Lifts the given thunk in the IO context, processing it synchronously when the task gets evaluated.

This is an alias for:

val thunk = () => 42
IO.eval(thunk())

WARN: behavior of IO.apply has changed since 3.0.0-RC2. Before the change (during Monix 2.x series), this operation was forcing a fork, being equivalent to the new IO.evalAsync.

Switch to IO.evalAsync if you wish the old behavior, or combine IO.eval with IO.executeAsync.

Create a non-cancelable Task from an asynchronous computation, which takes the form of a function with which we can register a callback to execute upon completion.

This operation is the implementation for cats.effect.Async and is thus yielding non-cancelable tasks, being the simplified version of IO.cancelable. This can be used to translate from a callback-based API to pure Task values that cannot be canceled.

See the the documentation for cats.effect.Async.

For example, in case we wouldn't have IO.deferFuture already defined, we could do this:

import scala.concurrent.{Future, ExecutionContext}
import scala.util._

def deferFuture[A](f: => Future[A])(implicit ec: ExecutionContext): Task[A] =
  Task.async { cb =>
    // N.B. we could do `f.onComplete(cb)` directly ;-)
    f.onComplete {
      case Success(a) => cb.onSuccess(a)
      case Failure(e) => cb.onError(e)
    }
  }

Note that this function needs an explicit ExecutionContext in order to trigger Future#complete, however Monix's Task can inject a Scheduler for you, thus allowing you to get rid of these pesky execution contexts being passed around explicitly. See IO.async0.

CONTRACT for register:

• the provided function is executed when the Task will be evaluated (via runAsync or when its turn comes in the flatMap chain, not before)
• the injected BiCallback can be called at most once, either with a successful result, or with an error; calling it more than once is a contract violation
• the injected callback is thread-safe and in case it gets called multiple times it will throw a monix.execution.exceptions.CallbackCalledMultipleTimesException; also see Callback.tryOnSuccess and Callback.tryOnError and monix.bio.BiCallback.tryOnTermination

Create a non-cancelable Task from an asynchronous computation, which takes the form of a function with which we can register a callback to execute upon completion, a function that also injects a Scheduler for managing async boundaries.

This operation is the implementation for cats.effect.Async and is thus yielding non-cancelable tasks, being the simplified version of IO.cancelable0. It can be used to translate from a callback-based API to pure Task values that cannot be canceled.

See the the documentation for cats.effect.Async.

For example, in case we wouldn't have IO.deferFuture already defined, we could do this:

import scala.concurrent.Future
import scala.util._

def deferFuture[A](f: => Future[A]): Task[A] =
  Task.async0 { (scheduler, cb) =>
    // We are being given an ExecutionContext ;-)
    implicit val ec = scheduler

    // N.B. we could do `f.onComplete(cb)` directly ;-)
    f.onComplete {
      case Success(a) => cb.onSuccess(a)
      case Failure(e) => cb.onError(e)
    }
  }

Note that this function doesn't need an implicit ExecutionContext. Compared with usage of IO.async, this function injects a Scheduler for us to use for managing async boundaries.

CONTRACT for register:

• the provided function is executed when the Task will be evaluated (via runAsync or when its turn comes in the flatMap chain, not before)
• the injected monix.bio.BiCallback can be called at most once, either with a successful result, or with an error; calling it more than once is a contract violation
• the injected callback is thread-safe and in case it gets called multiple times it will throw a monix.execution.exceptions.CallbackCalledMultipleTimesException; also see Callback.tryOnSuccess and Callback.tryOnError and BiCallback.tryOnTermination

NOTES on the naming:

• async comes from cats.effect.Async#async
• the 0 suffix is about overloading the simpler IO.async builder

Suspends an asynchronous side effect in IO, this being a variant of async that takes a pure registration function.

Implements cats.effect.Async.asyncF.

The difference versus async is that this variant can suspend side-effects via the provided function parameter. It's more relevant in polymorphic code making use of the cats.effect.Async type class, as it alleviates the need for cats.effect.Effect.

Contract for the returned IO[E, Unit] in the provided function:

• can be asynchronous
• can be cancelable, in which case it hooks into IO's cancelation mechanism such that the resulting task is cancelable
• it should not end in error, because the provided callback is the only way to signal the final result and it can only be called once, so invoking it twice would be a contract violation; so on errors thrown in IO, the task can become non-terminating, with the error being printed via Scheduler.reportFailure

Returns a cancelable boundary — a Task that checks for the cancellation status of the run-loop and does not allow for the bind continuation to keep executing in case cancellation happened.

This operation is very similar to Task.shift, as it can be dropped in flatMap chains in order to make loops cancelable.

Example:

import cats.syntax.all._

def fib(n: Int, a: Long, b: Long): Task[Long] =
  Task.suspend {
    if (n <= 0) Task.pure(a) else {
      val next = fib(n - 1, b, a + b)

      // Every 100-th cycle, check cancellation status
      if (n % 100 == 0)
        Task.cancelBoundary *> next
      else
        next
    }
  }

NOTE: that by default Task is configured to be auto-cancelable (see IO.Options), so this isn't strictly needed, unless you want to fine tune the cancelation boundaries.

Create a cancelable Task from an asynchronous computation that can be canceled, taking the form of a function with which we can register a callback to execute upon completion.

This operation is the implementation for cats.effect.Concurrent#cancelable and is thus yielding cancelable tasks. It can be used to translate from a callback-based API to pure Task values that can be canceled.

See the the documentation for cats.effect.Concurrent.

