Line data Source code
1 : // Copyright 2021 The LevelDB-Go and Pebble Authors. All rights reserved. Use
2 : // of this source code is governed by a BSD-style license that can be found in
3 : // the LICENSE file.
4 :
5 : package pebble
6 :
7 : import (
8 : "context"
9 :
10 : "github.com/cockroachdb/errors"
11 : "github.com/cockroachdb/pebble/internal/base"
12 : "github.com/cockroachdb/pebble/internal/invariants"
13 : "github.com/cockroachdb/pebble/internal/keyspan"
14 : "github.com/cockroachdb/pebble/internal/manifest"
15 : "github.com/cockroachdb/pebble/sstable"
16 : )
17 :
18 : // constructRangeKeyIter constructs the range-key iterator stack, populating
19 : // i.rangeKey.rangeKeyIter with the resulting iterator.
20 1 : func (i *Iterator) constructRangeKeyIter() {
21 1 : i.rangeKey.rangeKeyIter = i.rangeKey.iterConfig.Init(
22 1 : &i.comparer, i.seqNum, i.opts.LowerBound, i.opts.UpperBound,
23 1 : &i.hasPrefix, &i.prefixOrFullSeekKey, false /* internalKeys */, &i.rangeKey.rangeKeyBuffers.internal)
24 1 :
25 1 : if i.opts.DebugRangeKeyStack {
26 0 : // The default logger is preferable to i.opts.getLogger(), at least in the
27 0 : // metamorphic test.
28 0 : i.rangeKey.rangeKeyIter = keyspan.InjectLogging(i.rangeKey.rangeKeyIter, base.DefaultLogger)
29 0 : }
30 :
31 : // If there's an indexed batch with range keys, include it.
32 1 : if i.batch != nil {
33 1 : if i.batch.index == nil {
34 0 : // This isn't an indexed batch. We shouldn't have gotten this far.
35 0 : panic(errors.AssertionFailedf("creating an iterator over an unindexed batch"))
36 1 : } else {
37 1 : // Only include the batch's range key iterator if it has any keys.
38 1 : // NB: This can force reconstruction of the rangekey iterator stack
39 1 : // in SetOptions if subsequently range keys are added. See
40 1 : // SetOptions.
41 1 : if i.batch.countRangeKeys > 0 {
42 1 : i.batch.initRangeKeyIter(&i.opts, &i.batchRangeKeyIter, i.batchSeqNum)
43 1 : i.rangeKey.iterConfig.AddLevel(&i.batchRangeKeyIter)
44 1 : }
45 : }
46 : }
47 :
48 1 : if !i.batchOnlyIter {
49 1 : // Next are the flushables: memtables and large batches.
50 1 : if i.readState != nil {
51 1 : for j := len(i.readState.memtables) - 1; j >= 0; j-- {
52 1 : mem := i.readState.memtables[j]
53 1 : // We only need to read from memtables which contain sequence numbers older
54 1 : // than seqNum.
55 1 : if logSeqNum := mem.logSeqNum; logSeqNum >= i.seqNum {
56 1 : continue
57 : }
58 1 : if rki := mem.newRangeKeyIter(&i.opts); rki != nil {
59 1 : i.rangeKey.iterConfig.AddLevel(rki)
60 1 : }
61 : }
62 : }
63 :
64 1 : current := i.version
65 1 : if current == nil {
66 1 : current = i.readState.current
67 1 : }
68 : // Next are the file levels: L0 sub-levels followed by lower levels.
69 :
70 : // Add file-specific iterators for L0 files containing range keys. We
71 : // maintain a separate manifest.LevelMetadata for each level containing only
72 : // files that contain range keys, however we don't compute a separate
73 : // L0Sublevels data structure too.
74 : //
75 : // We first use L0's LevelMetadata to peek and see whether L0 contains any
76 : // range keys at all. If it does, we create a range key level iterator per
77 : // level that contains range keys using the information from L0Sublevels.
78 : // Some sublevels may not contain any range keys, and we need to iterate
79 : // through the fileMetadata to determine that. Since L0's file count should
80 : // not significantly exceed ~1000 files (see L0CompactionFileThreshold),
81 : // this should be okay.
82 1 : if !current.RangeKeyLevels[0].Empty() {
83 1 : // L0 contains at least 1 file containing range keys.
84 1 : // Add level iterators for the L0 sublevels, iterating from newest to
85 1 : // oldest.
