eleksir/bunny-bot
0
Fork 0
Code Pull Requests Projects Activity

Initial commit

Browse Source
This commit is contained in:
eleksir
2025-05-08 18:12:52 +03:00
commit 144da3f5c0
1519 changed files with 648343 additions and 0 deletions
Download Patch File Download Diff File Expand all files Collapse all files

201
vendor/github.com/cockroachdb/pebble/internal/arenaskl/LICENSE generated vendored Normal file
View File

@@ -0,0 +1,201 @@
Apache License
Version 2.0, January 2004
http://www.apache.org/licenses/
TERMS AND CONDITIONS FOR USE, REPRODUCTION, AND DISTRIBUTION
1. Definitions.
"License" shall mean the terms and conditions for use, reproduction,
and distribution as defined by Sections 1 through 9 of this document.
"Licensor" shall mean the copyright owner or entity authorized by
the copyright owner that is granting the License.
"Legal Entity" shall mean the union of the acting entity and all
other entities that control, are controlled by, or are under common
control with that entity. For the purposes of this definition,
"control" means (i) the power, direct or indirect, to cause the
direction or management of such entity, whether by contract or
otherwise, or (ii) ownership of fifty percent (50%) or more of the
outstanding shares, or (iii) beneficial ownership of such entity.
"You" (or "Your") shall mean an individual or Legal Entity
exercising permissions granted by this License.
"Source" form shall mean the preferred form for making modifications,
including but not limited to software source code, documentation
source, and configuration files.
"Object" form shall mean any form resulting from mechanical
transformation or translation of a Source form, including but
not limited to compiled object code, generated documentation,
and conversions to other media types.
"Work" shall mean the work of authorship, whether in Source or
Object form, made available under the License, as indicated by a
copyright notice that is included in or attached to the work
(an example is provided in the Appendix below).
"Derivative Works" shall mean any work, whether in Source or Object
form, that is based on (or derived from) the Work and for which the
editorial revisions, annotations, elaborations, or other modifications
represent, as a whole, an original work of authorship. For the purposes
of this License, Derivative Works shall not include works that remain
separable from, or merely link (or bind by name) to the interfaces of,
the Work and Derivative Works thereof.
"Contribution" shall mean any work of authorship, including
the original version of the Work and any modifications or additions
to that Work or Derivative Works thereof, that is intentionally
submitted to Licensor for inclusion in the Work by the copyright owner
or by an individual or Legal Entity authorized to submit on behalf of
the copyright owner. For the purposes of this definition, "submitted"
means any form of electronic, verbal, or written communication sent
to the Licensor or its representatives, including but not limited to
communication on electronic mailing lists, source code control systems,
and issue tracking systems that are managed by, or on behalf of, the
Licensor for the purpose of discussing and improving the Work, but
excluding communication that is conspicuously marked or otherwise
designated in writing by the copyright owner as "Not a Contribution."
"Contributor" shall mean Licensor and any individual or Legal Entity
on behalf of whom a Contribution has been received by Licensor and
subsequently incorporated within the Work.
2. Grant of Copyright License. Subject to the terms and conditions of
this License, each Contributor hereby grants to You a perpetual,
worldwide, non-exclusive, no-charge, royalty-free, irrevocable
copyright license to reproduce, prepare Derivative Works of,
publicly display, publicly perform, sublicense, and distribute the
Work and such Derivative Works in Source or Object form.
3. Grant of Patent License. Subject to the terms and conditions of
this License, each Contributor hereby grants to You a perpetual,
worldwide, non-exclusive, no-charge, royalty-free, irrevocable
(except as stated in this section) patent license to make, have made,
use, offer to sell, sell, import, and otherwise transfer the Work,
where such license applies only to those patent claims licensable
by such Contributor that are necessarily infringed by their
Contribution(s) alone or by combination of their Contribution(s)
with the Work to which such Contribution(s) was submitted. If You
institute patent litigation against any entity (including a
cross-claim or counterclaim in a lawsuit) alleging that the Work
or a Contribution incorporated within the Work constitutes direct
or contributory patent infringement, then any patent licenses
granted to You under this License for that Work shall terminate
as of the date such litigation is filed.
4. Redistribution. You may reproduce and distribute copies of the
Work or Derivative Works thereof in any medium, with or without
modifications, and in Source or Object form, provided that You
meet the following conditions:
(a) You must give any other recipients of the Work or
Derivative Works a copy of this License; and
(b) You must cause any modified files to carry prominent notices
stating that You changed the files; and
(c) You must retain, in the Source form of any Derivative Works
that You distribute, all copyright, patent, trademark, and
attribution notices from the Source form of the Work,
excluding those notices that do not pertain to any part of
the Derivative Works; and
(d) If the Work includes a "NOTICE" text file as part of its
distribution, then any Derivative Works that You distribute must
include a readable copy of the attribution notices contained
within such NOTICE file, excluding those notices that do not
pertain to any part of the Derivative Works, in at least one
of the following places: within a NOTICE text file distributed
as part of the Derivative Works; within the Source form or
documentation, if provided along with the Derivative Works; or,
within a display generated by the Derivative Works, if and
wherever such third-party notices normally appear. The contents
of the NOTICE file are for informational purposes only and
do not modify the License. You may add Your own attribution
notices within Derivative Works that You distribute, alongside
or as an addendum to the NOTICE text from the Work, provided
that such additional attribution notices cannot be construed
as modifying the License.
You may add Your own copyright statement to Your modifications and
may provide additional or different license terms and conditions
for use, reproduction, or distribution of Your modifications, or
for any such Derivative Works as a whole, provided Your use,
reproduction, and distribution of the Work otherwise complies with
the conditions stated in this License.
5. Submission of Contributions. Unless You explicitly state otherwise,
any Contribution intentionally submitted for inclusion in the Work
by You to the Licensor shall be under the terms and conditions of
this License, without any additional terms or conditions.
Notwithstanding the above, nothing herein shall supersede or modify
the terms of any separate license agreement you may have executed
with Licensor regarding such Contributions.
6. Trademarks. This License does not grant permission to use the trade
names, trademarks, service marks, or product names of the Licensor,
except as required for reasonable and customary use in describing the
origin of the Work and reproducing the content of the NOTICE file.
7. Disclaimer of Warranty. Unless required by applicable law or
agreed to in writing, Licensor provides the Work (and each
Contributor provides its Contributions) on an "AS IS" BASIS,
WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or
implied, including, without limitation, any warranties or conditions
of TITLE, NON-INFRINGEMENT, MERCHANTABILITY, or FITNESS FOR A
PARTICULAR PURPOSE. You are solely responsible for determining the
appropriateness of using or redistributing the Work and assume any
risks associated with Your exercise of permissions under this License.
8. Limitation of Liability. In no event and under no legal theory,
whether in tort (including negligence), contract, or otherwise,
unless required by applicable law (such as deliberate and grossly
negligent acts) or agreed to in writing, shall any Contributor be
liable to You for damages, including any direct, indirect, special,
incidental, or consequential damages of any character arising as a
result of this License or out of the use or inability to use the
Work (including but not limited to damages for loss of goodwill,
work stoppage, computer failure or malfunction, or any and all
other commercial damages or losses), even if such Contributor
has been advised of the possibility of such damages.
9. Accepting Warranty or Additional Liability. While redistributing
the Work or Derivative Works thereof, You may choose to offer,
and charge a fee for, acceptance of support, warranty, indemnity,
or other liability obligations and/or rights consistent with this
License. However, in accepting such obligations, You may act only
on Your own behalf and on Your sole responsibility, not on behalf
of any other Contributor, and only if You agree to indemnify,
defend, and hold each Contributor harmless for any liability
incurred by, or claims asserted against, such Contributor by reason
of your accepting any such warranty or additional liability.
END OF TERMS AND CONDITIONS
APPENDIX: How to apply the Apache License to your work.
To apply the Apache License to your work, attach the following
boilerplate notice, with the fields enclosed by brackets "{}"
replaced with your own identifying information. (Don't include
the brackets!) The text should be enclosed in the appropriate
comment syntax for the file format. We also recommend that a
file or class name and description of purpose be included on the
same "printed page" as the copyright notice for easier
identification within third-party archives.
Copyright {yyyy} {name of copyright owner}
Licensed under the Apache License, Version 2.0 (the "License");
you may not use this file except in compliance with the License.
You may obtain a copy of the License at
http://www.apache.org/licenses/LICENSE-2.0
Unless required by applicable law or agreed to in writing, software
distributed under the License is distributed on an "AS IS" BASIS,
WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
See the License for the specific language governing permissions and
limitations under the License.