For example, in case we wouldn't have IO.delayExecution already defined and we wanted to delay evaluation using a Java ScheduledExecutorService (no need for that because we've got Scheduler, but lets say for didactic purposes):

import java.util.concurrent.ScheduledExecutorService
import scala.concurrent.ExecutionContext
import scala.concurrent.duration._
import scala.util.control.NonFatal

def delayed[A](sc: ScheduledExecutorService, timespan: FiniteDuration)
  (thunk: => A)
  (implicit ec: ExecutionContext): Task[A] = {

  Task.cancelable { cb =>
    val future = sc.schedule(new Runnable { // scheduling delay
      def run() = ec.execute(new Runnable { // scheduling thunk execution
        def run() =
          try
            cb.onSuccess(thunk)
          catch { case NonFatal(e) =>
            cb.onError(e)
          }
        })
      },
      timespan.length,
      timespan.unit)

    // Returning the cancelation token that is able to cancel the
    // scheduling in case the active computation hasn't finished yet
    Task { future.cancel(false); () }
  }
}

Note in this sample we are passing an implicit ExecutionContext in order to do the actual processing, the ScheduledExecutorService being in charge just of scheduling. We don't need to do that, as Task affords to have a Scheduler injected instead via IO.cancelable0.

CONTRACT for register:

• the provided function is executed when the Task will be evaluated (via runAsync or when its turn comes in the flatMap chain, not before)
• the injected BiCallback can be called at most once, either with a successful result, or with an typed; calling it more than once is a contract violation
• the injected callback is thread-safe and in case it gets called multiple times it will throw a monix.execution.exceptions.CallbackCalledMultipleTimesException; also see Callback.tryOnSuccess and Callback.tryOnError and BiCallback.tryOnTermination

Create a cancelable Task from an asynchronous computation, which takes the form of a function with which we can register a callback to execute upon completion, a function that also injects a Scheduler for managing async boundaries.

This operation is the implementation for cats.effect.Concurrent#cancelable and is thus yielding cancelable tasks. It can be used to translate from a callback-based API to pure Task values that can be canceled.

See the the documentation for cats.effect.Concurrent.

For example, in case we wouldn't have IO.delayExecution already defined and we wanted to delay evaluation using a Java ScheduledExecutorService (no need for that because we've got Scheduler, but lets say for didactic purposes):

import java.util.concurrent.ScheduledExecutorService
import scala.concurrent.duration._
import scala.util.control.NonFatal

def delayed1[A](sc: ScheduledExecutorService, timespan: FiniteDuration)
  (thunk: => A): Task[A] = {

  Task.cancelable0 { (scheduler, cb) =>
    val future = sc.schedule(new Runnable { // scheduling delay
      def run = scheduler.execute(new Runnable { // scheduling thunk execution
        def run() =
          try
            cb.onSuccess(thunk)
          catch { case NonFatal(e) =>
            cb.onError(e)
          }
        })
      },
      timespan.length,
      timespan.unit)

    // Returning the cancel token that is able to cancel the
    // scheduling in case the active computation hasn't finished yet
    Task { future.cancel(false); () }
  }
}

As can be seen, the passed function needs to pass a Cancelable in order to specify cancelation logic.

This is a sample given for didactic purposes. Our cancelable0 is being injected a Scheduler and it is perfectly capable of doing such delayed execution without help from Java's standard library:

def delayed2[A](timespan: FiniteDuration)(thunk: => A): Task[A] =
  Task.cancelable0 { (scheduler, cb) =>
    // N.B. this already returns the Cancelable that we need!
    val cancelable = scheduler.scheduleOnce(timespan) {
      try cb.onSuccess(thunk)
      catch { case NonFatal(e) => cb.onError(e) }
    }
    // `scheduleOnce` above returns a Cancelable, which
    // has to be converted into a Task[Unit]
    Task(cancelable.cancel())
  }

CONTRACT for register:

• the provided function is executed when the Task will be evaluated (via runAsync or when its turn comes in the flatMap chain, not before)
• the injected BiCallback can be called at most once, either with a successful result, or with an error; calling it more than once is a contract violation
• the injected callback is thread-safe and in case it gets called multiple times it will throw a monix.execution.exceptions.CallbackCalledMultipleTimesException; also see Callback.tryOnSuccess and Callback.tryOnError and BiCallback.tryOnTermination

NOTES on the naming:

• cancelable comes from cats.effect.Concurrent#cancelable
• the 0 suffix is about overloading the simpler IO.cancelable builder

Global instance for cats.effect.Async and for cats.effect.Concurrent.

Implied are also cats.CoflatMap, cats.Applicative, cats.Monad, cats.MonadError and cats.effect.Sync.

As trivia, it's named "catsAsync" and not "catsConcurrent" because it represents the cats.effect.Async lineage, up until cats.effect.Effect, which imposes extra restrictions, in our case the need for a Scheduler to be in scope (see IO.catsEffect). So by naming the lineage, not the concrete sub-type implemented, we avoid breaking compatibility whenever a new type class (that we can implement) gets added into Cats.

Seek more info about Cats, the standard library for FP, at:

• typelevel/cats
• typelevel/cats-effect

Global instance for cats.effect.Effect and for cats.effect.ConcurrentEffect.

Implied are cats.CoflatMap, cats.Applicative, cats.Monad, cats.MonadError, cats.effect.Sync and cats.effect.Async.

Note this is different from IO.catsAsync because we need an implicit Scheduler in scope in order to trigger the execution of a Task. It's also lower priority in order to not trigger conflicts, because Effect <: Async and ConcurrentEffect <: Concurrent with Effect.

As trivia, it's named "catsEffect" and not "catsConcurrentEffect" because it represents the cats.effect.Effect lineage, as in the minimum that this value will support in the future. So by naming the lineage, not the concrete sub-type implemented, we avoid breaking compatibility whenever a new type class (that we can implement) gets added into Cats.

Seek more info about Cats, the standard library for FP, at:

• typelevel/cats
• typelevel/cats-effect

Given an A type that has a cats.Monoid[A] implementation, then this provides the evidence that IO[E, A] also has a Monoid[ IO[E, A] ] implementation.

Global instance for cats.Parallel.

The Parallel type class is useful for processing things in parallel in a generic way, usable with Cats' utils and syntax:

import cats.syntax.all._
import scala.concurrent.duration._

val taskA = Task.sleep(1.seconds).map(_ => "a")
val taskB = Task.sleep(2.seconds).map(_ => "b")
val taskC = Task.sleep(3.seconds).map(_ => "c")

// Returns "abc" after 3 seconds
(taskA, taskB, taskC).parMapN { (a, b, c) =>
  a + b + c
}

Seek more info about Cats, the standard library for FP, at:

• typelevel/cats
• typelevel/cats-effect

Given an A type that has a cats.Semigroup[A] implementation, then this provides the evidence that IO[E, A] also has a Semigroup[ IO[E, A] ] implementation.

