Getting started with Serenity BDD and Cucumber 4

John Ferguson Smart | Mentor | Author | Speaker - Author of 'BDD in Action'.
Helping teams deliver more valuable software sooner6th February 2019

Serenity BDD is a library that makes it easier to write high quality automated acceptance tests, with powerful reporting and living documentation features. It has strong support for both web testing with Selenium, and API testing using RestAssured.

Serenity strongly encourages good test automation design, and supports several design patterns, including classic Page Objects, the newer Lean Page Objects/ Action Classes approach, and the more sophisticated and flexible Screenplay pattern.

The latest version of Serenity supports both Cucumber 2.4 and the more recent Cucumber 4.x. Cucumber 4 is not backward compatible with Cucumber 2. This article walks you through how to get started with Serenity and Cucumber 4, and also gives you a quick introduction to some of Cucumber 4’s new features.

The starter project

The best place to start with Serenity and Cucumber is to clone or download the starter project on Github ( This project gives you a basic project setup, along with some sample tests and supporting classes. There are two versions to choose from. The master branch uses a more classic approach, using action classes and lightweight page objects, whereas the screenplay branch shows the same sample test implemented using Screenplay.

The project directory structure

The project has build scripts for both Maven and Gradle, and follows the standard directory structure used in most Serenity projects:

  + main
  + test
    + java                        Test runners and supporting code
    + resources
      + features                  Feature files
    + search                      Feature file subdirectories

  + webdriver                     Bundled webdriver binaries
    + linux
    + mac
    + windows
       chromedriver.exe           OS-specific Webdriver binaries

Adding the Cucumber 4 dependency

Serenity seamlessly supports both Cucumber 2.x and Cucumber 4. However, this flexibility requires a little tweaking in the build dependencies.

If you are using Maven, you need to do the following:

  • exclude the default cucumber-core dependency from your serenity-core dependency
  • Replace your serenity-cucumber dependency with the serenity-cucumber4 dependency
  • Add dependencies on the Cucumber 4.x version of cucumber-java and cucumber-junit into your project

An example of the correctly configured dependencies is shown below:


If you are using Gradle, you need to ensure that the 4.x version of cucumber-core is used using the resolutionStrategy element, and also add the Cucumber 4.x version of cucumber-java and cucumber-junit dependencies as mentioned above:

configurations.all {
    resolutionStrategy {
        force "io.cucumber:cucumber-core:4.2.0"

dependencies {
    testCompile "net.serenity-bdd:serenity-core:2.0.39",

In the rest of this article, we will walk through some of the highlights of both versions. Let’s start off with the version on the master branch, which uses lightweight page objects and actions.

The Cucumber 4 sample scenario

Both variations of the sample project uses the sample Cucumber scenario. In this scenario, Sergey (who likes to search for stuff) is performing a search on the DuckDuckGo search engine:

Feature: Search by keyword

  Scenario: Searching for a term
    Given Sergey is on the DuckDuckGo home page
    When he searches for "cucumber"
    Then all the result titles should contain the word "cucumber"

This scenario lets us explore a few of the new Cucumber 4 expressions. Cucumber 4 supports both the classic regular expressions, and the new Cucumber Expressions, which are more readable albeit not as powerful in some cases.

The glue code for this scenario uses both regular expressions and cucumber expressions. The glue code looks this this:

    @Given("^(?:.*) is on the DuckDuckGo home page")  
    public void i_am_on_the_DuckDuckGo_home_page() {

    @When("(s)he searches for {string}")               
    public void i_search_for(String term) {

    @Then("all the result titles should contain the word {string}")
    public void all_the_result_titles_should_contain_the_word(String term) {
                .matches(results -> results.size() > 0)
                .allMatch(title ->  

The @Given step uses a regular expression; the action class approach we use here is action-centric, not actor-centric, so we ignore the name of the actor.

The @When and @Then steps uses Cucumber expressions, and highlights two useful features. Rather than using a regular expression to match the search term, we use the more readable Cucumber expression {string}. This matches a single or double-quoted string (the quotes themselves are dropped). Cucumber 4 also supports other typed expressions, such as {int}, {word}, and {float}.

Parentheses can be used to indicate optional text, so “(s)he” will match both “he” and “she”. We could also write this using a slash: “she/he”.

Lean Page Objects and Action Classes

The glue code shown above uses Serenity step libraries as action classes to make the tests easier to read, and to improve maintainability.

These classes are declared using the Serenity @Steps annotation, shown below:

    NavigateTo navigateTo;

    SearchFor searchFor;

    SearchResult searchResult;

The @Stepsannotation tells Serenity to create a new instance of the class, and inject any other steps or page objects that this instance might need.

Each action class models a particular facet of user behaviour: navigating to a particular page, performing a search, or retrieving the results of a search. These classes are designed to be small and self-contained, which makes them more stable and easier to maintain.