93
vendor/github.com/cockroachdb/pebble/internal/arenaskl/README.md generated vendored Normal file
View File

@@ -0,0 +1,93 @@
# arenaskl
Fast, lock-free, arena-based Skiplist implementation in Go that supports iteration
in both directions.
## Advantages
Arenaskl offers several advantages over other skiplist implementations:
* High performance that linearly scales with the number of cores. This is
achieved by allocating from a fixed-size arena and by avoiding locks.
* Iterators that can be allocated on the stack and easily cloned by value.
* Simple-to-use and low overhead model for detecting and handling race conditions
with other threads.
* Support for iterating in reverse (i.e. previous links).
## Limitations
The advantages come at a cost that prevents arenaskl from being a general-purpose
skiplist implementation:
* The size of the arena sets a hard upper bound on the combined size of skiplist
nodes, keys, and values. This limit includes even the size of deleted nodes,
keys, and values.
* Deletion is not supported. Instead, higher-level code is expected to
add deletion tombstones and needs to process those tombstones
appropriately.
## Pedigree
This code is based on Andy Kimball's arenaskl code:
https://github.com/andy-kimball/arenaskl
The arenaskl code is based on the skiplist found in Badger, a Go-based
KV store:
https://github.com/dgraph-io/badger/tree/master/skl
The skiplist in Badger is itself based on a C++ skiplist built for
Facebook's RocksDB:
https://github.com/facebook/rocksdb/tree/master/memtable
## Benchmarks
The benchmarks consist of a mix of reads and writes executed in parallel. The
fraction of reads is indicated in the run name: "frac_X" indicates a run where
X percent of the operations are reads.
The results are much better than `skiplist` and `slist`.
```
name time/op
ReadWrite/frac_0-8 470ns ±11%
ReadWrite/frac_10-8 462ns ± 3%
ReadWrite/frac_20-8 436ns ± 2%
ReadWrite/frac_30-8 410ns ± 2%
ReadWrite/frac_40-8 385ns ± 2%
ReadWrite/frac_50-8 360ns ± 4%
ReadWrite/frac_60-8 386ns ± 1%
ReadWrite/frac_70-8 352ns ± 2%
ReadWrite/frac_80-8 306ns ± 3%
ReadWrite/frac_90-8 253ns ± 4%
ReadWrite/frac_100-8 28.1ns ± 2%
```
Note that the above numbers are for concurrent operations using 8x
parallelism. The same benchmarks without concurrency (use these
numbers when comparing vs batchskl):
```
name time/op
ReadWrite/frac_0 1.53µs ± 1%
ReadWrite/frac_10 1.46µs ± 2%
ReadWrite/frac_20 1.39µs ± 3%
ReadWrite/frac_30 1.28µs ± 3%
ReadWrite/frac_40 1.21µs ± 2%
ReadWrite/frac_50 1.11µs ± 3%
ReadWrite/frac_60 1.23µs ±17%
ReadWrite/frac_70 1.16µs ± 4%
ReadWrite/frac_80 959ns ± 3%
ReadWrite/frac_90 738ns ± 5%
ReadWrite/frac_100 81.9ns ± 2%
```
Forward and backward iteration are also fast:
```
name time/op
IterNext 3.97ns ± 5%
IterPrev 3.88ns ± 3%
```