This has a lower-level priority than IO.catsMonoid in order to avoid conflicts.

Builds a cats.effect.Clock instance, given a Scheduler reference.

Default, pure, globally visible cats.effect.Clock implementation that defers the evaluation to Task's default Scheduler (that's being injected in IO.runToFuture).

Global instance for cats.CommutativeApplicative

Builds a cats.effect.ContextShift instance, given a Scheduler reference.

Default, pure, globally visible cats.effect.ContextShift implementation that shifts the evaluation to Task's default Scheduler (that's being injected in IO.runToFuture).

Polymorphic Task builder that is able to describe asynchronous tasks depending on the type of the given callback.

Note that this function uses the Partially-Applied Type technique.

Calling create with a callback that returns Unit is equivalent with IO.async0:

Task.async0(f) <-> Task.create(f)

Example:

import scala.concurrent.Future

def deferFuture[A](f: => Future[A]): Task[A] =
  Task.create { (scheduler, cb) =>
    f.onComplete(cb(_))(scheduler)
  }

We could return a Cancelable reference and thus make a cancelable task. Example:

import monix.execution.Cancelable
import scala.concurrent.duration.FiniteDuration
import scala.util.Try

def delayResult1[A](timespan: FiniteDuration)(thunk: => A): Task[A] =
  Task.create { (scheduler, cb) =>
    val c = scheduler.scheduleOnce(timespan)(cb(Try(thunk)))
    // We can simply return `c`, but doing this for didactic purposes!
    Cancelable(() => c.cancel())
  }

Passed function can also return cats.effect.IO[Unit] as a task that describes a cancelation action:

import cats.effect.{IO => CIO}

def delayResult2[A](timespan: FiniteDuration)(thunk: => A): Task[A] =
  Task.create { (scheduler, cb) =>
    val c = scheduler.scheduleOnce(timespan)(cb(Try(thunk)))
    // We can simply return `c`, but doing this for didactic purposes!
    CIO(c.cancel())
  }

Passed function can also return Task[Unit] as a task that describes a cancelation action, thus for an f that can be passed to IO.cancelable0, and this equivalence holds:

Task.cancelable(f) <-> Task.create(f)

def delayResult3[A](timespan: FiniteDuration)(thunk: => A): Task[A] =
  Task.create { (scheduler, cb) =>
    val c = scheduler.scheduleOnce(timespan)(cb(Try(thunk)))
    // We can simply return `c`, but doing this for didactic purposes!
    Task(c.cancel())
  }

The supported types for the cancelation tokens are:

• Unit, yielding non-cancelable tasks
• Cancelable, the Monix standard
• Task[Unit]
• cats.effect.IO[Unit], see IO docs

Support for more might be added in the future.

Default Options to use for IO evaluation, thus:

• autoCancelableRunLoops is true by default
• localContextPropagation is false by default

On top of the JVM the default can be overridden by setting the following system properties:

• monix.environment.autoCancelableRunLoops (false, no or 0 for disabling)
• monix.environment.localContextPropagation (true, yes or 1 for enabling)

Defers the creation of a Task in case it is effectful.

It will catch any exceptions thrown in fa and expose them as a typed error.

Defers the creation of a Task by using the provided function, which has the ability to inject a needed Scheduler.

Example:

import scala.concurrent.duration.MILLISECONDS

def measureLatency[A](source: Task[A]): Task[(A, Long)] =
  Task.deferAction { implicit s =>
    // We have our Scheduler, which can inject time, we
    // can use it for side-effectful operations
    val start = s.clockRealTime(MILLISECONDS)

    source.map { a =>
      val finish = s.clockRealTime(MILLISECONDS)
      (a, finish - start)
    }
  }

Promote a non-strict Scala Future to a Task of the same type.

The equivalent of doing:

import scala.concurrent.Future
def mkFuture = Future.successful(27)

Task.defer(Task.fromFuture(mkFuture))

Wraps calls that generate Future results into Task, provided a callback with an injected Scheduler to act as the necessary ExecutionContext.

This builder helps with wrapping Future-enabled APIs that need an implicit ExecutionContext to work. Consider this example:

import scala.concurrent.{ExecutionContext, Future}

def sumFuture(list: Seq[Int])(implicit ec: ExecutionContext): Future[Int] =
  Future(list.sum)

We'd like to wrap this function into one that returns a lazy Task that evaluates this sum every time it is called, because that's how tasks work best. However in order to invoke this function an ExecutionContext is needed:

def sumTask(list: Seq[Int])(implicit ec: ExecutionContext): Task[Int] =
  Task.deferFuture(sumFuture(list))

But this is not only superfluous, but against the best practices of using Task. The difference is that Task takes a Scheduler (inheriting from ExecutionContext) only when runAsync happens. But with deferFutureAction we get to have an injected Scheduler in the passed callback:

def sumTask2(list: Seq[Int]): Task[Int] =
  Task.deferFutureAction { implicit scheduler =>
    sumFuture(list)
  }

Defers the creation of a IO in case it is effectful.

Promote a non-strict value, a thunk, to a IO, catching exceptions in the process.

Note that since IO is not memoized or strict, this will recompute the value each time the IO is executed, behaving like a function.

Lifts a non-strict value, a thunk, to a Task that will trigger a logical fork before evaluation.

Like eval, but the provided thunk will not be evaluated immediately. Equivalence:

Task.evalAsync(a) <-> Task.eval(a).executeAsync

Promote a non-strict value which does not throw any unexpected errors to UIO.

Note that since IO is not memoized or strict, this will recompute the value each time the IO is executed, behaving like a function.

Converts into a Task from any F[_] for which there exists a IOLike implementation.

Supported types include, but are not necessarily limited to:

• cats.Eval
• cats.effect.IO
• cats.effect.SyncIO
• cats.effect.Effect (Async)
• cats.effect.ConcurrentEffect
• scala.Either
• scala.util.Try
• scala.concurrent.Future

Builds a Task out of any data type that implements Concurrent and ConcurrentEffect.

