Workflows

Introduction

Workflows are the basic building blocks in Temporal.
A Workflow Definition contains the actual business logic. It's determenistic, those free from any side effects.
Refer to Temporal documentation for more information regarding the concept of workflows.

Defining workflows

Let's start with some basic imports that will be required for the whole demonstration:

import zio._
import zio.temporal._
import zio.temporal.worker._
import zio.temporal.workflow._

import java.util.UUID

Every Temporal application starts from Workflow Definitions.
A simple workflow consists of the following parts:

• A Scala trait with an abstract method
• The trait itself should be annotated with @workflowInterface
• The method should be annotated with @workflowMethod

@workflowInterface
trait EchoWorkflow {

  @workflowMethod
  def echo(str: String): String
}

Implementing simple workflows is as simple as just plain interfaces:

class EchoWorkflowImpl extends EchoWorkflow {
  override def echo(str: String): String = {
    println(s"Echo: $str")
    str
  }
}

That's it!

Executing workflows

First, you should connect to the Temporal cluster. This is done via the Workflow client.
Any workflow run requires providing some mandatory parameters, such as the unique Workflow ID, the Task Queue, and optionally others (such as timeouts).
The configuration is passed using ZWorkflowOptions:

val workflowOptions = ZWorkflowOptions
  .withWorkflowId(UUID.randomUUID().toString)
  .withTaskQueue("echo-queue")
  .withWorkflowRunTimeout(10.second)
// workflowOptions: ZWorkflowOptions = ZWorkflowOptions(
//   workflowId = "41a26949-bef0-4a94-981f-fea6a6204413",
//   taskQueue = "echo-queue",
//   workflowIdReusePolicy = None,
//   workflowRunTimeout = Some(value = PT10S),
//   workflowExecutionTimeout = None,
//   workflowTaskTimeout = None,
//   retryOptions = None,
//   memo = Map(),
//   searchAttributes = None,
//   contextPropagators = List(),
//   disableEagerExecution = None,
//   javaOptionsCustomization = zio.temporal.workflow.ZWorkflowOptions$SetTaskQueue$$$Lambda$8998/0x00007fea09a3e028@5b1b69a1
// )

In order to run a specific workflow, you should create a Workflow stub:

def createWorkflowStub(workflowClient: ZWorkflowClient): UIO[ZWorkflowStub.Of[EchoWorkflow]] = 
  workflowClient.newWorkflowStub[EchoWorkflow](workflowOptions)

Important notes:

• You need a ZWorkflowClient instance to build a Workflow stub. You should either access it via ZIO environment or have it already created somewhere
• In client.newWorkflowStub method you provide the specific workflow interface you will work with
  • The method requires specifying the Workflow Interface type and ZWorkflowOptions

Having a Workflow stub, you'll be able to execute the workflow. Executing means starting the workflow and waiting for its completion:

val workflowResultZIO = 
  for {
    echoWorkflow <- ZIO.serviceWithZIO[ZWorkflowClient](createWorkflowStub)
    result       <- ZWorkflowStub.execute(echoWorkflow.echo("Hello there"))
  } yield result

Important notes:

• You must always wrap the workflow method invocation into ZWorkflowStub.execute method.
  • The ZWorkflowStub.Of[EchoWorkflow] is a compile-time stub, so actual method invocations are only valid in compile-time
  • echoWorkflow.echo("Hello there") invocation would be re-written into an untyped Temporal's workflow invocation (see workflow Execution doc)
  • A direct method invocation will throw an exception
• The ZWorkflowStub is basically a proxy, which executes the Workflow via Temporal cluster

NOTE: Do not annotate workflow stubs with the workflow interface type. It must be ZWorkflowStub.Of[EchoWorkflow].
Otherwise, you'll get a compile-time error:

def doSomething(echoWorkflow: EchoWorkflow): TemporalIO[String] =
  ZWorkflowStub.execute(echoWorkflow.echo("Hello there"))
// error: zio.temporal.workflow.ZWorkflowStub.execute must be used only with typed zio.temporal.workflow.ZWorkflowStub.Of[A],
// but repl.MdocSession.MdocApp.EchoWorkflow found. Perhaps you added an explicit type annotation?
// The actual type must be zio.temporal.workflow.ZWorkflowStub.Of[repl.MdocSession.MdocApp.EchoWorkflow]
//   ZWorkflowStub.execute(echoWorkflow.echo("Hello there"))
//   ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^

Workers

In this section, we'll be referring to Worker Entity as a Worker.
The Worker runs a workflow which someone executed.

We assume that a Worker Process consists only of one Worker Entity listening to a single Task Queue.
Refer to Temporal documentation for more information regarding workers and task queues.

A Worker itself depends on a Workflow client which communicates with the Temporal cluster.
Using the Workflow client, it's now possible to create a Worker factory, which itself creates the Worker.

val worker: URLayer[ZWorkerFactory, Unit] = ZLayer.fromZIO {
  ZIO.serviceWithZIO[ZWorkerFactory] { workerFactory =>
    for {
      worker <- workerFactory.newWorker("echo-queue")
      _ <- worker
              .addWorkflow[EchoWorkflow]
              .from(new EchoWorkflowImpl)
    } yield ()
  }
}

There is also an alternative syntax for building workflows which relies on ZIO Aspects:

val workerAspects: URLayer[ZWorkerFactory, Unit] = ZLayer.fromZIO {
  ZWorkerFactory.newWorker("echo-queue") @@
    ZWorker.addWorkflow[EchoWorkflow].from(new EchoWorkflowImpl)
}.unit

This syntax allows avoiding syntactic noise of monadic composition and accessing ZIO's environment.
Therefore, it's a preferred one.

Important notes:

• You need a ZWorkerFactory instance to create a worker. You should either access it via ZIO environment or have it already created somewhere
• You must specify the task queue when creating a worker (echo-queue in this case).
• Make sure that worker uses the same task queue as the client code
• In worker.addWorkflow method you provide the specific workflow interface you run in this worker
• It's required to provide the workflow implementation instance there.
• Create the workflow implementation only inside the addWorkflow method, otherwise an exception will be thrown.

Running the Worker and the Workflow client

Let's bring all the parts into a program:

val program = 
  for {
    _              <- ZWorkerFactory.setup
    workflowResult <- workflowResultZIO
    _              <- ZIO.log(s"The workflow result: $workflowResult")
  } yield ()

When injecting, don't forget to provide the worker:

program.provide(
  worker,
  // Other dependencies
)