§What’s new in Play 2.6

This page highlights the new features of Play 2.6. If you want to learn about the changes you need to make when you migrate to Play 2.6, check out the Play 2.6 Migration Guide.

§Scala 2.12 support

Play 2.6 is the first release of Play to have been cross built against Scala 2.12 and 2.11. A number of dependencies were updated so that we can have support for both versions.

You can select which version of Scala you would like to use by setting the scalaVersion setting in your build.sbt.

For Scala 2.12:

scalaVersion := "2.12.2"

For Scala 2.11:

scalaVersion := "2.11.11"

§“Global-State-Free” Applications

The biggest under the hood change is that Play no longer relies on global state under the hood. You can still access the global application through play.api.Play.current / play.Play.application() in Play 2.6, but it is deprecated. This sets the stage for Play 3.0, where there is no global state at all.

You can disable access to global application entirely by setting the following configuration value:


The above setting will cause an exception on any invocation of Play.current.

§Akka HTTP Server Backend

Play now uses the Akka-HTTP server engine as the default backend. More detail about Play’s integration with Akka-HTTP can be found on the Akka HTTP Server page. There is an additional page on configuring Akka HTTP.

The Netty backend is still available, and has been upgraded to use Netty 4.1. You can explicitly configure your project to use Netty on the NettyServer page.

§HTTP/2 support (experimental)

Play now has HTTP/2 support on the Akka HTTP server using the PlayAkkaHttp2Support module:

lazy val root = (project in file("."))
  .enablePlugins(PlayJava, PlayAkkaHttp2Support)

This automates most of the process of setting up HTTP/2. However, it does not work with the run command by default. See the Akka HTTP Server page for more details.

§Request attributes

Requests in Play 2.6 now contain attributes. Attributes allow you to store extra information inside request objects. For example, you can write a filter that sets an attribute in the request and then access the attribute value later from within your actions.

Attributes are stored in a TypedMap that is attached to each request. TypedMaps are immutable maps that store type-safe keys and values. Attributes are indexed by a key and the key’s type indicates the type of the attribute.


// Create a TypedKey to store a User object
class Attrs {
  public static final TypedKey<User> USER = TypedKey.<User>create("user");

// Get the User object from the request
User user = req.attrs().get(Attrs.USER);
// Put a User object into the request
Request newReq = req.addAttr(Attrs.USER, newUser);


// Create a TypedKey to store a User object
object Attrs {
  val User: TypedKey[User] = TypedKey.apply[User]("user")

// Get the User object from the request
val user: User = req.attrs(Attrs.User)
// Put a User object into the request
val newReq = req.addAttr(Attrs.User, newUser)

Attributes are stored in a TypedMap. You can read more about attributes in the TypedMap documentation: Javadoc, Scaladoc.

Request tags have now been deprecated and you should migrate to use attributes instead. See the tags section in the migration docs for more information.

§Route modifier tags

The routes file syntax now allows you to add “modifiers” to each route that provide custom behavior. We have implemented one such tag in the CSRF filter, the “nocsrf” tag. By default, the following route will not have the CSRF filter applied.

+ nocsrf # Don't CSRF protect this route 
POST /api/foo/bar ApiController.foobar

You can also create your own modifiers: the + symbol can be followed by any number of whitespace-separated tags.

These are made available in the HandlerDef request attribute (which also contains other metadata on the handler definition in the routes file):


import java.util.List;
import play.routing.HandlerDef;
import play.routing.Router;

HandlerDef handler = req.attrs().get(Router.Attrs.HANDLER_DEF);
List<String> modifiers = handler.getModifiers();


import play.api.routing.{ HandlerDef, Router }
import play.api.mvc.RequestHeader

val handler = request.attrs(Router.Attrs.HandlerDef)
val modifiers = handler.modifiers

§Injectable Twirl Templates

Twirl templates can now be created with a constructor annotation using @this. The constructor annotation means that Twirl templates can be injected into templates directly and can manage their own dependencies, rather than the controller having to manage dependencies not only for itself, but also for the templates it has to render.