Example:

import cats.effect.{IO => CIO, _}
import cats.syntax.all._
import monix.execution.Scheduler.Implicits.global
import scala.concurrent.duration._

implicit val timer = CIO.timer(global)
implicit val cs = CIO.contextShift(global)

val cio = CIO.sleep(5.seconds) *> CIO(println("Hello!"))

// Resulting task is cancelable
val task: Task[Unit] = IO.fromConcurrentEffect(cio)

Cancellation / finalization behavior is carried over, so the resulting task can be safely cancelled.

Builds a IO out of any data type that implements Async and Effect.

Example:

import cats.effect.{IO => CIO}

val cio: CIO[Unit] = CIO(println("Hello!"))
val task: Task[Unit] = IO.fromEffect(cio)

WARNING: the resulting task might not carry the source's cancellation behavior if the source is cancelable! This is implicit in the usage of Effect.

Converts the given Scala Future into a Task.

There is an async boundary inserted at the end to guarantee that we stay on the main Scheduler.

NOTE: if you want to defer the creation of the future, use in combination with defer.

Builds a IO instance out of a Scala Option. If the Option is empty, the task fails with the provided fallback.

Builds a IO instance out of a Scala Option. If the Option is empty, the task fails with Unit.

Example:

IO.fromOption(Some(1)) // <-> IO.now(1))
IO.fromOption(None)    // <-> IO.raiseError(())

Builds a new IO instance out of a IO[E, Option[A]]. If the inner Option is empty, the task fails with the provided fallback.

Example:

type ErrorCode = Int
final case class Item()

def findItem(id: Int): IO[ErrorCode, Option[Item]] =
  UIO.now(Some(Item()))

IO.fromOptionEval(findItem(1), 404)

Converts an org.reactivestreams.Publisher into a IO.

See reactive-streams.org for the Reactive Streams specification.

Returns a F ~> Task (FunctionK) for transforming any supported data-type into Task.

Useful for mapK transformations, for example when working with Resource or Iterant:

import cats.effect.{IO => CIO, _}
import monix.bio._
import java.io._

def open(file: File) =
  Resource[CIO, InputStream](CIO {
    val in = new FileInputStream(file)
    (in, CIO(in.close()))
  })

// Lifting to a Resource of Task
val res: Resource[Task, InputStream] =
  open(new File("sample")).mapK(Task.liftFrom[CIO])

Returns a F ~> Task (FunctionK) for transforming any supported data-type, that implements cats.effect.ConcurrentEffect, into Task.

Useful for mapK transformations, for example when working with Resource or Iterant.

This is the less generic liftFrom operation, supplied in order order to force the usage of ConcurrentEffect for where it matters.

Returns a F ~> Task (FunctionK) for transforming any supported data-type, that implements Effect, into Task.

Useful for mapK transformations, for example when working with Resource or Iterant.

This is the less generic liftFrom operation, supplied in order order to force the usage of Effect for where it matters.

Generates cats.FunctionK values for converting from Task to supporting types (for which we have a IOLift instance).

See https://typelevel.org/cats/datatypes/functionk.html.

import cats.effect.{IO => CIO, _}
import monix.bio._
import java.io._

// Needed for converting from Task to something else, because we need
// ConcurrentEffect[Task] capabilities, also provided by [[BIOApp]]
import monix.execution.Scheduler.Implicits.global

def open(file: File) =
  Resource[Task, InputStream](Task {
    val in = new FileInputStream(file)
    (in, Task(in.close()))
  })

// Lifting to a Resource of cats.effect.IO
val res: Resource[CIO, InputStream] =
  open(new File("sample")).mapK(Task.liftTo[CIO])

// This was needed in order to process the resource
// with a Task, instead of a Coeval
res.use { in =>
  CIO {
    in.read()
  }
}

Generates cats.FunctionK values for converting from Task to supporting types (for which we have a cats.effect.Async) instance.

See https://typelevel.org/cats/datatypes/functionk.html.

Prefer to use liftTo, this alternative is provided in order to force the usage of cats.effect.Async, since IOLift is lawless.

Generates cats.FunctionK values for converting from Task to supporting types (for which we have a cats.effect.Concurrent) instance.

See https://typelevel.org/cats/datatypes/functionk.html.

Prefer to use liftTo, this alternative is provided in order to force the usage of cats.effect.Concurrent, since IOLift is lawless.

Pairs 2 IO values, applying the given mapping function.

Returns a new IO reference that completes with the result of mapping that function to their successful results, or in failure in case either of them fails.

This is a specialized IO.sequence operation and as such the tasks are evaluated in order, one after another, the operation being described in terms of .flatMap.

val fa1 = IO(1)
val fa2 = IO(2)

// Yields Success(3)
IO.map2(fa1, fa2) { (a, b) =>
  a + b
}

// Yields Failure(e), because the second arg is a failure
IO.map2(fa1, IO.raiseError(new RuntimeException("boo"))) { (a, b: Int) =>
  a + b
}

Pairs 3 IO values, applying the given mapping function.

Returns a new IO reference that completes with the result of mapping that function to their successful results, or in failure in case either of them fails.

This is a specialized IO.sequence operation and as such the tasks are evaluated in order, one after another, the operation being described in terms of .flatMap.

val fa1 = IO(1)
val fa2 = IO(2)
val fa3 = IO(3)

// Yields Success(6)
IO.map3(fa1, fa2, fa3) { (a, b, c) =>
  a + b + c
}

// Yields Failure(e), because the second arg is a failure
IO.map3(fa1, IO.raiseError(new RuntimeException("boo")), fa3) { (a, b: Int, c) =>
  a + b + c
}

Pairs 4 IO values, applying the given mapping function.

Returns a new IO reference that completes with the result of mapping that function to their successful results, or in failure in case either of them fails.

