[#13] Restore deleted copyright notice

Signed-off-by: Vitaliy Potyarkin <v.potyarkin@yadro.com>

LICENSE

@@ -1,6 +1,7 @@
 MIT License
 
-Copyright (c) 2019 NSPCC
+Copyright (c) 2023-2024 TrueCloudLab
+Copyright (c) 2019-2023 NSPCC
 
 Permission is hereby granted, free of charge, to any person obtaining a copy
 of this software and associated documentation files (the "Software"), to deal

README.md

@@ -1,15 +1,10 @@
 # Golang HRW implementation
 
-[![Build Status](https://travis-ci.org/nspcc-dev/hrw.svg?branch=master)](https://travis-ci.org/nspcc-dev/hrw)
-[![codecov](https://codecov.io/gh/nspcc-dev/hrw/badge.svg)](https://codecov.io/gh/nspcc-dev/hrw)
-[![Report](https://goreportcard.com/badge/github.com/nspcc-dev/hrw)](https://goreportcard.com/report/github.com/nspcc-dev/hrw)
-[![GitHub release](https://img.shields.io/github/release/nspcc-dev/hrw.svg)](https://github.com/nspcc-dev/hrw)
-
 [Rendezvous or highest random weight](https://en.wikipedia.org/wiki/Rendezvous_hashing) (HRW) hashing is an algorithm that allows clients to achieve distributed agreement on a set of k options out of a possible set of n options. A typical application is when clients need to agree on which sites (or proxies) objects are assigned to. When k is 1, it subsumes the goals of consistent hashing, using an entirely different method.
 
 ## Install
 
-`go get github.com/nspcc-dev/hrw`
+`go get git.frostfs.info/TrueCloudLab/hrw`
 
 ## Benchmark:
 

go.mod

@@ -3,8 +3,8 @@ module git.frostfs.info/TrueCloudLab/hrw
 go 1.18
 
 require (
-	github.com/spaolacci/murmur3 v1.1.0
 	github.com/stretchr/testify v1.3.0
+	github.com/twmb/murmur3 v1.1.8
 )
 
 require (

go.sum

@@ -2,8 +2,8 @@ github.com/davecgh/go-spew v1.1.0 h1:ZDRjVQ15GmhC3fiQ8ni8+OwkZQO4DARzQgrnXU1Liz8
 github.com/davecgh/go-spew v1.1.0/go.mod h1:J7Y8YcW2NihsgmVo/mv3lAwl/skON4iLHjSsI+c5H38=
 github.com/pmezard/go-difflib v1.0.0 h1:4DBwDE0NGyQoBHbLQYPwSUPoCMWR5BEzIk/f1lZbAQM=
 github.com/pmezard/go-difflib v1.0.0/go.mod h1:iKH77koFhYxTK1pcRnkKkqfTogsbg7gZNVY4sRDYZ/4=
-github.com/spaolacci/murmur3 v1.1.0 h1:7c1g84S4BPRrfL5Xrdp6fOJ206sU9y293DDHaoy0bLI=
-github.com/spaolacci/murmur3 v1.1.0/go.mod h1:JwIasOWyU6f++ZhiEuf87xNszmSA2myDM2Kzu9HwQUA=
 github.com/stretchr/objx v0.1.0/go.mod h1:HFkY916IF+rwdDfMAkV7OtwuqBVzrE8GR6GFx+wExME=
 github.com/stretchr/testify v1.3.0 h1:TivCn/peBQ7UY8ooIcPgZFpTNSz0Q2U6UrFlUfqbe0Q=
 github.com/stretchr/testify v1.3.0/go.mod h1:M5WIy9Dh21IEIfnGCwXGc5bZfKNJtfHm1UVUgZn+9EI=
+github.com/twmb/murmur3 v1.1.8 h1:8Yt9taO/WN3l08xErzjeschgZU2QSrwm1kclYq+0aRg=
+github.com/twmb/murmur3 v1.1.8/go.mod h1:Qq/R7NUyOfr65zD+6Q5IHKsJLwP7exErjN6lyyq3OSQ=

hrw.go

@@ -9,7 +9,7 @@ import (
 	"reflect"
 	"sort"
 
-	"github.com/spaolacci/murmur3"
+	"github.com/twmb/murmur3"
 )
 
 type (
@@ -21,6 +21,12 @@ type (
 		less func(i, j int) bool
 		swap func(i, j int)
 	}
+
+	hasherSorter[T Hasher, N interface{ ~uint64 | ~float64 }] struct {
+		slice []T
+		dist  []N
+		asc   bool
+	}
 )
 
