ston1th/docstore
1
1
Fork 0
Code Issues 3 Releases Activity

fix openbsd

Browse source
This commit is contained in:
ston1th 2025-01-17 22:06:08 +01:00
commit 6476ebc781
524 changed files with 36106 additions and 11790 deletions
Download patch file Download diff file Expand all files Collapse all files

1
vendor/github.com/RoaringBitmap/roaring/.drone.yml generated vendored
View file

@ -11,7 +11,6 @@ steps:
commands:
- go get -t
- go test
- go test -race -run TestConcurrent*
- go build -tags appengine
- go test -tags appengine
- GOARCH=386 go build

107
vendor/github.com/RoaringBitmap/roaring/Makefile generated vendored
View file

@ -1,107 +0,0 @@
.PHONY: help all test format fmtcheck vet lint qa deps clean nuke ser fetch-real-roaring-datasets
# Display general help about this command
help:
@echo ""
@echo "The following commands are available:"
@echo ""
@echo " make qa : Run all the tests"
@echo " make test : Run the unit tests"
@echo ""
@echo " make format : Format the source code"
@echo " make fmtcheck : Check if the source code has been formatted"
@echo " make vet : Check for suspicious constructs"
@echo " make lint : Check for style errors"
@echo ""
@echo " make deps : Get the dependencies"
@echo " make clean : Remove any build artifact"
@echo " make nuke : Deletes any intermediate file"
@echo ""
@echo " make fuzz-smat : Fuzzy testing with smat"
@echo " make fuzz-stream : Fuzzy testing with stream deserialization"
@echo " make fuzz-buffer : Fuzzy testing with buffer deserialization"
@echo ""
# Alias for help target
all: help
test:
go test
go test -race -run TestConcurrent*
# Format the source code
format:
@find ./ -type f -name "*.go" -exec gofmt -w {} \;
# Check if the source code has been formatted
fmtcheck:
@mkdir -p target
@find ./ -type f -name "*.go" -exec gofmt -d {} \; | tee target/format.diff
@test ! -s target/format.diff || { echo "ERROR: the source code has not been formatted - please use 'make format' or 'gofmt'"; exit 1; }
# Check for syntax errors
vet:
GOPATH=$(GOPATH) go vet ./...
# Check for style errors
lint:
GOPATH=$(GOPATH) PATH=$(GOPATH)/bin:$(PATH) golint ./...
# Alias to run all quality-assurance checks
qa: fmtcheck test vet lint
# --- INSTALL ---
# Get the dependencies
deps:
GOPATH=$(GOPATH) go get github.com/stretchr/testify
GOPATH=$(GOPATH) go get github.com/bits-and-blooms/bitset
GOPATH=$(GOPATH) go get github.com/golang/lint/golint
GOPATH=$(GOPATH) go get github.com/mschoch/smat
GOPATH=$(GOPATH) go get github.com/dvyukov/go-fuzz/go-fuzz
GOPATH=$(GOPATH) go get github.com/dvyukov/go-fuzz/go-fuzz-build
GOPATH=$(GOPATH) go get github.com/glycerine/go-unsnap-stream
GOPATH=$(GOPATH) go get github.com/philhofer/fwd
GOPATH=$(GOPATH) go get github.com/jtolds/gls
fuzz-smat:
go test -tags=gofuzz -run=TestGenerateSmatCorpus
go-fuzz-build -func FuzzSmat github.com/RoaringBitmap/roaring
go-fuzz -bin=./roaring-fuzz.zip -workdir=workdir/ -timeout=200
fuzz-stream:
go-fuzz-build -func FuzzSerializationStream github.com/RoaringBitmap/roaring
go-fuzz -bin=./roaring-fuzz.zip -workdir=workdir/ -timeout=200
fuzz-buffer:
go-fuzz-build -func FuzzSerializationBuffer github.com/RoaringBitmap/roaring
go-fuzz -bin=./roaring-fuzz.zip -workdir=workdir/ -timeout=200
# Remove any build artifact
clean:
GOPATH=$(GOPATH) go clean ./...
# Deletes any intermediate file
nuke:
rm -rf ./target
GOPATH=$(GOPATH) go clean -i ./...
cover:
go test -coverprofile=coverage.out
go tool cover -html=coverage.out
fetch-real-roaring-datasets:
# pull github.com/RoaringBitmap/real-roaring-datasets -> testdata/real-roaring-datasets
git submodule init
git submodule update

16
vendor/github.com/RoaringBitmap/roaring/README.md generated vendored
View file

