pax_global_header 0000666 0000000 0000000 00000000064 13441710145 0014512 g ustar 00root root 0000000 0000000 52 comment=a1d9ac4077fcad050b845856c51643c30795a171
golang-github-xiaq-persistent-0.0~git20190312.a1d9ac4/ 0000775 0000000 0000000 00000000000 13441710145 0022107 5 ustar 00root root 0000000 0000000 golang-github-xiaq-persistent-0.0~git20190312.a1d9ac4/.gitignore 0000664 0000000 0000000 00000000525 13441710145 0024101 0 ustar 00root root 0000000 0000000 /cover
[._]*.s[a-w][a-z]
[._]s[a-w][a-z]
*.un~
Session.vim
.netrwhist
*~
# Compiled Object files, Static and Dynamic libs (Shared Objects)
*.o
*.a
*.so
# Folders
_obj
_test
# Architecture specific extensions/prefixes
*.[568vq]
[568vq].out
*.cgo1.go
*.cgo2.c
_cgo_defun.c
_cgo_gotypes.go
_cgo_export.*
_testmain.go
*.exe
*.test
*.prof
golang-github-xiaq-persistent-0.0~git20190312.a1d9ac4/.travis.yml 0000664 0000000 0000000 00000000260 13441710145 0024216 0 ustar 00root root 0000000 0000000 language: go
go:
- 1.6
- 1.7
- 1.8
- 1.9
sudo: false
os:
- linux
- osx
matrix:
exclude:
- os: osx
go: 1.6
script:
make travis
golang-github-xiaq-persistent-0.0~git20190312.a1d9ac4/Makefile 0000664 0000000 0000000 00000002430 13441710145 0023546 0 ustar 00root root 0000000 0000000 PKG_BASE := github.com/xiaq/persistent
PKGS := $(shell go list ./... | grep -v /vendor/)
PKG_COVERS := $(shell go list ./... | grep -v '^$(PKG_BASE)/vendor/' | grep -v '^$(PKG_BASE)$$' | sed "s|^$(PKG_BASE)/|cover/|" | sed 's/$$/.cover/')
COVER_MODE := set
FIRST_GOPATH=$(shell go env GOPATH | cut -d: -f1)
default: get test
get:
go get .
generate:
go generate ./...
test:
go test $(PKGS)
cover/%.cover: %
mkdir -p $(dir $@)
go test -coverprofile=$@ -covermode=$(COVER_MODE) ./$<
cover/all: $(PKG_COVERS)
echo mode: $(COVER_MODE) > $@
for f in $(PKG_COVERS); do test -f $$f && sed 1d $$f >> $@ || true; done
# We would love to test for coverage in pull requests, but it's now
# bettered turned off for two reasons:
#
# 1) The goverall badge will always show the "latest" coverage, even if that
# comes from a PR.
#
# 2) Some of the tests have fluctuating coverage (the test against
# edit.tty.AsyncReader), and goveralls will put a big cross on the PR when the
# coverage happens to drop.
goveralls: cover/all
test "$(TRAVIS_PULL_REQUEST)" = false \
&& go get -u github.com/mattn/goveralls \
&& $(FIRST_GOPATH)/bin/goveralls -coverprofile=cover/all -service=travis-ci \
|| echo "not sending to coveralls"
travis: goveralls
.PHONY: default get generate test goveralls upload travis golang-github-xiaq-persistent-0.0~git20190312.a1d9ac4/README.md 0000664 0000000 0000000 00000010141 13441710145 0023363 0 ustar 00root root 0000000 0000000 # Persistent data structure in Go
[](https://godoc.org/github.com/xiaq/persistent)
[](https://travis-ci.org/xiaq/persistent)
[](https://coveralls.io/github/xiaq/persistent?branch=master)
This is a Go clone of Clojure's persistent data structures.
The API is not stable yet. **DO NOT USE** unless you are willing to cope with
API changes.
License is [Eclipse Public License 1.0](http://opensource.org/licenses/eclipse-1.0.php) (like Clojure). See [epl-v10.html](epl-v10.html) for a copy.
## Implementation notes
The list provided here is a singly-linked list and is very trivial to implement.
The implementation of persistent vector and hash map and based on a series of [excellent](http://blog.higher-order.net/2009/02/01/understanding-clojures-persistentvector-implementation) [blog](http://blog.higher-order.net/2009/09/08/understanding-clojures-persistenthashmap-deftwice) [posts](http://blog.higher-order.net/2010/08/16/assoc-and-clojures-persistenthashmap-part-ii.html) as well as the Clojure source code. Despite the hash map appearing more complicated, the vector is slightly harder to implement due to the "tail array" optimization and some tricky transformation of the tree structure, which is fully replicated here.
## Benchmarking results
### Vectors
Compared to native slices,
* Adding elements is anywhere from 2x to 8x as slow.
* Sequential read is about 9x as slow.
* Random read is about 7x as slow.
Benchmarked on an early 2015 MacBook Pro, with Go 1.9:
```
goos: darwin
goarch: amd64
pkg: github.com/xiaq/persistent/vector
BenchmarkConsNativeN1-4 1000000 2457 ns/op
BenchmarkConsNativeN2-4 300000 4418 ns/op
BenchmarkConsNativeN3-4 30000 55424 ns/op
BenchmarkConsNativeN4-4 300 4493289 ns/op
BenchmarkConsPersistentN1-4 100000 12250 ns/op 4.99x
BenchmarkConsPersistentN2-4 50000 26394 ns/op 5.97x
BenchmarkConsPersistentN3-4 3000 452146 ns/op 8.16x
BenchmarkConsPersistentN4-4 100 13057887 ns/op 2.91x
BenchmarkNthSeqNativeN4-4 30000 43156 ns/op
BenchmarkNthSeqPersistentN4-4 3000 399193 ns/op 9.25x
BenchmarkNthRandNative-4 20000 73860 ns/op
BenchmarkNthRandPersistent-4 3000 546124 ns/op 7.39x
BenchmarkEqualNative-4 50000 23828 ns/op
BenchmarkEqualPersistent-4 2000 1020893 ns/op 42.84x
```
### Hash map
Compared to native maps, adding elements is about 3-6x slow. Difference is
more pronunced when keys are sequential integers, but that workload is very
rare in the real world.
Benchmarked on an early 2015 MacBook Pro, with Go 1.9:
```
goos: darwin
goarch: amd64
pkg: github.com/xiaq/persistent/hashmap
BenchmarkSequentialConsNative1-4 300000 4143 ns/op
BenchmarkSequentialConsNative2-4 10000 130423 ns/op
BenchmarkSequentialConsNative3-4 300 4600842 ns/op
BenchmarkSequentialConsPersistent1-4 100000 14005 ns/op 3.38x
BenchmarkSequentialConsPersistent2-4 2000 641820 ns/op 4.92x
BenchmarkSequentialConsPersistent3-4 20 55180306 ns/op 11.99x
BenchmarkRandomStringsConsNative1-4 200000 7536 ns/op
BenchmarkRandomStringsConsNative2-4 5000 264489 ns/op
BenchmarkRandomStringsConsNative3-4 100 12132244 ns/op
BenchmarkRandomStringsConsPersistent1-4 50000 29109 ns/op 3.86x
BenchmarkRandomStringsConsPersistent2-4 1000 1327321 ns/op 5.02x
BenchmarkRandomStringsConsPersistent3-4 20 74204196 ns/op 6.12x
```
golang-github-xiaq-persistent-0.0~git20190312.a1d9ac4/add-slowdown 0000775 0000000 0000000 00000002021 13441710145 0024432 0 ustar 00root root 0000000 0000000 #!/usr/bin/env elvish
# Parse an output of "go test -bench .", annotating benchmark results for
# persistent operations with the slowdown ratio compared to their native
# counterparts.
use re
fn extract [line]{
# Extract the name and ns/op of a benchmark entry.
fields = [(re:split '\s+' $line)]
if (not (eq $fields[-1] ns/op)) {
fail 'Last column of '(repr $line)' not ns/op'
}
put $fields[0] $fields[-2]
}
native = [&]
each [line]{
if (re:match Native $line) {
# Remember the result so that it can be used later.
name data = (extract $line)
native[$name] = $data
} elif (re:match Persistent $line) {
# Calculate slowdown and append to the end of the line.
name data = (extract $line)
native-name = (re:replace Persistent Native $name)
if (not (has-key $native $native-name)) {
fail 'Native counterpart for '$name' not found'
}
line = $line' '(printf '%.2f' (/ $data $native[$native-name]))'x'
}
echo $line
}
golang-github-xiaq-persistent-0.0~git20190312.a1d9ac4/epl-v10.html 0000664 0000000 0000000 00000031165 13441710145 0024167 0 ustar 00root root 0000000 0000000
Eclipse Public License - Version 1.0
Eclipse Public License - v 1.0
THE ACCOMPANYING PROGRAM IS PROVIDED UNDER THE TERMS OF THIS ECLIPSE
PUBLIC LICENSE ("AGREEMENT"). ANY USE, REPRODUCTION OR
DISTRIBUTION OF THE PROGRAM CONSTITUTES RECIPIENT'S ACCEPTANCE OF THIS
AGREEMENT.
