retry

don't give up

install

With sbt, add the following to your project's build.sbt

libraryDependencies += "me.lessis" %% "retry-core" % "0.1.0"

usage

Applications can fail a runtime. We know this. Network connections drop. Connections timeout. Bad things happen.

Don't let bad things cause other bad things to happen. Give applications a second chance to retry.

Retry provides interfaces for common retry strategies that operate on scala.util.Futures.

Basic usage requires four things

• an implicit execution context for executing futures
• a definition of retry.Success[-T](pred: T => Boolean) to encode what "success" means for the type of your future
• a retry.Timer for asynchronously scheduling a followup attempt
• a block of code that results in a scala.concurrent.Future

Retry provides a set of defaults that uses scala.concurrent.ExecutionContext.Implicits.global as an execution context, retry.Success definitions for Option, Either, and scala.util.Try, and a retry.jdk.JdkTimer as a retry.Timer out of the box.

import scala.concurrent._
import retry.Defaults._
retry.Backoff()(Future {
  // something that can fail
})

Defining success.

Retry needs to know what failure or success means in order to know when to retry. It does this through a typed Success[-T](pred: T => Boolean) type. Where T matches the type your scala.util.Future[T] is defined with. Retry looks for this implicitly in within the scope of retrying. If you are using the retry.Defaults you will already have in scope a definition for success for Either, scala.util.Try, and Option types. If you wish to define an application-specific definition of what "success" means for your future, you may do so by specifying the following in scope of the retry.

implicit val success = new Success[Int](_ > 10)

If your future completes with an int less than or equal to 10. It is then considered a failure and will be retried.

Sleep schedules

Your application may run within a platform that provides its own way for scheduling tasks to happen in the future. If retry.jdk.JdkTimer isn't what you're looking for, you may wish to use the retry.Timer for netty, retry.netty.Timer in the retry-netty module or a retry.twitter.Timer available in the retry-twitter module. If these also aren't what you're looking for, you can define your own in the scope of the retry.

class YourTimer extends retry.Timer {
  def apply[T](len: Long, unit: TimeUnit, todo: => T) = new retry.Timeout {
    def cancel() {
      // return something which may be cancelled
    }
  }
}

implicit val timer = new YourTimer 

Doug Tangren (softprops) 2013