@ -1,5 +1,7 @@
roaring [![GoDoc](https://godoc.org/github.com/RoaringBitmap/roaring/roaring64?status.svg)](https://godoc.org/github.com/RoaringBitmap/roaring/roaring64) [![Go Report Card](https://goreportcard.com/badge/RoaringBitmap/roaring)](https://goreportcard.com/report/github.com/RoaringBitmap/roaring)
[![Build Status](https://cloud.drone.io/api/badges/RoaringBitmap/roaring/status.svg)](https://cloud.drone.io/RoaringBitmap/roaring)
# roaring
[![GoDoc](https://godoc.org/github.com/RoaringBitmap/roaring?status.svg)](https://godoc.org/github.com/RoaringBitmap/roaring) [![Go Report Card](https://goreportcard.com/badge/RoaringBitmap/roaring)](https://goreportcard.com/report/github.com/RoaringBitmap/roaring)
![Go-CI](https://github.com/RoaringBitmap/roaring/workflows/Go-CI/badge.svg)
![Go-ARM-CI](https://github.com/RoaringBitmap/roaring/workflows/Go-ARM-CI/badge.svg)
![Go-Windows-CI](https://github.com/RoaringBitmap/roaring/workflows/Go-Windows-CI/badge.svg)
@ -8,7 +10,7 @@ roaring [![GoDoc](https://godoc.org/github.com/RoaringBitmap/roaring/roaring64?s
This is a go version of the Roaring bitmap data structure.
Roaring bitmaps are used by several major systems such as [Apache Lucene][lucene] and derivative systems such as [Solr][solr] and
[Elasticsearch][elasticsearch], [Apache Druid (Incubating)][druid], [LinkedIn Pinot][pinot], [Netflix Atlas][atlas], [Apache Spark][spark], [OpenSearchServer][opensearchserver], [anacrolix/torrent][anacrolix/torrent], [Whoosh][whoosh], [Pilosa][pilosa], [Microsoft Visual Studio Team Services (VSTS)][vsts], and eBay's [Apache Kylin][kylin]. The YouTube SQL Engine, [Google Procella](https://research.google/pubs/pub48388/), uses Roaring bitmaps for indexing.
[Elasticsearch][elasticsearch], [Apache Druid (Incubating)][druid], [LinkedIn Pinot][pinot], [Netflix Atlas][atlas], [Apache Spark][spark], [OpenSearchServer][opensearchserver], [anacrolix/torrent][anacrolix/torrent], [Whoosh][whoosh], [Redpanda](https://github.com/redpanda-data/redpanda), [Pilosa][pilosa], [Microsoft Visual Studio Team Services (VSTS)][vsts], and eBay's [Apache Kylin][kylin]. The YouTube SQL Engine, [Google Procella](https://research.google/pubs/pub48388/), uses Roaring bitmaps for indexing.
[lucene]: https://lucene.apache.org/
[solr]: https://lucene.apache.org/solr/
@ -31,17 +33,17 @@ Roaring bitmaps are found to work well in many important applications:
The ``roaring`` Go library is used by
* [anacrolix/torrent]
* [runv](https://github.com/hyperhq/runv)
* [InfluxDB](https://www.influxdata.com)
* [Pilosa](https://www.pilosa.com/)
* [Bleve](http://www.blevesearch.com)
* [Weaviate](https://github.com/weaviate/weaviate)
* [lindb](https://github.com/lindb/lindb)
* [Elasticell](https://github.com/deepfabric/elasticell)
* [SourceGraph](https://github.com/sourcegraph/sourcegraph)
* [M3](https://github.com/m3db/m3)
* [trident](https://github.com/NetApp/trident)
* [Husky](https://www.datadoghq.com/blog/engineering/introducing-husky/)
* [FrostDB](https://github.com/polarsignals/frostdb)
This library is used in production in several systems, it is part of the [Awesome Go collection](https://awesome-go.com).
@ -99,7 +101,7 @@ whether you like it or not. That can become very wasteful.
This being said, there are definitively cases where attempting to use compressed bitmaps is wasteful.
For example, if you have a small universe size. E.g., your bitmaps represent sets of integers
from [0,n) where n is small (e.g., n=64 or n=128). If you are able to uncompressed BitSet and
from [0,n) where n is small (e.g., n=64 or n=128). If you can use uncompressed BitSet and
it does not blow up your memory usage, then compressed bitmaps are probably not useful
to you. In fact, if you do not need compression, then a BitSet offers remarkable speed.
@ -134,7 +136,7 @@ There is a big problem with these formats however that can hurt you badly in som
Roaring solves this problem. It works in the following manner. It divides the data into chunks of 2<sup>16</sup> integers
(e.g., [0, 2<sup>16</sup>), [2<sup>16</sup>, 2 x 2<sup>16</sup>), ...). Within a chunk, it can use an uncompressed bitmap, a simple list of integers,
or a list of runs. Whatever format it uses, they all allow you to check for the present of any one value quickly
or a list of runs. Whatever format it uses, they all allow you to check for the presence of any one value quickly
(e.g., with a binary search). The net result is that Roaring can compute many operations much faster than run-length-encoded
formats like WAH, EWAH, Concise... Maybe surprisingly, Roaring also generally offers better compression ratios.

63
vendor/github.com/RoaringBitmap/roaring/arraycontainer.go generated vendored
View file

@ -17,8 +17,17 @@ func (ac *arrayContainer) String() string {
}
func (ac *arrayContainer) fillLeastSignificant16bits(x []uint32, i int, mask uint32) int {
if i < 0 {
panic("negative index")
}
if len(ac.content) == 0 {
return i
}
_ = x[len(ac.content)-1+i]
_ = ac.content[len(ac.content)-1]
for k := 0; k < len(ac.content); k++ {
x[k+i] = uint32(ac.content[k]) | mask
x[k+i] =
uint32(ac.content[k]) | mask
}
return i + len(ac.content)
}
@ -655,10 +664,54 @@ func (ac *arrayContainer) iandNot(a container) container {
}
func (ac *arrayContainer) iandNotRun16(rc *runContainer16) container {
rcb := rc.toBitmapContainer()
acb := ac.toBitmapContainer()
acb.iandNotBitmapSurely(rcb)
*ac = *(acb.toArrayContainer())
// Fast path: if either the array container or the run container is empty, the result is the array.
if ac.isEmpty() || rc.isEmpty() {
// Empty
return ac
}
// Fast path: if the run container is full, the result is empty.
if rc.isFull() {
ac.content = ac.content[:0]
return ac
}
current_run := 0
// All values in [start_run, end_end] are part of the run
start_run := rc.iv[current_run].start
end_end := start_run + rc.iv[current_run].length
// We are going to read values in the array at index i, and we are
// going to write them at index pos. So we do in-place processing.
// We always have that pos <= i by construction. So we can either
// overwrite a value just read, or a value that was previous read.
pos := 0
i := 0
for ; i < len(ac.content); i++ {
if ac.content[i] < start_run {
// the value in the array appears before the run [start_run, end_end]
ac.content[pos] = ac.content[i]
pos++
} else if ac.content[i] <= end_end {
// nothing to do, the value is in the array but also in the run.
} else {
// We have the value in the array after the run. We cannot tell
// whether we need to keep it or not. So let us move to another run.
if current_run+1 < len(rc.iv) {
current_run++
start_run = rc.iv[current_run].start
end_end = start_run + rc.iv[current_run].length
i-- // retry with the same i
} else {
// We have exhausted the number of runs. We can keep the rest of the values
// from i to len(ac.content) - 1 inclusively.
break // We are done, the rest of the array will be kept
}
}
}
for ; i < len(ac.content); i++ {
ac.content[pos] = ac.content[i]
pos++
}
// We 'shink' the slice.
ac.content = ac.content[:pos]
return ac
}

63
vendor/github.com/RoaringBitmap/roaring/bitmapcontainer.go generated vendored
View file