 // Boundaries of valid normalized weights
@@ -33,6 +39,18 @@ func (s *sorter) Len() int           { return s.l }
 func (s *sorter) Less(i, j int) bool { return s.less(i, j) }
 func (s *sorter) Swap(i, j int)      { s.swap(i, j) }
 
+func (s *hasherSorter[T, N]) Len() int { return len(s.slice) }
+func (s *hasherSorter[T, N]) Less(i, j int) bool {
+	if s.asc {
+		return s.dist[i] < s.dist[j]
+	}
+	return s.dist[i] > s.dist[j]
+}
+func (s *hasherSorter[T, N]) Swap(i, j int) {
+	s.slice[i], s.slice[j] = s.slice[j], s.slice[i]
+	s.dist[i], s.dist[j] = s.dist[j], s.dist[i]
+}
+
 func distance(x uint64, y uint64) uint64 {
 	acc := x ^ y
 	// here used mmh3 64 bit finalizer
@@ -50,6 +68,11 @@ func Hash(key []byte) uint64 {
 	return murmur3.Sum64(key)
 }
 
+// StringHash uses murmur3 hash to return uint64
+func StringHash(key string) uint64 {
+	return murmur3.StringSum64(key)
+}
+
 // Sort receive nodes and hash, and sort it by distance
 func Sort(nodes []uint64, hash uint64) []uint64 {
 	l := len(nodes)
@@ -85,13 +108,15 @@ func SortSliceByValue(slice interface{}, hash uint64) {
 
 // SortHasherSliceByValue receives []Hasher and hash to sort by value-distance.
 func SortHasherSliceByValue[T Hasher](slice []T, hash uint64) {
-	rule := prepareHasherRule(slice)
-	if rule != nil {
-		swap := func(i, j int) {
-			slice[i], slice[j] = slice[j], slice[i]
+	if len(slice) == 0 {
+		return
 	}
-		sortByDistance(len(rule), false, rule, hash, swap)
+
+	dist := make([]uint64, len(slice))
+	for i := range dist {
+		dist[i] = distance(slice[i].Hash(), hash)
 	}
+	sortHasherByDistance(slice, false, dist)
 }
 
 // SortSliceByWeightValue received []T, weights and hash to sort by value-distance * weights
@@ -105,13 +130,41 @@ func SortSliceByWeightValue(slice interface{}, weights []float64, hash uint64) {
 
 // SortHasherSliceByWeightValue receives []Hasher, weights and hash to sort by value-distance * weights.
 func SortHasherSliceByWeightValue[T Hasher](slice []T, weights []float64, hash uint64) {
-	rule := prepareHasherRule(slice)
-	if rule != nil {
-		swap := func(i, j int) {
-			slice[i], slice[j] = slice[j], slice[i]
+	if len(slice) == 0 {
+		return
 	}
-		sortByWeight(len(slice), false, rule, weights, hash, swap)
+
+	if allSameF64(weights) {
+		dist := make([]uint64, len(slice))
+		for i := range dist {
+			dist[i] = distance(slice[i].Hash(), hash)
 		}
+		sortHasherByDistance(slice, false, dist)
+		return
+	}
+
+	dist := make([]float64, len(slice))
+	for i := range dist {
+		d := distance(slice[i].Hash(), hash)
+		// `maxUint64 - distance` makes the shorter distance more valuable
+		// it is necessary for operation with normalized values
+		dist[i] = float64(^uint64(0)-d) * weights[i]
+	}
+
+	sort.Sort(&hasherSorter[T, float64]{
+		slice: slice,
+		dist:  dist,
+		asc:   false,
+	})
+}
+
+// sortHasherByDistance is similar to sortByDistance but accepts slice directly.
+func sortHasherByDistance[T Hasher](slice []T, byIndex bool, dist []uint64) {
+	sort.Sort(&hasherSorter[T, uint64]{
+		slice: slice,
+		dist:  dist,
+		asc:   true,
+	})
 }
 
 // SortSliceByIndex received []T and hash to sort by index-distance
@@ -221,18 +274,6 @@ func prepareRule(slice interface{}) []uint64 {
 	return rule
 }
 
-func prepareHasherRule[T Hasher](hashers []T) []uint64 {
-	length := len(hashers)
-	if length == 0 {
-		return nil
-	}
-	result := make([]uint64, length)
-	for i := 0; i < length; i++ {
-		result[i] = hashers[i].Hash()
-	}
-	return result
-}
-
 // ValidateWeights checks if weights are normalized between 0.0 and 1.0
 func ValidateWeights(weights []float64) error {
 	for i := range weights {
@@ -243,24 +284,6 @@ func ValidateWeights(weights []float64) error {
 	return nil
 }
 