The NavigateTo class is an example of a very simple action class. In a larger application, it might have some other methods related to high level navigation, but in our sample project, it just needs to open the DuckDuckGo home page:

public class NavigateTo {

    DuckDuckGoHomePage duckDuckGoHomePage;

    @Step("Open the DuckDuckGo home page")
    public void theDuckDuckGoHomePage() {;

It does this using a standard Serenity Page Object. Page Objects are often very minimal, storing just the URL of the page itself:

class DuckDuckGoHomePage extends PageObject {}

The second class, SearchFor, is an interaction class. It needs to interact with the web page, and to enable this, we make the class extend the Serenity UIInteractionSteps. This gives the class full access to the powerful Serenity WebDriver API, including the $() method used below, which locates a web element using a By locator or an XPath or CSS expression:

public class SearchFor extends UIInteractionSteps {

    @Step("Search for term {0}")
    public void term(String term) {

The SearchForm class is typical of a light-weight Page Object: it is responsible uniquely for locating elements on the page, and it does this by defining locators or occasionally by resolving web elements dynamically.

class SearchForm {
    static By SEARCH_FIELD = By.cssSelector(".js-search-input");
    static By SEARCH_BUTTON = By.cssSelector(".js-search-button");

The last step library class used in the step definition code is the SearchResult class. The job of this class is to query the web page, and retrieve a list of search results that we can use in the AssertJ assertion at the end of the test. This class also extends UIInteractionSteps and

public class SearchResult extends UIInteractionSteps {
    public List<String> titles() {
        return findAll(SearchResultList.RESULT_TITLES)

The SearchResultList class is a lean Page Object that locates the search result titles on the results page:

class SearchResultList {
    static By RESULT_TITLES = By.cssSelector(".result__title");

The main advantage of the approach used in this example is not in the lines of code written, although Serenity does reduce a lot of the boilerplate code that you would normally need to write in a web test. The real advantage is in the use of many small, stable classes, each of which focuses on a single job. This application of the Single Responsibility Principle goes a long way to making the test code more stable, easier to understand, and easier to maintain.

The Screenplay starter project

If you prefer to use the Screenplay pattern, or want to try it out, check out the screenplay branch instead of the master branch. In this version of the starter project, the same scenario is implemented using the Screenplay pattern.

The Screenplay pattern describes tests in terms of actors and the tasks they perform. Tasks are represented as objects performed by an actor, rather than methods. This makes them more flexible and composable, at the cost of being a bit more wordy. Here is an example:

    public void setTheStage() {
        OnStage.setTheStage(new OnlineCast());

    @Given("^(.*) is on the DuckDuckGo home page")
    public void on_the_DuckDuckGo_home_page(String actor) {

    @When("she/he searches for {string}")
    public void search_for(String term) {

    @Then("all the result titles should contain the word {string}")
    public void all_the_result_titles_should_contain_the_word(String term) {
                seeThat("search result titles",
                seeThat("search result titles",

In both approaches, the Page Objects very close or identical. The differences are mainly in the action classes. Screenplay classes emphasise reusable components and a very readable declarative style, whereas Lean Page Objects and Action Classes opt for a more imperative style.

The NavigateTo class performs the same role as it’s equivalent in the Lean Page Object/Action Class version, and looks quite similar:

public class NavigateTo  {

    public static Performable theDuckDuckGoHomePage() {
        return Task.where("{0} opens the DuckDuckGo home page",

The SearchFor class is also similar: it is shown below:

public class SearchFor {

    public static Performable term(String term) {
        return Task.where("{0} attempts to search for #term",

In Screenplay, there is a clear distinction between actions (which change the system state) and questions (which read the system state). In Screenplay, we fetch the search results using a Question class, like this:

public class SearchResult {
    public static Question<List<String>> titles() {
        return actor ->  

The Screenplay DSL is rich and flexible, and well suited to teams working on large test automation projects with many team members, and who are reasonably comfortable with Java and design patterns. The Lean Page Objects/Action Classes approach proposes a gentler learning curve, but still provides significant advantages in terms of maintainability and reusability.

Executing the tests

To run the sample project, you can either just run the CucumberTestSuite test runner class, or run either mvn verify or gradle test from the command line.

By default, the tests will run using Chrome. You can run them in Firefox by overriding the driver system property, e.g.

$ mvn clean verify -Ddriver=firefox


$ gradle clean test -Pdriver=firefox

The test results will be recorded in the target/site/serenity directory.

Simplified WebDriver configuration and other Serenity extras

The sample projects both use some Serenity features which make configuring the tests easier. In particular, Serenity uses the serenity.conf file in the src/test/resources directory to configure test execution options.

Webdriver configuration

The WebDriver configuration is managed entirely from this file, as illustrated below:

webdriver {
    driver = chrome
headless.mode = true


The project also bundles some of the WebDriver binaries that you need to run Selenium tests in the src/test/resources/webdriver directories. These binaries are configured in the drivers section of the serenity.conf config file:

drivers {
  windows { = "src/test/resources/webdriver/windows/chromedriver.exe"
    webdriver.gecko.driver = "src/test/resources/webdriver/windows/geckodriver.exe"
  mac { = "src/test/resources/webdriver/mac/chromedriver"
    webdriver.gecko.driver = "src/test/resources/webdriver/mac/geckodriver"
  linux { = "src/test/resources/webdriver/linux/chromedriver"
    webdriver.gecko.driver = "src/test/resources/webdriver/linux/geckodriver"

This configuration means that development machines and build servers do not need to have a particular version of the WebDriver drivers installed for the tests to run correctly.

Environment-specific configurations

We can also configure environment-specific properties and options, so that the tests can be run in different environments. Here, we configure three environments, dev, staging and prod, with different starting URLs for each:

environments {
  default {
    webdriver.base.url = ""
  dev {
    webdriver.base.url = ""
  staging {
    webdriver.base.url = ""
  prod {
    webdriver.base.url = ""

You use the environment system property to determine which environment to run against. For example to run the tests in the staging environment, you could run:

$ mvn clean verify -Denvironment=staging

See this article for more details about this feature.

Want to learn more?

For more information about Serenity BDD, you can read the Serenity BDD Book, the official online Serenity documentation source. Other sources include:

© 2019 John Ferguson Smart