@ -888,13 +888,67 @@ func (bc *bitmapContainer) iandNot(a container) container {
}
func (bc *bitmapContainer) iandNotArray(ac *arrayContainer) container {
acb := ac.toBitmapContainer()
return bc.iandNotBitmapSurely(acb)
if ac.isEmpty() || bc.isEmpty() {
// Nothing to do.
return bc
}
// Word by word, we remove the elements in ac from bc. The approach is to build
// a mask of the elements to remove, and then apply it to the bitmap.
wordIdx := uint16(0)
mask := uint64(0)
for i, v := range ac.content {
if v/64 != wordIdx {
// Flush the current word.
if i != 0 {
// We're removing bits that are set in the mask and in the current word.
// To figure out the cardinality change, we count the number of bits that
// are set in the mask and in the current word.
mask &= bc.bitmap[wordIdx]
bc.bitmap[wordIdx] &= ^mask
bc.cardinality -= int(popcount(mask))
}
wordIdx = v / 64
mask = 0
}
mask |= 1 << (v % 64)
}
// Flush the last word.
mask &= bc.bitmap[wordIdx]
bc.bitmap[wordIdx] &= ^mask
bc.cardinality -= int(popcount(mask))
if bc.getCardinality() <= arrayDefaultMaxSize {
return bc.toArrayContainer()
}
return bc
}
func (bc *bitmapContainer) iandNotRun16(rc *runContainer16) container {
rcb := rc.toBitmapContainer()
return bc.iandNotBitmapSurely(rcb)
if rc.isEmpty() || bc.isEmpty() {
// Nothing to do.
return bc
}
wordRangeStart := rc.iv[0].start / 64
wordRangeEnd := (rc.iv[len(rc.iv)-1].last()) / 64 // inclusive
cardinalityChange := popcntSlice(bc.bitmap[wordRangeStart : wordRangeEnd+1]) // before cardinality - after cardinality (for word range)
for _, iv := range rc.iv {
resetBitmapRange(bc.bitmap, int(iv.start), int(iv.last())+1)
}
cardinalityChange -= popcntSlice(bc.bitmap[wordRangeStart : wordRangeEnd+1])
bc.cardinality -= int(cardinalityChange)
if bc.getCardinality() <= arrayDefaultMaxSize {
return bc.toArrayContainer()
}
return bc
}
func (bc *bitmapContainer) andNotArray(value2 *arrayContainer) container {
@ -1062,7 +1116,6 @@ func (bc *bitmapContainer) PrevSetBit(i int) int {
// reference the java implementation
// https://github.com/RoaringBitmap/RoaringBitmap/blob/master/src/main/java/org/roaringbitmap/BitmapContainer.java#L875-L892
//
func (bc *bitmapContainer) numberOfRuns() int {
if bc.cardinality == 0 {
return 0

63
vendor/github.com/RoaringBitmap/roaring/internal/byte_input.go generated vendored
View file

@ -10,6 +10,11 @@ type ByteInput interface {
// Next returns a slice containing the next n bytes from the buffer,
// advancing the buffer as if the bytes had been returned by Read.
Next(n int) ([]byte, error)
// NextReturnsSafeSlice returns true if Next() returns a safe slice as opposed
// to a slice that points to an underlying buffer possibly owned by another system.
// When NextReturnsSafeSlice returns false, the result from Next() should be copied
// before it is modified (i.e., it is immutable).
NextReturnsSafeSlice() bool
// ReadUInt32 reads uint32 with LittleEndian order
ReadUInt32() (uint32, error)
// ReadUInt16 reads uint16 with LittleEndian order
@ -42,6 +47,25 @@ type ByteBuffer struct {
off int
}
// NewByteBuffer creates a new ByteBuffer.
func NewByteBuffer(buf []byte) *ByteBuffer {
return &ByteBuffer{
buf: buf,
}
}
var _ io.Reader = (*ByteBuffer)(nil)
// Read implements io.Reader.
func (b *ByteBuffer) Read(p []byte) (int, error) {
data, err := b.Next(len(p))
if err != nil {
return 0, err
}
copy(p, data)
return len(data), nil
}
// Next returns a slice containing the next n bytes from the reader
// If there are fewer bytes than the given n, io.ErrUnexpectedEOF will be returned
func (b *ByteBuffer) Next(n int) ([]byte, error) {
@ -57,6 +81,12 @@ func (b *ByteBuffer) Next(n int) ([]byte, error) {
return data, nil
}
// NextReturnsSafeSlice returns false since ByteBuffer might hold
// an array owned by some other systems.
func (b *ByteBuffer) NextReturnsSafeSlice() bool {
return false
}
// ReadUInt32 reads uint32 with LittleEndian order
func (b *ByteBuffer) ReadUInt32() (uint32, error) {
if len(b.buf)-b.off < 4 {
@ -109,26 +139,45 @@ func (b *ByteBuffer) Reset(buf []byte) {
type ByteInputAdapter struct {
r io.Reader
readBytes int
buf [4]byte
}
var _ io.Reader = (*ByteInputAdapter)(nil)
// Read implements io.Reader.
func (b *ByteInputAdapter) Read(buf []byte) (int, error) {
m, err := io.ReadAtLeast(b.r, buf, len(buf))
b.readBytes += m
if err != nil {
return 0, err
}
return m, nil
}
// Next returns a slice containing the next n bytes from the buffer,
// advancing the buffer as if the bytes had been returned by Read.
func (b *ByteInputAdapter) Next(n int) ([]byte, error) {
buf := make([]byte, n)
m, err := io.ReadAtLeast(b.r, buf, n)
b.readBytes += m
_, err := b.Read(buf)
if err != nil {
return nil, err
}
return buf, nil
}
// NextReturnsSafeSlice returns true since ByteInputAdapter always returns a slice
// allocated with make([]byte, ...)
func (b *ByteInputAdapter) NextReturnsSafeSlice() bool {
return true
}
// ReadUInt32 reads uint32 with LittleEndian order
func (b *ByteInputAdapter) ReadUInt32() (uint32, error) {
buf, err := b.Next(4)
buf := b.buf[:4]
_, err := b.Read(buf)
if err != nil {
return 0, err
}
@ -138,8 +187,8 @@ func (b *ByteInputAdapter) ReadUInt32() (uint32, error) {
// ReadUInt16 reads uint16 with LittleEndian order
func (b *ByteInputAdapter) ReadUInt16() (uint16, error) {
buf, err := b.Next(2)
buf := b.buf[:2]
_, err := b.Read(buf)
if err != nil {
return 0, err
}

267
vendor/github.com/RoaringBitmap/roaring/roaring.go generated vendored
View file