-func newSorter(l int, byIndex bool, nodes []uint64, h uint64,
-	swap func(i, j int)) (*sorter, []int, []uint64) {
-	ind := make([]int, l)
-	dist := make([]uint64, l)
-	for i := 0; i < l; i++ {
-		ind[i] = i
-		dist[i] = getDistance(byIndex, i, nodes, h)
-	}
-
-	return &sorter{
-		l: l,
-		swap: func(i, j int) {
-			swap(i, j)
-			ind[i], ind[j] = ind[j], ind[i]
-		},
-	}, ind, dist
-}
-
 // sortByWeight sorts nodes by weight using provided swapper.
 // nodes contains hrw hashes. If it is nil, indices are used.
 func sortByWeight(l int, byIndex bool, nodes []uint64, weights []float64, hash uint64, swap func(i, j int)) {
@@ -270,14 +293,23 @@ func sortByWeight(l int, byIndex bool, nodes []uint64, weights []float64, hash u
 		return
 	}
 
-	s, ind, dist := newSorter(l, byIndex, nodes, hash, swap)
-	s.less = func(i, j int) bool {
-		ii, jj := ind[i], ind[j]
+	dist := make([]float64, l)
+	for i := 0; i < l; i++ {
+		d := getDistance(byIndex, i, nodes, hash)
 		// `maxUint64 - distance` makes the shorter distance more valuable
 		// it is necessary for operation with normalized values
-		wi := float64(^uint64(0)-dist[ii]) * weights[ii]
-		wj := float64(^uint64(0)-dist[jj]) * weights[jj]
-		return wi > wj // higher distance must be placed lower to be first
+		dist[i] = float64(^uint64(0)-d) * weights[i]
+	}
+
+	s := &sorter{
+		l: l,
+		swap: func(i, j int) {
+			swap(i, j)
+			dist[i], dist[j] = dist[j], dist[i]
+		},
+		less: func(i, j int) bool {
+			return dist[i] > dist[j] // higher distance must be placed lower to be first
+		},
 	}
 	sort.Sort(s)
 }
@@ -285,9 +317,20 @@ func sortByWeight(l int, byIndex bool, nodes []uint64, weights []float64, hash u
 // sortByDistance sorts nodes by hrw distance using provided swapper.
 // nodes contains hrw hashes. If it is nil, indices are used.
 func sortByDistance(l int, byIndex bool, nodes []uint64, hash uint64, swap func(i, j int)) {
-	s, ind, dist := newSorter(l, byIndex, nodes, hash, swap)
-	s.less = func(i, j int) bool {
-		return dist[ind[i]] < dist[ind[j]]
+	dist := make([]uint64, l)
+	for i := 0; i < l; i++ {
+		dist[i] = getDistance(byIndex, i, nodes, hash)
+	}
+
+	s := &sorter{
+		l: l,
+		swap: func(i, j int) {
+			swap(i, j)
+			dist[i], dist[j] = dist[j], dist[i]
+		},
+		less: func(i, j int) bool {
+			return dist[i] < dist[j]
+		},
 	}
 	sort.Sort(s)
 }

hrw_test.go

@@ -61,7 +61,7 @@ func Example() {
 }
 
 func (h hashString) Hash() uint64 {
-	return Hash([]byte(h))
+	return StringHash(string(h))
 }
 
 func TestSortSliceByIndex(t *testing.T) {
@@ -737,7 +737,7 @@ func BenchmarkSortHashersByWeightValueReflection_fnv_100(b *testing.B) {
 }
 
 func BenchmarkSortHashersByWeightValueReflection_fnv_1000(b *testing.B) {
-	benchmarkSortHashersByWeightValueRelection(b, 100, Hash(testKey))
+	benchmarkSortHashersByWeightValueRelection(b, 1000, Hash(testKey))
 }
 
 func BenchmarkSortHashersByWeightValueTyped_fnv_10(b *testing.B) {
@@ -749,7 +749,7 @@ func BenchmarkSortHashersByWeightValueTyped_fnv_100(b *testing.B) {
 }
 
 func BenchmarkSortHashersByWeightValueTyped_fnv_1000(b *testing.B) {
-	benchmarkSortHashersByWeightValueTyped(b, 100, Hash(testKey))
+	benchmarkSortHashersByWeightValueTyped(b, 1000, Hash(testKey))
 }
 
 func benchmarkSort(b *testing.B, n int, hash uint64) uint64 {