86 1 : for j := len(current.L0SublevelFiles) - 1; j >= 0; j-- {
87 1 : iter := current.L0SublevelFiles[j].Iter()
88 1 : if !containsAnyRangeKeys(iter) {
89 1 : continue
90 : }
91 :
92 1 : li := i.rangeKey.iterConfig.NewLevelIter()
93 1 : li.Init(
94 1 : i.opts.SpanIterOptions(),
95 1 : i.cmp,
96 1 : i.newIterRangeKey,
97 1 : iter.Filter(manifest.KeyTypeRange),
98 1 : manifest.L0Sublevel(j),
99 1 : manifest.KeyTypeRange,
100 1 : )
101 1 : i.rangeKey.iterConfig.AddLevel(li)
102 : }
103 : }
104 :
105 : // Add level iterators for the non-empty non-L0 levels.
106 1 : for level := 1; level < len(current.RangeKeyLevels); level++ {
107 1 : if current.RangeKeyLevels[level].Empty() {
108 1 : continue
109 : }
110 1 : li := i.rangeKey.iterConfig.NewLevelIter()
111 1 : spanIterOpts := i.opts.SpanIterOptions()
112 1 : li.Init(spanIterOpts, i.cmp, i.newIterRangeKey, current.RangeKeyLevels[level].Iter(),
113 1 : manifest.Level(level), manifest.KeyTypeRange)
114 1 : i.rangeKey.iterConfig.AddLevel(li)
115 : }
116 : }
117 : }
118 :
119 1 : func containsAnyRangeKeys(iter manifest.LevelIterator) bool {
120 1 : for f := iter.First(); f != nil; f = iter.Next() {
121 1 : if f.HasRangeKeys {
122 1 : return true
123 1 : }
124 : }
125 1 : return false
126 : }
127 :
128 : // Range key masking
129 : //
130 : // Pebble iterators may be configured such that range keys with suffixes mask
131 : // point keys with lower suffixes. The intended use is implementing a MVCC
132 : // delete range operation using range keys, when suffixes are MVCC timestamps.
133 : //
134 : // To enable masking, the user populates the IterOptions's RangeKeyMasking
135 : // field. The Suffix field configures which range keys act as masks. The
136 : // intended use is to hold a MVCC read timestamp. When implementing a MVCC
137 : // delete range operation, only range keys that are visible at the read
138 : // timestamp should be visible. If a range key has a suffix ≤
139 : // RangeKeyMasking.Suffix, it acts as a mask.
140 : //
141 : // Range key masking is facilitated by the keyspan.InterleavingIter. The
142 : // interleaving iterator interleaves range keys and point keys during combined
143 : // iteration. During user iteration, the interleaving iterator is configured
144 : // with a keyspan.SpanMask, implemented by the rangeKeyMasking struct below.
145 : // The SpanMask interface defines two methods: SpanChanged and SkipPoint.
146 : //
147 : // SpanChanged is used to keep the current mask up-to-date. Whenever the point
148 : // iterator has stepped into or out of the bounds of a range key, the
149 : // interleaving iterator invokes SpanChanged passing the current covering range
150 : // key. The below rangeKeyMasking implementation scans the range keys looking
151 : // for the range key with the largest suffix that's still ≤ the suffix supplied
152 : // to IterOptions.RangeKeyMasking.Suffix (the "read timestamp"). If it finds a
153 : // range key that meets the condition, the range key should act as a mask. The
154 : // span and the relevant range key's suffix are saved.
155 : //
156 : // The above ensures that `rangeKeyMasking.maskActiveSuffix` always contains the
157 : // current masking suffix such that any point keys with lower suffixes should be
158 : // skipped.
159 : //
160 : // There are two ways in which masked point keys are skipped.
161 : //
162 : // 1. Interleaving iterator SkipPoint
163 : //
164 : // Whenever the interleaving iterator encounters a point key that falls within
165 : // the bounds of a range key, it invokes SkipPoint. The interleaving iterator
166 : // guarantees that the SpanChanged method described above has already been
167 : // invoked with the covering range key. The below rangeKeyMasking implementation
168 : // of SkipPoint splits the key into prefix and suffix, compares the suffix to
169 : // the `maskActiveSuffix` updated by SpanChanged and returns true if
170 : // suffix(point) < maskActiveSuffix.
171 : //
172 : // The SkipPoint logic is sufficient to ensure that the Pebble iterator filters
173 : // out all masked point keys. However, it requires the iterator read each masked
174 : // point key. For broad range keys that mask many points, this may be expensive.
175 : //
176 : // 2. Block property filter
177 : //
178 : // For more efficient handling of braad range keys that mask many points, the
179 : // IterOptions.RangeKeyMasking field has an optional Filter option. This Filter
180 : // field takes a superset of the block-property filter interface, adding a
181 : // method to dynamically configure the filter's filtering criteria.