@ -13,6 +13,7 @@ import (
"strconv"
"github.com/RoaringBitmap/roaring/internal"
"github.com/bits-and-blooms/bitset"
)
// Bitmap represents a compressed bitmap where you can add integers.
@ -53,17 +54,186 @@ func (rb *Bitmap) ToBytes() ([]byte, error) {
return rb.highlowcontainer.toBytes()
}
const wordSize = uint64(64)
const log2WordSize = uint64(6)
const capacity = ^uint64(0)
const bitmapContainerSize = (1 << 16) / 64 // bitmap size in words
// DenseSize returns the size of the bitmap when stored as a dense bitmap.
func (rb *Bitmap) DenseSize() uint64 {
if rb.highlowcontainer.size() == 0 {
return 0
}
maximum := 1 + uint64(rb.Maximum())
if maximum > (capacity - wordSize + 1) {
return uint64(capacity >> log2WordSize)
}
return uint64((maximum + (wordSize - 1)) >> log2WordSize)
}
// ToDense returns a slice of uint64s representing the bitmap as a dense bitmap.
// Useful to convert a roaring bitmap to a format that can be used by other libraries
// like https://github.com/bits-and-blooms/bitset or https://github.com/kelindar/bitmap
func (rb *Bitmap) ToDense() []uint64 {
sz := rb.DenseSize()
if sz == 0 {
return nil
}
bitmap := make([]uint64, sz)
rb.WriteDenseTo(bitmap)
return bitmap
}
// FromDense creates a bitmap from a slice of uint64s representing the bitmap as a dense bitmap.
// Useful to convert bitmaps from libraries like https://github.com/bits-and-blooms/bitset or
// https://github.com/kelindar/bitmap into roaring bitmaps fast and with convenience.
//
// This function will not create any run containers, only array and bitmap containers. It's up to
// the caller to call RunOptimize if they want to further compress the runs of consecutive values.
//
// When doCopy is true, the bitmap is copied into a new slice for each bitmap container.
// This is useful when the bitmap is going to be modified after this function returns or if it's
// undesirable to hold references to large bitmaps which the GC would not be able to collect.
// One copy can still happen even when doCopy is false if the bitmap length is not divisible
// by bitmapContainerSize.
//
// See also FromBitSet.
func FromDense(bitmap []uint64, doCopy bool) *Bitmap {
sz := (len(bitmap) + bitmapContainerSize - 1) / bitmapContainerSize // round up
rb := &Bitmap{
highlowcontainer: roaringArray{
containers: make([]container, 0, sz),
keys: make([]uint16, 0, sz),
needCopyOnWrite: make([]bool, 0, sz),
},
}
rb.FromDense(bitmap, doCopy)
return rb
}
// FromDense unmarshalls from a slice of uint64s representing the bitmap as a dense bitmap.
// Useful to convert bitmaps from libraries like https://github.com/bits-and-blooms/bitset or
// https://github.com/kelindar/bitmap into roaring bitmaps fast and with convenience.
// Callers are responsible for ensuring that the bitmap is empty before calling this function.
//
// This function will not create any run containers, only array and bitmap containers. It is up to
// the caller to call RunOptimize if they want to further compress the runs of consecutive values.
//
// When doCopy is true, the bitmap is copied into a new slice for each bitmap container.
// This is useful when the bitmap is going to be modified after this function returns or if it's
// undesirable to hold references to large bitmaps which the GC would not be able to collect.
// One copy can still happen even when doCopy is false if the bitmap length is not divisible
// by bitmapContainerSize.
//
// See FromBitSet.
func (rb *Bitmap) FromDense(bitmap []uint64, doCopy bool) {
if len(bitmap) == 0 {
return
}
var k uint16
const size = bitmapContainerSize
for len(bitmap) > 0 {
hi := size
if len(bitmap) < size {
hi = len(bitmap)
}
words := bitmap[:hi]
count := int(popcntSlice(words))
switch {
case count > arrayDefaultMaxSize:
c := &bitmapContainer{cardinality: count, bitmap: words}
cow := true
if doCopy || len(words) < size {
c.bitmap = make([]uint64, size)
copy(c.bitmap, words)
cow = false
}
rb.highlowcontainer.appendContainer(k, c, cow)
case count > 0:
c := &arrayContainer{content: make([]uint16, count)}
var pos, base int
for _, w := range words {
for w != 0 {
t := w & -w
c.content[pos] = uint16(base + int(popcount(t-1)))
pos++
w ^= t
}
base += 64
}
rb.highlowcontainer.appendContainer(k, c, false)
}
bitmap = bitmap[hi:]
k++
}
}
// WriteDenseTo writes to a slice of uint64s representing the bitmap as a dense bitmap.
// Callers are responsible for allocating enough space in the bitmap using DenseSize.
// Useful to convert a roaring bitmap to a format that can be used by other libraries
// like https://github.com/bits-and-blooms/bitset or https://github.com/kelindar/bitmap
func (rb *Bitmap) WriteDenseTo(bitmap []uint64) {
for i, ct := range rb.highlowcontainer.containers {
hb := uint32(rb.highlowcontainer.keys[i]) << 16
switch c := ct.(type) {
case *arrayContainer:
for _, x := range c.content {
n := int(hb | uint32(x))
bitmap[n>>log2WordSize] |= uint64(1) << uint(x%64)
}
case *bitmapContainer:
copy(bitmap[int(hb)>>log2WordSize:], c.bitmap)
case *runContainer16:
for j := range c.iv {
start := uint32(c.iv[j].start)
end := start + uint32(c.iv[j].length) + 1
lo := int(hb|start) >> log2WordSize
hi := int(hb|(end-1)) >> log2WordSize
if lo == hi {
bitmap[lo] |= (^uint64(0) << uint(start%64)) &
(^uint64(0) >> (uint(-end) % 64))
continue
}
bitmap[lo] |= ^uint64(0) << uint(start%64)
for n := lo + 1; n < hi; n++ {
bitmap[n] = ^uint64(0)
}
bitmap[hi] |= ^uint64(0) >> (uint(-end) % 64)
}
default:
panic("unsupported container type")
}
}
}
// Checksum computes a hash (currently FNV-1a) for a bitmap that is suitable for
// using bitmaps as elements in hash sets or as keys in hash maps, as well as
// generally quicker comparisons.
// The implementation is biased towards efficiency in little endian machines, so
// expect some extra CPU cycles and memory to be used if your machine is big endian.
// Likewise, don't use this to verify integrity unless you're certain you'll load
// the bitmap on a machine with the same endianess used to create it.
// Likewise, do not use this to verify integrity unless you are certain you will load
// the bitmap on a machine with the same endianess used to create it. (Thankfully
// very few people use big endian machines these days.)
func (rb *Bitmap) Checksum() uint64 {
const (
offset = 14695981039346656037
prime = 1099511628211
prime = 1099511628211
)
var bytes []byte
@ -106,6 +276,20 @@ func (rb *Bitmap) Checksum() uint64 {
return hash
}
// FromUnsafeBytes reads a serialized version of this bitmap from the byte buffer without copy.
// It is the caller's responsibility to ensure that the input data is not modified and remains valid for the entire lifetime of this bitmap.
// This method avoids small allocations but holds references to the input data buffer. It is GC-friendly, but it may consume more memory eventually.
// The containers in the resulting bitmap are immutable containers tied to the provided byte array and they rely on
// copy-on-write which means that modifying them creates copies. Thus FromUnsafeBytes is more likely to be appropriate for read-only use cases,
// when the resulting bitmap can be considered immutable.
//
// See also the FromBuffer function.