This is a specialized IO.sequence operation and as such the tasks are evaluated in order, one after another, the operation being described in terms of .flatMap.

val fa1 = IO(1)
val fa2 = IO(2)
val fa3 = IO(3)
val fa4 = IO(4)

// Yields Success(10)
IO.map4(fa1, fa2, fa3, fa4) { (a, b, c, d) =>
  a + b + c + d
}

// Yields Failure(e), because the second arg is a failure
IO.map4(fa1, IO.raiseError(new RuntimeException("boo")), fa3, fa4) {
  (a, b: Int, c, d) => a + b + c + d
}

Pairs 5 IO values, applying the given mapping function.

Returns a new IO reference that completes with the result of mapping that function to their successful results, or in failure in case either of them fails.

This is a specialized IO.sequence operation and as such the tasks are evaluated in order, one after another, the operation being described in terms of .flatMap.

val fa1 = IO(1)
val fa2 = IO(2)
val fa3 = IO(3)
val fa4 = IO(4)
val fa5 = IO(5)

// Yields Success(15)
IO.map5(fa1, fa2, fa3, fa4, fa5) { (a, b, c, d, e) =>
  a + b + c + d + e
}

// Yields Failure(e), because the second arg is a failure
IO.map5(fa1, IO.raiseError(new RuntimeException("boo")), fa3, fa4, fa5) {
  (a, b: Int, c, d, e) => a + b + c + d + e
}

Pairs 6 IO values, applying the given mapping function.

Returns a new IO reference that completes with the result of mapping that function to their successful results, or in failure in case either of them fails.

This is a specialized IO.sequence operation and as such the tasks are evaluated in order, one after another, the operation being described in terms of .flatMap.

val fa1 = IO(1)
val fa2 = IO(2)
val fa3 = IO(3)
val fa4 = IO(4)
val fa5 = IO(5)
val fa6 = IO(6)

// Yields Success(21)
IO.map6(fa1, fa2, fa3, fa4, fa5, fa6) { (a, b, c, d, e, f) =>
  a + b + c + d + e + f
}

// Yields Failure(e), because the second arg is a failure
IO.map6(fa1, IO.raiseError(new RuntimeException("boo")), fa3, fa4, fa5, fa6) {
  (a, b: Int, c, d, e, f) => a + b + c + d + e + f
}

Yields a task that on evaluation will process the given tasks in parallel, then apply the given mapping function on their results.

Example:

val task1 = Task(1 + 1)
val task2 = Task(2 + 2)

// Yields 6
Task.mapBoth(task1, task2)((a, b) => a + b)

ADVICE: In a real life scenario the tasks should be expensive in order to warrant parallel execution. Parallelism doesn't magically speed up the code - it's usually fine for I/O-bound tasks, however for CPU-bound tasks it can make things worse. Performance improvements need to be verified.

NOTE: the tasks get forked automatically so there's no need to force asynchronous execution for immediate tasks, parallelism being guaranteed when multi-threading is available!

All specified tasks get evaluated in parallel, regardless of their execution model (IO.eval vs IO.evalAsync doesn't matter). Also the implementation tries to be smart about detecting forked tasks so it can eliminate extraneous forks for the very obvious cases.

Pairs 2 IO values, applying the given mapping function, ordering the results, but not the side effects, the evaluation being done in parallel.

This is a specialized IO.parSequence operation and as such the tasks are evaluated in parallel, ordering the results. In case one of the tasks fails, then all other tasks get cancelled and the final result will be a failure.

val fa1 = UIO(1)
val fa2 = UIO(2)

// Yields Success(3)
IO.parMap2(fa1, fa2) { (a, b) =>
  a + b
}

val ex: Task[Int] = IO.raiseError(new RuntimeException("boo"))

// Yields Failure(e), because the second arg is a failure
IO.parMap2(fa1, ex) { (a, b) =>
  a + b
}

ADVICE: In a real life scenario the tasks should be expensive in order to warrant parallel execution. Parallelism doesn't magically speed up the code - it's usually fine for I/O-bound tasks, however for CPU-bound tasks it can make things worse. Performance improvements need to be verified.

NOTE: the tasks get forked automatically so there's no need to force asynchronous execution for immediate tasks, parallelism being guaranteed when multi-threading is available!

All specified tasks get evaluated in parallel, regardless of their execution model (IO.eval vs IO.evalAsync doesn't matter). Also the implementation tries to be smart about detecting forked tasks so it can eliminate extraneous forks for the very obvious cases.

See IO.map2 for sequential processing.

Pairs 3 IO values, applying the given mapping function, ordering the results, but not the side effects, the evaluation being done in parallel.

This is a specialized IO.parSequence operation and as such the tasks are evaluated in parallel, ordering the results. In case one of the tasks fails, then all other tasks get cancelled and the final result will be a failure.

val fa1 = UIO(1)
val fa2 = UIO(2)
val fa3 = UIO(3)

// Yields Success(6)
IO.parMap3(fa1, fa2, fa3) { (a, b, c) =>
  a + b + c
}

val ex: Task[Int] = IO.raiseError(new RuntimeException("boo"))

// Yields Failure(e), because the second arg is a failure
IO.parMap3(fa1, ex, fa3) { (a, b, c) =>
  a + b + c
}

ADVICE: In a real life scenario the tasks should be expensive in order to warrant parallel execution. Parallelism doesn't magically speed up the code - it's usually fine for I/O-bound tasks, however for CPU-bound tasks it can make things worse. Performance improvements need to be verified.

NOTE: the tasks get forked automatically so there's no need to force asynchronous execution for immediate tasks, parallelism being guaranteed when multi-threading is available!

All specified tasks get evaluated in parallel, regardless of their execution model (IO.eval vs IO.evalAsync doesn't matter). Also the implementation tries to be smart about detecting forked tasks so it can eliminate extraneous forks for the very obvious cases.

See IO.map3 for sequential processing.

Pairs 4 IO values, applying the given mapping function, ordering the results, but not the side effects, the evaluation being done in parallel if the tasks are async.