182 : //
183 : // To make use of the Filter option, the user is required to define and
184 : // configure a block-property collector that collects a property containing at
185 : // least the maximum suffix of a key within a block.
186 : //
187 : // When the SpanChanged method described above is invoked, rangeKeyMasking also
188 : // reconfigures the user-provided filter. It invokes a SetSuffix method,
189 : // providing the `maskActiveSuffix`, requesting that from now on the
190 : // block-property filter return Intersects()=false for any properties indicating
191 : // that a block contains exclusively keys with suffixes greater than the
192 : // provided suffix.
193 : //
194 : // Note that unlike other block-property filters, the filter used for masking
195 : // must not apply across the entire keyspace. It must only filter blocks that
196 : // lie within the bounds of the range key that set the mask suffix. To
197 : // accommodate this, rangeKeyMasking implements a special interface:
198 : // sstable.BoundLimitedBlockPropertyFilter. This interface extends the block
199 : // property filter interface with two new methods: KeyIsWithinLowerBound and
200 : // KeyIsWithinUpperBound. The rangeKeyMasking type wraps the user-provided block
201 : // property filter, implementing these two methods and overriding Intersects to
202 : // always return true if there is no active mask.
203 : //
204 : // The logic to ensure that a mask block-property filter is only applied within
205 : // the bounds of the masking range key is subtle. The interleaving iterator
206 : // guarantees that it never invokes SpanChanged until the point iterator is
207 : // positioned within the range key. During forward iteration, this guarantees
208 : // that any block that a sstable reader might attempt to load contains only keys
209 : // greater than or equal to the range key's lower bound. During backward
210 : // iteration, it provides the analagous guarantee on the range key's upper
211 : // bound.
212 : //
213 : // The above ensures that an sstable reader only needs to verify that a block
214 : // that it skips meets the opposite bound. This is where the
215 : // KeyIsWithinLowerBound and KeyIsWithinUpperBound methods are used. When an
216 : // sstable iterator is configured with a BoundLimitedBlockPropertyFilter, it
217 : // checks for intersection with the block-property filter before every block
218 : // load, like ordinary block-property filters. However, if the bound-limited
219 : // block property filter indicates that it does NOT intersect, the filter's
220 : // relevant KeyIsWithin{Lower,Upper}Bound method is queried, using a block
221 : // index separator as the bound. If the method indicates that the provided index
222 : // separator does not fall within the range key bounds, the no-intersection
223 : // result is ignored, and the block is read.
224 :
225 : type rangeKeyMasking struct {
226 : cmp base.Compare
227 : split base.Split
228 : filter BlockPropertyFilterMask
229 : // maskActiveSuffix holds the suffix of a range key currently acting as a
230 : // mask, hiding point keys with suffixes greater than it. maskActiveSuffix
231 : // is only ever non-nil if IterOptions.RangeKeyMasking.Suffix is non-nil.
232 : // maskActiveSuffix is updated whenever the iterator passes over a new range
233 : // key. The maskActiveSuffix should only be used if maskSpan is non-nil.
234 : //
235 : // See SpanChanged.
236 : maskActiveSuffix []byte
237 : // maskSpan holds the span from which the active mask suffix was extracted.
238 : // The span is used for bounds comparisons, to ensure that a range-key mask
239 : // is not applied beyond the bounds of the range key.
240 : maskSpan *keyspan.Span
241 : parent *Iterator
242 : }
243 :
244 1 : func (m *rangeKeyMasking) init(parent *Iterator, cmp base.Compare, split base.Split) {
245 1 : m.cmp = cmp
246 1 : m.split = split
247 1 : if parent.opts.RangeKeyMasking.Filter != nil {
248 1 : m.filter = parent.opts.RangeKeyMasking.Filter()
249 1 : }
250 1 : m.parent = parent
251 : }
252 :
253 : // SpanChanged implements the keyspan.SpanMask interface, used during range key
254 : // iteration.
255 1 : func (m *rangeKeyMasking) SpanChanged(s *keyspan.Span) {
256 1 : if s == nil && m.maskSpan == nil {
257 1 : return
258 1 : }
259 1 : m.maskSpan = nil
260 1 : m.maskActiveSuffix = m.maskActiveSuffix[:0]
261 1 :
262 1 : // Find the smallest suffix of a range key contained within the Span,
263 1 : // excluding suffixes less than m.opts.RangeKeyMasking.Suffix.