// See https://github.com/RoaringBitmap/roaring/pull/395 for more details.
func (rb *Bitmap) FromUnsafeBytes(data []byte, cookieHeader ...byte) (p int64, err error) {
stream := internal.NewByteBuffer(data)
return rb.ReadFrom(stream)
}
// ReadFrom reads a serialized version of this bitmap from stream.
// The format is compatible with other RoaringBitmap
// implementations (Java, C) and is documented here:
@ -114,12 +298,18 @@ func (rb *Bitmap) Checksum() uint64 {
// So add cookieHeader to accept the 4-byte data that has been read in roaring64.ReadFrom.
// It is not necessary to pass cookieHeader when call roaring.ReadFrom to read the roaring32 data directly.
func (rb *Bitmap) ReadFrom(reader io.Reader, cookieHeader ...byte) (p int64, err error) {
stream := internal.ByteInputAdapterPool.Get().(*internal.ByteInputAdapter)
stream.Reset(reader)
stream, ok := reader.(internal.ByteInput)
if !ok {
byteInputAdapter := internal.ByteInputAdapterPool.Get().(*internal.ByteInputAdapter)
byteInputAdapter.Reset(reader)
stream = byteInputAdapter
}
p, err = rb.highlowcontainer.readFrom(stream, cookieHeader...)
internal.ByteInputAdapterPool.Put(stream)
if !ok {
internal.ByteInputAdapterPool.Put(stream.(*internal.ByteInputAdapter))
}
return
}
@ -139,12 +329,17 @@ func (rb *Bitmap) ReadFrom(reader io.Reader, cookieHeader ...byte) (p int64, err
// You should *not* change the copy-on-write status of the resulting
// bitmaps (SetCopyOnWrite).
//
// Thus FromBuffer is more likely to be appropriate for read-only use cases,
// when the resulting bitmap can be considered immutable.
//
// If buf becomes unavailable, then a bitmap created with
// FromBuffer would be effectively broken. Furthermore, any
// bitmap derived from this bitmap (e.g., via Or, And) might
// also be broken. Thus, before making buf unavailable, you should
// call CloneCopyOnWriteContainers on all such bitmaps.
//
// See also the FromUnsafeBytes function which can have better performance
// in some cases.
func (rb *Bitmap) FromBuffer(buf []byte) (p int64, err error) {
stream := internal.ByteBufferPool.Get().(*internal.ByteBuffer)
stream.Reset(buf)
@ -194,6 +389,16 @@ func (rb *Bitmap) Clear() {
rb.highlowcontainer.clear()
}
// ToBitSet copies the content of the RoaringBitmap into a bitset.BitSet instance
func (rb *Bitmap) ToBitSet() *bitset.BitSet {
return bitset.From(rb.ToDense())
}
// FromBitSet creates a new RoaringBitmap from a bitset.BitSet instance
func FromBitSet(bitset *bitset.BitSet) *Bitmap {
return FromDense(bitset.Bytes(), false)
}
// ToArray creates a new slice containing all of the integers stored in the Bitmap in sorted order
func (rb *Bitmap) ToArray() []uint32 {
array := make([]uint32, rb.GetCardinality())
@ -233,7 +438,7 @@ func BoundSerializedSizeInBytes(cardinality uint64, universeSize uint64) uint64
contnbr := (universeSize + uint64(65535)) / uint64(65536)
if contnbr > cardinality {
contnbr = cardinality
// we can't have more containers than we have values
// we cannot have more containers than we have values
}
headermax := 8*contnbr + 4
if 4 > (contnbr+7)/8 {
@ -276,9 +481,9 @@ type intIterator struct {
// This way, instead of making up-to 64k allocations per full iteration
// we get a single allocation and simply reinitialize the appropriate
// iterator and point to it in the generic `iter` member on each key bound.
shortIter shortIterator
runIter runIterator16
bitmapIter bitmapContainerShortIterator
shortIter shortIterator
runIter runIterator16
bitmapIter bitmapContainerShortIterator
}
// HasNext returns true if there are more integers to iterate over
@ -341,14 +546,13 @@ func (ii *intIterator) AdvanceIfNeeded(minval uint32) {
// IntIterator is meant to allow you to iterate through the values of a bitmap, see Initialize(a *Bitmap)
type IntIterator = intIterator
// Initialize configures the existing iterator so that it can iterate through the values of
// the provided bitmap.
// The iteration results are undefined if the bitmap is modified (e.g., with Add or Remove).
func (p *intIterator) Initialize(a *Bitmap) {
p.pos = 0
p.highlowcontainer = &a.highlowcontainer
p.init()
func (ii *intIterator) Initialize(a *Bitmap) {
ii.pos = 0
ii.highlowcontainer = &a.highlowcontainer
ii.init()
}
type intReverseIterator struct {
@ -357,9 +561,9 @@ type intReverseIterator struct {
iter shortIterable
highlowcontainer *roaringArray
shortIter reverseIterator
runIter runReverseIterator16
bitmapIter reverseBitmapContainerShortIterator
shortIter reverseIterator
runIter runReverseIterator16
bitmapIter reverseBitmapContainerShortIterator
}
// HasNext returns true if there are more integers to iterate over
@ -414,10 +618,10 @@ type IntReverseIterator = intReverseIterator
// Initialize configures the existing iterator so that it can iterate through the values of
// the provided bitmap.
// The iteration results are undefined if the bitmap is modified (e.g., with Add or Remove).
func (p *intReverseIterator) Initialize(a *Bitmap) {
p.highlowcontainer = &a.highlowcontainer
p.pos = a.highlowcontainer.size() - 1
p.init()
func (ii *intReverseIterator) Initialize(a *Bitmap) {
ii.highlowcontainer = &a.highlowcontainer
ii.pos = a.highlowcontainer.size() - 1
ii.init()
}
// ManyIntIterable allows you to iterate over the values in a Bitmap
@ -434,9 +638,9 @@ type manyIntIterator struct {
iter manyIterable
highlowcontainer *roaringArray
shortIter shortIterator
runIter runIterator16
bitmapIter bitmapContainerManyIterator
shortIter shortIterator
runIter runIterator16
bitmapIter bitmapContainerManyIterator
}
func (ii *manyIntIterator) init() {
@ -495,17 +699,16 @@ func (ii *manyIntIterator) NextMany64(hs64 uint64, buf []uint64) int {
return n
}
// ManyIntIterator is meant to allow you to iterate through the values of a bitmap, see Initialize(a *Bitmap)
type ManyIntIterator = manyIntIterator
// Initialize configures the existing iterator so that it can iterate through the values of
// the provided bitmap.
// The iteration results are undefined if the bitmap is modified (e.g., with Add or Remove).
func (p *manyIntIterator) Initialize(a *Bitmap) {
p.pos = 0
p.highlowcontainer = &a.highlowcontainer
p.init()
func (ii *manyIntIterator) Initialize(a *Bitmap) {
ii.pos = 0
ii.highlowcontainer = &a.highlowcontainer
ii.init()
}
// String creates a string representation of the Bitmap
@ -569,7 +772,7 @@ func (rb *Bitmap) Iterate(cb func(x uint32) bool) {
// Iterator creates a new IntPeekable to iterate over the integers contained in the bitmap, in sorted order;
// the iterator becomes invalid if the bitmap is modified (e.g., with Add or Remove).
func (rb *Bitmap) Iterator() IntPeekable {
p := new(intIterator)
p := new(intIterator)
p.Initialize(rb)
return p
}
@ -847,7 +1050,7 @@ func (rb *Bitmap) Select(x uint32) (uint32, error) {
return uint32(key)<<16 + uint32(c.selectInt(uint16(remaining))), nil
}
}
return 0, fmt.Errorf("can't find %dth integer in a bitmap with only %d items", x, rb.GetCardinality())
return 0, fmt.Errorf("cannot find %dth integer in a bitmap with only %d items", x, rb.GetCardinality())
}
// And computes the intersection between two bitmaps and stores the result in the current bitmap

