Line data Source code
1 : // Copyright 2022 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 keyspan
6 :
7 : import (
8 : "bytes"
9 : "fmt"
10 : "sort"
11 :
12 : "github.com/cockroachdb/pebble/internal/base"
13 : "github.com/cockroachdb/pebble/internal/invariants"
14 : "github.com/cockroachdb/pebble/internal/manifest"
15 : )
16 :
17 : // TODO(jackson): Consider implementing an optimization to seek lower levels
18 : // past higher levels' RANGEKEYDELs. This would be analaogous to the
19 : // optimization pebble.mergingIter performs for RANGEDELs during point key
20 : // seeks. It may not be worth it, because range keys are rare and cascading
21 : // seeks would require introducing key comparisons to switchTo{Min,Max}Heap
22 : // where there currently are none.
23 :
24 : // TODO(jackson): There are several opportunities to use base.Equal in the
25 : // MergingIter implementation, but will require a bit of plumbing to thread the
26 : // Equal function.
27 :
28 : // MergingIter merges spans across levels of the LSM, exposing an iterator over
29 : // spans that yields sets of spans fragmented at unique user key boundaries.
30 : //
31 : // A MergingIter is initialized with an arbitrary number of child iterators over
32 : // fragmented spans. Each child iterator exposes fragmented key spans, such that
33 : // overlapping keys are surfaced in a single Span. Key spans from one child
34 : // iterator may overlap key spans from another child iterator arbitrarily.
35 : //
36 : // The spans combined by MergingIter will return spans with keys sorted by
37 : // trailer descending. If the MergingIter is configured with a Transformer, it's
38 : // permitted to modify the ordering of the spans' keys returned by MergingIter.
39 : //
40 : // # Algorithm
41 : //
42 : // The merging iterator wraps child iterators, merging and fragmenting spans
43 : // across levels. The high-level algorithm is:
44 : //
45 : // 1. Initialize the heap with bound keys from child iterators' spans.
46 : // 2. Find the next [or previous] two unique user keys' from bounds.
47 : // 3. Consider the span formed between the two unique user keys a candidate
48 : // span.
49 : // 4. Determine if any of the child iterators' spans overlap the candidate
50 : // span.
51 : // 4a. If any of the child iterator's current bounds are end keys
52 : // (during forward iteration) or start keys (during reverse
53 : // iteration), then all the spans with that bound overlap the
54 : // candidate span.
55 : // 4b. Apply the configured transform, which may remove keys.
56 : // 4c. If no spans overlap, forget the smallest (forward iteration)
57 : // or largest (reverse iteration) unique user key and advance
58 : // the iterators to the next unique user key. Start again from 3.
59 : //
60 : // # Detailed algorithm
61 : //
62 : // Each level (i0, i1, ...) has a user-provided input FragmentIterator. The
63 : // merging iterator steps through individual boundaries of the underlying
64 : // spans separately. If the underlying FragmentIterator has fragments
65 : // [a,b){#2,#1} [b,c){#1} the mergingIterLevel.{next,prev} step through:
66 : //
67 : // (a, start), (b, end), (b, start), (c, end)
68 : //
69 : // Note that (a, start) and (b, end) are observed ONCE each, despite two keys
70 : // sharing those bounds. Also note that (b, end) and (b, start) are two distinct
71 : // iterator positions of a mergingIterLevel.
72 : //
73 : // The merging iterator maintains a heap (min during forward iteration, max
74 : // during reverse iteration) containing the boundKeys. Each boundKey is a
75 : // 3-tuple holding the bound user key, whether the bound is a start or end key
76 : // and the set of keys from that level that have that bound. The heap orders
77 : // based on the boundKey's user key only.
78 : //
79 : // The merging iterator is responsible for merging spans across levels to
80 : // determine which span is next, but it's also responsible for fragmenting
81 : // overlapping spans. Consider the example:
82 : //
83 : // i0: b---d e-----h
84 : // i1: a---c h-----k
85 : // i2: a------------------------------p
86 : //
87 : // fragments: a-b-c-d-e-----h-----k----------p
88 : //
89 : // None of the individual child iterators contain a span with the exact bounds
90 : // [c,d), but the merging iterator must produce a span [c,d). To accomplish
91 : // this, the merging iterator visits every span between unique boundary user
92 : // keys. In the above example, this is:
93 : //
94 : // [a,b), [b,c), [c,d), [d,e), [e, h), [h, k), [k, p)
95 : //
96 : // The merging iterator first initializes the heap to prepare for iteration.
97 : // The description below discusses the mechanics of forward iteration after a
98 : // call to First, but the mechanics are similar for reverse iteration and
99 : // other positioning methods.
100 : //
101 : // During a call to First, the heap is initialized by seeking every
102 : // mergingIterLevel to the first bound of the first fragment. In the above
103 : // example, this seeks the child iterators to:
104 : //
105 : // i0: (b, boundKindFragmentStart, [ [b,d) ])
106 : // i1: (a, boundKindFragmentStart, [ [a,c) ])
107 : // i2: (a, boundKindFragmentStart, [ [a,p) ])
108 : //
109 : // After fixing up the heap, the root of the heap is a boundKey with the
110 : // smallest user key ('a' in the example). Once the heap is setup for iteration
111 : // in the appropriate direction and location, the merging iterator uses
112 : // find{Next,Prev}FragmentSet to find the next/previous span bounds.
113 : //
114 : // During forward iteration, the root of the heap's user key is the start key
115 : // key of next merged span. findNextFragmentSet sets m.start to this user
116 : // key. The heap may contain other boundKeys with the same user key if another
117 : // level has a fragment starting or ending at the same key, so the
118 : // findNextFragmentSet method pulls from the heap until it finds the first key
119 : // greater than m.start. This key is used as the end key.
120 : //
121 : // In the above example, this results in m.start = 'a', m.end = 'b' and child
122 : // iterators in the following positions:
123 : //
124 : // i0: (b, boundKindFragmentStart, [ [b,d) ])
125 : // i1: (c, boundKindFragmentEnd, [ [a,c) ])
126 : // i2: (p, boundKindFragmentEnd, [ [a,p) ])
127 : //
128 : // With the user key bounds of the next merged span established,
129 : // findNextFragmentSet must determine which, if any, fragments overlap the span.
130 : // During forward iteration any child iterator that is now positioned at an end
131 : // boundary has an overlapping span. (Justification: The child iterator's end
132 : // boundary is ≥ m.end. The corresponding start boundary must be ≤ m.start since
133 : // there were no other user keys between m.start and m.end. So the fragments
134 : // associated with the iterator's current end boundary have start and end bounds
135 : // such that start ≤ m.start < m.end ≤ end).
