-
Notifications
You must be signed in to change notification settings - Fork 100
/
KafkaProducer.scala
314 lines (278 loc) · 10.5 KB
/
KafkaProducer.scala
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
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
/*
* Copyright 2018-2023 OVO Energy Limited
*
* SPDX-License-Identifier: Apache-2.0
*/
package fs2.kafka
import cats.{Apply, Functor}
import cats.effect._
import cats.syntax.all._
import fs2._
import fs2.kafka.internal._
import fs2.kafka.producer.MkProducer
import org.apache.kafka.clients.producer.RecordMetadata
import org.apache.kafka.common.{Metric, MetricName, PartitionInfo}
import scala.annotation.nowarn
import scala.concurrent.Promise
/**
* [[KafkaProducer]] represents a producer of Kafka records, with the
* ability to produce `ProducerRecord`s using [[produce]].
*/
abstract class KafkaProducer[F[_], K, V] {
/**
* Produces the specified [[ProducerRecords]] in two steps: the
* first effect puts the records in the buffer of the producer,
* and the second effect waits for the records to send.
*
* It's possible to `flatten` the result from this function to
* have an effect which both sends the records and waits for
* them to finish sending.
*
* Waiting for individual records to send can substantially
* limit performance. In some cases, this is necessary, and
* so we might want to consider the following alternatives.
*
* - Wait for the produced records in batches, improving
* the rate at which records are produced, but loosing
* the guarantee where `produce >> otherAction` means
* `otherAction` executes after the record has been
* sent.<br>
* - Run several `produce.flatten >> otherAction` concurrently,
* improving the rate at which records are produced, and still
* have `otherAction` execute after records have been sent,
* but losing the order of produced records.
*/
def produce(
records: ProducerRecords[K, V]
): F[F[ProducerResult[K, V]]]
}
object KafkaProducer {
implicit class ProducerOps[F[_], K, V](private val producer: KafkaProducer[F, K, V])
extends AnyVal {
/**
* Produce a single [[ProducerRecord]], see [[KafkaProducer.produce]] for general semantics.
*/
def produceOne_(record: ProducerRecord[K, V])(implicit F: Functor[F]): F[F[RecordMetadata]] =
produceOne(record).map(_.map { res =>
res.head.get._2 //Should always be present so get is ok
})
/**
* Produce a single record to the specified topic using the provided key and value,
* see [[KafkaProducer.produce]] for general semantics.
*/
def produceOne_(topic: String, key: K, value: V)(implicit F: Functor[F]): F[F[RecordMetadata]] =
produceOne_(ProducerRecord(topic, key, value))
/**
* Produce a single record to the specified topic using the provided key and value,
* see [[KafkaProducer.produce]] for general semantics.
*/
def produceOne(
topic: String,
key: K,
value: V
): F[F[ProducerResult[K, V]]] =
produceOne(ProducerRecord(topic, key, value))
/**
* Produce a single [[ProducerRecord]], see [[KafkaProducer.produce]] for general semantics.
*/
def produceOne(record: ProducerRecord[K, V]): F[F[ProducerResult[K, V]]] =
producer.produce(ProducerRecords.one(record))
}
/**
* [[KafkaProducer.Metrics]] extends [[KafkaProducer]] to provide
* access to the underlying producer metrics.
*/
abstract class Metrics[F[_], K, V] extends KafkaProducer[F, K, V] {
/**
* Returns producer metrics.
*
* @see org.apache.kafka.clients.producer.KafkaProducer#metrics
*/
def metrics: F[Map[MetricName, Metric]]
}
/**
* [[KafkaProducer.PartitionsFor]] extends [[KafkaProducer.Metrics]] to provide
* access to the underlying producer partitions.
*/
abstract class PartitionsFor[F[_], K, V] extends KafkaProducer.Metrics[F, K, V] {
/**
* Returns partition metadata for the given topic.
*
* @see org.apache.kafka.clients.producer.KafkaProducer#partitionsFor
*/
def partitionsFor(topic: String): F[List[PartitionInfo]]
}
/**
* Creates a new [[KafkaProducer]] in the `Resource` context,
* using the specified [[ProducerSettings]]. Note that there
* is another version where `F[_]` is specified explicitly and
* the key and value type can be inferred, which allows you
* to use the following syntax.
*
* {{{
* KafkaProducer.resource[F].using(settings)
* }}}
*/
def resource[F[_], K, V](
settings: ProducerSettings[F, K, V]
)(implicit F: Async[F], mk: MkProducer[F]): Resource[F, KafkaProducer.PartitionsFor[F, K, V]] =
KafkaProducerConnection.resource(settings)(F, mk).flatMap(_.withSerializersFrom(settings))
private[kafka] def from[F[_], K, V](
connection: KafkaProducerConnection[F],
keySerializer: KeySerializer[F, K],
valueSerializer: ValueSerializer[F, V]
): KafkaProducer.PartitionsFor[F, K, V] =
new KafkaProducer.PartitionsFor[F, K, V] {
override def produce(
records: ProducerRecords[K, V]
): F[F[ProducerResult[K, V]]] =
connection.produce(records)(keySerializer, valueSerializer)
override def metrics: F[Map[MetricName, Metric]] =
connection.metrics
override def toString: String =
"KafkaProducer$" + System.identityHashCode(this)
override def partitionsFor(topic: String): F[List[PartitionInfo]] =
connection.partitionsFor(topic)
}
/**
* Creates a new [[KafkaProducer]] in the `Stream` context,
* using the specified [[ProducerSettings]]. Note that there
* is another version where `F[_]` is specified explicitly and
* the key and value type can be inferred, which allows you
* to use the following syntax.