This is a specialized IO.parSequence operation and as such the tasks are evaluated in parallel, ordering the results. In case one of the tasks fails, then all other tasks get cancelled and the final result will be a failure.

val fa1 = UIO(1)
val fa2 = UIO(2)
val fa3 = UIO(3)
val fa4 = UIO(4)

// Yields Success(10)
IO.parMap4(fa1, fa2, fa3, fa4) { (a, b, c, d) =>
  a + b + c + d
}

val ex: Task[Int] = IO.raiseError(new RuntimeException("boo"))

// Yields Failure(e), because the second arg is a failure
IO.parMap4(fa1, ex, fa3, fa4) {
  (a, b, c, d) => a + b + c + d
}

ADVICE: In a real life scenario the tasks should be expensive in order to warrant parallel execution. Parallelism doesn't magically speed up the code - it's usually fine for I/O-bound tasks, however for CPU-bound tasks it can make things worse. Performance improvements need to be verified.

NOTE: the tasks get forked automatically so there's no need to force asynchronous execution for immediate tasks, parallelism being guaranteed when multi-threading is available!

All specified tasks get evaluated in parallel, regardless of their execution model (IO.eval vs IO.evalAsync doesn't matter). Also the implementation tries to be smart about detecting forked tasks so it can eliminate extraneous forks for the very obvious cases.

See IO.map4 for sequential processing.

Pairs 5 IO values, applying the given mapping function, ordering the results, but not the side effects, the evaluation being done in parallel if the tasks are async.

This is a specialized IO.parSequence operation and as such the tasks are evaluated in parallel, ordering the results. In case one of the tasks fails, then all other tasks get cancelled and the final result will be a failure.

val fa1 = UIO(1)
val fa2 = UIO(2)
val fa3 = UIO(3)
val fa4 = UIO(4)
val fa5 = UIO(5)

// Yields Success(15)
IO.parMap5(fa1, fa2, fa3, fa4, fa5) { (a, b, c, d, e) =>
  a + b + c + d + e
}

val ex: Task[Int] = IO.raiseError(new RuntimeException("boo"))

// Yields Failure(e), because the second arg is a failure
IO.parMap5(fa1, ex, fa3, fa4, fa5) {
  (a, b, c, d, e) => a + b + c + d + e
}

ADVICE: In a real life scenario the tasks should be expensive in order to warrant parallel execution. Parallelism doesn't magically speed up the code - it's usually fine for I/O-bound tasks, however for CPU-bound tasks it can make things worse. Performance improvements need to be verified.

NOTE: the tasks get forked automatically so there's no need to force asynchronous execution for immediate tasks, parallelism being guaranteed when multi-threading is available!

All specified tasks get evaluated in parallel, regardless of their execution model (IO.eval vs IO.evalAsync doesn't matter). Also the implementation tries to be smart about detecting forked tasks so it can eliminate extraneous forks for the very obvious cases.

See IO.map5 for sequential processing.

Pairs 6 IO values, applying the given mapping function, ordering the results, but not the side effects, the evaluation being done in parallel if the tasks are async.

This is a specialized IO.parSequence operation and as such the tasks are evaluated in parallel, ordering the results. In case one of the tasks fails, then all other tasks get cancelled and the final result will be a failure.

val fa1 = UIO(1)
val fa2 = UIO(2)
val fa3 = UIO(3)
val fa4 = UIO(4)
val fa5 = UIO(5)
val fa6 = UIO(6)

// Yields Success(21)
IO.parMap6(fa1, fa2, fa3, fa4, fa5, fa6) { (a, b, c, d, e, f) =>
  a + b + c + d + e + f
}

val ex: Task[Int] = IO.raiseError(new RuntimeException("boo"))

// Yields Failure(e), because the second arg is a failure
IO.parMap6(fa1, ex, fa3, fa4, fa5, fa6) {
  (a, b, c, d, e, f) => a + b + c + d + e + f
}

ADVICE: In a real life scenario the tasks should be expensive in order to warrant parallel execution. Parallelism doesn't magically speed up the code - it's usually fine for I/O-bound tasks, however for CPU-bound tasks it can make things worse. Performance improvements need to be verified.

NOTE: the tasks get forked automatically so there's no need to force asynchronous execution for immediate tasks, parallelism being guaranteed when multi-threading is available!

All specified tasks get evaluated in parallel, regardless of their execution model (IO.eval vs IO.evalAsync doesn't matter). Also the implementation tries to be smart about detecting forked tasks so it can eliminate extraneous forks for the very obvious cases.

See IO.map6 for sequential processing.

Executes the given sequence of tasks in parallel, non-deterministically gathering their results, returning a task that will signal the sequence of results once all tasks are finished.

This function is the nondeterministic analogue of sequence and should behave identically to sequence so long as there is no interaction between the effects being gathered. However, unlike sequence, which decides on a total order of effects, the effects in a parSequence are unordered with respect to each other, the tasks being execute in parallel, not in sequence.

Although the effects are unordered, we ensure the order of results matches the order of the input sequence. Also see parSequenceUnordered for the more efficient alternative.

Example:

val tasks = List(Task(1 + 1), Task(2 + 2), Task(3 + 3))

// Yields 2, 4, 6
Task.parSequence(tasks)

ADVICE: In a real life scenario the tasks should be expensive in order to warrant parallel execution. Parallelism doesn't magically speed up the code - it's usually fine for I/O-bound tasks, however for CPU-bound tasks it can make things worse. Performance improvements need to be verified.

NOTE: the tasks get forked automatically so there's no need to force asynchronous execution for immediate tasks, parallelism being guaranteed when multi-threading is available!

All specified tasks get evaluated in parallel, regardless of their execution model (IO.eval vs IO.evalAsync doesn't matter). Also the implementation tries to be smart about detecting forked tasks so it can eliminate extraneous forks for the very obvious cases.

Executes the given sequence of tasks in parallel, non-deterministically gathering their results, returning a task that will signal the sequence of results once all tasks are finished.

Implementation ensure there are at most n (= parallelism parameter) tasks running concurrently and the results are returned in order.

Example:

import scala.concurrent.duration._

val tasks = List(
  Task(1 + 1).delayExecution(1.second),
  Task(2 + 2).delayExecution(2.second),
  Task(3 + 3).delayExecution(3.second),
  Task(4 + 4).delayExecution(4.second)
 )

// Yields 2, 4, 6, 8 after around 6 seconds
Task.parSequenceN(2)(tasks)

ADVICE: In a real life scenario the tasks should be expensive in order to warrant parallel execution. Parallelism doesn't magically speed up the code - it's usually fine for I/O-bound tasks, however for CPU-bound tasks it can make things worse. Performance improvements need to be verified.