1. DEFINITIONS
"Contribution" means:
a) in the case of the initial Contributor, the initial
code and documentation distributed under this Agreement, and
b) in the case of each subsequent Contributor:
i) changes to the Program, and
ii) additions to the Program;
where such changes and/or additions to the Program
originate from and are distributed by that particular Contributor. A
Contribution 'originates' from a Contributor if it was added to the
Program by such Contributor itself or anyone acting on such
Contributor's behalf. Contributions do not include additions to the
Program which: (i) are separate modules of software distributed in
conjunction with the Program under their own license agreement, and (ii)
are not derivative works of the Program.
"Contributor" means any person or entity that distributes
the Program.
"Licensed Patents" mean patent claims licensable by a
Contributor which are necessarily infringed by the use or sale of its
Contribution alone or when combined with the Program.
"Program" means the Contributions distributed in accordance
with this Agreement.
"Recipient" means anyone who receives the Program under
this Agreement, including all Contributors.
2. GRANT OF RIGHTS
a) Subject to the terms of this Agreement, each
Contributor hereby grants Recipient a non-exclusive, worldwide,
royalty-free copyright license to reproduce, prepare derivative works
of, publicly display, publicly perform, distribute and sublicense the
Contribution of such Contributor, if any, and such derivative works, in
source code and object code form.
b) Subject to the terms of this Agreement, each
Contributor hereby grants Recipient a non-exclusive, worldwide,
royalty-free patent license under Licensed Patents to make, use, sell,
offer to sell, import and otherwise transfer the Contribution of such
Contributor, if any, in source code and object code form. This patent
license shall apply to the combination of the Contribution and the
Program if, at the time the Contribution is added by the Contributor,
such addition of the Contribution causes such combination to be covered
by the Licensed Patents. The patent license shall not apply to any other
combinations which include the Contribution. No hardware per se is
licensed hereunder.
c) Recipient understands that although each Contributor
grants the licenses to its Contributions set forth herein, no assurances
are provided by any Contributor that the Program does not infringe the
patent or other intellectual property rights of any other entity. Each
Contributor disclaims any liability to Recipient for claims brought by
any other entity based on infringement of intellectual property rights
or otherwise. As a condition to exercising the rights and licenses
granted hereunder, each Recipient hereby assumes sole responsibility to
secure any other intellectual property rights needed, if any. For
example, if a third party patent license is required to allow Recipient
to distribute the Program, it is Recipient's responsibility to acquire
that license before distributing the Program.
d) Each Contributor represents that to its knowledge it
has sufficient copyright rights in its Contribution, if any, to grant
the copyright license set forth in this Agreement.
3. REQUIREMENTS
A Contributor may choose to distribute the Program in object code
form under its own license agreement, provided that:
a) it complies with the terms and conditions of this
Agreement; and
b) its license agreement:
i) effectively disclaims on behalf of all Contributors
all warranties and conditions, express and implied, including warranties
or conditions of title and non-infringement, and implied warranties or
conditions of merchantability and fitness for a particular purpose;
ii) effectively excludes on behalf of all Contributors
all liability for damages, including direct, indirect, special,
incidental and consequential damages, such as lost profits;
iii) states that any provisions which differ from this
Agreement are offered by that Contributor alone and not by any other
party; and
iv) states that source code for the Program is available
from such Contributor, and informs licensees how to obtain it in a
reasonable manner on or through a medium customarily used for software
exchange.
When the Program is made available in source code form:
a) it must be made available under this Agreement; and
b) a copy of this Agreement must be included with each
copy of the Program.
Contributors may not remove or alter any copyright notices contained
within the Program.
Each Contributor must identify itself as the originator of its
Contribution, if any, in a manner that reasonably allows subsequent
Recipients to identify the originator of the Contribution.
4. COMMERCIAL DISTRIBUTION
Commercial distributors of software may accept certain
responsibilities with respect to end users, business partners and the
like. While this license is intended to facilitate the commercial use of
the Program, the Contributor who includes the Program in a commercial
product offering should do so in a manner which does not create
potential liability for other Contributors. Therefore, if a Contributor
includes the Program in a commercial product offering, such Contributor
("Commercial Contributor") hereby agrees to defend and
indemnify every other Contributor ("Indemnified Contributor")
against any losses, damages and costs (collectively "Losses")
arising from claims, lawsuits and other legal actions brought by a third
party against the Indemnified Contributor to the extent caused by the
acts or omissions of such Commercial Contributor in connection with its
distribution of the Program in a commercial product offering. The
obligations in this section do not apply to any claims or Losses
relating to any actual or alleged intellectual property infringement. In
order to qualify, an Indemnified Contributor must: a) promptly notify
the Commercial Contributor in writing of such claim, and b) allow the
Commercial Contributor to control, and cooperate with the Commercial
Contributor in, the defense and any related settlement negotiations. The
Indemnified Contributor may participate in any such claim at its own
expense.
For example, a Contributor might include the Program in a commercial
product offering, Product X. That Contributor is then a Commercial
Contributor. If that Commercial Contributor then makes performance
claims, or offers warranties related to Product X, those performance
claims and warranties are such Commercial Contributor's responsibility
alone. Under this section, the Commercial Contributor would have to
defend claims against the other Contributors related to those
performance claims and warranties, and if a court requires any other
Contributor to pay any damages as a result, the Commercial Contributor
must pay those damages.
5. NO WARRANTY
EXCEPT AS EXPRESSLY SET FORTH IN THIS AGREEMENT, THE PROGRAM IS
PROVIDED 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. Each Recipient is solely
responsible for determining the appropriateness of using and
distributing the Program and assumes all risks associated with its
exercise of rights under this Agreement , including but not limited to
the risks and costs of program errors, compliance with applicable laws,
damage to or loss of data, programs or equipment, and unavailability or
interruption of operations.
6. DISCLAIMER OF LIABILITY
EXCEPT AS EXPRESSLY SET FORTH IN THIS AGREEMENT, NEITHER RECIPIENT
NOR ANY CONTRIBUTORS SHALL HAVE ANY LIABILITY FOR ANY DIRECT, INDIRECT,
INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING
WITHOUT LIMITATION LOST PROFITS), HOWEVER CAUSED AND ON ANY THEORY OF
LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING
NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OR
DISTRIBUTION OF THE PROGRAM OR THE EXERCISE OF ANY RIGHTS GRANTED
HEREUNDER, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGES.
7. GENERAL
If any provision of this Agreement is invalid or unenforceable under
applicable law, it shall not affect the validity or enforceability of
the remainder of the terms of this Agreement, and without further action
by the parties hereto, such provision shall be reformed to the minimum
extent necessary to make such provision valid and enforceable.
If Recipient institutes patent litigation against any entity
(including a cross-claim or counterclaim in a lawsuit) alleging that the
Program itself (excluding combinations of the Program with other
software or hardware) infringes such Recipient's patent(s), then such
Recipient's rights granted under Section 2(b) shall terminate as of the
date such litigation is filed.
All Recipient's rights under this Agreement shall terminate if it
fails to comply with any of the material terms or conditions of this
Agreement and does not cure such failure in a reasonable period of time
after becoming aware of such noncompliance. If all Recipient's rights
under this Agreement terminate, Recipient agrees to cease use and
distribution of the Program as soon as reasonably practicable. However,
Recipient's obligations under this Agreement and any licenses granted by
Recipient relating to the Program shall continue and survive.
Everyone is permitted to copy and distribute copies of this
Agreement, but in order to avoid inconsistency the Agreement is
copyrighted and may only be modified in the following manner. The
Agreement Steward reserves the right to publish new versions (including
revisions) of this Agreement from time to time. No one other than the
Agreement Steward has the right to modify this Agreement. The Eclipse
Foundation is the initial Agreement Steward. The Eclipse Foundation may
assign the responsibility to serve as the Agreement Steward to a
suitable separate entity. Each new version of the Agreement will be
given a distinguishing version number. The Program (including
Contributions) may always be distributed subject to the version of the
Agreement under which it was received. In addition, after a new version
of the Agreement is published, Contributor may elect to distribute the
Program (including its Contributions) under the new version. Except as
expressly stated in Sections 2(a) and 2(b) above, Recipient receives no
rights or licenses to the intellectual property of any Contributor under
this Agreement, whether expressly, by implication, estoppel or
otherwise. All rights in the Program not expressly granted under this
Agreement are reserved.