As an example, suppose a template has a dependency on a component TemplateRenderingComponent, which is not used by the controller.

First create a file IndexTemplate.scala.html using the @this syntax for the constructor. Note that the constructor must be placed before the @() syntax used for the template’s parameters for the apply method:

@this(trc: TemplateRenderingComponent)
@(item: Item)


By default all generated Scala template classes Twirl creates with the @this syntax within Play will automatically be annotated with @javax.inject.Inject(). If desired you can change this behavior in build.sbt:

// Add one or more annotation(s):
TwirlKeys.constructorAnnotations += "@java.lang.Deprecated()"

// Or completely replace the default one with your own annotation(s):
TwirlKeys.constructorAnnotations := Seq("")

Now define the controller by injecting the template in the constructor:


public class MyController extends Controller {

  private final views.html.indexTemplate template;

  public MyController(views.html.indexTemplate template) {
    this.template = template;

  public Result index() {
    return ok(template.render());



class MyController @Inject()(indexTemplate: views.html.IndexTemplate,
                              cc: ControllerComponents)
  extends AbstractController(cc) {

  def index = Action { implicit request =>

Once the template is defined with its dependencies, then the controller can have the template injected into the controller, but the controller does not see TemplateRenderingComponent.

§Filters Enhancements

Play now comes with a default set of enabled filters, defined through configuration. This provides a “secure by default” experience for new Play applications, and tightens security on existing Play applications.

The following filters are enabled by default:

In addition, filters can now be configured through application.conf. To append to the defaults list, use the +=:


If you want to specifically disable a filter for testing, you can also do that from configuration:


Please see the Filters page for more details.

NOTE: If you are migrating from an existing project that does not use CSRF form helpers such as CSRF.formField, then you may see “403 Forbidden” on PUT and POST requests, from the CSRF filter. To check this behavior, please add <logger name="play.filters.csrf" value="TRACE"/> to your logback.xml. Likewise, if you are running a Play application on something other than localhost, you must configure the AllowedHostsFilter to specifically allow the hostname/ip you are connecting from.

§gzip filter

If you have the gzip filter enabled you can now also control which responses are and aren’t gzipped based on their content types via application.conf (instead of writing you own Filters class):

play.filters.gzip {

    contentType {

        # If non empty, then a response will only be compressed if its content type is in this list.
        whiteList = [ "text/*", "application/javascript", "application/json" ]

        # The black list is only used if the white list is empty.
        # Compress all responses except the ones whose content type is in this list.
        blackList = []

Please see the gzip filter page for more details.

§JWT Cookies

Play now uses JSON Web Token (JWT) format for session and flash cookies. This allows for a standardized signed cookie data format, cookie expiration (making replay attacks harder) and more flexibility in signing cookies.

Please see Scala or Java pages for more details.

§Logging Marker API

SLF4J Marker support has been added to play.Logger and play.api.Logger.

In the Java API, it is a straight port of the SLF4J Logger API. This is useful, but you may find an SLF4J wrapper like Godaddy Logger for a richer logging experience.

In the Scala API, markers are added through a MarkerContext trait, which is added as an implicit parameter to the logger methods, i.e.

import play.api._"some info message")(MarkerContext(someMarker))

This opens the door for implicit markers to be passed for logging in several statements, which makes adding context to logging much easier without resorting to MDC. For example, using Logstash Logback Encoder and an implicit conversion chain, request information can be encoded into logging statements automatically:

trait RequestMarkerContext {

  // Adding 'implicit request' enables implicit conversion chaining
  // See
  implicit def requestHeaderToMarkerContext(implicit request: RequestHeader): MarkerContext = {
    import net.logstash.logback.marker.LogstashMarker
    import net.logstash.logback.marker.Markers._

    val requestMarkers: LogstashMarker = append("host",
      .and(append("path", request.path))