264 1 : if s != nil {
265 1 : m.parent.rangeKey.stale = true
266 1 : if m.parent.opts.RangeKeyMasking.Suffix != nil {
267 1 : for j := range s.Keys {
268 1 : if s.Keys[j].Suffix == nil {
269 0 : continue
270 : }
271 1 : if m.cmp(s.Keys[j].Suffix, m.parent.opts.RangeKeyMasking.Suffix) < 0 {
272 1 : continue
273 : }
274 1 : if len(m.maskActiveSuffix) == 0 || m.cmp(m.maskActiveSuffix, s.Keys[j].Suffix) > 0 {
275 1 : m.maskSpan = s
276 1 : m.maskActiveSuffix = append(m.maskActiveSuffix[:0], s.Keys[j].Suffix...)
277 1 : }
278 : }
279 : }
280 : }
281 :
282 1 : if m.maskSpan != nil && m.parent.opts.RangeKeyMasking.Filter != nil {
283 1 : // Update the block-property filter to filter point keys with suffixes
284 1 : // greater than m.maskActiveSuffix.
285 1 : err := m.filter.SetSuffix(m.maskActiveSuffix)
286 1 : if err != nil {
287 0 : m.parent.err = err
288 0 : }
289 : }
290 : // If no span is active, we leave the inner block-property filter configured
291 : // with its existing suffix. That's okay, because Intersects calls are first
292 : // evaluated by iteratorRangeKeyState.Intersects, which considers all blocks
293 : // as intersecting if there's no active mask.
294 : }
295 :
296 : // SkipPoint implements the keyspan.SpanMask interface, used during range key
297 : // iteration. Whenever a point key is covered by a non-empty Span, the
298 : // interleaving iterator invokes SkipPoint. This function is responsible for
299 : // performing range key masking.
300 : //
301 : // If a non-nil IterOptions.RangeKeyMasking.Suffix is set, range key masking is
302 : // enabled. Masking hides point keys, transparently skipping over the keys.
303 : // Whether or not a point key is masked is determined by comparing the point
304 : // key's suffix, the overlapping span's keys' suffixes, and the user-configured
305 : // IterOption's RangeKeyMasking.Suffix. When configured with a masking threshold
306 : // _t_, and there exists a span with suffix _r_ covering a point key with suffix
307 : // _p_, and
308 : //
309 : // _t_ ≤ _r_ < _p_
310 : //
311 : // then the point key is elided. Consider the following rendering, where using
312 : // integer suffixes with higher integers sort before suffixes with lower
313 : // integers, (for example @7 ≤ @6 < @5):
314 : //
315 : // ^
316 : // @9 | •―――――――――――――――○ [e,m)@9
317 : // s 8 | • l@8
318 : // u 7 |------------------------------------ @7 RangeKeyMasking.Suffix
319 : // f 6 | [h,q)@6 •―――――――――――――――――○ (threshold)
320 : // f 5 | • h@5
321 : // f 4 | • n@4
322 : // i 3 | •―――――――――――○ [f,l)@3
323 : // x 2 | • b@2
324 : // 1 |
325 : // 0 |___________________________________
326 : // a b c d e f g h i j k l m n o p q
327 : //
328 : // An iterator scanning the entire keyspace with the masking threshold set to @7
329 : // will observe point keys b@2 and l@8. The span keys [h,q)@6 and [f,l)@3 serve
330 : // as masks, because cmp(@6,@7) ≥ 0 and cmp(@3,@7) ≥ 0. The span key [e,m)@9
331 : // does not serve as a mask, because cmp(@9,@7) < 0.
332 : //
333 : // Although point l@8 falls within the user key bounds of [e,m)@9, [e,m)@9 is
334 : // non-masking due to its suffix. The point key l@8 also falls within the user
335 : // key bounds of [h,q)@6, but since cmp(@6,@8) ≥ 0, l@8 is unmasked.
336 : //
337 : // Invariant: The userKey is within the user key bounds of the span most
338 : // recently provided to `SpanChanged`.
339 1 : func (m *rangeKeyMasking) SkipPoint(userKey []byte) bool {
340 1 : m.parent.stats.RangeKeyStats.ContainedPoints++
341 1 : if m.maskSpan == nil {
342 1 : // No range key is currently acting as a mask, so don't skip.
343 1 : return false
344 1 : }
345 : // Range key masking is enabled and the current span includes a range key
346 : // that is being used as a mask. (NB: SpanChanged already verified that the
347 : // range key's suffix is ≥ RangeKeyMasking.Suffix).
348 : //
349 : // This point key falls within the bounds of the range key (guaranteed by
350 : // the InterleavingIter). Skip the point key if the range key's suffix is
351 : // greater than the point key's suffix.