125
vendor/github.com/cockroachdb/pebble/internal/arenaskl/arena.go generated vendored Normal file
View File

@@ -0,0 +1,125 @@
/*
* Copyright 2017 Dgraph Labs, Inc. and Contributors
* Modifications copyright (C) 2017 Andy Kimball and Contributors
*
* Licensed under the Apache License, Version 2.0 (the "License");
* you may not use this file except in compliance with the License.
* You may obtain a copy of the License at
*
* http://www.apache.org/licenses/LICENSE-2.0
*
* Unless required by applicable law or agreed to in writing, software
* distributed under the License is distributed on an "AS IS" BASIS,
* WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
* See the License for the specific language governing permissions and
* limitations under the License.
*/
package arenaskl
import (
"sync/atomic"
"unsafe"
"github.com/cockroachdb/errors"
"github.com/cockroachdb/pebble/internal/constants"
"github.com/cockroachdb/pebble/internal/invariants"
)
// Arena is lock-free.
type Arena struct {
n atomic.Uint64
buf []byte
}
const nodeAlignment = 4
var (
// ErrArenaFull indicates that the arena is full and cannot perform any more
// allocations.
ErrArenaFull = errors.New("allocation failed because arena is full")
)
// NewArena allocates a new arena using the specified buffer as the backing
// store.
func NewArena(buf []byte) *Arena {
if len(buf) > constants.MaxUint32OrInt {
if invariants.Enabled {
panic(errors.AssertionFailedf("attempting to create arena of size %d", len(buf)))
}
buf = buf[:constants.MaxUint32OrInt]
}
a := &Arena{
buf: buf,
}
// We don't store data at position 0 in order to reserve offset=0 as a kind of
// nil pointer.
a.n.Store(1)
return a
}
// Size returns the number of bytes allocated by the arena.
func (a *Arena) Size() uint32 {
s := a.n.Load()
if s > constants.MaxUint32OrInt {
// The last failed allocation can push the size higher than len(a.buf).
// Saturate at the maximum representable offset.
return constants.MaxUint32OrInt
}
return uint32(s)
}
// Capacity returns the capacity of the arena.
func (a *Arena) Capacity() uint32 {
return uint32(len(a.buf))
}
// alloc allocates a buffer of the given size and with the given alignment
// (which must be a power of 2).
//
// If overflow is not 0, it also ensures that many bytes after the buffer are
// inside the arena (this is used for structures that are larger than the
// requested size but don't use those extra bytes).
func (a *Arena) alloc(size, alignment, overflow uint32) (uint32, uint32, error) {
if invariants.Enabled && (alignment&(alignment-1)) != 0 {
panic(errors.AssertionFailedf("invalid alignment %d", alignment))
}
// Verify that the arena isn't already full.
origSize := a.n.Load()
if int(origSize) > len(a.buf) {
return 0, 0, ErrArenaFull
}
// Pad the allocation with enough bytes to ensure the requested alignment.
padded := uint64(size) + uint64(alignment) - 1
newSize := a.n.Add(padded)
if newSize+uint64(overflow) > uint64(len(a.buf)) {
return 0, 0, ErrArenaFull
}
// Return the aligned offset.
offset := (uint32(newSize) - size) & ^(alignment - 1)
return offset, uint32(padded), nil
}
func (a *Arena) getBytes(offset uint32, size uint32) []byte {
if offset == 0 {
return nil
}
return a.buf[offset : offset+size : offset+size]
}
func (a *Arena) getPointer(offset uint32) unsafe.Pointer {
if offset == 0 {
return nil
}
return unsafe.Pointer(&a.buf[offset])
}
func (a *Arena) getPointerOffset(ptr unsafe.Pointer) uint32 {
if ptr == nil {
return 0
}
return uint32(uintptr(ptr) - uintptr(unsafe.Pointer(&a.buf[0])))
}

88
vendor/github.com/cockroachdb/pebble/internal/arenaskl/flush_iterator.go generated vendored Normal file
View File