And then used in a controller and carried through Future that may use different execution contexts:

def asyncIndex = Action.async { implicit request =>
  Future {
    methodInOtherExecutionContext() // implicit conversion here

def methodInOtherExecutionContext()(implicit mc: MarkerContext): Result = {
  logger.debug("index: ") // same as above

Note that marker contexts are also very useful for “tracer bullet” style logging, where you want to log on a specific request without explicitly changing log levels. For example, you can add a marker only when certain conditions are met:

trait TracerMarker {
  import TracerMarker._

  implicit def requestHeaderToMarkerContext(implicit request: RequestHeader): MarkerContext = {
    val marker = org.slf4j.MarkerFactory.getDetachedMarker("dynamic") // base do-nothing marker...
    if (request.getQueryString("trace").nonEmpty) {

object TracerMarker {
  private val tracerMarker = org.slf4j.MarkerFactory.getMarker("TRACER")

class TracerBulletController @Inject()(cc: ControllerComponents)
  extends AbstractController(cc) with TracerMarker {
  private val logger = play.api.Logger("application")

  def index = Action { implicit request: Request[AnyContent] =>
    logger.trace("Only logged if queryString contains trace=true")

    Ok("hello world")

And then trigger logging with the following TurboFilter in logback.xml:

<turboFilter class="ch.qos.logback.classic.turbo.MarkerFilter">

For more information, please see ScalaLogging or JavaLogging.

For more information about using Markers in logging, see TurboFilters and marker based triggering sections in the Logback manual.

§Configuration improvements

In the Java API, we have moved to the standard Config object from Lightbend’s Config library instead of play.Configuration. This brings the behavior in line with standard config behavior, as the methods now expect all keys to exist. See the Java config migration guide for migration details.

In the Scala API, we have introduced new methods to the play.api.Configuration class to simplify the API and allow loading of custom types. You can now use an implicit ConfigLoader to load any custom type you want. Like the Config API, the new Configuration#get[T] expects the key to exist by default and returns a value of type T, but there is also a ConfigLoader[Option[T]] that allows null config values. See the Scala configuration docs for more details.

§Security Logging

A security marker has been added for security related operations in Play, and failed security checks now log at WARN level, with the security marker set. This ensures that developers always know why a particular request is failing, which is important now that security filters are enabled by default in Play.

The security marker also allows security failures to be triggered or filtered distinct from normal logging. For example, to disable all logging with the SECURITY marker set, add the following lines to the logback.xml file:

<turboFilter class="ch.qos.logback.classic.turbo.MarkerFilter">

In addition, log events using the security marker can also trigger a message to a Security Information & Event Management (SIEM) engine for further processing.

§Configuring a Custom Logging Framework in Java

Before, if you want to use a custom logging framework, you had to configure it using Scala, even if the you have a Java project. Now it is possible to create custom LoggerConfigurator in both Java and Scala. To create a LoggerConfigurator in Java, you need to implement the given interface, for example, to configure Log4J:

import com.typesafe.config.Config;
import com.typesafe.config.ConfigFactory;
import org.slf4j.ILoggerFactory;
import play.Environment;
import play.LoggerConfigurator;
import play.Mode;
import play.api.PlayException;

import java.util.Collections;
import java.util.HashMap;
import java.util.Map;
import java.util.Optional;

import org.apache.logging.log4j.LogManager;
import org.apache.logging.log4j.core.*;
import org.apache.logging.log4j.core.config.Configurator;

public class JavaLog4JLoggerConfigurator implements LoggerConfigurator {

    private ILoggerFactory factory;

    public void init(File rootPath, Mode mode) {
        Map<String, String> properties = new HashMap<>();
        properties.put("application.home", rootPath.getAbsolutePath());