352 1 : pointSuffix := userKey[m.split(userKey):]
353 1 : if len(pointSuffix) > 0 && m.cmp(m.maskActiveSuffix, pointSuffix) < 0 {
354 1 : m.parent.stats.RangeKeyStats.SkippedPoints++
355 1 : return true
356 1 : }
357 1 : return false
358 : }
359 :
360 : // The iteratorRangeKeyState type implements the sstable package's
361 : // BoundLimitedBlockPropertyFilter interface in order to use block property
362 : // filters for range key masking. The iteratorRangeKeyState implementation wraps
363 : // the block-property filter provided in Options.RangeKeyMasking.Filter.
364 : //
365 : // Using a block-property filter for range-key masking requires limiting the
366 : // filter's effect to the bounds of the range key currently acting as a mask.
367 : // Consider the range key [a,m)@10, and an iterator positioned just before the
368 : // below block, bounded by index separators `c` and `z`:
369 : //
370 : // c z
371 : // x | c@9 c@5 c@1 d@7 e@4 y@4 | ...
372 : // iter pos
373 : //
374 : // The next block cannot be skipped, despite the range key suffix @10 is greater
375 : // than all the block's keys' suffixes, because it contains a key (y@4) outside
376 : // the bounds of the range key.
377 : //
378 : // This extended BoundLimitedBlockPropertyFilter interface adds two new methods,
379 : // KeyIsWithinLowerBound and KeyIsWithinUpperBound, for testing whether a
380 : // particular block is within bounds.
381 : //
382 : // The iteratorRangeKeyState implements these new methods by first checking if
383 : // the iterator is currently positioned within a range key. If not, the provided
384 : // key is considered out-of-bounds. If the iterator is positioned within a range
385 : // key, it compares the corresponding range key bound.
386 : var _ sstable.BoundLimitedBlockPropertyFilter = (*rangeKeyMasking)(nil)
387 :
388 : // Name implements the limitedBlockPropertyFilter interface defined in the
389 : // sstable package by passing through to the user-defined block property filter.
390 1 : func (m *rangeKeyMasking) Name() string {
391 1 : return m.filter.Name()
392 1 : }
393 :
394 : // Intersects implements the limitedBlockPropertyFilter interface defined in the
395 : // sstable package by passing the intersection decision to the user-provided
396 : // block property filter only if a range key is covering the current iterator
397 : // position.
398 1 : func (m *rangeKeyMasking) Intersects(prop []byte) (bool, error) {
399 1 : if m.maskSpan == nil {
400 1 : // No span is actively masking.
401 1 : return true, nil
402 1 : }
403 1 : return m.filter.Intersects(prop)
404 : }
405 :
406 1 : func (m *rangeKeyMasking) SyntheticSuffixIntersects(prop []byte, suffix []byte) (bool, error) {
407 1 : if m.maskSpan == nil {
408 1 : // No span is actively masking.
409 1 : return true, nil
410 1 : }
411 1 : return m.filter.SyntheticSuffixIntersects(prop, suffix)
412 : }
413 :
414 : // KeyIsWithinLowerBound implements the limitedBlockPropertyFilter interface
415 : // defined in the sstable package. It's used to restrict the masking block
416 : // property filter to only applying within the bounds of the active range key.
417 1 : func (m *rangeKeyMasking) KeyIsWithinLowerBound(key []byte) bool {
418 1 : // Invariant: m.maskSpan != nil
419 1 : //
420 1 : // The provided `key` is an inclusive lower bound of the block we're
421 1 : // considering skipping.
422 1 : return m.cmp(m.maskSpan.Start, key) <= 0
423 1 : }
424 :
425 : // KeyIsWithinUpperBound implements the limitedBlockPropertyFilter interface
426 : // defined in the sstable package. It's used to restrict the masking block
427 : // property filter to only applying within the bounds of the active range key.
428 1 : func (m *rangeKeyMasking) KeyIsWithinUpperBound(key []byte) bool {
429 1 : // Invariant: m.maskSpan != nil
430 1 : //
431 1 : // The provided `key` is an *inclusive* upper bound of the block we're
432 1 : // considering skipping, so the range key's end must be strictly greater
433 1 : // than the block bound for the block to be within bounds.
434 1 : return m.cmp(m.maskSpan.End, key) > 0
435 1 : }
436 :
437 : // lazyCombinedIter implements the internalIterator interface, wrapping a
438 : // pointIter. It requires the pointIter's the levelIters be configured with
439 : // pointers to its combinedIterState. When the levelIter observes a file
440 : // containing a range key, the lazyCombinedIter constructs the combined
441 : // range+point key iterator stack and switches to it.