@@ -0,0 +1,88 @@
/*
* Copyright 2017 Dgraph Labs, Inc. and Contributors
* Modifications copyright (C) 2017 Andy Kimball and Contributors
*
* Licensed under the Apache License, Version 2.0 (the "License");
* you may not use this file except in compliance with the License.
* You may obtain a copy of the License at
*
* http://www.apache.org/licenses/LICENSE-2.0
*
* Unless required by applicable law or agreed to in writing, software
* distributed under the License is distributed on an "AS IS" BASIS,
* WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
* See the License for the specific language governing permissions and
* limitations under the License.
*/
package arenaskl
import "github.com/cockroachdb/pebble/internal/base"
// flushIterator is an iterator over the skiplist object. Use Skiplist.NewFlushIter
// to construct an iterator. The current state of the iterator can be cloned by
// simply value copying the struct.
type flushIterator struct {
Iterator
bytesIterated *uint64
}
// flushIterator implements the base.InternalIterator interface.
var _ base.InternalIterator = (*flushIterator)(nil)
func (it *flushIterator) String() string {
return "memtable"
}
func (it *flushIterator) SeekGE(
key []byte, flags base.SeekGEFlags,
) (*base.InternalKey, base.LazyValue) {
panic("pebble: SeekGE unimplemented")
}
func (it *flushIterator) SeekPrefixGE(
prefix, key []byte, flags base.SeekGEFlags,
) (*base.InternalKey, base.LazyValue) {
panic("pebble: SeekPrefixGE unimplemented")
}
func (it *flushIterator) SeekLT(
key []byte, flags base.SeekLTFlags,
) (*base.InternalKey, base.LazyValue) {
panic("pebble: SeekLT unimplemented")
}
// First seeks position at the first entry in list. Returns the key and value
// if the iterator is pointing at a valid entry, and (nil, nil) otherwise. Note
// that First only checks the upper bound. It is up to the caller to ensure
// that key is greater than or equal to the lower bound.
func (it *flushIterator) First() (*base.InternalKey, base.LazyValue) {
key, val := it.Iterator.First()
if key == nil {
return nil, base.LazyValue{}
}
*it.bytesIterated += uint64(it.nd.allocSize)
return key, val
}
// Next advances to the next position. Returns the key and value if the
// iterator is pointing at a valid entry, and (nil, nil) otherwise.
// Note: flushIterator.Next mirrors the implementation of Iterator.Next
// due to performance. Keep the two in sync.
func (it *flushIterator) Next() (*base.InternalKey, base.LazyValue) {
it.nd = it.list.getNext(it.nd, 0)
if it.nd == it.list.tail {
return nil, base.LazyValue{}
}
it.decodeKey()
*it.bytesIterated += uint64(it.nd.allocSize)
return &it.key, base.MakeInPlaceValue(it.value())
}
func (it *flushIterator) NextPrefix(succKey []byte) (*base.InternalKey, base.LazyValue) {
panic("pebble: NextPrefix unimplemented")
}
func (it *flushIterator) Prev() (*base.InternalKey, base.LazyValue) {
panic("pebble: Prev unimplemented")
}

271
vendor/github.com/cockroachdb/pebble/internal/arenaskl/iterator.go generated vendored Normal file
View File