19
vendor/github.com/RoaringBitmap/roaring/roaringarray.go generated vendored
View file

@ -4,8 +4,9 @@ import (
"bytes"
"encoding/binary"
"fmt"
"github.com/RoaringBitmap/roaring/internal"
"io"
"github.com/RoaringBitmap/roaring/internal"
)
type container interface {
@ -112,9 +113,10 @@ func newRoaringArray() *roaringArray {
// runOptimize compresses the element containers to minimize space consumed.
// Q: how does this interact with copyOnWrite and needCopyOnWrite?
// A: since we aren't changing the logical content, just the representation,
// we don't bother to check the needCopyOnWrite bits. We replace
// (possibly all) elements of ra.containers in-place with space
// optimized versions.
//
// we don't bother to check the needCopyOnWrite bits. We replace
// (possibly all) elements of ra.containers in-place with space
// optimized versions.
func (ra *roaringArray) runOptimize() {
for i := range ra.containers {
ra.containers[i] = ra.containers[i].toEfficientContainer()
@ -465,9 +467,7 @@ func (ra *roaringArray) serializedSizeInBytes() uint64 {
return answer
}
//
// spec: https://github.com/RoaringBitmap/RoaringFormatSpec
//
func (ra *roaringArray) writeTo(w io.Writer) (n int64, err error) {
hasRun := ra.hasRunCompression()
isRunSizeInBytes := 0
@ -544,15 +544,14 @@ func (ra *roaringArray) writeTo(w io.Writer) (n int64, err error) {
return n, nil
}
//
// spec: https://github.com/RoaringBitmap/RoaringFormatSpec
//
func (ra *roaringArray) toBytes() ([]byte, error) {
var buf bytes.Buffer
_, err := ra.writeTo(&buf)
return buf.Bytes(), err
}
// Reads a serialized roaringArray from a byte slice.
func (ra *roaringArray) readFrom(stream internal.ByteInput, cookieHeader ...byte) (int64, error) {
var cookie uint32
var err error
@ -567,6 +566,8 @@ func (ra *roaringArray) readFrom(stream internal.ByteInput, cookieHeader ...byte
return stream.GetReadBytes(), fmt.Errorf("error in roaringArray.readFrom: could not read initial cookie: %s", err)
}
}
// If NextReturnsSafeSlice is false, then willNeedCopyOnWrite should be true
willNeedCopyOnWrite := !stream.NextReturnsSafeSlice()
var size uint32
var isRunBitmap []byte
@ -631,7 +632,7 @@ func (ra *roaringArray) readFrom(stream internal.ByteInput, cookieHeader ...byte
key := keycard[2*i]
card := int(keycard[2*i+1]) + 1
ra.keys[i] = key
ra.needCopyOnWrite[i] = true
ra.needCopyOnWrite[i] = willNeedCopyOnWrite
if isRunBitmap != nil && isRunBitmap[i/8]&(1<<(i%8)) != 0 {
// run container

49
vendor/github.com/RoaringBitmap/roaring/runcontainer.go generated vendored
View file

@ -47,6 +47,7 @@ import (
// runContainer16 does run-length encoding of sets of
// uint16 integers.
type runContainer16 struct {
// iv is a slice of sorted, non-overlapping, non-adjacent intervals.
iv []interval16
}
@ -253,10 +254,8 @@ func newRunContainer16FromBitmapContainer(bc *bitmapContainer) *runContainer16 {
}
//
// newRunContainer16FromArray populates a new
// runContainer16 from the contents of arr.
//
func newRunContainer16FromArray(arr *arrayContainer) *runContainer16 {
// keep this in sync with newRunContainer16FromVals above
@ -834,24 +833,23 @@ func (rc *runContainer16) numIntervals() int {
// If key is not already present, then whichInterval16 is
// set as follows:
//
// a) whichInterval16 == len(rc.iv)-1 if key is beyond our
// last interval16 in rc.iv;
// a) whichInterval16 == len(rc.iv)-1 if key is beyond our
// last interval16 in rc.iv;
//
// b) whichInterval16 == -1 if key is before our first
// interval16 in rc.iv;
// b) whichInterval16 == -1 if key is before our first
// interval16 in rc.iv;
//
// c) whichInterval16 is set to the minimum index of rc.iv
// which comes strictly before the key;
// so rc.iv[whichInterval16].last < key,
// and if whichInterval16+1 exists, then key < rc.iv[whichInterval16+1].start
// (Note that whichInterval16+1 won't exist when
// whichInterval16 is the last interval.)
// c) whichInterval16 is set to the minimum index of rc.iv
// which comes strictly before the key;
// so rc.iv[whichInterval16].last < key,
// and if whichInterval16+1 exists, then key < rc.iv[whichInterval16+1].start
// (Note that whichInterval16+1 won't exist when
// whichInterval16 is the last interval.)
//
// runContainer16.search always returns whichInterval16 < len(rc.iv).
//
// The search space is from startIndex to endxIndex. If endxIndex is set to zero, then there
// no upper bound.
//
func (rc *runContainer16) searchRange(key int, startIndex int, endxIndex int) (whichInterval16 int, alreadyPresent bool, numCompares int) {
n := int(len(rc.iv))
if n == 0 {
@ -937,21 +935,20 @@ func (rc *runContainer16) searchRange(key int, startIndex int, endxIndex int) (w
// If key is not already present, then whichInterval16 is
// set as follows:
//
// a) whichInterval16 == len(rc.iv)-1 if key is beyond our
// last interval16 in rc.iv;
// a) whichInterval16 == len(rc.iv)-1 if key is beyond our
// last interval16 in rc.iv;
//
// b) whichInterval16 == -1 if key is before our first
// interval16 in rc.iv;
// b) whichInterval16 == -1 if key is before our first
// interval16 in rc.iv;
//
// c) whichInterval16 is set to the minimum index of rc.iv
// which comes strictly before the key;
// so rc.iv[whichInterval16].last < key,
// and if whichInterval16+1 exists, then key < rc.iv[whichInterval16+1].start
// (Note that whichInterval16+1 won't exist when
// whichInterval16 is the last interval.)
// c) whichInterval16 is set to the minimum index of rc.iv
// which comes strictly before the key;
// so rc.iv[whichInterval16].last < key,
// and if whichInterval16+1 exists, then key < rc.iv[whichInterval16+1].start
// (Note that whichInterval16+1 won't exist when
// whichInterval16 is the last interval.)
//
// runContainer16.search always returns whichInterval16 < len(rc.iv).
//
func (rc *runContainer16) search(key int) (whichInterval16 int, alreadyPresent bool, numCompares int) {
return rc.searchRange(key, 0, 0)
}
@ -994,7 +991,6 @@ func newRunContainer16() *runContainer16 {
// newRunContainer16CopyIv creates a run container, initializing
// with a copy of the supplied iv slice.
//
func newRunContainer16CopyIv(iv []interval16) *runContainer16 {
rc := &runContainer16{
iv: make([]interval16, len(iv)),
@ -1011,7 +1007,6 @@ func (rc *runContainer16) Clone() *runContainer16 {
// newRunContainer16TakeOwnership returns a new runContainer16
// backed by the provided iv slice, which we will
// assume exclusive control over from now on.
//
func newRunContainer16TakeOwnership(iv []interval16) *runContainer16 {
rc := &runContainer16{
iv: iv,
@ -2006,7 +2001,6 @@ func (rc *runContainer16) not(firstOfRange, endx int) container {
// Current routine is correct but
// makes 2 more passes through the arrays than should be
// strictly necessary. Measure both ways though--this may not matter.
//
func (rc *runContainer16) Not(firstOfRange, endx int) *runContainer16 {
if firstOfRange > endx {
@ -2329,7 +2323,6 @@ func runArrayUnionToRuns(rc *runContainer16, ac *arrayContainer) ([]interval16,
// the backing array, and then you write
// the answer at the beginning. What this
// trick does is minimize memory allocations.
//
func (rc *runContainer16) lazyIOR(a container) container {
// not lazy at the moment
return rc.ior(a)