442 : type lazyCombinedIter struct {
443 : // parent holds a pointer to the root *pebble.Iterator containing this
444 : // iterator. It's used to mutate the internalIterator in use when switching
445 : // to combined iteration.
446 : parent *Iterator
447 : pointIter internalIterator
448 : combinedIterState combinedIterState
449 : }
450 :
451 : // combinedIterState encapsulates the current state of combined iteration.
452 : // Various low-level iterators (mergingIter, leveliter) hold pointers to the
453 : // *pebble.Iterator's combinedIterState. This allows them to check whether or
454 : // not they must monitor for files containing range keys (!initialized), or not.
455 : //
456 : // When !initialized, low-level iterators watch for files containing range keys.
457 : // When one is discovered, they set triggered=true and key to the smallest
458 : // (forward direction) or largest (reverse direction) range key that's been
459 : // observed.
460 : type combinedIterState struct {
461 : // key holds the smallest (forward direction) or largest (backward
462 : // direction) user key from a range key bound discovered during the iterator
463 : // operation that triggered the switch to combined iteration.
464 : //
465 : // Slices stored here must be stable. This is possible because callers pass
466 : // a Smallest/Largest bound from a fileMetadata, which are immutable. A key
467 : // slice's bytes must not be overwritten.
468 : key []byte
469 : triggered bool
470 : initialized bool
471 : }
472 :
473 : // Assert that *lazyCombinedIter implements internalIterator.
474 : var _ internalIterator = (*lazyCombinedIter)(nil)
475 :
476 : // initCombinedIteration is invoked after a pointIter positioning operation
477 : // resulted in i.combinedIterState.triggered=true.
478 : //
479 : // The `dir` parameter is `+1` or `-1` indicating forward iteration or backward
480 : // iteration respectively.
481 : //
482 : // The `pointKey` and `pointValue` parameters provide the new point key-value
483 : // pair that the iterator was just positioned to. The combined iterator should
484 : // be seeded with this point key-value pair and return the smaller (forward
485 : // iteration) or largest (backward iteration) of the two.
486 : //
487 : // The `seekKey` parameter is non-nil only if the iterator operation that
488 : // triggered the switch to combined iteration was a SeekGE, SeekPrefixGE or
489 : // SeekLT. It provides the seek key supplied and is used to seek the range-key
490 : // iterator using the same key. This is necessary for SeekGE/SeekPrefixGE
491 : // operations that land in the middle of a range key and must truncate to the
492 : // user-provided seek key.
493 : func (i *lazyCombinedIter) initCombinedIteration(
494 : dir int8, pointKey *InternalKey, pointValue base.LazyValue, seekKey []byte,
495 1 : ) (*InternalKey, base.LazyValue) {
496 1 : // Invariant: i.parent.rangeKey is nil.
497 1 : // Invariant: !i.combinedIterState.initialized.
498 1 : if invariants.Enabled {
499 1 : if i.combinedIterState.initialized {
500 0 : panic("pebble: combined iterator already initialized")
501 : }
502 1 : if i.parent.rangeKey != nil {
503 0 : panic("pebble: iterator already has a range-key iterator stack")
504 : }
505 : }
506 :
507 : // We need to determine the key to seek the range key iterator to. If
508 : // seekKey is not nil, the user-initiated operation that triggered the
509 : // switch to combined iteration was itself a seek, and we can use that key.
510 : // Otherwise, a First/Last or relative positioning operation triggered the
511 : // switch to combined iteration.
512 : //
513 : // The levelIter that observed a file containing range keys populated
514 : // combinedIterState.key with the smallest (forward) or largest (backward)
515 : // range key it observed. If multiple levelIters observed files with range
516 : // keys during the same operation on the mergingIter, combinedIterState.key
517 : // is the smallest [during forward iteration; largest in reverse iteration]
518 : // such key.
519 1 : if seekKey == nil {
520 1 : // Use the levelIter-populated key.
521 1 : seekKey = i.combinedIterState.key
522 1 :
523 1 : // We may need to adjust the levelIter-populated seek key to the
524 1 : // surfaced point key. If the key observed is beyond [in the iteration
525 1 : // direction] the current point key, there may still exist a range key
526 1 : // at an earlier key. Consider the following example:
527 1 : //
528 1 : // L5: 000003:[bar.DEL.5, foo.RANGEKEYSET.9]
529 1 : // L6: 000001:[bar.SET.2] 000002:[bax.RANGEKEYSET.8]
530 1 : //
531 1 : // A call to First() seeks the levels to files L5.000003 and L6.000001.
532 1 : // The L5 levelIter observes that L5.000003 contains the range key with
533 1 : // start key `foo`, and triggers a switch to combined iteration, setting
534 1 : // `combinedIterState.key` = `foo`.