@@ -0,0 +1,271 @@
/*
* Copyright 2017 Dgraph Labs, Inc. and Contributors
* Modifications copyright (C) 2017 Andy Kimball and Contributors
*
* Licensed under the Apache License, Version 2.0 (the "License");
* you may not use this file except in compliance with the License.
* You may obtain a copy of the License at
*
* http://www.apache.org/licenses/LICENSE-2.0
*
* Unless required by applicable law or agreed to in writing, software
* distributed under the License is distributed on an "AS IS" BASIS,
* WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
* See the License for the specific language governing permissions and
* limitations under the License.
*/
package arenaskl
import (
"sync"
"github.com/cockroachdb/pebble/internal/base"
)
type splice struct {
prev *node
next *node
}
func (s *splice) init(prev, next *node) {
s.prev = prev
s.next = next
}
// Iterator is an iterator over the skiplist object. Use Skiplist.NewIter
// to construct an iterator. The current state of the iterator can be cloned by
// simply value copying the struct. All iterator methods are thread-safe.
type Iterator struct {
list *Skiplist
nd *node
key base.InternalKey
lower []byte
upper []byte
}
// Iterator implements the base.InternalIterator interface.
var _ base.InternalIterator = (*Iterator)(nil)
var iterPool = sync.Pool{
New: func() interface{} {
return &Iterator{}
},
}
// Close resets the iterator.
func (it *Iterator) Close() error {
it.list = nil
it.nd = nil
it.lower = nil
it.upper = nil
iterPool.Put(it)
return nil
}
func (it *Iterator) String() string {
return "memtable"
}
// Error returns any accumulated error.
func (it *Iterator) Error() error {
return nil
}
// SeekGE moves the iterator to the first entry whose key is greater than or
// equal to the given key. Returns the key and value if the iterator is
// pointing at a valid entry, and (nil, nil) otherwise. Note that SeekGE only
// checks the upper bound. It is up to the caller to ensure that key is greater
// than or equal to the lower bound.
func (it *Iterator) SeekGE(key []byte, flags base.SeekGEFlags) (*base.InternalKey, base.LazyValue) {
if flags.TrySeekUsingNext() {
if it.nd == it.list.tail {
// Iterator is done.
return nil, base.LazyValue{}
}
less := it.list.cmp(it.key.UserKey, key) < 0
// Arbitrary constant. By measuring the seek cost as a function of the
// number of elements in the skip list, and fitting to a model, we
// could adjust the number of nexts based on the current size of the
// skip list.
const numNexts = 5
for i := 0; less && i < numNexts; i++ {
k, _ := it.Next()
if k == nil {
// Iterator is done.
return nil, base.LazyValue{}
}
less = it.list.cmp(it.key.UserKey, key) < 0
}
if !less {
return &it.key, base.MakeInPlaceValue(it.value())
}
}
_, it.nd, _ = it.seekForBaseSplice(key)
if it.nd == it.list.tail {
return nil, base.LazyValue{}
}
it.decodeKey()
if it.upper != nil && it.list.cmp(it.upper, it.key.UserKey) <= 0 {
it.nd = it.list.tail
return nil, base.LazyValue{}
}
return &it.key, base.MakeInPlaceValue(it.value())
}
// SeekPrefixGE moves the iterator to the first entry whose key is greater than
// or equal to the given key. This method is equivalent to SeekGE and is
// provided so that an arenaskl.Iterator implements the
// internal/base.InternalIterator interface.
func (it *Iterator) SeekPrefixGE(
prefix, key []byte, flags base.SeekGEFlags,
) (*base.InternalKey, base.LazyValue) {
return it.SeekGE(key, flags)
}
// SeekLT moves the iterator to the last entry whose key is less than the given
// key. Returns the key and value if the iterator is pointing at a valid entry,
// and (nil, nil) otherwise. Note that SeekLT only checks the lower bound. It
// is up to the caller to ensure that key is less than the upper bound.
func (it *Iterator) SeekLT(key []byte, flags base.SeekLTFlags) (*base.InternalKey, base.LazyValue) {
// NB: the top-level Iterator has already adjusted key based on
// the upper-bound.
it.nd, _, _ = it.seekForBaseSplice(key)
if it.nd == it.list.head {
return nil, base.LazyValue{}
}
it.decodeKey()
if it.lower != nil && it.list.cmp(it.lower, it.key.UserKey) > 0 {
it.nd = it.list.head
return nil, base.LazyValue{}
}
return &it.key, base.MakeInPlaceValue(it.value())
}
// First seeks position at the first entry in list. Returns the key and value
// if the iterator is pointing at a valid entry, and (nil, nil) otherwise. Note
// that First only checks the upper bound. It is up to the caller to ensure
// that key is greater than or equal to the lower bound (e.g. via a call to SeekGE(lower)).
func (it *Iterator) First() (*base.InternalKey, base.LazyValue) {
it.nd = it.list.getNext(it.list.head, 0)
if it.nd == it.list.tail {
return nil, base.LazyValue{}
}
it.decodeKey()
if it.upper != nil && it.list.cmp(it.upper, it.key.UserKey) <= 0 {
it.nd = it.list.tail
return nil, base.LazyValue{}
}
return &it.key, base.MakeInPlaceValue(it.value())
}
// Last seeks position at the last entry in list. Returns the key and value if
// the iterator is pointing at a valid entry, and (nil, nil) otherwise. Note
// that Last only checks the lower bound. It is up to the caller to ensure that
// key is less than the upper bound (e.g. via a call to SeekLT(upper)).
func (it *Iterator) Last() (*base.InternalKey, base.LazyValue) {
it.nd = it.list.getPrev(it.list.tail, 0)
if it.nd == it.list.head {
return nil, base.LazyValue{}
}
it.decodeKey()
if it.lower != nil && it.list.cmp(it.lower, it.key.UserKey) > 0 {
it.nd = it.list.head
return nil, base.LazyValue{}
}
return &it.key, base.MakeInPlaceValue(it.value())
}
// Next advances to the next position. Returns the key and value if the
// iterator is pointing at a valid entry, and (nil, nil) otherwise.
// Note: flushIterator.Next mirrors the implementation of Iterator.Next
// due to performance. Keep the two in sync.
func (it *Iterator) Next() (*base.InternalKey, base.LazyValue) {
it.nd = it.list.getNext(it.nd, 0)
if it.nd == it.list.tail {
return nil, base.LazyValue{}
}
it.decodeKey()
if it.upper != nil && it.list.cmp(it.upper, it.key.UserKey) <= 0 {
it.nd = it.list.tail
return nil, base.LazyValue{}
}
return &it.key, base.MakeInPlaceValue(it.value())
}
// NextPrefix advances to the next position with a new prefix. Returns the key
// and value if the iterator is pointing at a valid entry, and (nil, nil)
// otherwise.
func (it *Iterator) NextPrefix(succKey []byte) (*base.InternalKey, base.LazyValue) {
return it.SeekGE(succKey, base.SeekGEFlagsNone.EnableTrySeekUsingNext())
}
// Prev moves to the previous position. Returns the key and value if the
// iterator is pointing at a valid entry, and (nil, nil) otherwise.
func (it *Iterator) Prev() (*base.InternalKey, base.LazyValue) {
it.nd = it.list.getPrev(it.nd, 0)
if it.nd == it.list.head {
return nil, base.LazyValue{}
}
it.decodeKey()
if it.lower != nil && it.list.cmp(it.lower, it.key.UserKey) > 0 {
it.nd = it.list.head
return nil, base.LazyValue{}
}
return &it.key, base.MakeInPlaceValue(it.value())
}
// value returns the value at the current position.
func (it *Iterator) value() []byte {
return it.nd.getValue(it.list.arena)
}
// Head true iff the iterator is positioned at the sentinel head node.
func (it *Iterator) Head() bool {
return it.nd == it.list.head
}
// Tail true iff the iterator is positioned at the sentinel tail node.
func (it *Iterator) Tail() bool {
return it.nd == it.list.tail
}
// SetBounds sets the lower and upper bounds for the iterator. Note that the
// result of Next and Prev will be undefined until the iterator has been
// repositioned with SeekGE, SeekPrefixGE, SeekLT, First, or Last.
func (it *Iterator) SetBounds(lower, upper []byte) {
it.lower = lower
it.upper = upper
}
func (it *Iterator) decodeKey() {
it.key.UserKey = it.list.arena.getBytes(it.nd.keyOffset, it.nd.keySize)
it.key.Trailer = it.nd.keyTrailer
}
func (it *Iterator) seekForBaseSplice(key []byte) (prev, next *node, found bool) {
ikey := base.MakeSearchKey(key)
level := int(it.list.Height() - 1)
prev = it.list.head
for {
prev, next, found = it.list.findSpliceForLevel(ikey, level, prev)
if found {
if level != 0 {
// next is pointing at the target node, but we need to find previous on
// the bottom level.
prev = it.list.getPrev(next, 0)
}
break
}
if level == 0 {
break
}
level--
}
return
}

133
vendor/github.com/cockroachdb/pebble/internal/arenaskl/node.go generated vendored Normal file
View File

