forked from ava-labs/avalanchego
-
Notifications
You must be signed in to change notification settings - Fork 4
/
weighted_heap.go
106 lines (92 loc) · 2.76 KB
/
weighted_heap.go
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
// Copyright (C) 2019-2023, Ava Labs, Inc. All rights reserved.
// See the file LICENSE for licensing terms.
package sampler
import (
"github.com/MetalBlockchain/metalgo/utils"
"github.com/MetalBlockchain/metalgo/utils/math"
)
var (
_ Weighted = (*weightedHeap)(nil)
_ utils.Sortable[weightedHeapElement] = weightedHeapElement{}
)
type weightedHeapElement struct {
weight uint64
cumulativeWeight uint64
index int
}
func (e weightedHeapElement) Less(other weightedHeapElement) bool {
// By accounting for the initial index of the weights, this results in a
// stable sort. We do this rather than using `sort.Stable` because of the
// reported change in performance of the sort used.
if e.weight > other.weight {
return true
}
if e.weight < other.weight {
return false
}
return e.index < other.index
}
// Sampling is performed by executing a search over a tree of elements in the
// order of their probabilistic occurrence.
//
// Initialization takes O(n * log(n)) time, where n is the number of elements
// that can be sampled.
// Sampling can take up to O(log(n)) time. As the distribution becomes more
// biased, sampling will become faster in expectation.
type weightedHeap struct {
heap []weightedHeapElement
}
func (s *weightedHeap) Initialize(weights []uint64) error {
numWeights := len(weights)
if numWeights <= cap(s.heap) {
s.heap = s.heap[:numWeights]
} else {
s.heap = make([]weightedHeapElement, numWeights)
}
for i, weight := range weights {
s.heap[i] = weightedHeapElement{
weight: weight,
cumulativeWeight: weight,
index: i,
}
}
// Optimize so that the most probable values are at the top of the heap
utils.Sort(s.heap)
// Initialize the heap
for i := len(s.heap) - 1; i > 0; i-- {
// Explicitly performing a shift here allows the compiler to avoid
// checking for negative numbers, which saves a couple cycles
parentIndex := (i - 1) >> 1
newWeight, err := math.Add64(
s.heap[parentIndex].cumulativeWeight,
s.heap[i].cumulativeWeight,
)
if err != nil {
return err
}
s.heap[parentIndex].cumulativeWeight = newWeight
}
return nil
}
func (s *weightedHeap) Sample(value uint64) (int, error) {
if len(s.heap) == 0 || s.heap[0].cumulativeWeight <= value {
return 0, ErrOutOfRange
}
index := 0
for {
currentElement := s.heap[index]
currentWeight := currentElement.weight
if value < currentWeight {
return currentElement.index, nil
}
value -= currentWeight
// We shouldn't return the root, so check the left child
index = index*2 + 1
if leftWeight := s.heap[index].cumulativeWeight; leftWeight <= value {
// If the weight is greater than the left weight, you should move to
// the right child
value -= leftWeight
index++
}
}
}