This Agreement is governed by the laws of the State of New York and
the intellectual property laws of the United States of America. No party
to this Agreement will bring a legal action under this Agreement more
than one year after the cause of action arose. Each party waives its
rights to a jury trial in any resulting litigation.
golang-github-xiaq-persistent-0.0~git20190312.a1d9ac4/go.mod 0000664 0000000 0000000 00000000053 13441710145 0023213 0 ustar 00root root 0000000 0000000 module github.com/xiaq/persistent
go 1.12
golang-github-xiaq-persistent-0.0~git20190312.a1d9ac4/hash/ 0000775 0000000 0000000 00000000000 13441710145 0023032 5 ustar 00root root 0000000 0000000 golang-github-xiaq-persistent-0.0~git20190312.a1d9ac4/hash/hash.go 0000664 0000000 0000000 00000001712 13441710145 0024305 0 ustar 00root root 0000000 0000000 // Package hash contains some common hash functions suitable for use in hash
// maps.
package hash
import "unsafe"
const DJBInit uint32 = 5381
func DJBCombine(acc, h uint32) uint32 {
return mul33(acc) + h
}
func DJB(hs ...uint32) uint32 {
acc := DJBInit
for _, h := range hs {
acc = DJBCombine(acc, h)
}
return acc
}
func UInt32(u uint32) uint32 {
return u
}
func UInt64(u uint64) uint32 {
return mul33(uint32(u>>32)) + uint32(u&0xffffffff)
}
func Pointer(p unsafe.Pointer) uint32 {
if unsafe.Sizeof(p) == 4 {
return UInt32(uint32(uintptr(p)))
} else {
return UInt64(uint64(uintptr(p)))
}
// NOTE: We don't care about 128-bit archs yet.
}
func UIntPtr(u uintptr) uint32 {
if unsafe.Sizeof(u) == 4 {
return UInt32(uint32(u))
} else {
return UInt64(uint64(u))
}
}
func String(s string) uint32 {
h := DJBInit
for i := 0; i < len(s); i++ {
h = DJBCombine(h, uint32(s[i]))
}
return h
}
func mul33(u uint32) uint32 {
return u<<5 + u
}
golang-github-xiaq-persistent-0.0~git20190312.a1d9ac4/hashmap/ 0000775 0000000 0000000 00000000000 13441710145 0023530 5 ustar 00root root 0000000 0000000 golang-github-xiaq-persistent-0.0~git20190312.a1d9ac4/hashmap/hashmap.go 0000664 0000000 0000000 00000032757 13441710145 0025516 0 ustar 00root root 0000000 0000000 // Package hashmap implements persistent hashmap.
package hashmap
import (
"bytes"
"encoding"
"encoding/json"
"fmt"
"reflect"
"strconv"
)
const (
chunkBits = 5
nodeCap = 1 << chunkBits
chunkMask = nodeCap - 1
)
// Equal is the type of a function that reports whether two keys are equal.
type Equal func(k1, k2 interface{}) bool
// Hash is the type of a function that returns the hash code of a key.
type Hash func(k interface{}) uint32
// New takes an equality function and a hash function, and returns an empty
// Map.
func New(e Equal, h Hash) Map {
return &hashMap{0, emptyBitmapNode, e, h}
}
type hashMap struct {
count int
root node
equal Equal
hash Hash
}
func (m *hashMap) Len() int {
return m.count
}
func (m *hashMap) Index(k interface{}) (interface{}, bool) {
return m.root.find(0, m.hash(k), k, m.equal)
}
func (m *hashMap) Assoc(k, v interface{}) Map {
newRoot, added := m.root.assoc(0, m.hash(k), k, v, m.hash, m.equal)
newCount := m.count
if added {
newCount++
}
return &hashMap{newCount, newRoot, m.equal, m.hash}
}
func (m *hashMap) Dissoc(k interface{}) Map {
newRoot, deleted := m.root.without(0, m.hash(k), k, m.equal)
newCount := m.count
if deleted {
newCount--
}
return &hashMap{newCount, newRoot, m.equal, m.hash}
}
func (m *hashMap) Iterator() Iterator {
return m.root.iterator()
}
func (m *hashMap) MarshalJSON() ([]byte, error) {
var buf bytes.Buffer
buf.WriteByte('{')
first := true
for it := m.Iterator(); it.HasElem(); it.Next() {
if first {
first = false
} else {
buf.WriteByte(',')
}
k, v := it.Elem()
kString, err := convertKey(k)
if err != nil {
return nil, err
}
kBytes, err := json.Marshal(kString)
if err != nil {
return nil, err
}
vBytes, err := json.Marshal(v)
if err != nil {
return nil, err
}
buf.Write(kBytes)
buf.WriteByte(':')
buf.Write(vBytes)
}
buf.WriteByte('}')
return buf.Bytes(), nil
}
// convertKey converts a map key to a string. The implementation matches the
// behavior of how json.Marshal encodes keys of the builtin map type.
func convertKey(k interface{}) (string, error) {
kref := reflect.ValueOf(k)
if kref.Kind() == reflect.String {
return kref.String(), nil
}
if t, ok := k.(encoding.TextMarshaler); ok {
b2, err := t.MarshalText()
if err != nil {
return "", err
}
return string(b2), nil
}
switch kref.Kind() {
case reflect.Int, reflect.Int8, reflect.Int16, reflect.Int32, reflect.Int64:
return strconv.FormatInt(kref.Int(), 10), nil
case reflect.Uint, reflect.Uint8, reflect.Uint16, reflect.Uint32, reflect.Uint64, reflect.Uintptr:
return strconv.FormatUint(kref.Uint(), 10), nil
}
return "", fmt.Errorf("unsupported key type %T", k)
}
// node is an interface for all nodes in the hash map tree.
type node interface {
// assoc adds a new pair of key and value. It returns the new node, and
// whether the key did not exist before (i.e. a new pair has been added,
// instead of replaced).
assoc(shift, hash uint32, k, v interface{}, h Hash, eq Equal) (node, bool)
// without removes a key. It returns the new node and whether the key did
// not exist before (i.e. a key was indeed removed).
without(shift, hash uint32, k interface{}, eq Equal) (node, bool)
// find finds the value for a key. It returns the found value (if any) and
// whether such a pair exists.
find(shift, hash uint32, k interface{}, eq Equal) (interface{}, bool)
// iterator returns an iterator.
iterator() Iterator
}
// arrayNode stores all of its children in an array. The array is always at
// least 1/4 full, otherwise it will be packed into a bitmapNode.
type arrayNode struct {
nChildren int
children [nodeCap]node
}
func (n *arrayNode) withNewChild(i uint32, newChild node, d int) *arrayNode {
newChildren := n.children
newChildren[i] = newChild
return &arrayNode{n.nChildren + d, newChildren}
}
func (n *arrayNode) assoc(shift, hash uint32, k, v interface{}, h Hash, eq Equal) (node, bool) {
idx := chunk(shift, hash)
child := n.children[idx]
if child == nil {
newChild, _ := emptyBitmapNode.assoc(shift+chunkBits, hash, k, v, h, eq)
return n.withNewChild(idx, newChild, 1), true
}
newChild, added := child.assoc(shift+chunkBits, hash, k, v, h, eq)
return n.withNewChild(idx, newChild, 0), added
}
func (n *arrayNode) without(shift, hash uint32, k interface{}, eq Equal) (node, bool) {
idx := chunk(shift, hash)
child := n.children[idx]
if child == nil {
return n, false
}
newChild, _ := child.without(shift+chunkBits, hash, k, eq)
if newChild == child {
return n, false
}
if newChild == emptyBitmapNode {
if n.nChildren <= nodeCap/4 {
// less than 1/4 full; shrink
return n.pack(int(idx)), true
}
return n.withNewChild(idx, nil, -1), true
}
return n.withNewChild(idx, newChild, 0), true
}
func (n *arrayNode) pack(skip int) *bitmapNode {
newNode := bitmapNode{0, make([]mapEntry, n.nChildren-1)}
j := 0
for i, child := range n.children {
// TODO(xiaq): Benchmark performance difference after unrolling this
// into two loops without the if
if i != skip && child != nil {
newNode.bitmap |= 1 << uint(i)
newNode.entries[j].value = child
j++
}
}
return &newNode
}
func (n *arrayNode) find(shift, hash uint32, k interface{}, eq Equal) (interface{}, bool) {
idx := chunk(shift, hash)
child := n.children[idx]
if child == nil {
return nil, false
}
return child.find(shift+chunkBits, hash, k, eq)
}
func (n *arrayNode) iterator() Iterator {
it := &arrayNodeIterator{n, 0, nil}
it.fixCurrent()
return it
}
type arrayNodeIterator struct {
n *arrayNode
index int
current Iterator
}
func (it *arrayNodeIterator) fixCurrent() {
for ; it.index < nodeCap && it.n.children[it.index] == nil; it.index++ {
}
if it.index < nodeCap {
it.current = it.n.children[it.index].iterator()
} else {
it.current = nil
}
}
func (it *arrayNodeIterator) Elem() (interface{}, interface{}) {
return it.current.Elem()
}
func (it *arrayNodeIterator) HasElem() bool {
return it.current != nil
}
func (it *arrayNodeIterator) Next() {
it.current.Next()
if !it.current.HasElem() {
it.index++
it.fixCurrent()
}
}
var emptyBitmapNode = &bitmapNode{}
type bitmapNode struct {
bitmap uint32
entries []mapEntry
}
// mapEntry is a map entry. When used in a collisionNode, it is also an entry
// with non-nil key. When used in a bitmapNode, it is also abused to represent
// children when the key is nil.