@@ -0,0 +1,133 @@
/*
* Copyright 2017 Dgraph Labs, Inc. and Contributors
* Modifications copyright (C) 2017 Andy Kimball and Contributors
*
* Licensed under the Apache License, Version 2.0 (the "License");
* you may not use this file except in compliance with the License.
* You may obtain a copy of the License at
*
* http://www.apache.org/licenses/LICENSE-2.0
*
* Unless required by applicable law or agreed to in writing, software
* distributed under the License is distributed on an "AS IS" BASIS,
* WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
* See the License for the specific language governing permissions and
* limitations under the License.
*/
package arenaskl
import (
"math"
"sync/atomic"
"github.com/cockroachdb/pebble/internal/base"
)
// MaxNodeSize returns the maximum space needed for a node with the specified
// key and value sizes. This could overflow a uint32, which is why a uint64
// is used here. If a key/value overflows a uint32, it should not be added to
// the skiplist.
func MaxNodeSize(keySize, valueSize uint32) uint64 {
const maxPadding = nodeAlignment - 1
return uint64(maxNodeSize) + uint64(keySize) + uint64(valueSize) + maxPadding
}
type links struct {
nextOffset atomic.Uint32
prevOffset atomic.Uint32
}
func (l *links) init(prevOffset, nextOffset uint32) {
l.nextOffset.Store(nextOffset)
l.prevOffset.Store(prevOffset)
}
type node struct {
// Immutable fields, so no need to lock to access key.
keyOffset uint32
keySize uint32
keyTrailer uint64
valueSize uint32
allocSize uint32
// Most nodes do not need to use the full height of the tower, since the
// probability of each successive level decreases exponentially. Because
// these elements are never accessed, they do not need to be allocated.
// Therefore, when a node is allocated in the arena, its memory footprint
// is deliberately truncated to not include unneeded tower elements.
//
// All accesses to elements should use CAS operations, with no need to lock.
tower [maxHeight]links
}
func newNode(
arena *Arena, height uint32, key base.InternalKey, value []byte,
) (nd *node, err error) {
if height < 1 || height > maxHeight {
panic("height cannot be less than one or greater than the max height")
}
keySize := len(key.UserKey)
if int64(keySize) > math.MaxUint32 {
panic("key is too large")
}
valueSize := len(value)
if int64(len(value)) > math.MaxUint32 {
panic("value is too large")
}
if int64(len(value))+int64(keySize)+int64(maxNodeSize) > math.MaxUint32 {
panic("combined key and value size is too large")
}
nd, err = newRawNode(arena, height, uint32(keySize), uint32(valueSize))
if err != nil {
return
}
nd.keyTrailer = key.Trailer
copy(nd.getKeyBytes(arena), key.UserKey)
copy(nd.getValue(arena), value)
return
}
func newRawNode(arena *Arena, height uint32, keySize, valueSize uint32) (nd *node, err error) {
// Compute the amount of the tower that will never be used, since the height
// is less than maxHeight.
unusedSize := uint32((maxHeight - int(height)) * linksSize)
nodeSize := uint32(maxNodeSize) - unusedSize
nodeOffset, allocSize, err := arena.alloc(nodeSize+keySize+valueSize, nodeAlignment, unusedSize)
if err != nil {
return
}
nd = (*node)(arena.getPointer(nodeOffset))
nd.keyOffset = nodeOffset + nodeSize
nd.keySize = keySize
nd.valueSize = valueSize
nd.allocSize = allocSize
return
}
func (n *node) getKeyBytes(arena *Arena) []byte {
return arena.getBytes(n.keyOffset, n.keySize)
}
func (n *node) getValue(arena *Arena) []byte {
return arena.getBytes(n.keyOffset+n.keySize, uint32(n.valueSize))
}
func (n *node) nextOffset(h int) uint32 {
return n.tower[h].nextOffset.Load()
}
func (n *node) prevOffset(h int) uint32 {
return n.tower[h].prevOffset.Load()
}
func (n *node) casNextOffset(h int, old, val uint32) bool {
return n.tower[h].nextOffset.CompareAndSwap(old, val)
}
func (n *node) casPrevOffset(h int, old, val uint32) bool {
return n.tower[h].prevOffset.CompareAndSwap(old, val)
}

464
vendor/github.com/cockroachdb/pebble/internal/arenaskl/skl.go generated vendored Normal file
View File