NOTE: the tasks get forked automatically so there's no need to force asynchronous execution for immediate tasks, parallelism being guaranteed when multi-threading is available!

All specified tasks get evaluated in parallel, regardless of their execution model (IO.eval vs IO.evalAsync doesn't matter). Also the implementation tries to be smart about detecting forked tasks so it can eliminate extraneous forks for the very obvious cases.

Processes the given collection of tasks in parallel and nondeterministically gather the results without keeping the original ordering of the given tasks.

This function is similar to parSequence, but neither the effects nor the results will be ordered. Useful when you don't need ordering because:

• it has non-blocking behavior (but not wait-free)
• it can be more efficient (compared with parSequence), but not necessarily (if you care about performance, then test)

Example:

val tasks = List(Task(1 + 1), Task(2 + 2), Task(3 + 3))

// Yields 2, 4, 6 (but order is NOT guaranteed)
Task.parSequenceUnordered(tasks)

ADVICE: In a real life scenario the tasks should be expensive in order to warrant parallel execution. Parallelism doesn't magically speed up the code - it's usually fine for I/O-bound tasks, however for CPU-bound tasks it can make things worse. Performance improvements need to be verified.

NOTE: the tasks get forked automatically so there's no need to force asynchronous execution for immediate tasks, parallelism being guaranteed when multi-threading is available!

All specified tasks get evaluated in parallel, regardless of their execution model (IO.eval vs IO.evalAsync doesn't matter). Also the implementation tries to be smart about detecting forked tasks so it can eliminate extraneous forks for the very obvious cases.

Given a Iterable[A] and a function A => IO[E, B], nondeterministically apply the function to each element of the collection and return a task that will signal a collection of the results once all tasks are finished.

This function is the nondeterministic analogue of traverse and should behave identically to traverse so long as there is no interaction between the effects being gathered. However, unlike traverse, which decides on a total order of effects, the effects in a parTraverse are unordered with respect to each other.

Although the effects are unordered, we ensure the order of results matches the order of the input sequence. Also see parTraverseUnordered for the more efficient alternative.

It's a generalized version of parSequence.

ADVICE: In a real life scenario the tasks should be expensive in order to warrant parallel execution. Parallelism doesn't magically speed up the code - it's usually fine for I/O-bound tasks, however for CPU-bound tasks it can make things worse. Performance improvements need to be verified.

NOTE: the tasks get forked automatically so there's no need to force asynchronous execution for immediate tasks, parallelism being guaranteed when multi-threading is available!

All specified tasks get evaluated in parallel, regardless of their execution model (IO.eval vs IO.evalAsync doesn't matter). Also the implementation tries to be smart about detecting forked tasks so it can eliminate extraneous forks for the very obvious cases.

Applies the provided function in a non-deterministic way to each element of the input collection. The result will be signalled once all tasks are finished with a success, or as soon as some task finishes with a typed or terminal error.

Note that his method has a fail-fast semantics: as soon as one of the tasks fails (either in a typed or terminal manner), no subsequent tasks will be executed and they will be cancelled.

The final result will be a collection of success values, or a typed/fatal error if at least one of the tasks finished without a success.

This method allows specifying the parallelism level of the execution, i.e. the maximum number of how many tasks should be running concurrently.

Although the execution of the effects is unordered and non-deterministic, the collection of results will preserve the order of the input collection.

Example:

import scala.concurrent.duration._

val numbers = List(1, 2, 3, 4)

// Yields 2, 4, 6, 8 after around 6 seconds
IO.parTraverseN(2)(numbers)(n => IO(n + n).delayExecution(n.second))

ADVICE: In a real life scenario the tasks should be expensive in order to warrant parallel execution. Parallelism doesn't magically speed up the code - it's usually fine for I/O-bound tasks, however for CPU-bound tasks it can make things worse. Performance improvements need to be verified.

NOTE: the tasks get forked automatically so there's no need to force asynchronous execution for immediate tasks, parallelism being guaranteed when multi-threading is available!

All specified tasks get evaluated in parallel, regardless of their execution model (IO.eval vs IO.evalAsync doesn't matter). Also the implementation tries to be smart about detecting forked tasks so it can eliminate extraneous forks for the very obvious cases.

Given a Iterable[A] and a function A => IO[E, B], nondeterministically apply the function to each element of the collection without keeping the original ordering of the results.

This function is similar to parTraverse, but neither the effects nor the results will be ordered. Useful when you don't need ordering because:

• it has non-blocking behavior (but not wait-free)
• it can be more efficient (compared with parTraverse), but not necessarily (if you care about performance, then test)

It's a generalized version of parSequenceUnordered.

ADVICE: In a real life scenario the tasks should be expensive in order to warrant parallel execution. Parallelism doesn't magically speed up the code - it's usually fine for I/O-bound tasks, however for CPU-bound tasks it can make things worse. Performance improvements need to be verified.

NOTE: the tasks get forked automatically so there's no need to force asynchronous execution for immediate tasks, parallelism being guaranteed when multi-threading is available!

All specified tasks get evaluated in parallel, regardless of their execution model (IO.eval vs IO.evalAsync doesn't matter). Also the implementation tries to be smart about detecting forked tasks so it can eliminate extraneous forks for the very obvious cases.

Lifts a value into the task context. Alias for now.

Run two Task actions concurrently, and return the first to finish, either in success or error. The loser of the race is cancelled.

The two tasks are executed in parallel, the winner being the first that signals a result.

As an example, this would be equivalent with IO.timeout:

import scala.concurrent.duration._
import scala.concurrent.TimeoutException

// some long running task
val myTask = Task(42)

val timeoutError = Task
  .raiseError(new TimeoutException)
  .delayExecution(5.seconds)

Task.race(myTask, timeoutError)

Similarly IO.timeoutTo is expressed in terms of race.

