Documentation

You are viewing the documentation for the 3.0.5 release. The latest stable release series is 3.0.x.

§Filters

Play provides a simple filter API for applying global filters to each request.

§Filters vs action composition

The filter API is intended for cross cutting concerns that are applied indiscriminately to all routes. For example, here are some common use cases for filters:

In contrast, action composition is intended for route specific concerns, such as authentication and authorization, caching and so on. If your filter is not one that you want applied to every route, consider using action composition instead, it is far more powerful. And don’t forget that you can create your own action builders that compose your own custom defined sets of actions to each route, to minimize boilerplate.

§A simple logging filter

The following is a simple filter that times and logs how long a request takes to execute in Play Framework, which implements the Filter trait:

import javax.inject.Inject

import scala.concurrent.ExecutionContext
import scala.concurrent.Future

import org.apache.pekko.stream.Materializer
import play.api.mvc._
import play.api.Logging

class LoggingFilter @Inject() (implicit val mat: Materializer, ec: ExecutionContext) extends Filter with Logging {
  def apply(nextFilter: RequestHeader => Future[Result])(requestHeader: RequestHeader): Future[Result] = {
    val startTime = System.currentTimeMillis

    nextFilter(requestHeader).map { result =>
      val endTime     = System.currentTimeMillis
      val requestTime = endTime - startTime

      logger.info(
        s"${requestHeader.method} ${requestHeader.uri} took ${requestTime}ms and returned ${result.header.status}"
      )

      result.withHeaders("Request-Time" -> requestTime.toString)
    }
  }
}

Let’s understand what’s happening here. The first thing to notice is the signature of the apply method. It’s a curried function, with the first parameter, nextFilter, being a function that takes a request header and produces a result, and the second parameter, requestHeader, being the actual request header of the incoming request.

The nextFilter parameter represents the next action in the filter chain. Invoking it will cause the action to be invoked. In most cases you will probably want to invoke this at some point in your future. You may decide to not invoke it if for some reason you want to block the request.

We save a timestamp before invoking the next filter in the chain. Invoking the next filter returns a Future[Result] that will redeemed eventually. Take a look at the Handling asynchronous results chapter for more details on asynchronous results. We then manipulate the Result in the Future by calling the map method with a closure that takes a Result. We calculate the time it took for the request, log it and send it back to the client in the response headers by calling result.withHeaders("Request-Time" -> requestTime.toString).

§Using filters

The simplest way to use a filter is to provide an implementation of the HttpFilters trait in the root package. If you’re using Play’s runtime dependency injection support (such as Guice) you can extend the DefaultHttpFilters class and pass your filters to the varargs constructor:

import javax.inject.Inject

import play.api.http.DefaultHttpFilters
import play.api.http.EnabledFilters
import play.filters.gzip.GzipFilter

class Filters @Inject() (
    defaultFilters: EnabledFilters,
    gzip: GzipFilter,
    log: LoggingFilter
) extends DefaultHttpFilters(defaultFilters.filters :+ gzip :+ log: _*)

If you want to have different filters in different environments, or would prefer not putting this class in the root package, you can configure where Play should find the class by setting play.http.filters in application.conf to the fully qualified class name of the class. For example:

play.http.filters=com.example.MyFilters

If you’re using BuiltInComponents for compile-time dependency injection, you can simply override the httpFilters lazy val:


import play.api._
import play.filters.gzip._
import play.filters.HttpFiltersComponents
import router.Routes

class MyComponents(context: ApplicationLoader.Context)
    extends BuiltInComponentsFromContext(context)
    with HttpFiltersComponents
    with GzipFilterComponents {
  // implicit executionContext and materializer are defined in BuiltInComponents
  lazy val loggingFilter: LoggingFilter = new LoggingFilter()

  // gzipFilter is defined in GzipFilterComponents
  override lazy val httpFilters = Seq(gzipFilter, loggingFilter)

  lazy val router: Routes = new Routes( /* ... */ )
}

The filters provided by Play all provide traits that work with BuiltInComponents:
- GzipFilterComponents
- CSRFComponents
- CORSComponents
- SecurityHeadersComponents
- AllowedHostsComponents

§Where do filters fit in?

