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
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
|
package com
import (
"context"
"github.com/icinga/icingadb/pkg/contracts"
"golang.org/x/sync/errgroup"
"sync"
"time"
)
// BulkChunkSplitPolicy is a state machine which tracks the items of a chunk a bulker assembles.
// A call takes an item for the current chunk into account.
// Output true indicates that the state machine was reset first and the bulker
// shall finish the current chunk now (not e.g. once $size is reached) without the given item.
type BulkChunkSplitPolicy[T any] func(T) bool
type BulkChunkSplitPolicyFactory[T any] func() BulkChunkSplitPolicy[T]
// NeverSplit returns a pseudo state machine which never demands splitting.
func NeverSplit[T any]() BulkChunkSplitPolicy[T] {
return neverSplit[T]
}
// SplitOnDupId returns a state machine which tracks the inputs' IDs.
// Once an already seen input arrives, it demands splitting.
func SplitOnDupId[T contracts.IDer]() BulkChunkSplitPolicy[T] {
seenIds := map[string]struct{}{}
return func(ider T) bool {
id := ider.ID().String()
_, ok := seenIds[id]
if ok {
seenIds = map[string]struct{}{id: {}}
} else {
seenIds[id] = struct{}{}
}
return ok
}
}
func neverSplit[T any](T) bool {
return false
}
// Bulker reads all values from a channel and streams them in chunks into a Bulk channel.
type Bulker[T any] struct {
ch chan []T
ctx context.Context
mu sync.Mutex
}
// NewBulker returns a new Bulker and starts streaming.
func NewBulker[T any](
ctx context.Context, ch <-chan T, count int, splitPolicyFactory BulkChunkSplitPolicyFactory[T],
) *Bulker[T] {
b := &Bulker[T]{
ch: make(chan []T),
ctx: ctx,
mu: sync.Mutex{},
}
go b.run(ch, count, splitPolicyFactory)
return b
}
// Bulk returns the channel on which the bulks are delivered.
func (b *Bulker[T]) Bulk() <-chan []T {
return b.ch
}
func (b *Bulker[T]) run(ch <-chan T, count int, splitPolicyFactory BulkChunkSplitPolicyFactory[T]) {
defer close(b.ch)
bufCh := make(chan T, count)
splitPolicy := splitPolicyFactory()
g, ctx := errgroup.WithContext(b.ctx)
g.Go(func() error {
defer close(bufCh)
for {
select {
case v, ok := <-ch:
if !ok {
return nil
}
bufCh <- v
case <-ctx.Done():
return ctx.Err()
}
}
})
g.Go(func() error {
for done := false; !done; {
buf := make([]T, 0, count)
timeout := time.After(256 * time.Millisecond)
for drain := true; drain && len(buf) < count; {
select {
case v, ok := <-bufCh:
if !ok {
drain = false
done = true
break
}
if splitPolicy(v) {
if len(buf) > 0 {
b.ch <- buf
buf = make([]T, 0, count)
}
timeout = time.After(256 * time.Millisecond)
}
buf = append(buf, v)
case <-timeout:
drain = false
case <-ctx.Done():
return ctx.Err()
}
}
if len(buf) > 0 {
b.ch <- buf
}
splitPolicy = splitPolicyFactory()
}
return nil
})
// We don't expect an error here.
// We only use errgroup for the encapsulated use of sync.WaitGroup.
_ = g.Wait()
}
// Bulk reads all values from a channel and streams them in chunks into a returned channel.
func Bulk[T any](
ctx context.Context, ch <-chan T, count int, splitPolicyFactory BulkChunkSplitPolicyFactory[T],
) <-chan []T {
if count <= 1 {
return oneBulk(ctx, ch)
}
return NewBulker(ctx, ch, count, splitPolicyFactory).Bulk()
}
// oneBulk operates just as NewBulker(ctx, ch, 1, splitPolicy).Bulk(),
// but without the overhead of the actual bulk creation with a buffer channel, timeout and BulkChunkSplitPolicy.
func oneBulk[T any](ctx context.Context, ch <-chan T) <-chan []T {
out := make(chan []T)
go func() {
defer close(out)
for {
select {
case item, ok := <-ch:
if !ok {
return
}
select {
case out <- []T{item}:
case <-ctx.Done():
return
}
case <-ctx.Done():
return
}
}
}()
return out
}
var (
_ BulkChunkSplitPolicyFactory[struct{}] = NeverSplit[struct{}]
_ BulkChunkSplitPolicyFactory[contracts.Entity] = SplitOnDupId[contracts.Entity]
)
|