        String resourceName = "log4j2.xml";
        URL resourceUrl = this.getClass().getClassLoader().getResource(resourceName);
        configure(properties, Optional.ofNullable(resourceUrl));

    public void configure(Environment env) {
        Map<String, String> properties = LoggerConfigurator.generateProperties(env, ConfigFactory.empty(), Collections.emptyMap());
        URL resourceUrl = env.resource("log4j2.xml");
        configure(properties, Optional.ofNullable(resourceUrl));

    public void configure(Environment env, Config configuration, Map<String, String> optionalProperties) {
        // LoggerConfigurator.generateProperties enables play.logger.includeConfigProperties=true
        Map<String, String> properties = LoggerConfigurator.generateProperties(env, configuration, optionalProperties);
        URL resourceUrl = env.resource("log4j2.xml");
        configure(properties, Optional.ofNullable(resourceUrl));

    public void configure(Map<String, String> properties, Optional<URL> config) {
        try {
            LoggerContext loggerContext = (LoggerContext) LogManager.getContext(false);

            factory = org.slf4j.impl.StaticLoggerBinder.getSingleton().getLoggerFactory();
        } catch (URISyntaxException ex) {
            throw new PlayException(
                "log4j2.xml resource was not found",
                "Could not parse the location for log4j2.xml resource",

    public ILoggerFactory loggerFactory() {
        return factory;

    public void shutdown() {
        LoggerContext loggerContext = (LoggerContext) LogManager.getContext();

Note: this implementation is fully compatible with Scala version LoggerConfigurator and can even be used in Scala projects if necessary, which means that module creators can provide a Java or Scala implementation of LoggerConfigurator and they will be usable in both Java and Scala projects.

§Separate Java Forms module and PlayMinimalJava plugin

The Java forms functionality has been split out into a separate module. The forms functionality depends on a few Spring modules and the Hibernate validator, so if you are not using forms, you may wish to remove the Java forms module to avoid those unnecessary dependencies.

This module is automatically included by the PlayJava plugin, but can be disabled by using the PlayMinimalJava plugin instead:

lazy val root = (project in file("."))

§Java Compile Time Components

Just as in Scala, Play now has components to enable Java Compile Time Dependency Injection. The components were created as interfaces that you should implements and they provide default implementations. There are components for all the types that could be injected when using Runtime Dependency Injection. To create an application using Compile Time Dependency Injection, you just need to provide an implementation of play.ApplicationLoader that uses a custom implementation of play.BuiltInComponents, for example:

import play.routing.Router;
import play.ApplicationLoader;
import play.BuiltInComponentsFromContext;
import play.filters.components.HttpFiltersComponents;

public class MyComponents extends BuiltInComponentsFromContext
        implements HttpFiltersComponents {

    public MyComponents(ApplicationLoader.Context context) {

    public Router router() {
        return Router.empty();

The play.ApplicationLoader:

import play.ApplicationLoader;

public class MyApplicationLoader implements ApplicationLoader {

    public Application load(Context context) {
        return new MyComponents(context).application();


And configure MyApplicationLoader as explained in Java Compile-Time Dependency Injection docs.

§Improved Form Handling I18N support

The MessagesApi and Lang classes are used for internationalization in Play, and are required to display error messages in forms.

In the past, putting together a form in Play has required multiple steps, and the creation of a Messages instance from a request was not discussed in the context of form handling.

In addition, it was inconvenient to have a Messages instance passed through all template fragments when form handling was required, and Messages implicit support was provided directly through the controller trait. The I18N API has been refined with the addition of a MessagesProvider trait, implicits that are tied directly to requests, and the forms documentation has been improved.

The MessagesActionBuilder has been added. This action builder provides a MessagesRequest, which is a WrappedRequest that extends MessagesProvider, only a single implicit parameter needs to be made available to templates, and you don’t need to extend Controller with I18nSupport. This is also useful because to use CSRF with forms, both a Request (technically a RequestHeader) and a Messages object must be available to the template.

class FormController @Inject()(messagesAction: MessagesActionBuilder, components: ControllerComponents)
  extends AbstractController(components) {


  val userForm = Form(
      "name" -> text,
      "age" -> number

  def index = messagesAction { implicit request: MessagesRequest[AnyContent] =>

  def post = ...

where displayForm.scala.html is defined as:

@(userForm: Form[UserData])(implicit request: MessagesRequestHeader)

@import helper._

@helper.form(action = {
  @CSRF.formField                     @* <- takes a RequestHeader    *@
  @helper.inputText(userForm("name")) @* <- takes a MessagesProvider *@
  @helper.inputText(userForm("age"))  @* <- takes a MessagesProvider *@

For more information, please see ScalaI18N.