Filters wrap the action after the action has been looked up by the router. This means you cannot use a filter to transform a path, method or query parameter to impact the router. However you can direct the request to a different action by invoking that action directly from the filter, though be aware that this will bypass the rest of the filter chain. If you do need to modify the request before the router is invoked, a better way to do this would be to place your logic in `HttpRequestHandler` instead.

Since filters are applied after routing is done, it is possible to access routing information from the request, via the attrs map on the RequestHeader. For example, you might want to log the time against the action method. In that case, you might update the filter to look like this:

import javax.inject.Inject

import scala.concurrent.ExecutionContext
import scala.concurrent.Future

import org.apache.pekko.stream.Materializer
import play.api.mvc.Filter
import play.api.mvc.RequestHeader
import play.api.mvc.Result
import play.api.routing.HandlerDef
import play.api.routing.Router
import play.api.Logging

class LoggingFilter @Inject() (implicit val mat: Materializer, ec: ExecutionContext) extends Filter with Logging {
  def apply(nextFilter: RequestHeader => Future[Result])(requestHeader: RequestHeader): Future[Result] = {
    val startTime = System.currentTimeMillis

    nextFilter(requestHeader).map { result =>
      val handlerDef: HandlerDef = requestHeader.attrs(Router.Attrs.HandlerDef)
      val action                 = handlerDef.controller + "." + handlerDef.method
      val endTime                = System.currentTimeMillis
      val requestTime            = endTime - startTime

      logger.info(s"${action} took ${requestTime}ms and returned ${result.header.status}")

      result.withHeaders("Request-Time" -> requestTime.toString)
    }
  }
}

Routing attributes are a feature of the Play router. If you use a custom router, or return a custom action through a custom request handler, these parameters may not be available.

§More powerful filters

Play provides a lower level filter API called EssentialFilter which gives you full access to the body of the request. This API allows you to wrap EssentialAction with another action.

Here is the above filter example rewritten as an EssentialFilter:

import javax.inject.Inject

import scala.concurrent.ExecutionContext

import org.apache.pekko.util.ByteString
import play.api.libs.streams.Accumulator
import play.api.mvc._
import play.api.Logging

class LoggingFilter @Inject() (implicit ec: ExecutionContext) extends EssentialFilter with Logging {
  def apply(nextFilter: EssentialAction): EssentialAction = new EssentialAction {
    def apply(requestHeader: RequestHeader) = {
      val startTime = System.currentTimeMillis

      val accumulator: Accumulator[ByteString, Result] = nextFilter(requestHeader)

      accumulator.map { result =>
        val endTime     = System.currentTimeMillis
        val requestTime = endTime - startTime

        logger.info(
          s"${requestHeader.method} ${requestHeader.uri} took ${requestTime}ms and returned ${result.header.status}"
        )
        result.withHeaders("Request-Time" -> requestTime.toString)
      }
    }
  }
}

The key difference here, apart from creating a new EssentialAction to wrap the passed in next action, is when we invoke next, we get back an Accumulator.

You could compose the Accumulator with an Pekko Streams Flow using the through method with some transformations to the stream if you wished. We then map the result of the iteratee and thus handle it.

class AccumulatorFlowFilter @Inject() (actorSystem: ActorSystem)(implicit ec: ExecutionContext)
    extends EssentialFilter {
  private val logger = org.slf4j.LoggerFactory.getLogger("application.AccumulatorFlowFilter")

  private implicit val logging: LoggingAdapter = Logging(actorSystem.eventStream, logger.getName)

  override def apply(next: EssentialAction): EssentialAction = new EssentialAction {
    override def apply(request: RequestHeader): Accumulator[ByteString, Result] = {
      val accumulator: Accumulator[ByteString, Result] = next(request)

      val flow: Flow[ByteString, ByteString, NotUsed] = Flow[ByteString].log("byteflow")
      val accumulatorWithResult = accumulator.through(flow).map { result =>
        logger.info(s"The flow has completed and the result is $result")
        result
      }

      accumulatorWithResult
    }
  }
}

Although it may seem that there are two different filter APIs, there is only one, EssentialFilter. The simpler Filter API in the earlier examples extends EssentialFilter, and implements it by creating a new EssentialAction. The passed in callback makes it appear to skip the body parsing by creating a promise for the Result, while the body parsing and the rest of the action are executed asynchronously.

Next: Testing your application