136 : //
137 : // findNextFragmentSet iterates over the levels, collecting keys from any child
138 : // iterators positioned at end boundaries. In the above example, i1 and i2 are
139 : // positioned at end boundaries, so findNextFragmentSet collects the keys of
140 : // [a,c) and [a,p). These spans contain the merging iterator's [m.start, m.end)
141 : // span, but they may also extend beyond the m.start and m.end. The merging
142 : // iterator returns the keys with the merging iter's m.start and m.end bounds,
143 : // preserving the underlying keys' sequence numbers, key kinds and values.
144 : //
145 : // A MergingIter is configured with a Transform that's applied to the span
146 : // before surfacing it to the iterator user. A Transform may remove keys
147 : // arbitrarily, but it may not modify the values themselves.
148 : //
149 : // It may be the case that findNextFragmentSet finds no levels positioned at end
150 : // boundaries, or that there are no spans remaining after applying a transform,
151 : // in which case the span [m.start, m.end) overlaps with nothing. In this case
152 : // findNextFragmentSet loops, repeating the above process again until it finds a
153 : // span that does contain keys.
154 : //
155 : // # Memory safety
156 : //
157 : // The FragmentIterator interface only guarantees stability of a Span and its
158 : // associated slices until the next positioning method is called. Adjacent Spans
159 : // may be contained in different sstables, requring the FragmentIterator
160 : // implementation to close one sstable, releasing its memory, before opening the
161 : // next. Most of the state used by the MergingIter is derived from spans at
162 : // current child iterator positions only, ensuring state is stable. The one
163 : // exception is the start bound during forward iteration and the end bound
164 : // during reverse iteration.
165 : //
166 : // If the heap root originates from an end boundary when findNextFragmentSet
167 : // begins, a Next on the heap root level may invalidate the end boundary. To
168 : // accommodate this, find{Next,Prev}FragmentSet copy the initial boundary if the
169 : // subsequent Next/Prev would move to the next span.
170 : type MergingIter struct {
171 : *MergingBuffers
172 : // start and end hold the bounds for the span currently under the
173 : // iterator position.
174 : //
175 : // Invariant: None of the levels' iterators contain spans with a bound
176 : // between start and end. For all bounds b, b ≤ start || b ≥ end.
177 : start, end []byte
178 :
179 : // transformer defines a transformation to be applied to a span before it's
180 : // yielded to the user. Transforming may filter individual keys contained
181 : // within the span.
182 : transformer Transformer
183 : // span holds the iterator's current span. This span is used as the
184 : // destination for transforms. Every tranformed span overwrites the
185 : // previous.
186 : span Span
187 : err error
188 : dir int8
189 :
190 : // alloc preallocates mergingIterLevel and mergingIterItems for use by the
191 : // merging iterator. As long as the merging iterator is used with
192 : // manifest.NumLevels+3 and fewer fragment iterators, the merging iterator
193 : // will not need to allocate upon initialization. The value NumLevels+3
194 : // mirrors the preallocated levels in iterAlloc used for point iterators.
195 : // Invariant: cap(levels) == cap(items)
196 : alloc struct {
197 : levels [manifest.NumLevels + 3]mergingIterLevel
198 : items [manifest.NumLevels + 3]mergingIterItem
199 : }
200 : }
201 :
202 : // MergingBuffers holds buffers used while merging keyspans.
203 : type MergingBuffers struct {
204 : // keys holds all of the keys across all levels that overlap the key span
205 : // [start, end), sorted by Trailer descending. This slice is reconstituted
206 : // in synthesizeKeys from each mergingIterLevel's keys every time the
207 : // [start, end) bounds change.
208 : //
209 : // Each element points into a child iterator's memory, so the keys may not
210 : // be directly modified.
211 : keys keysBySeqNumKind
212 : // levels holds levels allocated by MergingIter.init. The MergingIter will
213 : // prefer use of its `manifest.NumLevels+3` array, so this slice will be
214 : // longer if set.
215 : levels []mergingIterLevel
216 : // heap holds a slice for the merging iterator heap allocated by
217 : // MergingIter.init. The MergingIter will prefer use of its
218 : // `manifest.NumLevels+3` items array, so this slice will be longer if set.
219 : heap mergingIterHeap
220 : // buf is a buffer used to save [start, end) boundary keys.
221 : buf []byte
222 : }
223 :
224 : // PrepareForReuse discards any excessively large buffers.
225 1 : func (bufs *MergingBuffers) PrepareForReuse() {
226 1 : if cap(bufs.buf) > bufferReuseMaxCapacity {
227 0 : bufs.buf = nil
228 0 : }
229 : }
230 :
231 : // MergingIter implements the FragmentIterator interface.
232 : var _ FragmentIterator = (*MergingIter)(nil)
233 :
234 : type mergingIterLevel struct {
235 : iter FragmentIterator
236 :
237 : // heapKey holds the current key at this level for use within the heap.
238 : heapKey boundKey
239 : }
240 :
241 1 : func (l *mergingIterLevel) next() {
242 1 : if l.heapKey.kind == boundKindFragmentStart {
243 1 : l.heapKey = boundKey{
244 1 : kind: boundKindFragmentEnd,
245 1 : key: l.heapKey.span.End,
246 1 : span: l.heapKey.span,
247 1 : }
248 1 : return
249 1 : }
250 1 : if s := l.iter.Next(); s == nil {
251 1 : l.heapKey = boundKey{kind: boundKindInvalid}
252 1 : } else {
253 1 : l.heapKey = boundKey{
254 1 : kind: boundKindFragmentStart,
255 1 : key: s.Start,
256 1 : span: s,
257 1 : }
258 1 : }
259 : }
260 :
261 1 : func (l *mergingIterLevel) prev() {
262 1 : if l.heapKey.kind == boundKindFragmentEnd {
263 1 : l.heapKey = boundKey{
264 1 : kind: boundKindFragmentStart,
265 1 : key: l.heapKey.span.Start,
266 1 : span: l.heapKey.span,
267 1 : }
268 1 : return
269 1 : }
270 1 : if s := l.iter.Prev(); s == nil {
271 1 : l.heapKey = boundKey{kind: boundKindInvalid}
272 1 : } else {
273 1 : l.heapKey = boundKey{
274 1 : kind: boundKindFragmentEnd,
275 1 : key: s.End,
276 1 : span: s,
277 1 : }
278 1 : }
279 : }
280 :
281 : // Init initializes the merging iterator with the provided fragment iterators.
282 : func (m *MergingIter) Init(
283 : cmp base.Compare, transformer Transformer, bufs *MergingBuffers, iters ...FragmentIterator,
284 1 : ) {
285 1 : *m = MergingIter{
286 1 : MergingBuffers: bufs,
287 1 : transformer: transformer,
288 1 : }
289 1 : m.heap.cmp = cmp
290 1 : levels, items := m.levels, m.heap.items
291 1 :
292 1 : // Invariant: cap(levels) >= cap(items)
293 1 : // Invariant: cap(alloc.levels) == cap(alloc.items)
294 1 : if len(iters) <= len(m.alloc.levels) {
295 1 : // The slices allocated on the MergingIter struct are large enough.