§Testing Support

Support for creating MessagesApi instances has been improved. Now, when you want to create a MessagesApi instance, you can create DefaultMessagesApi() or DefaultLangs() with default arguments. If you want to specify test messages from configuration or from another source, you can pass in those values:

val messagesApi: MessagesApi = {
    val env = new Environment(new File("."), this.getClass.getClassLoader, Mode.Dev)
    val config = Configuration.reference ++ Configuration.from(Map("play.i18n.langs" -> Seq("en", "fr", "fr-CH")))
    val langs = new DefaultLangsProvider(config).get
    new DefaultMessagesApi(testMessages, langs)

§Future Timeout and Delayed Support

Play’s support for futures in asynchronous operations has been improved, using the Futures trait.

You can use the play.libs.concurrent.Futures interface to wrap a CompletionStage in a non-blocking timeout:

class MyClass {
    public MyClass(Futures futures) {
        this.futures = futures;

    CompletionStage<Double> callWithOneSecondTimeout() {
        return futures.timeout(computePIAsynchronously(), Duration.ofSeconds(1));

or use play.api.libs.concurrent.Futures trait in the Scala API:

import play.api.libs.concurrent.Futures._

class MyController @Inject()(cc: ControllerComponents)(implicit futures: Futures) extends AbstractController(cc) {

  def index = Action.async {
    // withTimeout is an implicit type enrichment provided by importing Futures._
    intensiveComputation().withTimeout(1.seconds).map { i =>
      Ok("Got result: " + i)
    }.recover {
      case e: TimeoutException =>

There is also a delayed method which only executes a Future after a specified delay, which works similarly to timeout.

For more information, please see ScalaAsync or JavaAsync.

§CustomExecutionContext and Thread Pool Sizing

This class defines a custom execution context that delegates to an It is very useful for situations in which the default execution context should not be used, for example if a database or blocking I/O is being used. Detailed information can be found in the ThreadPools page, but Play 2.6.x adds a CustomExecutionContext class that handles the underlying Akka dispatcher lookup.

§Updated Templates with Preconfigured CustomExecutionContexts

All of the Play example templates on Play’s download page that use blocking APIs (i.e. Anorm, JPA) have been updated to use custom execution contexts where appropriate. For example, going to shows that the JPAPersonRepository class takes a DatabaseExecutionContext that wraps all the database operations.

For thread pool sizing involving JDBC connection pools, you want a fixed thread pool size matching the connection pool, using a thread pool executor. Following the advice in HikariCP’s pool sizing page, you should configure your JDBC connection pool to double the number of physical cores, plus the number of disk spindles.

The dispatcher settings used here come from Akka dispatcher:

# db connections = ((physical_core_count * 2) + effective_spindle_count)
fixedConnectionPool = 9

database.dispatcher {
  executor = "thread-pool-executor"
  throughput = 1
  thread-pool-executor {
    fixed-pool-size = ${fixedConnectionPool}

§Defining a CustomExecutionContext in Scala

To define a custom execution context, subclass CustomExecutionContext with the dispatcher name:

class DatabaseExecutionContext @Inject()(system: ActorSystem)
   extends CustomExecutionContext(system, "database.dispatcher")

Then have the execution context passed in as an implicit parameter:

class DatabaseService @Inject()(implicit executionContext: DatabaseExecutionContext) {

§Defining a CustomExecutionContext in Java

To define a custom execution context, subclass CustomExecutionContext with the dispatcher name:

import play.libs.concurrent.CustomExecutionContext;

public class DatabaseExecutionContext
        extends CustomExecutionContext {

    public DatabaseExecutionContext(ActorSystem actorSystem) {
        // uses a custom thread pool defined in application.conf
        super(actorSystem, "database.dispatcher");

