coordinode-lsm-tree 4.2.0

A K.I.S.S. implementation of log-structured merge trees (LSM-trees/LSMTs) — CoordiNode fork
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
// Copyright (c) 2025-present, fjall-rs
// This source code is licensed under both the Apache 2.0 and MIT License
// (found in the LICENSE-* files in the repository)

use super::{Block, BlockHandle, GlobalTableId};
use crate::{
    Cache, CompressionType, KeyRange, Table, encryption::EncryptionProvider,
    file_accessor::FileAccessor, fs::FsFile, table::block::BlockType, version::run::Ranged,
};
use std::{path::Path, sync::Arc};

#[cfg(feature = "metrics")]
use crate::metrics::Metrics;

/// Returns the bounding key range of a table slice.
///
/// Takes `first().min()` and `last().max()` — no comparison needed because
/// callers pass tables that are already sorted in comparator order (via
/// `push_cmp` / `sort_by_cmp`). Works correctly for any comparator.
#[must_use]
pub fn aggregate_run_key_range(tables: &[Table]) -> KeyRange {
    #[expect(clippy::expect_used, reason = "runs are never empty by definition")]
    let lo = tables.first().expect("run should never be empty");
    #[expect(clippy::expect_used, reason = "runs are never empty by definition")]
    let hi = tables.last().expect("run should never be empty");
    KeyRange::new((lo.key_range().min().clone(), hi.key_range().max().clone()))
}

/// [start, end] slice indexes
#[derive(Debug)]
pub struct SliceIndexes(pub usize, pub usize);

/// Loads a block from disk or block cache, if cached.
///
/// Also handles file descriptor opening and caching.
#[expect(
    clippy::too_many_arguments,
    reason = "block loading requires table id, path, file accessor, cache, handle, block type, compression, and encryption context"
)]
pub fn load_block(
    table_id: GlobalTableId,
    path: &Path,
    file_accessor: &FileAccessor,
    cache: &Cache,
    handle: &BlockHandle,
    block_type: BlockType,
    compression: CompressionType,
    encryption: Option<&dyn EncryptionProvider>,
    #[cfg(zstd_any)] zstd_dict: Option<&crate::compression::ZstdDictionary>,
    #[cfg(feature = "metrics")] metrics: &Metrics,
) -> crate::Result<Block> {
    #[cfg(feature = "metrics")]
    use std::sync::atomic::Ordering::Relaxed;

    log::trace!("load {block_type:?} block {handle:?}");

    if let Some(block) = cache.get_block(table_id, handle.offset()) {
        #[cfg(feature = "metrics")]
        match block_type {
            BlockType::Filter => {
                metrics.filter_block_load_cached.fetch_add(1, Relaxed);
            }
            BlockType::Index => {
                metrics.index_block_load_cached.fetch_add(1, Relaxed);
            }
            BlockType::RangeTombstone => {
                metrics
                    .range_tombstone_block_load_cached
                    .fetch_add(1, Relaxed);
            }
            BlockType::Data | BlockType::Meta => {
                metrics.data_block_load_cached.fetch_add(1, Relaxed);
            }
        }

        return Ok(block);
    }

    let (fd, fd_cache_miss) = if let Some(cached_fd) = file_accessor.access_for_table(&table_id) {
        #[cfg(feature = "metrics")]
        metrics.table_file_opened_cached.fetch_add(1, Relaxed);

        (cached_fd, false)
    } else {
        let file = std::fs::File::open(path)?;

        #[cfg(feature = "metrics")]
        metrics.table_file_opened_uncached.fetch_add(1, Relaxed);

        // The if-branch returns Arc<dyn FsFile> from the descriptor
        // table, so the else-branch needs an explicit type annotation
        // to trigger unsizing coercion.
        let fd: Arc<dyn FsFile> = Arc::new(file);
        (fd, true)
    };

    let block = Block::from_file(
        fd.as_ref(),
        *handle,
        compression,
        encryption,
        #[cfg(zstd_any)]
        zstd_dict,
    )?;

    if block.header.block_type != block_type {
        return Err(crate::Error::InvalidTag((
            "BlockType",
            block.header.block_type.into(),
        )));
    }