type mapEntry struct {
key interface{}
value interface{}
}
func chunk(shift, hash uint32) uint32 {
return (hash >> shift) & chunkMask
}
func bitpos(shift, hash uint32) uint32 {
return 1 << chunk(shift, hash)
}
func index(bitmap, bit uint32) uint32 {
return popCount(bitmap & (bit - 1))
}
const (
m1 uint32 = 0x55555555
m2 = 0x33333333
m4 = 0x0f0f0f0f
m8 = 0x00ff00ff
m16 = 0x0000ffff
)
// TODO(xiaq): Use an optimized implementation.
func popCount(u uint32) uint32 {
u = (u & m1) + ((u >> 1) & m1)
u = (u & m2) + ((u >> 2) & m2)
u = (u & m4) + ((u >> 4) & m4)
u = (u & m8) + ((u >> 8) & m8)
u = (u & m16) + ((u >> 16) & m16)
return u
}
func createNode(shift uint32, k1 interface{}, v1 interface{}, h2 uint32, k2 interface{}, v2 interface{}, h Hash, eq Equal) node {
h1 := h(k1)
if h1 == h2 {
return &collisionNode{h1, []mapEntry{{k1, v1}, {k2, v2}}}
}
n, _ := emptyBitmapNode.assoc(shift, h1, k1, v1, h, eq)
n, _ = n.assoc(shift, h2, k2, v2, h, eq)
return n
}
func (n *bitmapNode) unpack(shift, idx uint32, newChild node, h Hash, eq Equal) *arrayNode {
var newNode arrayNode
newNode.nChildren = len(n.entries) + 1
newNode.children[idx] = newChild
j := 0
for i := uint(0); i < nodeCap; i++ {
if (n.bitmap>>i)&1 != 0 {
entry := n.entries[j]
j++
if entry.key == nil {
newNode.children[i] = entry.value.(node)
} else {
newNode.children[i], _ = emptyBitmapNode.assoc(
shift+chunkBits, h(entry.key), entry.key, entry.value, h, eq)
}
}
}
return &newNode
}
func (n *bitmapNode) withoutEntry(bit, idx uint32) *bitmapNode {
if n.bitmap == bit {
return emptyBitmapNode
}
return &bitmapNode{n.bitmap ^ bit, withoutEntry(n.entries, idx)}
}
func withoutEntry(entries []mapEntry, idx uint32) []mapEntry {
newEntries := make([]mapEntry, len(entries)-1)
copy(newEntries[:idx], entries[:idx])
copy(newEntries[idx:], entries[idx+1:])
return newEntries
}
func (n *bitmapNode) withReplacedEntry(i uint32, entry mapEntry) *bitmapNode {
return &bitmapNode{n.bitmap, replaceEntry(n.entries, i, entry.key, entry.value)}
}
func replaceEntry(entries []mapEntry, i uint32, k, v interface{}) []mapEntry {
newEntries := append([]mapEntry(nil), entries...)
newEntries[i] = mapEntry{k, v}
return newEntries
}
func (n *bitmapNode) assoc(shift, hash uint32, k, v interface{}, h Hash, eq Equal) (node, bool) {
bit := bitpos(shift, hash)
idx := index(n.bitmap, bit)
if n.bitmap&bit == 0 {
// Entry does not exist yet
nEntries := len(n.entries)
if nEntries >= nodeCap/2 {
// Unpack into an arrayNode
newNode, _ := emptyBitmapNode.assoc(shift+chunkBits, hash, k, v, h, eq)
return n.unpack(shift, chunk(shift, hash), newNode, h, eq), true
}
// Add a new entry
newEntries := make([]mapEntry, len(n.entries)+1)
copy(newEntries[:idx], n.entries[:idx])
newEntries[idx] = mapEntry{k, v}
copy(newEntries[idx+1:], n.entries[idx:])
return &bitmapNode{n.bitmap | bit, newEntries}, true
}
// Entry exists
entry := n.entries[idx]
if entry.key == nil {
// Non-leaf child
child := entry.value.(node)
newChild, added := child.assoc(shift+chunkBits, hash, k, v, h, eq)
return n.withReplacedEntry(idx, mapEntry{nil, newChild}), added
}
// Leaf
if eq(k, entry.key) {
// Identical key, replace
return n.withReplacedEntry(idx, mapEntry{k, v}), false
}
// Create and insert new inner node
newNode := createNode(shift+chunkBits, entry.key, entry.value, hash, k, v, h, eq)
return n.withReplacedEntry(idx, mapEntry{nil, newNode}), true
}
func (n *bitmapNode) without(shift, hash uint32, k interface{}, eq Equal) (node, bool) {
bit := bitpos(shift, hash)
if n.bitmap&bit == 0 {
return n, false
}
idx := index(n.bitmap, bit)
entry := n.entries[idx]
if entry.key == nil {
// Non-leaf child
child := entry.value.(node)
newChild, deleted := child.without(shift+chunkBits, hash, k, eq)
if newChild == child {
return n, false
}
if newChild == emptyBitmapNode {
return n.withoutEntry(bit, idx), true
}
return n.withReplacedEntry(idx, mapEntry{nil, newChild}), deleted
} else if eq(entry.key, k) {
// Leaf, and this is the entry to delete.
return n.withoutEntry(bit, idx), true
}
// Nothing to delete.
return n, false
}
func (n *bitmapNode) find(shift, hash uint32, k interface{}, eq Equal) (interface{}, bool) {
bit := bitpos(shift, hash)
if n.bitmap&bit == 0 {
return nil, false
}
idx := index(n.bitmap, bit)
entry := n.entries[idx]
if entry.key == nil {
child := entry.value.(node)
return child.find(shift+chunkBits, hash, k, eq)
} else if eq(entry.key, k) {
return entry.value, true
}
return nil, false
}
func (n *bitmapNode) iterator() Iterator {
it := &bitmapNodeIterator{n, 0, nil}
it.fixCurrent()
return it
}
type bitmapNodeIterator struct {
n *bitmapNode
index int
current Iterator
}
func (it *bitmapNodeIterator) fixCurrent() {
if it.index < len(it.n.entries) {
entry := it.n.entries[it.index]
if entry.key == nil {
it.current = entry.value.(node).iterator()
} else {
it.current = nil
}
} else {
it.current = nil
}
}
func (it *bitmapNodeIterator) Elem() (interface{}, interface{}) {
if it.current != nil {
return it.current.Elem()
}
entry := it.n.entries[it.index]
return entry.key, entry.value
}
func (it *bitmapNodeIterator) HasElem() bool {
return it.index < len(it.n.entries)
}
func (it *bitmapNodeIterator) Next() {
if it.current != nil {
it.current.Next()
}
if it.current == nil || !it.current.HasElem() {
it.index++
it.fixCurrent()
}
}
type collisionNode struct {
hash uint32
entries []mapEntry
}
func (n *collisionNode) assoc(shift, hash uint32, k, v interface{}, h Hash, eq Equal) (node, bool) {
if hash == n.hash {
idx := n.findIndex(k, eq)
if idx != -1 {
return &collisionNode{
n.hash, replaceEntry(n.entries, uint32(idx), k, v)}, false
}
newEntries := make([]mapEntry, len(n.entries)+1)
copy(newEntries[:len(n.entries)], n.entries[:])
newEntries[len(n.entries)] = mapEntry{k, v}
return &collisionNode{n.hash, newEntries}, true
}
// Wrap in a bitmapNode and add the entry
wrap := bitmapNode{bitpos(shift, n.hash), []mapEntry{{nil, n}}}
return wrap.assoc(shift, hash, k, v, h, eq)
}
func (n *collisionNode) without(shift, hash uint32, k interface{}, eq Equal) (node, bool) {
idx := n.findIndex(k, eq)
if idx == -1 {
return n, false
}
if len(n.entries) == 1 {
return emptyBitmapNode, true
}
return &collisionNode{n.hash, withoutEntry(n.entries, uint32(idx))}, true
}
func (n *collisionNode) find(shift, hash uint32, k interface{}, eq Equal) (interface{}, bool) {
idx := n.findIndex(k, eq)
if idx == -1 {
return nil, false
}
return n.entries[idx].value, true
}
func (n *collisionNode) findIndex(k interface{}, eq Equal) int {
for i, entry := range n.entries {
if eq(k, entry.key) {
return i
}
}
return -1
}
func (n *collisionNode) iterator() Iterator {
return &collisionNodeIterator{n, 0}
}
type collisionNodeIterator struct {
n *collisionNode
index int
}
func (it *collisionNodeIterator) Elem() (interface{}, interface{}) {
entry := it.n.entries[it.index]
return entry.key, entry.value
}
func (it *collisionNodeIterator) HasElem() bool {
return it.index < len(it.n.entries)
}
func (it *collisionNodeIterator) Next() {
it.index++
}
golang-github-xiaq-persistent-0.0~git20190312.a1d9ac4/hashmap/hashmap_test.go 0000664 0000000 0000000 00000020303 13441710145 0026535 0 ustar 00root root 0000000 0000000 package hashmap
import (
"math/rand"
"strconv"
"testing"
"time"
"github.com/xiaq/persistent/hash"
)
const (
NSequential = 0x1000
NCollision = 0x100
NRandom = 0x4000
NReplace = 0x200
SmallRandomPass = 0x100
NSmallRandom = 0x400
SmallRandomHighBound = 0x50
SmallRandomLowBound = 0x200
NArrayNode = 0x100
NIneffectiveDissoc = 0x200
N1 = nodeCap + 1
N2 = nodeCap*nodeCap + 1
N3 = nodeCap*nodeCap*nodeCap + 1
)
type testKey uint64
type anotherTestKey uint32
func equalFunc(k1, k2 interface{}) bool {
switch k1 := k1.(type) {
case testKey:
t2, ok := k2.(testKey)
return ok && k1 == t2
case anotherTestKey:
return false
default:
return k1 == k2
}
}
func hashFunc(k interface{}) uint32 {
switch k := k.(type) {
case uint32:
return k
case string:
return hash.String(k)
case testKey:
// Return the lower 32 bits for testKey. This is intended so that hash
// collisions can be easily constructed.