296 1 : m.levels = m.alloc.levels[:len(iters)]
297 1 : m.heap.items = m.alloc.items[:0]
298 1 : } else if len(iters) <= cap(levels) {
299 0 : // The existing heap-allocated slices are large enough, so reuse them.
300 0 : m.levels = levels[:len(iters)]
301 0 : m.heap.items = items[:0]
302 1 : } else {
303 1 : // Heap allocate new slices.
304 1 : m.levels = make([]mergingIterLevel, len(iters))
305 1 : m.heap.items = make([]mergingIterItem, 0, len(iters))
306 1 : }
307 1 : for i := range m.levels {
308 1 : m.levels[i] = mergingIterLevel{iter: iters[i]}
309 1 : }
310 : }
311 :
312 : // AddLevel adds a new level to the bottom of the merging iterator. AddLevel
313 : // must be called after Init and before any other method.
314 1 : func (m *MergingIter) AddLevel(iter FragmentIterator) {
315 1 : m.levels = append(m.levels, mergingIterLevel{iter: iter})
316 1 : }
317 :
318 : // SeekGE moves the iterator to the first span covering a key greater than
319 : // or equal to the given key. This is equivalent to seeking to the first
320 : // span with an end key greater than the given key.
321 1 : func (m *MergingIter) SeekGE(key []byte) *Span {
322 1 : m.invalidate() // clear state about current position
323 1 :
324 1 : // SeekGE(k) seeks to the first span with an end key greater than the given
325 1 : // key. The merged span M that we're searching for might straddle the seek
326 1 : // `key`. In this case, the M.Start may be a key ≤ the seek key.
327 1 : //
328 1 : // Consider a SeekGE(dog) in the following example.
329 1 : //
330 1 : // i0: b---d e-----h
331 1 : // i1: a---c h-----k
332 1 : // i2: a------------------------------p
333 1 : // merged: a-b-c-d-e-----h-----k----------p
334 1 : //
335 1 : // The merged span M containing 'dog' is [d,e). The 'd' of the merged span
336 1 : // comes from i0's [b,d)'s end boundary. The [b,d) span does not cover any
337 1 : // key >= dog, so we cannot find the span by positioning the child iterators
338 1 : // using a SeekGE(dog).
339 1 : //
340 1 : // Instead, if we take all the child iterators' spans bounds:
341 1 : // a b c d e h k p
342 1 : // We want to partition them into keys ≤ `key` and keys > `key`.
343 1 : // dog
344 1 : // │
345 1 : // a b c d│e h k p
346 1 : // │
347 1 : // The largest key on the left of the partition forms the merged span's
348 1 : // start key, and the smallest key on the right of the partition forms the
349 1 : // merged span's end key. Recharacterized:
350 1 : //
351 1 : // M.Start: the largest boundary ≤ k of any child span
352 1 : // M.End: the smallest boundary > k of any child span
353 1 : //
354 1 : // The FragmentIterator interface doesn't implement seeking by all bounds,
355 1 : // it implements seeking by containment. A SeekGE(k) will ensure we observe
356 1 : // all start boundaries ≥ k and all end boundaries > k but does not ensure
357 1 : // we observe end boundaries = k or any boundaries < k. A SeekLT(k) will
358 1 : // ensure we observe all start boundaries < k and all end boundaries ≤ k but
359 1 : // does not ensure we observe any start boundaries = k or any boundaries >
360 1 : // k. This forces us to seek in one direction and step in the other.
361 1 : //
362 1 : // In a SeekGE, we want to end up oriented in the forward direction when
363 1 : // complete, so we begin with searching for M.Start by SeekLT-ing every
364 1 : // child iterator to `k`. For every child span found, we determine the
365 1 : // largest bound ≤ `k` and use it to initialize our max heap. The resulting
366 1 : // root of the max heap is a preliminary value for `M.Start`.
367 1 : for i := range m.levels {
368 1 : l := &m.levels[i]
369 1 : s := l.iter.SeekLT(key)
370 1 : if s == nil {
371 1 : l.heapKey = boundKey{kind: boundKindInvalid}
372 1 : } else if m.cmp(s.End, key) <= 0 {
373 1 : l.heapKey = boundKey{
374 1 : kind: boundKindFragmentEnd,
375 1 : key: s.End,
376 1 : span: s,
377 1 : }
378 1 : } else {
379 1 : // s.End > key && s.Start < key
380 1 : // We need to use this span's start bound, since that's the largest
381 1 : // bound ≤ key.
382 1 : l.heapKey = boundKey{
383 1 : kind: boundKindFragmentStart,
384 1 : key: s.Start,
385 1 : span: s,
386 1 : }
387 1 : }
388 : }
389 1 : m.initMaxHeap()
390 1 : if m.err != nil {
391 0 : return nil
392 1 : } else if len(m.heap.items) == 0 {
393 1 : // There are no spans covering any key < `key`. There is no span that
394 1 : // straddles the seek key. Reorient the heap into a min heap and return
395 1 : // the first span we find in the forward direction.
396 1 : m.switchToMinHeap()
397 1 : return m.findNextFragmentSet()
398 1 : }
399 :
400 : // The heap root is now the largest boundary key b such that:
401 : // 1. b < k
402 : // 2. b = k, and b is an end boundary
403 : // There's a third case that we will need to consider later, after we've
404 : // switched to a min heap:
405 : // 3. there exists a start boundary key b such that b = k.
406 : // A start boundary key equal to k would not be surfaced when we seeked all
407 : // the levels using SeekLT(k), since no key <k would be covered within a
408 : // span within an inclusive `k` start boundary.
409 : //
410 : // Assume that the tightest boundary ≤ k is the current heap root (cases 1 &
411 : // 2). After we switch to a min heap, we'll check for the third case and
412 : // adjust the start boundary if necessary.
413 1 : m.start = m.heap.items[0].boundKey.key
414 1 :
415 1 : // Before switching the direction of the heap, save a copy of the start
416 1 : // boundary if it's the end boundary of some child span. Next-ing the child
417 1 : // iterator might switch files and invalidate the memory of the bound.
418 1 : if m.heap.items[0].boundKey.kind == boundKindFragmentEnd {
419 1 : m.buf = append(m.buf[:0], m.start...)
420 1 : m.start = m.buf
421 1 : }
422 :
423 : // Switch to a min heap. This will move each level to the next bound in
424 : // every level, and then establish a min heap. This allows us to obtain the
425 : // smallest boundary key > `key`, which will serve as our candidate end
426 : // bound.
