Programming language: Scala
License: MIT License
Tags: Misc     Table Of Contents    
Latest version: v3.0.6
Add another 'Misc' Package



Build Release Maven Central Coverage Status

fs2 Streaming utilities for interacting with AWS

Scope of the project

fs2-aws provides an fs2 interface to AWS services

The design goals are the same as fs2:

compositionality, expressiveness, resource safety, and speed


Find the latest release version and add the following dependency:

libraryDependencies +=  "io.laserdisc" %% "fs2-aws" % "VERSION"


The module fs2-aws-s3 provides a purely functional API to operate with the AWS-S3 API. It defines four functions:

trait S3[F[_]] {
  def delete(bucket: BucketName, key: FileKey): F[Unit]
  def uploadFile(bucket: BucketName, key: FileKey): Pipe[F, Byte, ETag]
  def uploadFileMultipart(bucket: BucketName, key: FileKey, partSize: PartSizeMB): Pipe[F, Byte, ETag]
  def readFile(bucket: BucketName, key: FileKey): Stream[F, Byte]
  def readFileMultipart(bucket: BucketName, key: FileKey, partSize: PartSizeMB): Stream[F, Byte]

You can find out more in the scaladocs for each function, but as a rule of thumb for:

  • Small files: use readFile and uploadFile.
  • Big files: use readFileMultipart and uploadFileMultipart.

You can also combine them as you see fit. For example, use uploadFileMultipart and then read it in one shot using readFile.

Getting started with the S3 module

In order to create an instance of S3 we need to first create an S3Client, as well as a cats.effect.Blocker. Here's an example of the former:

def s3StreamResource: Resource[IO, (S3AsyncClientOp[IO], Blocker)] =
  for {
    blocker     <- Blocker[IO]
    credentials = AwsBasicCredentials.create("accesskey", "secretkey")
    port        = 4566
    s3 <- S3Interpreter[IO](blocker).S3AsyncClientOpResource(
  } yield s3 -> blocker

Now we can create our S3[IO] instance:

s3StreamResource.use {
  case (s3Ops, blocker) => S3.create(s3Ops, blocker).flatMap{ s3 =>
    // do stuff with s3 here (or just share it with other functions)

Create it once and share it as an argument, as any other resource.

For more details on how to work with S3 streams follow [link](fs2-aws-examples/src/main/scala/S3Example.scala)

Reading a file from S3

The simple way:

s3.readFile(BucketName("test"), FileKey("foo"))
  .evalMap(line => IO(println(line)))

The streaming way in a multipart fashion (part size is indicated in MBs and must be 5 or higher):

s3.readFileMultipart(BucketName("test"), FileKey("foo"), partSize = 5)
  .evalMap(line => IO(println(line)))

Writing to a file in S3

The simple way:

Stream.emits("test data".getBytes("UTF-8"))
  .through(s3.uploadFile(BucketName("foo"), FileKey("bar"), partSize = 5))
  .evalMap(t => IO(println(s"eTag: $t")))

The streaming way in a multipart fashion. Again, part size is indicated in MBs and must be 5 or higher.

Stream.emits("test data".getBytes("UTF-8"))
  .through(s3.uploadFileMultipart(BucketName("foo"), FileKey("bar"), partSize = 5))
  .evalMap(t => IO(println(s"eTag: $t")))

Deleting a file in S3

There is a simple function to delete a file.

s3.delete(BucketName("foo"), FileKey("bar"))


Streaming records from Kinesis with KCL

Example using IO for effects (any monad F <: ConcurrentEffect can be used):

val stream: Stream[IO, CommittableRecord] = readFromKinesisStream[IO]("appName", "streamName")

There are a number of other stream constructors available where you can provide more specific configuration for the KCL worker.


TODO: Implement better test consumer

For now, you can stub CommitableRecord and create a fs2.Stream to emit these records:

val record = new Record()
  .withApproximateArrivalTimestamp(new Date())

val testRecord = CommittableRecord(

Checkpointing records

Records must be checkpointed in Kinesis to keep track of which messages each consumer has received. Checkpointing a record in the KCL will automatically checkpoint all records upto that record. To checkpoint records, a Pipe and Sink are available. To help distinguish whether a record has been checkpointed or not, a CommittableRecord class exists to denote a record that hasn't been checkpointed, while the base Record class denotes a committed record.

readFromKinesisStream[IO]("appName", "streamName")

Publishing records to Kinesis with KPL

A Pipe and Sink allow for writing a stream of tuple2 (partitionKey, ByteBuffer) to a Kinesis stream.


  .map { d => ("partitionKey", ByteBuffer.wrap(d.getBytes))}

AWS credential chain and region can be configured by overriding the respective fields in the KinesisProducerClient parameter to writeToKinesis. Defaults to using the default AWS credentials chain and us-east-1 for region.

Use with LocalStack

In some situations (e.g. local dev and automated tests), it may be desirable to be able to consume from and publish to Kinesis running inside LocalStack. Ensure that the endpoint setting is set correctly (e.g. http://localhost:4566) and that retrievalMode is set to Polling (LocalStack doesn't support FanOut).

Kinesis Firehose

TODO: Stream get data, Stream send data



implicit val messageDecoder: Message => Either[Throwable, Quote] = { sqs_msg =>
      .sqsStream[IO, Quote](
        (config, callback) => SQSConsumerBuilder(config, callback))


//create stream for testing
def stream(deferredListener: Deferred[IO, MessageListener]) =
              .sqsStream[IO, Quote](deferredListener)

//create the program for testing the stream
import io.circe.fs2.aws.examples.syntax._
import io.circe.generic.auto._
val quote = Quote(...)
val program : IO[List[(Quote, MessageListener)]] = for {
            d <- Deferred[IO, MessageListener]
            r <- IO.racePair(stream(d), d.get).flatMap {
              case Right((streamFiber, listener)) =>
                //simulate SQS stream fan-in here
                listener.onMessage(new SQSTextMessage(Printer.noSpaces.pretty(quote.asJson)))
              case _ => IO(Nil)
          } yield r

//Assert results
val result = program
result should be(...)

TODO: Stream send SQS messages


YourKit Image

This project is supported by YourKit with monitoring and profiling Tools. YourKit supports open source with innovative and intelligent tools for monitoring and profiling Java and .NET applications. YourKit is the creator of YourKit Java Profiler, YourKit .NET Profiler, and YourKit YouMonitor.