return uint32(k & 0xffffffff)
case anotherTestKey:
return uint32(k)
default:
return 0
}
}
var empty = New(equalFunc, hashFunc)
type refEntry struct {
k testKey
v string
}
func hex(i uint64) string {
return "0x" + strconv.FormatUint(i, 16)
}
func init() {
rand.Seed(time.Now().UTC().UnixNano())
}
var randomStrings []string
// getRandomStrings returns a slice of N3 random strings. It builds the slice
// once and caches it. If the slice is built for the first time, it stops the
// timer of the benchmark.
func getRandomStrings(b *testing.B) []string {
if randomStrings == nil {
b.StopTimer()
defer b.StartTimer()
randomStrings = make([]string, N3)
for i := 0; i < N3; i++ {
randomStrings[i] = makeRandomString()
}
}
return randomStrings
}
// makeRandomString builds a random string consisting of n bytes (randomized
// between 0 and 99) and each byte is randomized between 0 and 255. The string
// need not be valid UTF-8.
func makeRandomString() string {
bytes := make([]byte, rand.Intn(100))
for i := range bytes {
bytes[i] = byte(rand.Intn(256))
}
return string(bytes)
}
func TestHashMap(t *testing.T) {
var refEntries []refEntry
add := func(k testKey, v string) {
refEntries = append(refEntries, refEntry{k, v})
}
for i := 0; i < NSequential; i++ {
add(testKey(i), hex(uint64(i)))
}
for i := 0; i < NCollision; i++ {
add(testKey(uint64(i+1)<<32), "collision "+hex(uint64(i)))
}
for i := 0; i < NRandom; i++ {
// Avoid rand.Uint64 for compatibility with pre 1.8 Go
k := uint64(rand.Int63())>>31 | uint64(rand.Int63())<<32
add(testKey(k), "random "+hex(k))
}
for i := 0; i < NReplace; i++ {
k := uint64(rand.Int31n(NSequential))
add(testKey(k), "replace "+hex(k))
}
testHashMapWithRefEntries(t, refEntries)
}
func TestHashMapSmallRandom(t *testing.T) {
for p := 0; p < SmallRandomPass; p++ {
var refEntries []refEntry
add := func(k testKey, v string) {
refEntries = append(refEntries, refEntry{k, v})
}
for i := 0; i < NSmallRandom; i++ {
k := uint64(uint64(rand.Int31n(SmallRandomHighBound))<<32 |
uint64(rand.Int31n(SmallRandomLowBound)))
add(testKey(k), "random "+hex(k))
}
testHashMapWithRefEntries(t, refEntries)
}
}
var marshalJSONTests = []struct {
in Map
wantOut string
wantErr bool
}{
{makeHashMap(uint32(1), "a", "2", "b"), `{"1":"a","2":"b"}`, false},
// Invalid key type
{makeHashMap([]interface{}{}, "x"), "", true},
}
func TestMarshalJSON(t *testing.T) {
for i, test := range marshalJSONTests {
out, err := test.in.MarshalJSON()
if string(out) != test.wantOut {
t.Errorf("m%d.MarshalJSON -> out %s, want %s", i, out, test.wantOut)
}
if (err != nil) != test.wantErr {
var wantErr string
if test.wantErr {
wantErr = "non-nil"
} else {
wantErr = "nil"
}
t.Errorf("m%d.MarshalJSON -> err %v, want %s", i, err, wantErr)
}
}
}
func makeHashMap(data ...interface{}) Map {
m := empty
for i := 0; i+1 < len(data); i += 2 {
k, v := data[i], data[i+1]
m = m.Assoc(k, v)
}
return m
}
// testHashMapWithRefEntries tests the operations of a Map. It uses the supplied
// list of entries to build the map, and then test all its operations.
func testHashMapWithRefEntries(t *testing.T, refEntries []refEntry) {
m := empty
// Len of Empty should be 0.
if m.Len() != 0 {
t.Errorf("m.Len = %d, want %d", m.Len(), 0)
}
// Assoc and Len, test by building a map simutaneously.
ref := make(map[testKey]string, len(refEntries))
for _, e := range refEntries {
ref[e.k] = e.v
m = m.Assoc(e.k, e.v)
if m.Len() != len(ref) {
t.Errorf("m.Len = %d, want %d", m.Len(), len(ref))
}
}
// Index.
testMapContent(t, m, ref)
got, in := m.Index(anotherTestKey(0))
if in {
t.Errorf("m.Index returns entry %v", got)
}
// Iterator.
testIterator(t, m, ref)
// Dissoc.
// Ineffective ones.
for i := 0; i < NIneffectiveDissoc; i++ {
k := anotherTestKey(uint32(rand.Int31())>>15 | uint32(rand.Int31())<<16)
m = m.Dissoc(k)
if m.Len() != len(ref) {
t.Errorf("m.Dissoc removes item when it shouldn't")
}
}
// Effective ones.
for x := 0; x < len(refEntries); x++ {
i := rand.Intn(len(refEntries))
k := refEntries[i].k
delete(ref, k)
m = m.Dissoc(k)
if m.Len() != len(ref) {
t.Errorf("m.Len() = %d after removing, should be %v", m.Len(), len(ref))
}
_, in := m.Index(k)
if in {
t.Errorf("m.Index(%v) still returns item after removal", k)
}
// Checking all elements is expensive. Only do this 1% of the time.
if rand.Float64() < 0.01 {
testMapContent(t, m, ref)
testIterator(t, m, ref)
}
}
}
func testMapContent(t *testing.T, m Map, ref map[testKey]string) {
for k, v := range ref {
got, in := m.Index(k)
if !in {
t.Errorf("m.Index 0x%x returns no entry", k)
}
if got != v {
t.Errorf("m.Index(0x%x) = %v, want %v", k, got, v)
}
}
}
func testIterator(t *testing.T, m Map, ref map[testKey]string) {
ref2 := map[interface{}]interface{}{}
for k, v := range ref {
ref2[k] = v
}
for it := m.Iterator(); it.HasElem(); it.Next() {
k, v := it.Elem()
if ref2[k] != v {
t.Errorf("iterator yields unexpected pair %v, %v", k, v)
}
delete(ref2, k)
}
if len(ref2) != 0 {
t.Errorf("iterating was not exhaustive")
}
}
func BenchmarkSequentialConsNative1(b *testing.B) { nativeSequentialAdd(b.N, N1) }
func BenchmarkSequentialConsNative2(b *testing.B) { nativeSequentialAdd(b.N, N2) }
func BenchmarkSequentialConsNative3(b *testing.B) { nativeSequentialAdd(b.N, N3) }
// nativeSequntialAdd starts with an empty native map and adds elements 0...n-1
// to the map, using the same value as the key, repeating for N times.
func nativeSequentialAdd(N int, n uint32) {
for r := 0; r < N; r++ {
m := make(map[uint32]uint32)
for i := uint32(0); i < n; i++ {
m[i] = i
}
}
}
func BenchmarkSequentialConsPersistent1(b *testing.B) { sequentialCons(b.N, N1) }
func BenchmarkSequentialConsPersistent2(b *testing.B) { sequentialCons(b.N, N2) }
func BenchmarkSequentialConsPersistent3(b *testing.B) { sequentialCons(b.N, N3) }
// sequentialCons starts with an empty hash map and adds elements 0...n-1 to the
// map, using the same value as the key, repeating for N times.