Then pass the JPA context in explicitly:

public class JPAPersonRepository implements PersonRepository {

    private final JPAApi jpaApi;
    private final DatabaseExecutionContext executionContext;

    public JPAPersonRepository(JPAApi jpaApi, DatabaseExecutionContext executionContext) {
        this.jpaApi = jpaApi;
        this.executionContext = executionContext;


§Play WSClient Improvements

There are substantial improvements to Play WSClient. Play WSClient is now a wrapper around the standalone play-ws implementation, which can be used outside of Play. In addition, the underlying libraries involved in play-ws have been shaded, so that the Netty implementation used in it does not conflict with Spark, Play or any other library that uses a different version of Netty.

Finally, there is now support for HTTP Caching if a cache implementation is present. Using an HTTP cache means savings on repeated requests to backend REST services, and is especially useful when combined with resiliency features such as stale-on-error and stale-while-revalidate.

For more details, please see WsCache and the WS Migration Guide.

§Play JSON improvements

§Ability to serialize tuples

Now, tuples are able to be serialized by play-json, and there are Reads and Writes implementations in the implicit scope. Tuples are serialized to arrays, so ("foo", 2, "bar") will render as ["foo", 2, "bar"] in the JSON.

§Scala.js support

Play JSON 2.6.0 now supports Scala.js. You can add the dependency with:

libraryDependencies += "" %%% "play-json" % version

where version is the version you wish to use. The library should effectively work the same as it does on the JVM, except without support for JVM types.

§Testing Improvements

Some utility classes have been added to the play.api.test package in 2.6.x to make functional testing easier with dependency injected components.


There are many functional tests that use the injector directly through the implicit app:

"test" in new WithApplication() {
  val executionContext = app.injector.instanceOf[ExecutionContext]

Now with the Injecting trait, you can elide this:

"test" in new WithApplication() with Injecting {
  val executionContext = inject[ExecutionContext]


The StubControllerComponentsFactory creates a stub ControllerComponents that can be used for unit testing a controller:

val controller = new MyController(stubControllerComponents())


The StubBodyParserFactory creates a stub BodyParser that can be used for unit testing content:

val stubParser = stubBodyParser(AnyContent("hello"))

§File Upload Improvements

Uploading files uses a TemporaryFile API which relies on storing files in a temporary filesystem, as specified in ScalaFileUpload / JavaFileUpload, accessible through the ref attribute.

Uploading files is an inherently dangerous operation, because unbounded file upload can cause the filesystem to fill up – as such, the idea behind TemporaryFile is that it’s only in scope at completion and should be moved out of the temporary file system as soon as possible. Any temporary files that are not moved are deleted.

In 2.5.x, TemporaryFile were deleted as the file references were garbage collected, using finalize. However, under certain conditions, garbage collection did not occur in a timely fashion. The background cleanup has been moved to use FinalizableReferenceQueue and PhantomReferences rather than use finalize.

The Java and Scala APIs for TemporaryFile has been reworked so that all TemporaryFile references come from a TemporaryFileCreator trait, and the implementation can be swapped out as necessary, and there’s now an atomicMoveWithFallback method that uses StandardCopyOption.ATOMIC_MOVE if available.


There’s also now a play.api.libs.Files.TemporaryFileReaper that can be enabled to delete temporary files on a scheduled basis using the Akka scheduler, distinct from the garbage collection method.

The reaper is disabled by default, and is enabled through application.conf:

play.temporaryFile {
  reaper {
    enabled = true
    initialDelay = "5 minutes"
    interval = "30 seconds"
    olderThan = "30 minutes"

The above configuration will delete files that are more than 30 minutes old, using the “olderThan” property. It will start the reaper five minutes after the application starts, and will check the filesystem every 30 seconds thereafter. The reaper is not aware of any existing file uploads, so protracted file uploads may run into the reaper if the system is not carefully configured.

Next: Migration Guides

Found an error in this documentation? The source code for this page can be found here. After reading the documentation guidelines, please feel free to contribute a pull request.