427 1 : m.switchToMinHeap()
428 1 : if m.err != nil {
429 0 : return nil
430 1 : } else if len(m.heap.items) == 0 {
431 1 : return nil
432 1 : }
433 :
434 : // Check for the case 3 described above. It's possible that when we switch
435 : // heap directions, we discover a start boundary of some child span that is
436 : // equal to the seek key `key`. In this case, we want this key to be our
437 : // start boundary.
438 1 : if m.heap.items[0].boundKey.kind == boundKindFragmentStart &&
439 1 : m.cmp(m.heap.items[0].boundKey.key, key) == 0 {
440 1 : // Call findNextFragmentSet, which will set m.start to the heap root and
441 1 : // proceed forward.
442 1 : return m.findNextFragmentSet()
443 1 : }
444 :
445 1 : m.end = m.heap.items[0].boundKey.key
446 1 : if found, s := m.synthesizeKeys(+1); found && s != nil {
447 1 : return s
448 1 : }
449 1 : return m.findNextFragmentSet()
450 :
451 : }
452 :
453 : // SeekLT moves the iterator to the last span covering a key less than the
454 : // given key. This is equivalent to seeking to the last span with a start
455 : // key less than the given key.
456 1 : func (m *MergingIter) SeekLT(key []byte) *Span {
457 1 : m.invalidate() // clear state about current position
458 1 :
459 1 : // SeekLT(k) seeks to the last span with a start key less than the given
460 1 : // key. The merged span M that we're searching for might straddle the seek
461 1 : // `key`. In this case, the M.End may be a key ≥ the seek key.
462 1 : //
463 1 : // Consider a SeekLT(dog) in the following example.
464 1 : //
465 1 : // i0: b---d e-----h
466 1 : // i1: a---c h-----k
467 1 : // i2: a------------------------------p
468 1 : // merged: a-b-c-d-e-----h-----k----------p
469 1 : //
470 1 : // The merged span M containing the largest key <'dog' is [d,e). The 'e' of
471 1 : // the merged span comes from i0's [e,h)'s start boundary. The [e,h) span
472 1 : // does not cover any key < dog, so we cannot find the span by positioning
473 1 : // the child iterators using a SeekLT(dog).
474 1 : //
475 1 : // Instead, if we take all the child iterators' spans bounds:
476 1 : // a b c d e h k p
477 1 : // We want to partition them into keys < `key` and keys ≥ `key`.
478 1 : // dog
479 1 : // │
480 1 : // a b c d│e h k p
481 1 : // │
482 1 : // The largest key on the left of the partition forms the merged span's
483 1 : // start key, and the smallest key on the right of the partition forms the
484 1 : // merged span's end key. Recharacterized:
485 1 : //
486 1 : // M.Start: the largest boundary < k of any child span
487 1 : // M.End: the smallest boundary ≥ k of any child span
488 1 : //
489 1 : // The FragmentIterator interface doesn't implement seeking by all bounds,
490 1 : // it implements seeking by containment. A SeekGE(k) will ensure we observe
491 1 : // all start boundaries ≥ k and all end boundaries > k but does not ensure
492 1 : // we observe end boundaries = k or any boundaries < k. A SeekLT(k) will
493 1 : // ensure we observe all start boundaries < k and all end boundaries ≤ k but
494 1 : // does not ensure we observe any start boundaries = k or any boundaries >
495 1 : // k. This forces us to seek in one direction and step in the other.
496 1 : //
497 1 : // In a SeekLT, we want to end up oriented in the backward direction when
498 1 : // complete, so we begin with searching for M.End by SeekGE-ing every
499 1 : // child iterator to `k`. For every child span found, we determine the
500 1 : // smallest bound ≥ `k` and use it to initialize our min heap. The resulting
501 1 : // root of the min heap is a preliminary value for `M.End`.
502 1 : for i := range m.levels {
503 1 : l := &m.levels[i]
504 1 : s := l.iter.SeekGE(key)
505 1 : if s == nil {
506 1 : l.heapKey = boundKey{kind: boundKindInvalid}
507 1 : } else if m.cmp(s.Start, key) >= 0 {
508 1 : l.heapKey = boundKey{
509 1 : kind: boundKindFragmentStart,
510 1 : key: s.Start,
511 1 : span: s,
512 1 : }
513 1 : } else {
514 1 : // s.Start < key
515 1 : // We need to use this span's end bound, since that's the smallest
516 1 : // bound > key.
517 1 : l.heapKey = boundKey{
518 1 : kind: boundKindFragmentEnd,
519 1 : key: s.End,
520 1 : span: s,
521 1 : }
522 1 : }
523 : }
524 1 : m.initMinHeap()
525 1 : if m.err != nil {
526 0 : return nil
527 1 : } else if len(m.heap.items) == 0 {
528 1 : // There are no spans covering any key ≥ `key`. There is no span that
529 1 : // straddles the seek key. Reorient the heap into a max heap and return
530 1 : // the first span we find in the reverse direction.
531 1 : m.switchToMaxHeap()
532 1 : return m.findPrevFragmentSet()
533 1 : }
534 :
535 : // The heap root is now the smallest boundary key b such that:
536 : // 1. b > k
537 : // 2. b = k, and b is a start boundary
538 : // There's a third case that we will need to consider later, after we've
539 : // switched to a max heap:
540 : // 3. there exists an end boundary key b such that b = k.
541 : // An end boundary key equal to k would not be surfaced when we seeked all
542 : // the levels using SeekGE(k), since k would not be contained within the
543 : // exclusive end boundary.
544 : //
545 : // Assume that the tightest boundary ≥ k is the current heap root (cases 1 &
546 : // 2). After we switch to a max heap, we'll check for the third case and
547 : // adjust the end boundary if necessary.
548 1 : m.end = m.heap.items[0].boundKey.key
549 1 :
550 1 : // Before switching the direction of the heap, save a copy of the end
551 1 : // boundary if it's the start boundary of some child span. Prev-ing the
552 1 : // child iterator might switch files and invalidate the memory of the bound.
553 1 : if m.heap.items[0].boundKey.kind == boundKindFragmentStart {
554 1 : m.buf = append(m.buf[:0], m.end...)
555 1 : m.end = m.buf
556 1 : }
557 :
558 : // Switch to a max heap. This will move each level to the previous bound in
559 : // every level, and then establish a max heap. This allows us to obtain the
560 : // largest boundary key < `key`, which will serve as our candidate start
561 : // bound.
562 1 : m.switchToMaxHeap()
563 1 : if m.err != nil {
564 0 : return nil
565 1 : } else if len(m.heap.items) == 0 {
566 1 : return nil
567 1 : }
568 : // Check for the case 3 described above. It's possible that when we switch
569 : // heap directions, we discover an end boundary of some child span that is
570 : // equal to the seek key `key`. In this case, we want this key to be our end
571 : // boundary.