func sequentialCons(N int, n uint32) {
for r := 0; r < N; r++ {
m := empty
for i := uint32(0); i < n; i++ {
m = m.Assoc(i, i)
}
}
}
func BenchmarkRandomStringsConsNative1(b *testing.B) { nativeRandomStringsAdd(b, N1) }
func BenchmarkRandomStringsConsNative2(b *testing.B) { nativeRandomStringsAdd(b, N2) }
func BenchmarkRandomStringsConsNative3(b *testing.B) { nativeRandomStringsAdd(b, N3) }
// nativeSequntialAdd starts with an empty native map and adds n random strings
// to the map, using the same value as the key, repeating for b.N times.
func nativeRandomStringsAdd(b *testing.B, n int) {
ss := getRandomStrings(b)
for r := 0; r < b.N; r++ {
m := make(map[string]string)
for i := 0; i < n; i++ {
s := ss[i]
m[s] = s
}
}
}
func BenchmarkRandomStringsConsPersistent1(b *testing.B) { randomStringsCons(b, N1) }
func BenchmarkRandomStringsConsPersistent2(b *testing.B) { randomStringsCons(b, N2) }
func BenchmarkRandomStringsConsPersistent3(b *testing.B) { randomStringsCons(b, N3) }
func randomStringsCons(b *testing.B, n int) {
ss := getRandomStrings(b)
for r := 0; r < b.N; r++ {
m := empty
for i := 0; i < n; i++ {
s := ss[i]
m = m.Assoc(s, s)
}
}
}
golang-github-xiaq-persistent-0.0~git20190312.a1d9ac4/hashmap/map.go 0000664 0000000 0000000 00000002677 13441710145 0024650 0 ustar 00root root 0000000 0000000 package hashmap
import "encoding/json"
// Map is a persistent associative data structure mapping keys to values. It
// is immutable, and supports near-O(1) operations to create modified version of
// the map that shares the underlying data structure. Because it is immutable,
// all of its methods are safe for concurrent use.
type Map interface {
json.Marshaler
// Len returns the length of the map.
Len() int
// Index returns whether there is a value associated with the given key, and
// that value or nil.
Index(k interface{}) (interface{}, bool)
// Assoc returns an almost identical map, with the given key associated with
// the given value.
Assoc(k, v interface{}) Map
// Dissoc returns an almost identical map, with the given key associated
// with no value.
Dissoc(k interface{}) Map
// Iterator returns an iterator over the map.
Iterator() Iterator
}
// Iterator is an iterator over map elements. It can be used like this:
//
// for it := m.Iterator(); it.HasElem(); it.Next() {
// key, value := it.Elem()
// // do something with elem...
// }
type Iterator interface {
// Elem returns the current key-value pair.
Elem() (interface{}, interface{})
// HasElem returns whether the iterator is pointing to an element.
HasElem() bool
// Next moves the iterator to the next position.
Next()
}
// HasKey reports whether a Map has the given key.
func HasKey(m Map, k interface{}) bool {
_, ok := m.Index(k)
return ok
}
golang-github-xiaq-persistent-0.0~git20190312.a1d9ac4/list/ 0000775 0000000 0000000 00000000000 13441710145 0023062 5 ustar 00root root 0000000 0000000 golang-github-xiaq-persistent-0.0~git20190312.a1d9ac4/list/list.go 0000664 0000000 0000000 00000001347 13441710145 0024371 0 ustar 00root root 0000000 0000000 // Package list implements persistent list.
package list
// List is a persistent list.
type List interface {
// Len returns the number of values in the list.
Len() int
// Cons returns a new list with an additional value in the front.
Cons(interface{}) List
// First returns the first value in the list.
First() interface{}
// Rest returns the list after the first value.
Rest() List
}
// Empty is an empty list.
var Empty List = &list{}
type list struct {
first interface{}
rest *list
count int
}
func (l *list) Len() int {
return l.count
}
func (l *list) Cons(val interface{}) List {
return &list{val, l, l.count + 1}
}
func (l *list) First() interface{} {
return l.first
}
func (l *list) Rest() List {
return l.rest
}
golang-github-xiaq-persistent-0.0~git20190312.a1d9ac4/persistent.go 0000664 0000000 0000000 00000000176 13441710145 0024642 0 ustar 00root root 0000000 0000000 // Package persistent contains subpackages for persistent data structures,
// similar to those of Clojure.
package persistent
golang-github-xiaq-persistent-0.0~git20190312.a1d9ac4/vector/ 0000775 0000000 0000000 00000000000 13441710145 0023411 5 ustar 00root root 0000000 0000000 golang-github-xiaq-persistent-0.0~git20190312.a1d9ac4/vector/vector.go 0000664 0000000 0000000 00000024147 13441710145 0025252 0 ustar 00root root 0000000 0000000 // Package vector implements persistent vector.
package vector
import (
"bytes"
"encoding/json"
"fmt"
)
const (
chunkBits = 5
nodeSize = 1 << chunkBits
tailMaxLen = nodeSize
chunkMask = nodeSize - 1
)
// Vector is a persistent sequential container for arbitrary values. It supports
// O(1) lookup by index, modification by index, and insertion and removal
// operations at the end. Being a persistent variant of the data structure, it
// is immutable, and provides O(1) operations to create modified versions of the
// vector that shares the underlying data structure, making it suitable for
// concurrent access. The empty value is a valid empty vector.
type Vector interface {
json.Marshaler
// Len returns the length of the vector.
Len() int
// Index returns the i-th element of the vector, if it exists. The second
// return value indicates whether the element exists.
Index(i int) (interface{}, bool)
// Assoc returns an almost identical Vector, with the i-th element
// replaced. If the index is smaller than 0 or greater than the length of
// the vector, it returns nil. If the index is equal to the size of the
// vector, it is equivalent to Cons.
Assoc(i int, val interface{}) Vector
// Cons returns an almost identical Vector, with an additional element
// appended to the end.
Cons(val interface{}) Vector
// Pop returns an almost identical Vector, with the last element removed. It
// returns nil if the vector is already empty.
Pop() Vector
// SubVector returns a subvector containing the elements from i up to but
// not including j.
SubVector(i, j int) Vector
// Iterator returns an iterator over the vector.
Iterator() Iterator
}
// Iterator is an iterator over vector elements. It can be used like this:
//
// for it := v.Iterator(); it.HasElem(); it.Next() {
// elem := it.Elem()
// // do something with elem...
// }
type Iterator interface {
// Elem returns the element at the current position.
Elem() interface{}
// HasElem returns whether the iterator is pointing to an element.
HasElem() bool
// Next moves the iterator to the next position.
Next()
}
type vector struct {
count int
// height of the tree structure, defined to be 0 when root is a leaf.
height uint
root node
tail []interface{}
}
// Empty is an empty Vector.
var Empty Vector = &vector{}
// node is a node in the vector tree. It is always of the size nodeSize.
type node *[nodeSize]interface{}
func newNode() node {
return node(&[nodeSize]interface{}{})
}
func clone(n node) node {
a := *n
return node(&a)
}
func nodeFromSlice(s []interface{}) node {
var n [nodeSize]interface{}
copy(n[:], s)
return &n
}
// Count returns the number of elements in a Vector.
func (v *vector) Len() int {
return v.count
}
// treeSize returns the number of elements stored in the tree (as opposed to the
// tail).
func (v *vector) treeSize() int {
if v.count < tailMaxLen {
return 0
}
return ((v.count - 1) >> chunkBits) << chunkBits
}
func (v *vector) Index(i int) (interface{}, bool) {
if i < 0 || i >= v.count {
return nil, false
}
// The following is very similar to sliceFor, but is implemented separately
// to avoid unncessary copying.
if i >= v.treeSize() {
return v.tail[i&chunkMask], true
}
n := v.root
for shift := v.height * chunkBits; shift > 0; shift -= chunkBits {
n = n[(i>>shift)&chunkMask].(node)
}
return n[i&chunkMask], true
}
// sliceFor returns the slice where the i-th element is stored. The index must
// be in bound.
func (v *vector) sliceFor(i int) []interface{} {
if i >= v.treeSize() {
return v.tail
}
n := v.root
for shift := v.height * chunkBits; shift > 0; shift -= chunkBits {
n = n[(i>>shift)&chunkMask].(node)
}
return n[:]
}
func (v *vector) Assoc(i int, val interface{}) Vector {
if i < 0 || i > v.count {
return nil
} else if i == v.count {
return v.Cons(val)
}
if i >= v.treeSize() {
newTail := append([]interface{}(nil), v.tail...)
copy(newTail, v.tail)
newTail[i&chunkMask] = val
return &vector{v.count, v.height, v.root, newTail}
}
return &vector{v.count, v.height, doAssoc(v.height, v.root, i, val), v.tail}
}
// doAssoc returns an almost identical tree, with the i-th element replaced by
// val.
func doAssoc(height uint, n node, i int, val interface{}) node {
m := clone(n)
if height == 0 {
m[i&chunkMask] = val
} else {
sub := (i >> (height * chunkBits)) & chunkMask
m[sub] = doAssoc(height-1, m[sub].(node), i, val)
}
return m
}
func (v *vector) Cons(val interface{}) Vector {
// Room in tail?