535 1 : //
536 1 : // The L6 levelIter did not observe the true first range key
537 1 : // (bax.RANGEKEYSET.8), because it appears in a later sstable. When the
538 1 : // combined iterator is initialized, the range key iterator must be
539 1 : // seeked to a key that will find `bax`. To accomplish this, we seek the
540 1 : // key instead to `bar`. It is guaranteed that no range key exists
541 1 : // earlier than `bar`, otherwise a levelIter would've observed it and
542 1 : // set `combinedIterState.key` to its start key.
543 1 : if pointKey != nil {
544 1 : if dir == +1 && i.parent.cmp(i.combinedIterState.key, pointKey.UserKey) > 0 {
545 1 : seekKey = pointKey.UserKey
546 1 : } else if dir == -1 && i.parent.cmp(seekKey, pointKey.UserKey) < 0 {
547 1 : seekKey = pointKey.UserKey
548 1 : }
549 : }
550 : }
551 :
552 : // An operation on the point iterator observed a file containing range keys,
553 : // so we must switch to combined interleaving iteration. First, construct
554 : // the range key iterator stack. It must not exist, otherwise we'd already
555 : // be performing combined iteration.
556 1 : i.parent.rangeKey = iterRangeKeyStateAllocPool.Get().(*iteratorRangeKeyState)
557 1 : i.parent.rangeKey.init(i.parent.comparer.Compare, i.parent.comparer.Split, &i.parent.opts)
558 1 : i.parent.constructRangeKeyIter()
559 1 :
560 1 : // Initialize the Iterator's interleaving iterator.
561 1 : i.parent.rangeKey.iiter.Init(
562 1 : &i.parent.comparer, i.parent.pointIter, i.parent.rangeKey.rangeKeyIter,
563 1 : keyspan.InterleavingIterOpts{
564 1 : Mask: &i.parent.rangeKeyMasking,
565 1 : LowerBound: i.parent.opts.LowerBound,
566 1 : UpperBound: i.parent.opts.UpperBound,
567 1 : })
568 1 :
569 1 : // Set the parent's primary iterator to point to the combined, interleaving
570 1 : // iterator that's now initialized with our current state.
571 1 : i.parent.iter = &i.parent.rangeKey.iiter
572 1 : i.combinedIterState.initialized = true
573 1 : i.combinedIterState.key = nil
574 1 :
575 1 : // All future iterator operations will go directly through the combined
576 1 : // iterator.
577 1 : //
578 1 : // Initialize the interleaving iterator. We pass the point key-value pair so
579 1 : // that the interleaving iterator knows where the point iterator is
580 1 : // positioned. Additionally, we pass the seek key to which the range-key
581 1 : // iterator should be seeked in order to initialize its position.
582 1 : //
583 1 : // In the forward direction (invert for backwards), the seek key is a key
584 1 : // guaranteed to find the smallest range key that's greater than the last
585 1 : // key the iterator returned. The range key may be less than pointKey, in
586 1 : // which case the range key will be interleaved next instead of the point
587 1 : // key.
588 1 : if dir == +1 {
589 1 : var prefix []byte
590 1 : if i.parent.hasPrefix {
591 1 : prefix = i.parent.prefixOrFullSeekKey
592 1 : }
593 1 : return i.parent.rangeKey.iiter.InitSeekGE(prefix, seekKey, pointKey, pointValue)
594 : }
595 1 : return i.parent.rangeKey.iiter.InitSeekLT(seekKey, pointKey, pointValue)
596 : }
597 :
598 : func (i *lazyCombinedIter) SeekGE(
599 : key []byte, flags base.SeekGEFlags,
600 1 : ) (*InternalKey, base.LazyValue) {
601 1 : if i.combinedIterState.initialized {
602 0 : return i.parent.rangeKey.iiter.SeekGE(key, flags)
603 0 : }
604 1 : k, v := i.pointIter.SeekGE(key, flags)
605 1 : if i.combinedIterState.triggered {
606 1 : return i.initCombinedIteration(+1, k, v, key)
607 1 : }
608 1 : return k, v
609 : }
610 :
611 : func (i *lazyCombinedIter) SeekPrefixGE(
612 : prefix, key []byte, flags base.SeekGEFlags,
613 1 : ) (*InternalKey, base.LazyValue) {