572 1 : if m.heap.items[0].boundKey.kind == boundKindFragmentEnd &&
573 1 : m.cmp(m.heap.items[0].boundKey.key, key) == 0 {
574 1 : // Call findPrevFragmentSet, which will set m.end to the heap root and
575 1 : // proceed backwards.
576 1 : return m.findPrevFragmentSet()
577 1 : }
578 :
579 1 : m.start = m.heap.items[0].boundKey.key
580 1 : if found, s := m.synthesizeKeys(-1); found && s != nil {
581 1 : return s
582 1 : }
583 1 : return m.findPrevFragmentSet()
584 : }
585 :
586 : // First seeks the iterator to the first span.
587 1 : func (m *MergingIter) First() *Span {
588 1 : m.invalidate() // clear state about current position
589 1 : for i := range m.levels {
590 1 : if s := m.levels[i].iter.First(); s == nil {
591 1 : m.levels[i].heapKey = boundKey{kind: boundKindInvalid}
592 1 : } else {
593 1 : m.levels[i].heapKey = boundKey{
594 1 : kind: boundKindFragmentStart,
595 1 : key: s.Start,
596 1 : span: s,
597 1 : }
598 1 : }
599 : }
600 1 : m.initMinHeap()
601 1 : return m.findNextFragmentSet()
602 : }
603 :
604 : // Last seeks the iterator to the last span.
605 1 : func (m *MergingIter) Last() *Span {
606 1 : m.invalidate() // clear state about current position
607 1 : for i := range m.levels {
608 1 : if s := m.levels[i].iter.Last(); s == nil {
609 0 : m.levels[i].heapKey = boundKey{kind: boundKindInvalid}
610 1 : } else {
611 1 : m.levels[i].heapKey = boundKey{
612 1 : kind: boundKindFragmentEnd,
613 1 : key: s.End,
614 1 : span: s,
615 1 : }
616 1 : }
617 : }
618 1 : m.initMaxHeap()
619 1 : return m.findPrevFragmentSet()
620 : }
621 :
622 : // Next advances the iterator to the next span.
623 1 : func (m *MergingIter) Next() *Span {
624 1 : if m.err != nil {
625 0 : return nil
626 0 : }
627 1 : if m.dir == +1 && (m.end == nil || m.start == nil) {
628 0 : return nil
629 0 : }
630 1 : if m.dir != +1 {
631 1 : m.switchToMinHeap()
632 1 : }
633 1 : return m.findNextFragmentSet()
634 : }
635 :
636 : // Prev advances the iterator to the previous span.
637 1 : func (m *MergingIter) Prev() *Span {
638 1 : if m.err != nil {
639 0 : return nil
640 0 : }
641 1 : if m.dir == -1 && (m.end == nil || m.start == nil) {
642 0 : return nil
643 0 : }
644 1 : if m.dir != -1 {
645 1 : m.switchToMaxHeap()
646 1 : }
647 1 : return m.findPrevFragmentSet()
648 : }
649 :
650 : // Error returns any accumulated error.
651 1 : func (m *MergingIter) Error() error {
652 1 : if m.heap.len() == 0 || m.err != nil {
653 1 : return m.err
654 1 : }
655 1 : return m.levels[m.heap.items[0].index].iter.Error()
656 : }
657 :
658 : // Close closes the iterator, releasing all acquired resources.
659 1 : func (m *MergingIter) Close() error {
660 1 : for i := range m.levels {
661 1 : if err := m.levels[i].iter.Close(); err != nil && m.err == nil {
662 0 : m.err = err
663 0 : }
664 : }
665 1 : m.levels = nil
666 1 : m.heap.items = m.heap.items[:0]
667 1 : return m.err
668 : }
669 :
670 : // String implements fmt.Stringer.
671 0 : func (m *MergingIter) String() string {
672 0 : return "merging-keyspan"
673 0 : }
674 :
675 1 : func (m *MergingIter) initMinHeap() {
676 1 : m.dir = +1
677 1 : m.heap.reverse = false
678 1 : m.initHeap()
679 1 : }
680 :
681 1 : func (m *MergingIter) initMaxHeap() {
682 1 : m.dir = -1
683 1 : m.heap.reverse = true
684 1 : m.initHeap()
685 1 : }
686 :
687 1 : func (m *MergingIter) initHeap() {
688 1 : m.heap.items = m.heap.items[:0]
689 1 : for i := range m.levels {
690 1 : if l := &m.levels[i]; l.heapKey.kind != boundKindInvalid {
691 1 : m.heap.items = append(m.heap.items, mergingIterItem{
692 1 : index: i,
693 1 : boundKey: &l.heapKey,
694 1 : })
695 1 : } else {
696 1 : m.err = firstError(m.err, l.iter.Error())
697 1 : if m.err != nil {
698 0 : return
699 0 : }
700 : }
701 : }
702 1 : m.heap.init()
703 : }
704 :
705 1 : func (m *MergingIter) switchToMinHeap() {
706 1 : // switchToMinHeap reorients the heap for forward iteration, without moving
707 1 : // the current MergingIter position.
708 1 :
709 1 : // The iterator is currently positioned at the span [m.start, m.end),
710 1 : // oriented in the reverse direction, so each level's iterator is positioned
711 1 : // to the largest key ≤ m.start. To reorient in the forward direction, we
712 1 : // must advance each level's iterator to the smallest key ≥ m.end. Consider
713 1 : // this three-level example.
714 1 : //
715 1 : // i0: b---d e-----h
716 1 : // i1: a---c h-----k
717 1 : // i2: a------------------------------p
718 1 : //
719 1 : // merged: a-b-c-d-e-----h-----k----------p
720 1 : //
721 1 : // If currently positioned at the merged span [c,d), then the level
722 1 : // iterators' heap keys are:
723 1 : //
724 1 : // i0: (b, [b, d)) i1: (c, [a,c)) i2: (a, [a,p))
725 1 : //
726 1 : // Reversing the heap should not move the merging iterator and should not
727 1 : // change the current [m.start, m.end) bounds. It should only prepare for
728 1 : // forward iteration by updating the child iterators' heap keys to:
729 1 : //
730 1 : // i0: (d, [b, d)) i1: (h, [h,k)) i2: (p, [a,p))
731 1 : //
732 1 : // In every level the first key ≥ m.end is the next in the iterator.
733 1 : // Justification: Suppose not and a level iterator's next key was some key k
734 1 : // such that k < m.end. The max-heap invariant dictates that the current
735 1 : // iterator position is the largest entry with a user key ≥ m.start. This
736 1 : // means k > m.start. We started with the assumption that k < m.end, so
737 1 : // m.start < k < m.end. But then k is between our current span bounds,
738 1 : // and reverse iteration would have constructed the current interval to be
739 1 : // [k, m.end) not [m.start, m.end).