if v.count-v.treeSize() < tailMaxLen {
newTail := make([]interface{}, len(v.tail)+1)
copy(newTail, v.tail)
newTail[len(v.tail)] = val
return &vector{v.count + 1, v.height, v.root, newTail}
}
// Full tail; push into tree.
tailNode := nodeFromSlice(v.tail)
newHeight := v.height
var newRoot node
// Overflow root?
if (v.count >> chunkBits) > (1 << (v.height * chunkBits)) {
newRoot = newNode()
newRoot[0] = v.root
newRoot[1] = newPath(v.height, tailNode)
newHeight++
} else {
newRoot = v.pushTail(v.height, v.root, tailNode)
}
return &vector{v.count + 1, newHeight, newRoot, []interface{}{val}}
}
// pushTail returns a tree with tail appended.
func (v *vector) pushTail(height uint, n node, tail node) node {
if height == 0 {
return tail
}
idx := ((v.count - 1) >> (height * chunkBits)) & chunkMask
m := clone(n)
child := n[idx]
if child == nil {
m[idx] = newPath(height-1, tail)
} else {
m[idx] = v.pushTail(height-1, child.(node), tail)
}
return m
}
// newPath creates a left-branching tree of specified height and leaf.
func newPath(height uint, leaf node) node {
if height == 0 {
return leaf
}
ret := newNode()
ret[0] = newPath(height-1, leaf)
return ret
}
func (v *vector) Pop() Vector {
switch v.count {
case 0:
return nil
case 1:
return Empty
}
if v.count-v.treeSize() > 1 {
newTail := make([]interface{}, len(v.tail)-1)
copy(newTail, v.tail)
return &vector{v.count - 1, v.height, v.root, newTail}
}
newTail := v.sliceFor(v.count - 2)
newRoot := v.popTail(v.height, v.root)
newHeight := v.height
if v.height > 0 && newRoot[1] == nil {
newRoot = newRoot[0].(node)
newHeight--
}
return &vector{v.count - 1, newHeight, newRoot, newTail}
}
// popTail returns a new tree with the last leaf removed.
func (v *vector) popTail(level uint, n node) node {
idx := ((v.count - 2) >> (level * chunkBits)) & chunkMask
if level > 1 {
newChild := v.popTail(level-1, n[idx].(node))
if newChild == nil && idx == 0 {
return nil
}
m := clone(n)
if newChild == nil {
// This is needed since `m[idx] = newChild` would store an
// interface{} with a non-nil type part, which is non-nil
m[idx] = nil
} else {
m[idx] = newChild
}
return m
} else if idx == 0 {
return nil
} else {
m := clone(n)
m[idx] = nil
return m
}
}
func (v *vector) SubVector(begin, end int) Vector {
if begin < 0 || begin > end || end > v.count {
return nil
}
return &subVector{v, begin, end}
}
func (v *vector) Iterator() Iterator {
return newIterator(v)
}
func (v *vector) MarshalJSON() ([]byte, error) {
return marshalJSON(v.Iterator())
}
type subVector struct {
v *vector
begin int
end int
}
func (s *subVector) Len() int {
return s.end - s.begin
}
func (s *subVector) Index(i int) (interface{}, bool) {
if i < 0 || s.begin+i >= s.end {
return nil, false
}
return s.v.Index(s.begin + i)
}
func (s *subVector) Assoc(i int, val interface{}) Vector {
if i < 0 || s.begin+i > s.end {
return nil
} else if s.begin+i == s.end {
return s.Cons(val)
}
return s.v.Assoc(s.begin+i, val).SubVector(s.begin, s.end)
}
func (s *subVector) Cons(val interface{}) Vector {
return s.v.Assoc(s.end, val).SubVector(s.begin, s.end+1)
}
func (s *subVector) Pop() Vector {
switch s.Len() {
case 0:
return nil
case 1:
return Empty
default:
return s.v.SubVector(s.begin, s.end-1)
}
}
func (s *subVector) SubVector(i, j int) Vector {
return s.v.SubVector(s.begin+i, s.begin+j)
}
func (s *subVector) Iterator() Iterator {
return newIteratorWithRange(s.v, s.begin, s.end)
}
func (s *subVector) MarshalJSON() ([]byte, error) {
return marshalJSON(s.Iterator())
}
type iterator struct {
v *vector
treeSize int
index int
end int
path []pathEntry
}
type pathEntry struct {
node node
index int
}
func (e pathEntry) current() interface{} {
return e.node[e.index]
}
func newIterator(v *vector) *iterator {
return newIteratorWithRange(v, 0, v.Len())
}
func newIteratorWithRange(v *vector, begin, end int) *iterator {
it := &iterator{v, v.treeSize(), begin, end, nil}
// Find the node for begin, remembering all nodes along the path.
n := v.root
for shift := v.height * chunkBits; shift > 0; shift -= chunkBits {
idx := (begin >> shift) & chunkMask
it.path = append(it.path, pathEntry{n, idx})
n = n[idx].(node)
}
it.path = append(it.path, pathEntry{n, begin & chunkMask})
return it
}
func (it *iterator) Elem() interface{} {
if it.index >= it.treeSize {
return it.v.tail[it.index-it.treeSize]
}
return it.path[len(it.path)-1].current()
}
func (it *iterator) HasElem() bool {
return it.index < it.end
}
func (it *iterator) Next() {
if it.index+1 >= it.treeSize {
// Next element is in tail. Just increment the index.
it.index++
return
}
// Find the deepest level that can be advanced.
var i int
for i = len(it.path) - 1; i >= 0; i-- {
e := it.path[i]
if e.index+1 < len(e.node) && e.node[e.index+1] != nil {
break
}
}
if i == -1 {
panic("cannot advance; vector iterator bug")
}
// Advance on this node, and re-populate all deeper levels.
it.path[i].index++
for i++; i < len(it.path); i++ {
it.path[i] = pathEntry{it.path[i-1].current().(node), 0}
}
it.index++
}
type marshalError struct {
index int
cause error
}
func (err *marshalError) Error() string {
return fmt.Sprintf("element %d: %s", err.index, err.cause)
}
func marshalJSON(it Iterator) ([]byte, error) {
var buf bytes.Buffer
buf.WriteByte('[')
index := 0
for ; it.HasElem(); it.Next() {
if index > 0 {
buf.WriteByte(',')
}
elemBytes, err := json.Marshal(it.Elem())
if err != nil {
return nil, &marshalError{index, err}
}
buf.Write(elemBytes)
index++
}
buf.WriteByte(']')
return buf.Bytes(), nil
}
golang-github-xiaq-persistent-0.0~git20190312.a1d9ac4/vector/vector_test.go 0000664 0000000 0000000 00000020623 13441710145 0026304 0 ustar 00root root 0000000 0000000 package vector
import (
"errors"
"math/rand"
"testing"
"time"
)
// Nx is the minimum number of elements for the internal tree of the vector to
// be x levels deep.
const (
N1 = tailMaxLen + 1 // 33
N2 = nodeSize + tailMaxLen + 1 // 65
N3 = nodeSize*nodeSize + tailMaxLen + 1 // 1057
N4 = nodeSize*nodeSize*nodeSize + tailMaxLen + 1 // 32801
)
func init() {
rand.Seed(time.Now().UTC().UnixNano())
}
func TestVector(t *testing.T) {
const (
subst = "233"
n = N4
)
v := testCons(t, n)
testIndex(t, v, 0, n)
testAssoc(t, v, subst)
testIterator(t, v.Iterator(), 0, n)
testPop(t, v)
}
// testCons creates a vector containing 0...n-1 with Cons, and ensures that the
// length of the old and new vectors are expected after each Cons. It returns
// the created vector.
func testCons(t *testing.T, n int) Vector {
v := Empty
for i := 0; i < n; i++ {
oldv := v
v = v.Cons(i)
if count := oldv.Len(); count != i {
t.Errorf("oldv.Count() == %v, want %v", count, i)
}
if count := v.Len(); count != i+1 {
t.Errorf("v.Count() == %v, want %v", count, i+1)
}
}
return v
}
// testIndex tests Index, assuming that the vector contains begin...int-1.
func testIndex(t *testing.T, v Vector, begin, end int) {
n := v.Len()
for i := 0; i < n; i++ {
elem, _ := v.Index(i)
if elem != i {
t.Errorf("v.Index(%v) == %v, want %v", i, elem, i)
}
}
for _, i := range []int{-2, -1, n, n + 1, n * 2} {
if elem, _ := v.Index(i); elem != nil {
t.Errorf("v.Index(%d) == %v, want nil", i, elem)
}
}
}
// testIterator tests the iterator, assuming that the result is begin...end-1.