    #[cfg(feature = "metrics")]
    match block_type {
        BlockType::Filter => {
            metrics.filter_block_load_io.fetch_add(1, Relaxed);

            metrics
                .filter_block_io_requested
                .fetch_add(handle.size().into(), Relaxed);
        }
        BlockType::Index => {
            metrics.index_block_load_io.fetch_add(1, Relaxed);

            metrics
                .index_block_io_requested
                .fetch_add(handle.size().into(), Relaxed);
        }
        BlockType::RangeTombstone => {
            metrics.range_tombstone_block_load_io.fetch_add(1, Relaxed);

            metrics
                .range_tombstone_block_io_requested
                .fetch_add(handle.size().into(), Relaxed);
        }
        BlockType::Data | BlockType::Meta => {
            metrics.data_block_load_io.fetch_add(1, Relaxed);

            metrics
                .data_block_io_requested
                .fetch_add(handle.size().into(), Relaxed);
        }
    }

    // Cache FD
    if fd_cache_miss {
        file_accessor.insert_for_table(table_id, fd);
    }

    cache.insert_block(table_id, handle.offset(), block.clone());

    Ok(block)
}

#[must_use]
pub fn longest_shared_prefix_length(s1: &[u8], s2: &[u8]) -> usize {
    s1.iter()
        .zip(s2.iter())
        .take_while(|(c1, c2)| c1 == c2)
        .count()
}

/// Compares the conceptual concatenation `prefix + suffix` against `needle`
/// using the given comparator.
///
/// For the default lexicographic comparator this performs a zero-allocation
/// bytewise comparison. Custom comparators fall back to concatenating prefix
/// and suffix into a temporary `Vec` so that `UserComparator::compare` always
/// receives a complete key.
#[must_use]
pub fn compare_prefixed_slice(
    prefix: &[u8],
    suffix: &[u8],
    needle: &[u8],
    cmp: &dyn crate::comparator::UserComparator,
) -> std::cmp::Ordering {
    // Fast path: zero-allocation bytewise comparison for the default
    // (lexicographic) comparator. This is the hot path for block index
    // and data block binary searches.
    if cmp.is_lexicographic() {
        return compare_prefixed_slice_lexicographic(prefix, suffix, needle);
    }

    // Slow path: materialize prefix+suffix into a contiguous buffer for
    // custom comparators. Uses a stack buffer for typical key sizes to
    // avoid heap allocation on the hot binary-search path.
    let total_len = prefix.len() + suffix.len();

    if total_len <= 256 {
        let mut buf = [0_u8; 256];

        // SAFETY (indexing): total_len <= 256 == buf.len(), and
        // prefix.len() + suffix.len() == total_len, so all slices are in bounds.
        #[expect(clippy::indexing_slicing, reason = "total_len <= 256 checked above")]
        {
            buf[..prefix.len()].copy_from_slice(prefix);
            buf[prefix.len()..total_len].copy_from_slice(suffix);
        }

        #[expect(clippy::indexing_slicing, reason = "total_len <= 256 checked above")]
        return cmp.compare(&buf[..total_len], needle);
    }

    // Fallback for unusually large keys: allocate a temporary Vec.
    let mut full_key = Vec::with_capacity(total_len);
    full_key.extend_from_slice(prefix);
    full_key.extend_from_slice(suffix);
    cmp.compare(&full_key, needle)
}