1
vendor/github.com/RoaringBitmap/roaring/serialization.go generated vendored
View file

@ -7,7 +7,6 @@ import (
// writeTo for runContainer16 follows this
// spec: https://github.com/RoaringBitmap/RoaringFormatSpec
//
func (b *runContainer16) writeTo(stream io.Writer) (int, error) {
buf := make([]byte, 2+4*len(b.iv))
binary.LittleEndian.PutUint16(buf[0:], uint16(len(b.iv)))

94
vendor/github.com/RoaringBitmap/roaring/serialization_littleendian.go generated vendored
View file

@ -79,12 +79,12 @@ func (bc *bitmapContainer) asLittleEndianByteSlice() []byte {
// Deserialization code follows
////
// //
// These methods (byteSliceAsUint16Slice,...) do not make copies,
// they are pointer-based (unsafe). The caller is responsible to
// ensure that the input slice does not get garbage collected, deleted
// or modified while you hold the returned slince.
////
// //
func byteSliceAsUint16Slice(slice []byte) (result []uint16) { // here we create a new slice holder
if len(slice)%2 != 0 {
panic("Slice size should be divisible by 2")
@ -295,7 +295,6 @@ func byteSliceAsBoolSlice(slice []byte) (result []bool) {
// bitmap derived from this bitmap (e.g., via Or, And) might
// also be broken. Thus, before making buf unavailable, you should
// call CloneCopyOnWriteContainers on all such bitmaps.
//
func (rb *Bitmap) FrozenView(buf []byte) error {
return rb.highlowcontainer.frozenView(buf)
}
@ -313,7 +312,7 @@ func (rb *Bitmap) FrozenView(buf []byte) error {
* <typecodes> uint8_t[num_containers]
* <header> uint32_t
*
* <header> is a 4-byte value which is a bit union of FROZEN_COOKIE (15 bits)
* <header> is a 4-byte value which is a bit union of frozenCookie (15 bits)
* and the number of containers (17 bits).
*
* <counts> stores number of elements for every container.
@ -329,43 +328,50 @@ func (rb *Bitmap) FrozenView(buf []byte) error {
* All members have their native alignments during deserilization except <header>,
* which is not guaranteed to be aligned by 4 bytes.
*/
const FROZEN_COOKIE = 13766
const frozenCookie = 13766
var (
FrozenBitmapInvalidCookie = errors.New("header does not contain the FROZEN_COOKIE")
FrozenBitmapBigEndian = errors.New("loading big endian frozen bitmaps is not supported")
FrozenBitmapIncomplete = errors.New("input buffer too small to contain a frozen bitmap")
FrozenBitmapOverpopulated = errors.New("too many containers")
FrozenBitmapUnexpectedData = errors.New("spurious data in input")
FrozenBitmapInvalidTypecode = errors.New("unrecognized typecode")
FrozenBitmapBufferTooSmall = errors.New("buffer too small")
// ErrFrozenBitmapInvalidCookie is returned when the header does not contain the frozenCookie.
ErrFrozenBitmapInvalidCookie = errors.New("header does not contain the frozenCookie")
// ErrFrozenBitmapBigEndian is returned when the header is big endian.
ErrFrozenBitmapBigEndian = errors.New("loading big endian frozen bitmaps is not supported")
// ErrFrozenBitmapIncomplete is returned when the buffer is too small to contain a frozen bitmap.
ErrFrozenBitmapIncomplete = errors.New("input buffer too small to contain a frozen bitmap")
// ErrFrozenBitmapOverpopulated is returned when the number of containers is too large.
ErrFrozenBitmapOverpopulated = errors.New("too many containers")
// ErrFrozenBitmapUnexpectedData is returned when the buffer contains unexpected data.
ErrFrozenBitmapUnexpectedData = errors.New("spurious data in input")
// ErrFrozenBitmapInvalidTypecode is returned when the typecode is invalid.
ErrFrozenBitmapInvalidTypecode = errors.New("unrecognized typecode")
// ErrFrozenBitmapBufferTooSmall is returned when the buffer is too small.
ErrFrozenBitmapBufferTooSmall = errors.New("buffer too small")
)
func (ra *roaringArray) frozenView(buf []byte) error {
if len(buf) < 4 {
return FrozenBitmapIncomplete
return ErrFrozenBitmapIncomplete
}
headerBE := binary.BigEndian.Uint32(buf[len(buf)-4:])
if headerBE&0x7fff == FROZEN_COOKIE {
return FrozenBitmapBigEndian
if headerBE&0x7fff == frozenCookie {
return ErrFrozenBitmapBigEndian
}
header := binary.LittleEndian.Uint32(buf[len(buf)-4:])
buf = buf[:len(buf)-4]
if header&0x7fff != FROZEN_COOKIE {
return FrozenBitmapInvalidCookie
if header&0x7fff != frozenCookie {
return ErrFrozenBitmapInvalidCookie
}
nCont := int(header >> 15)
if nCont > (1 << 16) {
return FrozenBitmapOverpopulated
return ErrFrozenBitmapOverpopulated
}
// 1 byte per type, 2 bytes per key, 2 bytes per count.
if len(buf) < 5*nCont {
return FrozenBitmapIncomplete
return ErrFrozenBitmapIncomplete