740 1 :
741 1 : if invariants.Enabled {
742 1 : for i := range m.levels {
743 1 : l := &m.levels[i]
744 1 : if l.heapKey.kind != boundKindInvalid && m.cmp(l.heapKey.key, m.start) > 0 {
745 0 : panic("pebble: invariant violation: max-heap key > m.start")
746 : }
747 : }
748 : }
749 :
750 1 : for i := range m.levels {
751 1 : m.levels[i].next()
752 1 : }
753 1 : m.initMinHeap()
754 : }
755 :
756 1 : func (m *MergingIter) switchToMaxHeap() {
757 1 : // switchToMaxHeap reorients the heap for reverse iteration, without moving
758 1 : // the current MergingIter position.
759 1 :
760 1 : // The iterator is currently positioned at the span [m.start, m.end),
761 1 : // oriented in the forward direction. Each level's iterator is positioned at
762 1 : // the smallest bound ≥ m.end. To reorient in the reverse direction, we must
763 1 : // move each level's iterator to the largest key ≤ m.start. Consider this
764 1 : // three-level example.
765 1 : //
766 1 : // i0: b---d e-----h
767 1 : // i1: a---c h-----k
768 1 : // i2: a------------------------------p
769 1 : //
770 1 : // merged: a-b-c-d-e-----h-----k----------p
771 1 : //
772 1 : // If currently positioned at the merged span [c,d), then the level
773 1 : // iterators' heap keys are:
774 1 : //
775 1 : // i0: (d, [b, d)) i1: (h, [h,k)) i2: (p, [a,p))
776 1 : //
777 1 : // Reversing the heap should not move the merging iterator and should not
778 1 : // change the current [m.start, m.end) bounds. It should only prepare for
779 1 : // reverse iteration by updating the child iterators' heap keys to:
780 1 : //
781 1 : // i0: (b, [b, d)) i1: (c, [a,c)) i2: (a, [a,p))
782 1 : //
783 1 : // In every level the largest key ≤ m.start is the prev in the iterator.
784 1 : // Justification: Suppose not and a level iterator's prev key was some key k
785 1 : // such that k > m.start. The min-heap invariant dictates that the current
786 1 : // iterator position is the smallest entry with a user key ≥ m.end. This
787 1 : // means k < m.end, otherwise the iterator would be positioned at k. We
788 1 : // started with the assumption that k > m.start, so m.start < k < m.end. But
789 1 : // then k is between our current span bounds, and reverse iteration
790 1 : // would have constructed the current interval to be [m.start, k) not
791 1 : // [m.start, m.end).
792 1 :
793 1 : if invariants.Enabled {
794 1 : for i := range m.levels {
795 1 : l := &m.levels[i]
796 1 : if l.heapKey.kind != boundKindInvalid && m.cmp(l.heapKey.key, m.end) < 0 {
797 0 : panic("pebble: invariant violation: min-heap key < m.end")
798 : }
799 : }
800 : }
801 :
802 1 : for i := range m.levels {
803 1 : m.levels[i].prev()
804 1 : }
805 1 : m.initMaxHeap()
806 : }
807 :
808 1 : func (m *MergingIter) cmp(a, b []byte) int {
809 1 : return m.heap.cmp(a, b)
810 1 : }
811 :
812 1 : func (m *MergingIter) findNextFragmentSet() *Span {
813 1 : // Each iteration of this loop considers a new merged span between unique
814 1 : // user keys. An iteration may find that there exists no overlap for a given
815 1 : // span, (eg, if the spans [a,b), [d, e) exist within level iterators, the
816 1 : // below loop will still consider [b,d) before continuing to [d, e)). It
817 1 : // returns when it finds a span that is covered by at least one key.
818 1 :
819 1 : for m.heap.len() > 0 && m.err == nil {
820 1 : // Initialize the next span's start bound. SeekGE and First prepare the
821 1 : // heap without advancing. Next leaves the heap in a state such that the
822 1 : // root is the smallest bound key equal to the returned span's end key,
823 1 : // so the heap is already positioned at the next merged span's start key.
824 1 :
825 1 : // NB: m.heapRoot() might be either an end boundary OR a start boundary
826 1 : // of a level's span. Both end and start boundaries may still be a start
827 1 : // key of a span in the set of fragmented spans returned by MergingIter.
828 1 : // Consider the scenario:
829 1 : // a----------l #1
830 1 : // b-----------m #2
831 1 : //
832 1 : // The merged, fully-fragmented spans that MergingIter exposes to the caller
833 1 : // have bounds:
834 1 : // a-b #1
835 1 : // b--------l #1
836 1 : // b--------l #2
837 1 : // l-m #2
838 1 : //
839 1 : // When advancing to l-m#2, we must set m.start to 'l', which originated
840 1 : // from [a,l)#1's end boundary.
841 1 : m.start = m.heap.items[0].boundKey.key
842 1 :
843 1 : // Before calling nextEntry, consider whether it might invalidate our
844 1 : // start boundary. If the start boundary key originated from an end
845 1 : // boundary, then we need to copy the start key before advancing the
846 1 : // underlying iterator to the next Span.
847 1 : if m.heap.items[0].boundKey.kind == boundKindFragmentEnd {
848 1 : m.buf = append(m.buf[:0], m.start...)
849 1 : m.start = m.buf
850 1 : }
851 :
852 : // There may be many entries all with the same user key. Spans in other
853 : // levels may also start or end at this same user key. For eg:
854 : // L1: [a, c) [c, d)
855 : // L2: [c, e)
856 : // If we're positioned at L1's end(c) end boundary, we want to advance
857 : // to the first bound > c.
858 1 : m.nextEntry()
859 1 : for len(m.heap.items) > 0 && m.err == nil && m.cmp(m.heapRoot(), m.start) == 0 {
860 1 : m.nextEntry()
861 1 : }
862 1 : if len(m.heap.items) == 0 || m.err != nil {
863 1 : break
864 : }
865 :
866 : // The current entry at the top of the heap is the first key > m.start.
867 : // It must become the end bound for the span we will return to the user.
868 : // In the above example, the root of the heap is L1's end(d).
869 1 : m.end = m.heap.items[0].boundKey.key
870 1 :
871 1 : // Each level within m.levels may have a span that overlaps the
872 1 : // fragmented key span [m.start, m.end). Update m.keys to point to them
873 1 : // and sort them by kind, sequence number. There may not be any keys
874 1 : // defined over [m.start, m.end) if we're between the end of one span
875 1 : // and the start of the next, OR if the configured transform filters any
876 1 : // keys out. We allow empty spans that were emitted by child iterators, but
877 1 : // we elide empty spans created by the mergingIter itself that don't overlap
878 1 : // with any child iterator returned spans (i.e. empty spans that bridge two
879 1 : // distinct child-iterator-defined spans).
880 1 : if found, s := m.synthesizeKeys(+1); found && s != nil {
881 1 : return s
882 1 : }
883 : }
884 : // Exhausted.