*
* {{{
* KafkaProducer.stream[F].using(settings)
* }}}
*/
def stream[F[_], K, V](
settings: ProducerSettings[F, K, V]
)(implicit F: Async[F], mk: MkProducer[F]): Stream[F, KafkaProducer.PartitionsFor[F, K, V]] =
Stream.resource(KafkaProducer.resource(settings)(F, mk))
private[kafka] def produce[F[_]: Async, K, V](
withProducer: WithProducer[F],
keySerializer: KeySerializer[F, K],
valueSerializer: ValueSerializer[F, V],
records: ProducerRecords[K, V]
): F[F[ProducerResult[K, V]]] =
withProducer { (producer, blocking) =>
records
.traverse(produceRecord(keySerializer, valueSerializer, producer, blocking))
.map(_.sequence)
}
private[kafka] def produceRecord[F[_], K, V](
keySerializer: KeySerializer[F, K],
valueSerializer: ValueSerializer[F, V],
producer: KafkaByteProducer,
blocking: Blocking[F]
)(
implicit F: Async[F]
): ProducerRecord[K, V] => F[F[(ProducerRecord[K, V], RecordMetadata)]] =
record =>
asJavaRecord(keySerializer, valueSerializer, record).flatMap { javaRecord =>
F.delay(Promise[(ProducerRecord[K, V], RecordMetadata)]()).flatMap { promise =>
blocking {
producer.send(
javaRecord, { (metadata, exception) =>
if (exception == null)
promise.success((record, metadata))
else promise.failure(exception)
}
)
}.as(F.fromFuture(F.delay(promise.future)))
}
}
/**
* Creates a [[KafkaProducer]] using the provided settings and
* produces record in batches.
*/
def pipe[F[_], K, V](
settings: ProducerSettings[F, K, V]
)(
implicit F: Async[F],
mk: MkProducer[F]
): Pipe[F, ProducerRecords[K, V], ProducerResult[K, V]] =
records => stream(settings)(F, mk).flatMap(pipe(_).apply(records))
/**
* Produces records in batches using the provided [[KafkaProducer]].
*/
def pipe[F[_]: Concurrent, K, V](
producer: KafkaProducer[F, K, V]
): Pipe[F, ProducerRecords[K, V], ProducerResult[K, V]] =
_.evalMap(producer.produce).parEvalMap(Int.MaxValue)(identity)
private[this] def serializeToBytes[F[_], K, V](
keySerializer: KeySerializer[F, K],
valueSerializer: ValueSerializer[F, V],
record: ProducerRecord[K, V]
)(implicit F: Apply[F]): F[(Array[Byte], Array[Byte])] = {
val keyBytes =
keySerializer.serialize(record.topic, record.headers, record.key)
val valueBytes =
valueSerializer.serialize(record.topic, record.headers, record.value)
keyBytes.product(valueBytes)
}
private[this] def asJavaRecord[F[_], K, V](
keySerializer: KeySerializer[F, K],
valueSerializer: ValueSerializer[F, V],
record: ProducerRecord[K, V]
)(implicit F: Apply[F]): F[KafkaByteProducerRecord] =
serializeToBytes(keySerializer, valueSerializer, record).map {
case (keyBytes, valueBytes) =>
new KafkaByteProducerRecord(
record.topic,
record.partition.fold[java.lang.Integer](null)(identity),
record.timestamp.fold[java.lang.Long](null)(identity),
keyBytes,
valueBytes,
record.headers.asJava
)
}
def apply[F[_]]: ProducerPartiallyApplied[F] =
new ProducerPartiallyApplied
private[kafka] final class ProducerPartiallyApplied[F[_]](val dummy: Boolean = true)
extends AnyVal {
/**
* Alternative version of `resource` where the `F[_]` is
* specified explicitly, and where the key and value type can
* be inferred from the [[ProducerSettings]]. This allows you
* to use the following syntax.
*
* {{{
* KafkaProducer[F].resource(settings)
* }}}
*/
def resource[K, V](settings: ProducerSettings[F, K, V])(
implicit F: Async[F],
mk: MkProducer[F]
): Resource[F, KafkaProducer[F, K, V]] =
KafkaProducer.resource(settings)(F, mk)
/**
* Alternative version of `stream` where the `F[_]` is
* specified explicitly, and where the key and value type can
* be inferred from the [[ProducerSettings]]. This allows you
* to use the following syntax.
*
* {{{
* KafkaProducer[F].stream(settings)
* }}}
*/
def stream[K, V](settings: ProducerSettings[F, K, V])(
implicit F: Async[F],
mk: MkProducer[F]
): Stream[F, KafkaProducer[F, K, V]] =
KafkaProducer.stream(settings)(F, mk)
override def toString: String =
"ProducerPartiallyApplied$" + System.identityHashCode(this)
}
/*
* Prevents the default `MkProducer` instance from being implicitly available
* to code defined in this object, ensuring factory methods require an instance
* to be provided at the call site.
*/
@nowarn("msg=never used")
implicit private def mkAmbig1[F[_]]: MkProducer[F] =
throw new AssertionError("should not be used")
@nowarn("msg=never used")
implicit private def mkAmbig2[F[_]]: MkProducer[F] =
throw new AssertionError("should not be used")
}