}
types := buf[len(buf)-nCont:]
@ -390,12 +396,12 @@ func (ra *roaringArray) frozenView(buf []byte) error {
nRun++
nRunEl += int(counts[i])
default:
return FrozenBitmapInvalidTypecode
return ErrFrozenBitmapInvalidTypecode
}
}
if len(buf) < (1<<13)*nBitmap+4*nRunEl+2*nArrayEl {
return FrozenBitmapIncomplete
return ErrFrozenBitmapIncomplete
}
bitsetsArena := byteSliceAsUint64Slice(buf[:(1<<13)*nBitmap])
@ -408,15 +414,15 @@ func (ra *roaringArray) frozenView(buf []byte) error {
buf = buf[2*nArrayEl:]
if len(buf) != 0 {
return FrozenBitmapUnexpectedData
return ErrFrozenBitmapUnexpectedData
}
var c container
containersSz := int(unsafe.Sizeof(c))*nCont
bitsetsSz := int(unsafe.Sizeof(bitmapContainer{}))*nBitmap
arraysSz := int(unsafe.Sizeof(arrayContainer{}))*nArray
runsSz := int(unsafe.Sizeof(runContainer16{}))*nRun
needCOWSz := int(unsafe.Sizeof(true))*nCont
containersSz := int(unsafe.Sizeof(c)) * nCont
bitsetsSz := int(unsafe.Sizeof(bitmapContainer{})) * nBitmap
arraysSz := int(unsafe.Sizeof(arrayContainer{})) * nArray
runsSz := int(unsafe.Sizeof(runContainer16{})) * nRun
needCOWSz := int(unsafe.Sizeof(true)) * nCont
bitmapArenaSz := containersSz + bitsetsSz + arraysSz + runsSz + needCOWSz
bitmapArena := make([]byte, bitmapArenaSz)
@ -475,9 +481,10 @@ func (ra *roaringArray) frozenView(buf []byte) error {
return nil
}
func (bm *Bitmap) GetFrozenSizeInBytes() uint64 {
// GetFrozenSizeInBytes returns the size in bytes of the frozen bitmap.
func (rb *Bitmap) GetFrozenSizeInBytes() uint64 {
nBits, nArrayEl, nRunEl := uint64(0), uint64(0), uint64(0)
for _, c := range bm.highlowcontainer.containers {
for _, c := range rb.highlowcontainer.containers {
switch v := c.(type) {
case *bitmapContainer:
nBits++
@ -487,19 +494,21 @@ func (bm *Bitmap) GetFrozenSizeInBytes() uint64 {
nRunEl += uint64(len(v.iv))
}
}
return 4 + 5*uint64(len(bm.highlowcontainer.containers)) +
return 4 + 5*uint64(len(rb.highlowcontainer.containers)) +
(nBits << 13) + 2*nArrayEl + 4*nRunEl
}
func (bm *Bitmap) Freeze() ([]byte, error) {
sz := bm.GetFrozenSizeInBytes()
// Freeze serializes the bitmap in the CRoaring's frozen format.
func (rb *Bitmap) Freeze() ([]byte, error) {
sz := rb.GetFrozenSizeInBytes()
buf := make([]byte, sz)
_, err := bm.FreezeTo(buf)
_, err := rb.FreezeTo(buf)
return buf, err
}
func (bm *Bitmap) FreezeTo(buf []byte) (int, error) {
containers := bm.highlowcontainer.containers
// FreezeTo serializes the bitmap in the CRoaring's frozen format.
func (rb *Bitmap) FreezeTo(buf []byte) (int, error) {
containers := rb.highlowcontainer.containers
nCont := len(containers)
nBits, nArrayEl, nRunEl := 0, 0, 0
@ -516,7 +525,7 @@ func (bm *Bitmap) FreezeTo(buf []byte) (int, error) {
serialSize := 4 + 5*nCont + (1<<13)*nBits + 4*nRunEl + 2*nArrayEl
if len(buf) < serialSize {
return 0, FrozenBitmapBufferTooSmall
return 0, ErrFrozenBitmapBufferTooSmall
}
bitsArena := byteSliceAsUint64Slice(buf[:(1<<13)*nBits])
@ -537,10 +546,10 @@ func (bm *Bitmap) FreezeTo(buf []byte) (int, error) {
types := buf[:nCont]
buf = buf[nCont:]
header := uint32(FROZEN_COOKIE | (nCont << 15))
header := uint32(frozenCookie | (nCont << 15))
binary.LittleEndian.PutUint32(buf[:4], header)
copy(keys, bm.highlowcontainer.keys[:])
copy(keys, rb.highlowcontainer.keys[:])
for i, c := range containers {
switch v := c.(type) {
@ -567,11 +576,12 @@ func (bm *Bitmap) FreezeTo(buf []byte) (int, error) {
return serialSize, nil
}
func (bm *Bitmap) WriteFrozenTo(wr io.Writer) (int, error) {
// WriteFrozenTo serializes the bitmap in the CRoaring's frozen format.
func (rb *Bitmap) WriteFrozenTo(wr io.Writer) (int, error) {
// FIXME: this is a naive version that iterates 4 times through the
// containers and allocates 3*len(containers) bytes; it's quite likely
// it can be done more efficiently.
containers := bm.highlowcontainer.containers
containers := rb.highlowcontainer.containers
written := 0
for _, c := range containers {
@ -610,7 +620,7 @@ func (bm *Bitmap) WriteFrozenTo(wr io.Writer) (int, error) {
}
}
n, err := wr.Write(uint16SliceAsByteSlice(bm.highlowcontainer.keys))
n, err := wr.Write(uint16SliceAsByteSlice(rb.highlowcontainer.keys))
written += n
if err != nil {
return written, err
@ -642,7 +652,7 @@ func (bm *Bitmap) WriteFrozenTo(wr io.Writer) (int, error) {
return written, err
}
header := uint32(FROZEN_COOKIE | (len(containers) << 15))
header := uint32(frozenCookie | (len(containers) << 15))
if err := binary.Write(wr, binary.LittleEndian, header); err != nil {
return written, err
}