885 1 : m.clear()
886 1 : return nil
887 : }
888 :
889 1 : func (m *MergingIter) findPrevFragmentSet() *Span {
890 1 : // Each iteration of this loop considers a new merged span between unique
891 1 : // user keys. An iteration may find that there exists no overlap for a given
892 1 : // span, (eg, if the spans [a,b), [d, e) exist within level iterators, the
893 1 : // below loop will still consider [b,d) before continuing to [a, b)). It
894 1 : // returns when it finds a span that is covered by at least one key.
895 1 :
896 1 : for m.heap.len() > 0 && m.err == nil {
897 1 : // Initialize the next span's end bound. SeekLT and Last prepare the
898 1 : // heap without advancing. Prev leaves the heap in a state such that the
899 1 : // root is the largest bound key equal to the returned span's start key,
900 1 : // so the heap is already positioned at the next merged span's end key.
901 1 :
902 1 : // NB: m.heapRoot() might be either an end boundary OR a start boundary
903 1 : // of a level's span. Both end and start boundaries may still be a start
904 1 : // key of a span returned by MergingIter. Consider the scenario:
905 1 : // a----------l #2
906 1 : // b-----------m #1
907 1 : //
908 1 : // The merged, fully-fragmented spans that MergingIter exposes to the caller
909 1 : // have bounds:
910 1 : // a-b #2
911 1 : // b--------l #2
912 1 : // b--------l #1
913 1 : // l-m #1
914 1 : //
915 1 : // When Preving to a-b#2, we must set m.end to 'b', which originated
916 1 : // from [b,m)#1's start boundary.
917 1 : m.end = m.heap.items[0].boundKey.key
918 1 :
919 1 : // Before calling prevEntry, consider whether it might invalidate our
920 1 : // end boundary. If the end boundary key originated from a start
921 1 : // boundary, then we need to copy the end key before advancing the
922 1 : // underlying iterator to the previous Span.
923 1 : if m.heap.items[0].boundKey.kind == boundKindFragmentStart {
924 1 : m.buf = append(m.buf[:0], m.end...)
925 1 : m.end = m.buf
926 1 : }
927 :
928 : // There may be many entries all with the same user key. Spans in other
929 : // levels may also start or end at this same user key. For eg:
930 : // L1: [a, c) [c, d)
931 : // L2: [c, e)
932 : // If we're positioned at L1's start(c) start boundary, we want to prev
933 : // to move to the first bound < c.
934 1 : m.prevEntry()
935 1 : for len(m.heap.items) > 0 && m.err == nil && m.cmp(m.heapRoot(), m.end) == 0 {
936 1 : m.prevEntry()
937 1 : }
938 1 : if len(m.heap.items) == 0 || m.err != nil {
939 1 : break
940 : }
941 :
942 : // The current entry at the top of the heap is the first key < m.end.
943 : // It must become the start bound for the span we will return to the
944 : // user. In the above example, the root of the heap is L1's start(a).
945 1 : m.start = m.heap.items[0].boundKey.key
946 1 :
947 1 : // Each level within m.levels may have a set of keys that overlap the
948 1 : // fragmented key span [m.start, m.end). Update m.keys to point to them
949 1 : // and sort them by kind, sequence number. There may not be any keys
950 1 : // spanning [m.start, m.end) if we're between the end of one span and
951 1 : // the start of the next, OR if the configured transform filters any
952 1 : // keys out. We allow empty spans that were emitted by child iterators, but
953 1 : // we elide empty spans created by the mergingIter itself that don't overlap
954 1 : // with any child iterator returned spans (i.e. empty spans that bridge two
955 1 : // distinct child-iterator-defined spans).
956 1 : if found, s := m.synthesizeKeys(-1); found && s != nil {
957 1 : return s
958 1 : }
959 : }
960 : // Exhausted.
961 1 : m.clear()
962 1 : return nil
963 : }
964 :
965 1 : func (m *MergingIter) heapRoot() []byte {
966 1 : return m.heap.items[0].boundKey.key
967 1 : }
968 :
969 : // synthesizeKeys is called by find{Next,Prev}FragmentSet to populate and
970 : // sort the set of keys overlapping [m.start, m.end).
971 : //
972 : // During forward iteration, if the current heap item is a fragment end,
973 : // then the fragment's start must be ≤ m.start and the fragment overlaps the
974 : // current iterator position of [m.start, m.end).
975 : //
976 : // During reverse iteration, if the current heap item is a fragment start,
977 : // then the fragment's end must be ≥ m.end and the fragment overlaps the
978 : // current iteration position of [m.start, m.end).
979 : //
980 : // The boolean return value, `found`, is true if the returned span overlaps
981 : // with a span returned by a child iterator.
982 1 : func (m *MergingIter) synthesizeKeys(dir int8) (bool, *Span) {
983 1 : if invariants.Enabled {
984 1 : if m.cmp(m.start, m.end) >= 0 {
985 0 : panic(fmt.Sprintf("pebble: invariant violation: span start ≥ end: %s >= %s", m.start, m.end))
986 : }
987 : }
988 :
989 1 : m.keys = m.keys[:0]
990 1 : found := false
991 1 : for i := range m.levels {
992 1 : if dir == +1 && m.levels[i].heapKey.kind == boundKindFragmentEnd ||
993 1 : dir == -1 && m.levels[i].heapKey.kind == boundKindFragmentStart {
994 1 : m.keys = append(m.keys, m.levels[i].heapKey.span.Keys...)
995 1 : found = true
996 1 : }
997 : }
998 : // TODO(jackson): We should be able to remove this sort and instead
999 : // guarantee that we'll return keys in the order of the levels they're from.
1000 : // With careful iterator construction, this would guarantee that they're
1001 : // sorted by trailer descending for the range key iteration use case.
1002 1 : sort.Sort(&m.keys)
1003 1 :
1004 1 : // Apply the configured transform. See VisibleTransform.
1005 1 : m.span = Span{
1006 1 : Start: m.start,
1007 1 : End: m.end,
1008 1 : Keys: m.keys,
1009 1 : KeysOrder: ByTrailerDesc,
1010 1 : }
1011 1 : // NB: m.heap.cmp is a base.Compare, whereas m.cmp is a method on
1012 1 : // MergingIter.
1013 1 : if err := m.transformer.Transform(m.heap.cmp, m.span, &m.span); err != nil {
1014 0 : m.err = err
1015 0 : return false, nil
1016 0 : }
1017 1 : return found, &m.span
1018 : }
1019 :
1020 1 : func (m *MergingIter) invalidate() {
1021 1 : m.err = nil
1022 1 : }
1023 :
1024 1 : func (m *MergingIter) clear() {
1025 1 : for fi := range m.keys {
1026 1 : m.keys[fi] = Key{}
1027 1 : }
1028 1 : m.keys = m.keys[:0]
1029 : }
1030 :
1031 : // nextEntry steps to the next entry.