/// Zero-allocation lexicographic comparison of `prefix + suffix` against `needle`.
#[must_use]
fn compare_prefixed_slice_lexicographic(
    prefix: &[u8],
    suffix: &[u8],
    needle: &[u8],
) -> std::cmp::Ordering {
    use std::cmp::Ordering::{Equal, Greater};

    if needle.is_empty() {
        let combined_len = prefix.len() + suffix.len();
        return if combined_len > 0 { Greater } else { Equal };
    }

    let max_pfx_len = prefix.len().min(needle.len());

    {
        // SAFETY: max_pfx_len = min(prefix.len(), needle.len()), so both
        // slices [0..max_pfx_len] are within bounds by construction.
        #[expect(
            unsafe_code,
            reason = "max_pfx_len <= prefix.len() && max_pfx_len <= needle.len()"
        )]
        let pfx = unsafe { prefix.get_unchecked(0..max_pfx_len) };

        #[expect(
            unsafe_code,
            reason = "max_pfx_len <= prefix.len() && max_pfx_len <= needle.len()"
        )]
        let ndl = unsafe { needle.get_unchecked(0..max_pfx_len) };

        match pfx.cmp(ndl) {
            Equal => {}
            ordering => return ordering,
        }
    }

    let rest_len = prefix.len().saturating_sub(needle.len());
    if rest_len > 0 {
        return Greater;
    }

    // SAFETY: rest_len == 0 means prefix.len() <= needle.len(), so
    // max_pfx_len == prefix.len() <= needle.len() and needle[max_pfx_len..] is in-bounds.
    #[expect(
        unsafe_code,
        reason = "max_pfx_len <= needle.len() guaranteed by rest_len == 0 guard above"
    )]
    let remaining_needle = unsafe { needle.get_unchecked(max_pfx_len..) };
    suffix.cmp(remaining_needle)
}

#[cfg(test)]
mod tests {
    use super::*;
    use crate::comparator::DefaultUserComparator;
    use test_log::test;

    #[test]
    fn test_longest_shared_prefix_length() {
        assert_eq!(3, longest_shared_prefix_length(b"abc", b"abc"));
        assert_eq!(1, longest_shared_prefix_length(b"abc", b"a"));
        assert_eq!(1, longest_shared_prefix_length(b"a", b"abc"));
        assert_eq!(0, longest_shared_prefix_length(b"abc", b""));
        assert_eq!(0, longest_shared_prefix_length(b"", b"abc"));
        assert_eq!(0, longest_shared_prefix_length(b"", b""));
        assert_eq!(0, longest_shared_prefix_length(b"", b""));
        assert_eq!(0, longest_shared_prefix_length(b"abc", b"def"));
        assert_eq!(1, longest_shared_prefix_length(b"abc", b"acc"));
    }

    #[test]
    fn test_compare_prefixed_slice() {
        use std::cmp::Ordering::{Equal, Greater, Less};

        assert_eq!(
            Greater,
            compare_prefixed_slice(&[0, 161], &[], &[0], &DefaultUserComparator)
        );