NOTE: the tasks get forked automatically so there's no need to force asynchronous execution for immediate tasks, parallelism being guaranteed when multi-threading is available!

All specified tasks get evaluated in parallel, regardless of their execution model (IO.eval vs IO.evalAsync doesn't matter). Also the implementation tries to be smart about detecting forked tasks so it can eliminate extraneous forks for the very obvious cases.

Runs multiple tasks in a concurrent way and returns the fastest of them, regardless whether it's a success, a typed error or a terminal error. Every task losing the race gets cancelled.

import scala.concurrent.duration._

val tasks: List[UIO[Int]] =
  List(1, 2, 3).map(i => IO.sleep(i.seconds).map(_ => i))

val winner: UIO[Int] = IO.raceMany(tasks)

NOTE: the tasks get forked automatically so there's no need to force asynchronous execution for immediate tasks, parallelism being guaranteed when multi-threading is available!

All specified tasks get evaluated in parallel, regardless of their execution model (IO.eval vs IO.evalAsync doesn't matter). Also the implementation tries to be smart about detecting forked tasks so it can eliminate extraneous forks for the very obvious cases.

Run two Task actions concurrently, and returns a pair containing both the winner's successful value and the loser represented as a still-unfinished task.

If the first task completes in error, then the result will complete in error, the other task being cancelled.

On usage the user has the option of cancelling the losing task, this being equivalent with plain race:

import scala.concurrent.duration._

val ta = Task.sleep(2.seconds).map(_ => "a")
val tb = Task.sleep(3.seconds).map(_ => "b")

// `tb` is going to be cancelled as it returns 1 second after `ta`
Task.racePair(ta, tb).flatMap {
  case Left((a, taskB)) =>
    taskB.cancel.map(_ => a)
  case Right((taskA, b)) =>
    taskA.cancel.map(_ => b)
}

NOTE: the tasks get forked automatically so there's no need to force asynchronous execution for immediate tasks, parallelism being guaranteed when multi-threading is available!

All specified tasks get evaluated in parallel, regardless of their execution model (IO.eval vs IO.evalAsync doesn't matter). Also the implementation tries to be smart about detecting forked tasks so it can eliminate extraneous forks for the very obvious cases.

Returns raiseError when cond is false, otherwise IO.unit

Returns raiseError when the cond is true, otherwise IO.unit

Returns the current IO.Options configuration, which determine the task's run-loop behavior.

Inverse of attempt. Creates a new IO that absorbs Either.

IO.rethrow(IO.now(Right(42))) <-> IO.now(42)

IO.rethrow(IO.now(Left("error"))) <-> IO.raiseError("error")

Given a Iterable of tasks, transforms it to a task signaling the collection, executing the tasks one by one and gathering their results in the same collection.

This operation will execute the tasks one by one, in order, which means that both effects and results will be ordered. See parSequence and parSequenceUnordered for unordered results or effects, and thus potential of running in parallel.

It's a simple version of traverse.

Asynchronous boundary described as an effectful Task that can be used in flatMap chains to "shift" the continuation of the run-loop to another call stack or thread, managed by the given execution context.

This is the equivalent of IO.shift.

For example we can introduce an asynchronous boundary in the flatMap chain before a certain task, this being literally the implementation of executeAsync:

val task = IO.eval(35)

IO.shift.flatMap(_ => task)

And this can also be described with >> from Cats:

import cats.syntax.all._

IO.shift >> task

Or we can specify an asynchronous boundary after the evaluation of a certain task, this being literally the implementation of .asyncBoundary:

task.flatMap(a => IO.shift.map(_ => a))

And again we can also describe this with <* from Cats:

task <* IO.shift

Asynchronous boundary described as an effectful Task that can be used in flatMap chains to "shift" the continuation of the run-loop to another thread or call stack, managed by the default Scheduler.

This is the equivalent of IO.shift, except that Monix's Task gets executed with an injected Scheduler in IO.runAsync or in IO.runToFuture and that's going to be the Scheduler responsible for the "shift".

For example we can introduce an asynchronous boundary in the flatMap chain before a certain task, this being literally the implementation of executeAsync:

val task = IO.eval(35)

IO.shift.flatMap(_ => task)

And this can also be described with >> from Cats:

import cats.syntax.all._

IO.shift >> task

Or we can specify an asynchronous boundary after the evaluation of a certain task, this being literally the implementation of .asyncBoundary:

task.flatMap(a => IO.shift.map(_ => a))

And again we can also describe this with <* from Cats:

task <* IO.shift

Creates a new Task that will sleep for the given duration, emitting a tick when that time span is over.

As an example on evaluation this will print "Hello!" after 3 seconds:

import scala.concurrent.duration._

IO.sleep(3.seconds).flatMap { _ =>
  IO.eval(println("Hello!"))
}

See IO.delayExecution for this operation described as a method on Task references or IO.delayResult for the helper that triggers the evaluation of the source on time, but then delays the result.

Keeps calling f until it returns a Right result.

Based on Phil Freeman's Stack Safety for Free.

Returns a task that on execution is always finishing in a fatal (unexpected) error emitting the specified exception.

This type of errors is not reflected in the type signature and it skips all regular error handlers, except for IO.redeemCause and IO.redeemCauseWith.

Builds a cats.effect.Timer instance, given a Scheduler reference.

Default, pure, globally visible cats.effect.Timer implementation that defers the evaluation to Task's default Scheduler (that's being injected in IO.runToFuture).

Given a Iterable[A] and a function A => Task[B], sequentially apply the function to each element of the collection and gather their results in the same collection.

It's a generalized version of sequence.

Returns the given argument if cond is false, otherwise IO.Unit

Returns the given argument if cond is true, otherwise IO.Unit

Newtype encoding, see the IO.Par type alias for more details.

DEPRECATED — renamed to IO.parSequence.

DEPRECATED — renamed to IO.parSequenceN

DEPRECATED — renamed to IO.parSequenceUnordered

DEPRECATED — renamed to IO.parTraverse

DEPRECATED — renamed to IO.parTraverseN

DEPRECATED — renamed to IO.parTraverseUnordered