func testIterator(t *testing.T, it Iterator, begin, end int) {
i := begin
for ; it.HasElem(); it.Next() {
elem := it.Elem()
if elem != i {
t.Errorf("iterator produce %v, want %v", elem, i)
}
i++
}
if i != end {
t.Errorf("iterator produces up to %v, want %v", i, end)
}
}
// testAssoc tests Assoc by replacing each element.
func testAssoc(t *testing.T, v Vector, subst interface{}) {
n := v.Len()
for i := 0; i <= n; i++ {
oldv := v
v = v.Assoc(i, subst)
if i < n {
elem, _ := oldv.Index(i)
if elem != i {
t.Errorf("oldv.Index(%v) == %v, want %v", i, elem, i)
}
}
elem, _ := v.Index(i)
if elem != subst {
t.Errorf("v.Index(%v) == %v, want %v", i, elem, subst)
}
}
n++
for _, i := range []int{-1, n + 1, n + 2, n * 2} {
newv := v.Assoc(i, subst)
if newv != nil {
t.Errorf("v.Assoc(%d) = %v, want nil", i, newv)
}
}
}
// testPop tests Pop by removing each element.
func testPop(t *testing.T, v Vector) {
n := v.Len()
for i := 0; i < n; i++ {
oldv := v
v = v.Pop()
if count := oldv.Len(); count != n-i {
t.Errorf("oldv.Count() == %v, want %v", count, n-i)
}
if count := v.Len(); count != n-i-1 {
t.Errorf("oldv.Count() == %v, want %v", count, n-i-1)
}
}
newv := v.Pop()
if newv != nil {
t.Errorf("v.Pop() = %v, want nil", newv)
}
}
func TestSubVector(t *testing.T) {
v := Empty
for i := 0; i < 10; i++ {
v = v.Cons(i)
}
sv := v.SubVector(0, 4)
testIndex(t, sv, 0, 4)
testAssoc(t, sv, "233")
testIterator(t, sv.Iterator(), 0, 4)
testPop(t, sv)
sv = v.SubVector(1, 4)
if !checkVector(sv, 1, 2, 3) {
t.Errorf("v[0:4] is not expected")
}
if !checkVector(sv.Assoc(1, "233"), 1, "233", 3) {
t.Errorf("v[0:4].Assoc is not expected")
}
if !checkVector(sv.Cons("233"), 1, 2, 3, "233") {
t.Errorf("v[0:4].Cons is not expected")
}
if !checkVector(sv.Pop(), 1, 2) {
t.Errorf("v[0:4].Pop is not expected")
}
if !checkVector(sv.SubVector(1, 2), 2) {
t.Errorf("v[0:4][1:2] is not expected")
}
testIterator(t, sv.Iterator(), 1, 4)
if !checkVector(v.SubVector(1, 1)) {
t.Errorf("v[1:1] is not expected")
}
// Begin is allowed to be equal to n if end is also n
if !checkVector(v.SubVector(10, 10)) {
t.Errorf("v[10:10] is not expected")
}
bad := v.SubVector(-1, 0)
if bad != nil {
t.Errorf("v.SubVector(-1, 0) = %v, want nil", bad)
}
bad = v.SubVector(5, 100)
if bad != nil {
t.Errorf("v.SubVector(5, 100) = %v, want nil", bad)
}
bad = v.SubVector(-1, 100)
if bad != nil {
t.Errorf("v.SubVector(-1, 100) = %v, want nil", bad)
}
bad = v.SubVector(4, 2)
if bad != nil {
t.Errorf("v.SubVector(4, 2) = %v, want nil", bad)
}
}
func checkVector(v Vector, values ...interface{}) bool {
if v.Len() != len(values) {
return false
}
for i, a := range values {
if x, _ := v.Index(i); x != a {
return false
}
}
return true
}
func TestVectorEqual(t *testing.T) {
v1, v2 := Empty, Empty
for i := 0; i < N3; i++ {
elem := rand.Int63()
v1 = v1.Cons(elem)
v2 = v2.Cons(elem)
if !eqVector(v1, v2) {
t.Errorf("Not equal after Cons'ing %d elements", i+1)
}
}
}
func eqVector(v1, v2 Vector) bool {
if v1.Len() != v2.Len() {
return false
}
for i := 0; i < v1.Len(); i++ {
a1, _ := v1.Index(i)
a2, _ := v2.Index(i)
if a1 != a2 {
return false
}
}
return true
}
var marshalJSONTests = []struct {
in Vector
wantOut string
wantErr error
}{
{makeVector("1", 2, nil), `["1",2,null]`, nil},
{makeVector("1", makeVector(2)), `["1",[2]]`, nil},
{makeVector(0, 1, 2, 3, 4, 5).SubVector(1, 5), `[1,2,3,4]`, nil},
{makeVector(0, func() {}), "", errors.New("element 1: json: unsupported type: func()")},
}
func TestMarshalJSON(t *testing.T) {
for i, test := range marshalJSONTests {
out, err := test.in.MarshalJSON()
if string(out) != test.wantOut {
t.Errorf("v%d.MarshalJSON -> out %q, want %q", i, out, test.wantOut)
}
if err == nil || test.wantErr == nil {
if err != test.wantErr {
t.Errorf("v%d.MarshalJSON -> err %v, want %v", i, err, test.wantErr)
}
} else {
if err.Error() != test.wantErr.Error() {
t.Errorf("v%d.MarshalJSON -> err %v, want %v", i, err, test.wantErr)
}
}
}
}
func makeVector(elements ...interface{}) Vector {
v := Empty
for _, element := range elements {
v = v.Cons(element)
}
return v
}
func BenchmarkConsNativeN1(b *testing.B) { benchmarkNativeAppend(b, N1) }
func BenchmarkConsNativeN2(b *testing.B) { benchmarkNativeAppend(b, N2) }
func BenchmarkConsNativeN3(b *testing.B) { benchmarkNativeAppend(b, N3) }
func BenchmarkConsNativeN4(b *testing.B) { benchmarkNativeAppend(b, N4) }
func benchmarkNativeAppend(b *testing.B, n int) {
for r := 0; r < b.N; r++ {
var s []interface{}
for i := 0; i < n; i++ {
s = append(s, i)
}
}
}
func BenchmarkConsPersistentN1(b *testing.B) { benchmarkCons(b, N1) }
func BenchmarkConsPersistentN2(b *testing.B) { benchmarkCons(b, N2) }
func BenchmarkConsPersistentN3(b *testing.B) { benchmarkCons(b, N3) }
func BenchmarkConsPersistentN4(b *testing.B) { benchmarkCons(b, N4) }
func benchmarkCons(b *testing.B, n int) {
for r := 0; r < b.N; r++ {
v := Empty
for i := 0; i < n; i++ {
v = v.Cons(i)
}
}
}
var (
sliceN4 = make([]interface{}, N4)
vectorN4 = Empty
)
func init() {
for i := 0; i < N4; i++ {
vectorN4 = vectorN4.Cons(i)
}
}
var x interface{}
func BenchmarkIndexSeqNativeN4(b *testing.B) { benchmarkIndexSeqNative(b, N4) }
func benchmarkIndexSeqNative(b *testing.B, n int) {
for r := 0; r < b.N; r++ {
for i := 0; i < n; i++ {
x = sliceN4[i]
}
}
}
func BenchmarkIndexSeqPersistentN4(b *testing.B) { benchmarkIndexSeqPersistent(b, N4) }
func benchmarkIndexSeqPersistent(b *testing.B, n int) {
for r := 0; r < b.N; r++ {
for i := 0; i < n; i++ {
x, _ = vectorN4.Index(i)
}
}
}
var randIndicies []int
func init() {
randIndicies = make([]int, N4)
for i := 0; i < N4; i++ {
randIndicies[i] = rand.Intn(N4)
}
}
func BenchmarkIndexRandNative(b *testing.B) {
for r := 0; r < b.N; r++ {
for _, i := range randIndicies {
x = sliceN4[i]
}
}
}
func BenchmarkIndexRandPersistent(b *testing.B) {
for r := 0; r < b.N; r++ {
for _, i := range randIndicies {
x, _ = vectorN4.Index(i)
}
}
}
func nativeEqual(s1, s2 []int) bool {
if len(s1) != len(s2) {
return false
}
for i, v1 := range s1 {
if v1 != s2[i] {
return false
}
}
return true
}
func BenchmarkEqualNative(b *testing.B) {
b.StopTimer()
var s1, s2 []int
for i := 0; i < N4; i++ {
s1 = append(s1, i)
s2 = append(s2, i)
}
b.StartTimer()
for r := 0; r < b.N; r++ {
eq := nativeEqual(s1, s2)
if !eq {
panic("not equal")
}
}
}
func BenchmarkEqualPersistent(b *testing.B) {
b.StopTimer()
v1, v2 := Empty, Empty
for i := 0; i < N4; i++ {
v1 = v1.Cons(i)
v2 = v2.Cons(i)
}
b.StartTimer()
for r := 0; r < b.N; r++ {
eq := eqVector(v1, v2)
if !eq {
panic("not equal")
}
}
}