        assert_eq!(
            Equal,
            compare_prefixed_slice(b"abc", b"xyz", b"abcxyz", &DefaultUserComparator)
        );
        assert_eq!(
            Equal,
            compare_prefixed_slice(b"abc", b"", b"abc", &DefaultUserComparator)
        );
        assert_eq!(
            Equal,
            compare_prefixed_slice(b"abc", b"abc", b"abcabc", &DefaultUserComparator)
        );
        assert_eq!(
            Equal,
            compare_prefixed_slice(b"", b"", b"", &DefaultUserComparator)
        );
        assert_eq!(
            Less,
            compare_prefixed_slice(b"a", b"", b"y", &DefaultUserComparator)
        );
        assert_eq!(
            Less,
            compare_prefixed_slice(b"a", b"", b"yyy", &DefaultUserComparator)
        );
        assert_eq!(
            Less,
            compare_prefixed_slice(b"a", b"", b"yyy", &DefaultUserComparator)
        );
        assert_eq!(
            Less,
            compare_prefixed_slice(b"yyyy", b"a", b"yyyyb", &DefaultUserComparator)
        );
        assert_eq!(
            Less,
            compare_prefixed_slice(b"yyy", b"b", b"yyyyb", &DefaultUserComparator)
        );
        assert_eq!(
            Less,
            compare_prefixed_slice(b"abc", b"d", b"abce", &DefaultUserComparator)
        );
        assert_eq!(
            Less,
            compare_prefixed_slice(b"ab", b"", b"ac", &DefaultUserComparator)
        );
        assert_eq!(
            Greater,
            compare_prefixed_slice(b"a", b"", b"", &DefaultUserComparator)
        );
        assert_eq!(
            Greater,
            compare_prefixed_slice(b"", b"a", b"", &DefaultUserComparator)
        );
        assert_eq!(
            Greater,
            compare_prefixed_slice(b"a", b"a", b"", &DefaultUserComparator)
        );
        assert_eq!(
            Greater,
            compare_prefixed_slice(b"b", b"a", b"a", &DefaultUserComparator)
        );
        assert_eq!(
            Greater,
            compare_prefixed_slice(b"a", b"b", b"a", &DefaultUserComparator)
        );
        assert_eq!(
            Greater,
            compare_prefixed_slice(b"abc", b"xy", b"abcw", &DefaultUserComparator)
        );
        assert_eq!(
            Greater,
            compare_prefixed_slice(b"ab", b"cde", b"a", &DefaultUserComparator)
        );
        assert_eq!(
            Greater,
            compare_prefixed_slice(b"abcd", b"zz", b"abc", &DefaultUserComparator)
        );
        assert_eq!(
            Greater,
            compare_prefixed_slice(b"abc", b"d", b"abc", &DefaultUserComparator)
        );
        assert_eq!(
            Greater,
            compare_prefixed_slice(b"aaaa", b"aaab", b"aaaaaaaa", &DefaultUserComparator)
        );
        assert_eq!(
            Greater,
            compare_prefixed_slice(b"aaaa", b"aaba", b"aaaaaaaa", &DefaultUserComparator)
        );
        assert_eq!(
            Greater,
            compare_prefixed_slice(b"abcd", b"x", b"abc", &DefaultUserComparator)
        );

        assert_eq!(
            Less,
            compare_prefixed_slice(&[0x7F], &[], &[0x80], &DefaultUserComparator)
        );
        assert_eq!(
            Greater,
            compare_prefixed_slice(&[0xFF], &[], &[0x10], &DefaultUserComparator)
        );
    }

    /// Reverse comparator to exercise the Vec-allocation slow path.
    struct ReverseComparator;
    impl crate::comparator::UserComparator for ReverseComparator {
        fn name(&self) -> &'static str {
            "test-reverse"
        }

        fn compare(&self, a: &[u8], b: &[u8]) -> std::cmp::Ordering {
            b.cmp(a)
        }
    }

    #[test]
    fn test_compare_prefixed_slice_custom_comparator() {
        use std::cmp::Ordering::{Equal, Greater, Less};

        use crate::comparator::UserComparator as _;
        assert_eq!(ReverseComparator.name(), "test-reverse");

        // With reverse comparator, "abc" > "xyz" (reversed)
        assert_eq!(
            Greater,
            compare_prefixed_slice(b"ab", b"c", b"xyz", &ReverseComparator)
        );
        assert_eq!(
            Less,
            compare_prefixed_slice(b"xy", b"z", b"abc", &ReverseComparator)
        );
        assert_eq!(
            Equal,
            compare_prefixed_slice(b"ab", b"c", b"abc", &ReverseComparator)
        );
        // Empty cases
        assert_eq!(
            Equal,
            compare_prefixed_slice(b"", b"", b"", &ReverseComparator)
        );
        assert_eq!(
            Less, // reversed: non-empty > empty
            compare_prefixed_slice(b"a", b"", b"", &ReverseComparator)
        );
    }
}