@@ -0,0 +1,464 @@
/*
* Copyright 2017 Dgraph Labs, Inc. and Contributors
* Modifications copyright (C) 2017 Andy Kimball and Contributors
*
* Licensed under the Apache License, Version 2.0 (the "License");
* you may not use this file except in compliance with the License.
* You may obtain a copy of the License at
*
* http://www.apache.org/licenses/LICENSE-2.0
*
* Unless required by applicable law or agreed to in writing, software
* distributed under the License is distributed on an "AS IS" BASIS,
* WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
* See the License for the specific language governing permissions and
* limitations under the License.
*/
/*
Adapted from RocksDB inline skiplist.
Key differences:
- No optimization for sequential inserts (no "prev").
- No custom comparator.
- Support overwrites. This requires care when we see the same key when inserting.
For RocksDB or LevelDB, overwrites are implemented as a newer sequence number in the key, so
there is no need for values. We don't intend to support versioning. In-place updates of values
would be more efficient.
- We discard all non-concurrent code.
- We do not support Splices. This simplifies the code a lot.
- No AllocateNode or other pointer arithmetic.
- We combine the findLessThan, findGreaterOrEqual, etc into one function.
*/
/*
Further adapted from Badger: https://github.com/dgraph-io/badger.
Key differences:
- Support for previous pointers - doubly linked lists. Note that it's up to higher
level code to deal with the intermediate state that occurs during insertion,
where node A is linked to node B, but node B is not yet linked back to node A.
- Iterator includes mutator functions.
*/
package arenaskl // import "github.com/cockroachdb/pebble/internal/arenaskl"
import (
"math"
"runtime"
"sync/atomic"
"unsafe"
"github.com/cockroachdb/errors"
"github.com/cockroachdb/pebble/internal/base"
"github.com/cockroachdb/pebble/internal/fastrand"
)
const (
maxHeight = 20
maxNodeSize = int(unsafe.Sizeof(node{}))
linksSize = int(unsafe.Sizeof(links{}))
pValue = 1 / math.E
)
// ErrRecordExists indicates that an entry with the specified key already
// exists in the skiplist. Duplicate entries are not directly supported and
// instead must be handled by the user by appending a unique version suffix to
// keys.
var ErrRecordExists = errors.New("record with this key already exists")
// Skiplist is a fast, concurrent skiplist implementation that supports forward
// and backward iteration. See batchskl.Skiplist for a non-concurrent
// skiplist. Keys and values are immutable once added to the skiplist and
// deletion is not supported. Instead, higher-level code is expected to add new
// entries that shadow existing entries and perform deletion via tombstones. It
// is up to the user to process these shadow entries and tombstones
// appropriately during retrieval.
type Skiplist struct {
arena *Arena
cmp base.Compare
head *node
tail *node
height atomic.Uint32 // Current height. 1 <= height <= maxHeight. CAS.
// If set to true by tests, then extra delays are added to make it easier to
// detect unusual race conditions.
testing bool
}
// Inserter TODO(peter)
type Inserter struct {
spl [maxHeight]splice
height uint32
}
// Add TODO(peter)
func (ins *Inserter) Add(list *Skiplist, key base.InternalKey, value []byte) error {
return list.addInternal(key, value, ins)
}
var (
probabilities [maxHeight]uint32
)
func init() {
// Precompute the skiplist probabilities so that only a single random number
// needs to be generated and so that the optimal pvalue can be used (inverse
// of Euler's number).
p := float64(1.0)
for i := 0; i < maxHeight; i++ {
probabilities[i] = uint32(float64(math.MaxUint32) * p)
p *= pValue
}
}
// NewSkiplist constructs and initializes a new, empty skiplist. All nodes, keys,
// and values in the skiplist will be allocated from the given arena.
func NewSkiplist(arena *Arena, cmp base.Compare) *Skiplist {
skl := &Skiplist{}
skl.Reset(arena, cmp)
return skl
}
// Reset the skiplist to empty and re-initialize.
func (s *Skiplist) Reset(arena *Arena, cmp base.Compare) {
// Allocate head and tail nodes.
head, err := newRawNode(arena, maxHeight, 0, 0)
if err != nil {
panic("arenaSize is not large enough to hold the head node")
}
head.keyOffset = 0
tail, err := newRawNode(arena, maxHeight, 0, 0)
if err != nil {
panic("arenaSize is not large enough to hold the tail node")
}
tail.keyOffset = 0
// Link all head/tail levels together.
headOffset := arena.getPointerOffset(unsafe.Pointer(head))
tailOffset := arena.getPointerOffset(unsafe.Pointer(tail))
for i := 0; i < maxHeight; i++ {
head.tower[i].nextOffset.Store(tailOffset)
tail.tower[i].prevOffset.Store(headOffset)
}
*s = Skiplist{
arena: arena,
cmp: cmp,
head: head,
tail: tail,
}
s.height.Store(1)
}
// Height returns the height of the highest tower within any of the nodes that
// have ever been allocated as part of this skiplist.
func (s *Skiplist) Height() uint32 { return s.height.Load() }
// Arena returns the arena backing this skiplist.
func (s *Skiplist) Arena() *Arena { return s.arena }
// Size returns the number of bytes that have allocated from the arena.
func (s *Skiplist) Size() uint32 { return s.arena.Size() }
// Add adds a new key if it does not yet exist. If the key already exists, then
// Add returns ErrRecordExists. If there isn't enough room in the arena, then
// Add returns ErrArenaFull.
func (s *Skiplist) Add(key base.InternalKey, value []byte) error {
var ins Inserter
return s.addInternal(key, value, &ins)
}
func (s *Skiplist) addInternal(key base.InternalKey, value []byte, ins *Inserter) error {
if s.findSplice(key, ins) {
// Found a matching node, but handle case where it's been deleted.
return ErrRecordExists
}
if s.testing {
// Add delay to make it easier to test race between this thread
// and another thread that sees the intermediate state between
// finding the splice and using it.
runtime.Gosched()
}
nd, height, err := s.newNode(key, value)
if err != nil {
return err
}
ndOffset := s.arena.getPointerOffset(unsafe.Pointer(nd))
// We always insert from the base level and up. After you add a node in base
// level, we cannot create a node in the level above because it would have
// discovered the node in the base level.
var found bool
var invalidateSplice bool
for i := 0; i < int(height); i++ {
prev := ins.spl[i].prev
next := ins.spl[i].next
if prev == nil {
// New node increased the height of the skiplist, so assume that the
// new level has not yet been populated.
if next != nil {
panic("next is expected to be nil, since prev is nil")
}
prev = s.head
next = s.tail
}
// +----------------+ +------------+ +----------------+