1032 1 : func (m *MergingIter) nextEntry() {
1033 1 : l := &m.levels[m.heap.items[0].index]
1034 1 : l.next()
1035 1 : if !l.heapKey.valid() {
1036 1 : // l.iter is exhausted.
1037 1 : m.err = l.iter.Error()
1038 1 : if m.err == nil {
1039 1 : m.heap.pop()
1040 1 : }
1041 1 : return
1042 : }
1043 :
1044 1 : if m.heap.len() > 1 {
1045 1 : m.heap.fix(0)
1046 1 : }
1047 : }
1048 :
1049 : // prevEntry steps to the previous entry.
1050 1 : func (m *MergingIter) prevEntry() {
1051 1 : l := &m.levels[m.heap.items[0].index]
1052 1 : l.prev()
1053 1 : if !l.heapKey.valid() {
1054 1 : // l.iter is exhausted.
1055 1 : m.err = l.iter.Error()
1056 1 : if m.err == nil {
1057 1 : m.heap.pop()
1058 1 : }
1059 1 : return
1060 : }
1061 :
1062 1 : if m.heap.len() > 1 {
1063 1 : m.heap.fix(0)
1064 1 : }
1065 : }
1066 :
1067 : // DebugString returns a string representing the current internal state of the
1068 : // merging iterator and its heap for debugging purposes.
1069 0 : func (m *MergingIter) DebugString() string {
1070 0 : var buf bytes.Buffer
1071 0 : fmt.Fprintf(&buf, "Current bounds: [%q, %q)\n", m.start, m.end)
1072 0 : for i := range m.levels {
1073 0 : fmt.Fprintf(&buf, "%d: heap key %s\n", i, m.levels[i].heapKey)
1074 0 : }
1075 0 : return buf.String()
1076 : }
1077 :
1078 : type mergingIterItem struct {
1079 : // boundKey points to the corresponding mergingIterLevel's `iterKey`.
1080 : *boundKey
1081 : // index is the index of this level within the MergingIter's levels field.
1082 : index int
1083 : }
1084 :
1085 : // mergingIterHeap is copied from mergingIterHeap defined in the root pebble
1086 : // package for use with point keys.
1087 :
1088 : type mergingIterHeap struct {
1089 : cmp base.Compare
1090 : reverse bool
1091 : items []mergingIterItem
1092 : }
1093 :
1094 1 : func (h *mergingIterHeap) len() int {
1095 1 : return len(h.items)
1096 1 : }
1097 :
1098 1 : func (h *mergingIterHeap) less(i, j int) bool {
1099 1 : // This key comparison only uses the user key and not the boundKind. Bound
1100 1 : // kind doesn't matter because when stepping over a user key,
1101 1 : // findNextFragmentSet and findPrevFragmentSet skip past all heap items with
1102 1 : // that user key, and makes no assumptions on ordering. All other heap
1103 1 : // examinations only consider the user key.
1104 1 : ik, jk := h.items[i].key, h.items[j].key
1105 1 : c := h.cmp(ik, jk)
1106 1 : if h.reverse {
1107 1 : return c > 0
1108 1 : }
1109 1 : return c < 0
1110 : }
1111 :
1112 1 : func (h *mergingIterHeap) swap(i, j int) {
1113 1 : h.items[i], h.items[j] = h.items[j], h.items[i]
1114 1 : }
1115 :
1116 : // init, fix, up and down are copied from the go stdlib.
1117 1 : func (h *mergingIterHeap) init() {
1118 1 : // heapify
1119 1 : n := h.len()
1120 1 : for i := n/2 - 1; i >= 0; i-- {
1121 1 : h.down(i, n)
1122 1 : }
1123 : }
1124 :
1125 1 : func (h *mergingIterHeap) fix(i int) {
1126 1 : if !h.down(i, h.len()) {
1127 1 : h.up(i)
1128 1 : }
1129 : }
1130 :
1131 1 : func (h *mergingIterHeap) pop() *mergingIterItem {
1132 1 : n := h.len() - 1
1133 1 : h.swap(0, n)
1134 1 : h.down(0, n)
1135 1 : item := &h.items[n]
1136 1 : h.items = h.items[:n]
1137 1 : return item
1138 1 : }
1139 :
1140 1 : func (h *mergingIterHeap) up(j int) {
1141 1 : for {
1142 1 : i := (j - 1) / 2 // parent
1143 1 : if i == j || !h.less(j, i) {
1144 1 : break
1145 : }
1146 0 : h.swap(i, j)
1147 0 : j = i
1148 : }
1149 : }
1150 :
1151 1 : func (h *mergingIterHeap) down(i0, n int) bool {
1152 1 : i := i0
1153 1 : for {
1154 1 : j1 := 2*i + 1
1155 1 : if j1 >= n || j1 < 0 { // j1 < 0 after int overflow
1156 1 : break
1157 : }
1158 1 : j := j1 // left child
1159 1 : if j2 := j1 + 1; j2 < n && h.less(j2, j1) {
1160 1 : j = j2 // = 2*i + 2 // right child
1161 1 : }
1162 1 : if !h.less(j, i) {
1163 1 : break
1164 : }
1165 1 : h.swap(i, j)
1166 1 : i = j
1167 : }
1168 1 : return i > i0
1169 : }
1170 :
1171 : type boundKind int8
1172 :
1173 : const (
1174 : boundKindInvalid boundKind = iota
1175 : boundKindFragmentStart
1176 : boundKindFragmentEnd
1177 : )
1178 :
1179 : type boundKey struct {
1180 : kind boundKind
1181 : key []byte
1182 : // span holds the span the bound key comes from.
1183 : //
1184 : // If kind is boundKindFragmentStart, then key is span.Start. If kind is
1185 : // boundKindFragmentEnd, then key is span.End.
1186 : span *Span
1187 : }
1188 :
1189 1 : func (k boundKey) valid() bool {
1190 1 : return k.kind != boundKindInvalid
1191 1 : }
1192 :
1193 0 : func (k boundKey) String() string {
1194 0 : var buf bytes.Buffer
1195 0 : switch k.kind {
1196 0 : case boundKindInvalid:
1197 0 : fmt.Fprint(&buf, "invalid")
1198 0 : case boundKindFragmentStart:
1199 0 : fmt.Fprint(&buf, "fragment-start")
1200 0 : case boundKindFragmentEnd:
1201 0 : fmt.Fprint(&buf, "fragment-end ")
1202 0 : default:
1203 0 : fmt.Fprintf(&buf, "unknown-kind(%d)", k.kind)
1204 : }
1205 0 : fmt.Fprintf(&buf, " %s [", k.key)
1206 0 : fmt.Fprintf(&buf, "%s", k.span)
1207 0 : fmt.Fprint(&buf, "]")
1208 0 : return buf.String()
1209 : }
|