snap_entries: Vec<_> = snap.iter().map(|e| e.key.to_vec()).collect();
    assert_eq!(snap_entries.len(), 100);

    // Verify no post-snapshot keys leaked
    for key in &snap_entries {
        let s = std::str::from_utf8(key).unwrap();
        assert!(s.starts_with("key_"), "snapshot leaked: {}", s);
    }

    for h in handles {
        h.join().unwrap();
    }

    // Live iterator now sees everything
    assert_eq!(sl.len(), 2100); // 100 + 4*500
}