// | prev | | nd | | next |
// | prevNextOffset |---->| | | |
// | |<----| prevOffset | | |
// | | | nextOffset |---->| |
// | | | |<----| nextPrevOffset |
// +----------------+ +------------+ +----------------+
//
// 1. Initialize prevOffset and nextOffset to point to prev and next.
// 2. CAS prevNextOffset to repoint from next to nd.
// 3. CAS nextPrevOffset to repoint from prev to nd.
for {
prevOffset := s.arena.getPointerOffset(unsafe.Pointer(prev))
nextOffset := s.arena.getPointerOffset(unsafe.Pointer(next))
nd.tower[i].init(prevOffset, nextOffset)
// Check whether next has an updated link to prev. If it does not,
// that can mean one of two things:
// 1. The thread that added the next node hasn't yet had a chance
// to add the prev link (but will shortly).
// 2. Another thread has added a new node between prev and next.
nextPrevOffset := next.prevOffset(i)
if nextPrevOffset != prevOffset {
// Determine whether #1 or #2 is true by checking whether prev
// is still pointing to next. As long as the atomic operations
// have at least acquire/release semantics (no need for
// sequential consistency), this works, as it is equivalent to
// the "publication safety" pattern.
prevNextOffset := prev.nextOffset(i)
if prevNextOffset == nextOffset {
// Ok, case #1 is true, so help the other thread along by
// updating the next node's prev link.
next.casPrevOffset(i, nextPrevOffset, prevOffset)
}
}
if prev.casNextOffset(i, nextOffset, ndOffset) {
// Managed to insert nd between prev and next, so update the next
// node's prev link and go to the next level.
if s.testing {
// Add delay to make it easier to test race between this thread
// and another thread that sees the intermediate state between
// setting next and setting prev.
runtime.Gosched()
}
next.casPrevOffset(i, prevOffset, ndOffset)
break
}
// CAS failed. We need to recompute prev and next. It is unlikely to
// be helpful to try to use a different level as we redo the search,
// because it is unlikely that lots of nodes are inserted between prev
// and next.
prev, next, found = s.findSpliceForLevel(key, i, prev)
if found {
if i != 0 {
panic("how can another thread have inserted a node at a non-base level?")
}
return ErrRecordExists
}
invalidateSplice = true
}
}
// If we had to recompute the splice for a level, invalidate the entire
// cached splice.
if invalidateSplice {
ins.height = 0
} else {
// The splice was valid. We inserted a node between spl[i].prev and
// spl[i].next. Optimistically update spl[i].prev for use in a subsequent
// call to add.
for i := uint32(0); i < height; i++ {
ins.spl[i].prev = nd
}
}
return nil
}
// NewIter returns a new Iterator object. The lower and upper bound parameters
// control the range of keys the iterator will return. Specifying for nil for
// lower or upper bound disables the check for that boundary. Note that lower
// bound is not checked on {SeekGE,First} and upper bound is not check on
// {SeekLT,Last}. The user is expected to perform that check. Note that it is
// safe for an iterator to be copied by value.
func (s *Skiplist) NewIter(lower, upper []byte) *Iterator {
it := iterPool.Get().(*Iterator)
*it = Iterator{list: s, nd: s.head, lower: lower, upper: upper}
return it
}
// NewFlushIter returns a new flushIterator, which is similar to an Iterator
// but also sets the current number of the bytes that have been iterated
// through.
func (s *Skiplist) NewFlushIter(bytesFlushed *uint64) base.InternalIterator {
return &flushIterator{
Iterator: Iterator{list: s, nd: s.head},
bytesIterated: bytesFlushed,
}
}
func (s *Skiplist) newNode(
key base.InternalKey, value []byte,
) (nd *node, height uint32, err error) {
height = s.randomHeight()
nd, err = newNode(s.arena, height, key, value)
if err != nil {
return
}
// Try to increase s.height via CAS.
listHeight := s.Height()
for height > listHeight {
if s.height.CompareAndSwap(listHeight, height) {
// Successfully increased skiplist.height.
break
}
listHeight = s.Height()
}
return
}
func (s *Skiplist) randomHeight() uint32 {
rnd := fastrand.Uint32()
h := uint32(1)
for h < maxHeight && rnd <= probabilities[h] {
h++
}
return h
}
func (s *Skiplist) findSplice(key base.InternalKey, ins *Inserter) (found bool) {
listHeight := s.Height()
var level int
prev := s.head
if ins.height < listHeight {
// Our cached height is less than the list height, which means there were
// inserts that increased the height of the list. Recompute the splice from
// scratch.
ins.height = listHeight
level = int(ins.height)
} else {
// Our cached height is equal to the list height.
for ; level < int(listHeight); level++ {
spl := &ins.spl[level]
if s.getNext(spl.prev, level) != spl.next {
// One or more nodes have been inserted between the splice at this
// level.
continue
}
if spl.prev != s.head && !s.keyIsAfterNode(spl.prev, key) {
// Key lies before splice.
level = int(listHeight)
break
}
if spl.next != s.tail && s.keyIsAfterNode(spl.next, key) {
// Key lies after splice.
level = int(listHeight)
break
}
// The splice brackets the key!
prev = spl.prev
break
}
}
for level = level - 1; level >= 0; level-- {
var next *node
prev, next, found = s.findSpliceForLevel(key, level, prev)
if next == nil {
next = s.tail
}
ins.spl[level].init(prev, next)
}
return
}
func (s *Skiplist) findSpliceForLevel(
key base.InternalKey, level int, start *node,
) (prev, next *node, found bool) {
prev = start
for {
// Assume prev.key < key.
next = s.getNext(prev, level)
if next == s.tail {
// Tail node, so done.
break
}
offset, size := next.keyOffset, next.keySize
nextKey := s.arena.buf[offset : offset+size]
cmp := s.cmp(key.UserKey, nextKey)
if cmp < 0 {
// We are done for this level, since prev.key < key < next.key.
break
}
if cmp == 0 {
// User-key equality.
if key.Trailer == next.keyTrailer {
// Internal key equality.
found = true
break
}
if key.Trailer > next.keyTrailer {
// We are done for this level, since prev.key < key < next.key.
break
}
}
// Keep moving right on this level.
prev = next
}
return
}
func (s *Skiplist) keyIsAfterNode(nd *node, key base.InternalKey) bool {
ndKey := s.arena.buf[nd.keyOffset : nd.keyOffset+nd.keySize]
cmp := s.cmp(ndKey, key.UserKey)
if cmp < 0 {
return true
}
if cmp > 0 {
return false
}
// User-key equality.
if key.Trailer == nd.keyTrailer {
// Internal key equality.
return false
}
return key.Trailer < nd.keyTrailer
}
func (s *Skiplist) getNext(nd *node, h int) *node {
offset := nd.tower[h].nextOffset.Load()
return (*node)(s.arena.getPointer(offset))
}
func (s *Skiplist) getPrev(nd *node, h int) *node {
offset := nd.tower[h].prevOffset.Load()
return (*node)(s.arena.getPointer(offset))
}