614 1 : if i.combinedIterState.initialized {
615 0 : return i.parent.rangeKey.iiter.SeekPrefixGE(prefix, key, flags)
616 0 : }
617 1 : k, v := i.pointIter.SeekPrefixGE(prefix, key, flags)
618 1 : if i.combinedIterState.triggered {
619 1 : return i.initCombinedIteration(+1, k, v, key)
620 1 : }
621 1 : return k, v
622 : }
623 :
624 : func (i *lazyCombinedIter) SeekLT(
625 : key []byte, flags base.SeekLTFlags,
626 1 : ) (*InternalKey, base.LazyValue) {
627 1 : if i.combinedIterState.initialized {
628 0 : return i.parent.rangeKey.iiter.SeekLT(key, flags)
629 0 : }
630 1 : k, v := i.pointIter.SeekLT(key, flags)
631 1 : if i.combinedIterState.triggered {
632 1 : return i.initCombinedIteration(-1, k, v, key)
633 1 : }
634 1 : return k, v
635 : }
636 :
637 1 : func (i *lazyCombinedIter) First() (*InternalKey, base.LazyValue) {
638 1 : if i.combinedIterState.initialized {
639 0 : return i.parent.rangeKey.iiter.First()
640 0 : }
641 1 : k, v := i.pointIter.First()
642 1 : if i.combinedIterState.triggered {
643 1 : return i.initCombinedIteration(+1, k, v, nil)
644 1 : }
645 1 : return k, v
646 : }
647 :
648 1 : func (i *lazyCombinedIter) Last() (*InternalKey, base.LazyValue) {
649 1 : if i.combinedIterState.initialized {
650 0 : return i.parent.rangeKey.iiter.Last()
651 0 : }
652 1 : k, v := i.pointIter.Last()
653 1 : if i.combinedIterState.triggered {
654 1 : return i.initCombinedIteration(-1, k, v, nil)
655 1 : }
656 1 : return k, v
657 : }
658 :
659 1 : func (i *lazyCombinedIter) Next() (*InternalKey, base.LazyValue) {
660 1 : if i.combinedIterState.initialized {
661 0 : return i.parent.rangeKey.iiter.Next()
662 0 : }
663 1 : k, v := i.pointIter.Next()
664 1 : if i.combinedIterState.triggered {
665 1 : return i.initCombinedIteration(+1, k, v, nil)
666 1 : }
667 1 : return k, v
668 : }
669 :
670 1 : func (i *lazyCombinedIter) NextPrefix(succKey []byte) (*InternalKey, base.LazyValue) {
671 1 : if i.combinedIterState.initialized {
672 0 : return i.parent.rangeKey.iiter.NextPrefix(succKey)
673 0 : }
674 1 : k, v := i.pointIter.NextPrefix(succKey)
675 1 : if i.combinedIterState.triggered {
676 0 : return i.initCombinedIteration(+1, k, v, nil)
677 0 : }
678 1 : return k, v
679 : }
680 :
681 1 : func (i *lazyCombinedIter) Prev() (*InternalKey, base.LazyValue) {
682 1 : if i.combinedIterState.initialized {
683 0 : return i.parent.rangeKey.iiter.Prev()
684 0 : }
685 1 : k, v := i.pointIter.Prev()
686 1 : if i.combinedIterState.triggered {
687 1 : return i.initCombinedIteration(-1, k, v, nil)
688 1 : }
689 1 : return k, v
690 : }
691 :
692 1 : func (i *lazyCombinedIter) Error() error {
693 1 : if i.combinedIterState.initialized {
694 0 : return i.parent.rangeKey.iiter.Error()
695 0 : }
696 1 : return i.pointIter.Error()
697 : }
698 :
699 1 : func (i *lazyCombinedIter) Close() error {
700 1 : if i.combinedIterState.initialized {
701 0 : return i.parent.rangeKey.iiter.Close()
702 0 : }
703 1 : return i.pointIter.Close()
704 : }
705 :
706 1 : func (i *lazyCombinedIter) SetBounds(lower, upper []byte) {
707 1 : if i.combinedIterState.initialized {
708 0 : i.parent.rangeKey.iiter.SetBounds(lower, upper)
709 0 : return
710 0 : }
711 1 : i.pointIter.SetBounds(lower, upper)
712 : }
713 :
714 0 : func (i *lazyCombinedIter) SetContext(ctx context.Context) {
715 0 : if i.combinedIterState.initialized {
716 0 : i.parent.rangeKey.iiter.SetContext(ctx)
717 0 : return
718 0 : }
719 0 : i.pointIter.SetContext(ctx)
720 : }
721 :
722 0 : func (i *lazyCombinedIter) String() string {
723 0 : if i.combinedIterState.initialized {
724 0 : return i.parent.rangeKey.iiter.String()
725 0 : }
726 0 : return i.pointIter.String()
727 : }
|