Playwright and Cucumber.js BDD tutorial

A tutorial for setting up automated BDD cross-browser functional tests using Cucumber.js, Playwright and TypeScript

Published
Words
1465
Reading time
6 minutes

We're going write automated, functional, browser-based tests in Gherkin syntax. We'll run them using Cucumber.js. And we'll write step definitions in TypeScript that use Playwright to automate browsers. That's a lot of words and tech, so first let's explain some things.

What is Playwright?

Well, to quote the docs:

Playwright is a Node.js library to automate Chromium, Firefox and WebKit with a single API

But what does that mean? On a really basic level it means you can write JavaScript code to control a browser. Playwright is bit like a modern alternative to selenium and a more cross-browser alternative to puppeteer. Selenium (and WebDriver) has been the go-to tool for functional, browser-based test automation for a while now. But Playwright is a bit different and is definitely worth a look. It's backed by Microsoft so will be around for a while and it's now fairly stable.

It's got some interesting features baked in like emulating mobile devices along with things like geolocation. You can also intercept network activity for stubbing, modifying and mocking network requests. If you've ever tried to do stuff like that with selenium (WebDriver) you'll have found it's hacky. We resorted to installing ModHeader in Chrome in one scenario!

It's worth noting Playwright is a browser automation tool and not a full test automation framework. It's easy enough to hook Playwright up to a test runner like Jest, AVA or Mocha.

What is Cucumber.js?

Let's quote the docs again:

Cucumber is a tool for running automated tests written in plain language...Cucumber.js is the JavaScript implementation of Cucumber and runs on Node and modern web browsers.

It allows you to write and run automated tests written in Gherkin, a human readable language. These tests consist of feature files written in Gherkin. The features consist of multiple steps. These steps are then translated into actions by step definitions. In Cucumber.js, these step definitions are witten in JavaScript, or TypeScript.

It's these step definitions we'll use to translate Cucumber.js steps into Playwright commands.

Alternatives

I mentioned above that Playwright is a browser automation tool and not a full test automation framework. If you're new to test automation then it's worth mentioning a few frameworks you should look at as alternatives.

We're going to run cucumber directly, but there's also jest-cucumber. This is a project that aims for the best of both between Jest and Cucumber. This could be used to hook up step definitions to Playwright.

If you're looking at BDD but aren't bothered about Playwright then there are plenty of Selenium, WebDriver API or Puppeteer based frameworks. Things like WebdriverIO, Nightwatch.js, Cypress and Testcafe. These are all worth looking at if you're new to testing, and there are no doubt others, and these are just the JavaScript based ones!

The code

Enough talk. Let's write some code. We're going to do the following:

  • Use Cucumber.js as a test runner
  • Write feature files in Gherkin syntax
  • Write custom step definitions in TypeScript
  • Use Playwright from the step definitions to automate the browser
  • Use Node's built in assert to module to make assertions.

So let's get started.

npm project setup

Install Node 10.17 or above, this is what Playwright requires. It's usually best to install the LTS (Long Term Support) version of Node.

Create a new, empty folder and run npm init to start a new npm project. Or use the -y option (shorthand for --yes) to skip the questions and just use the default config: npm init -y. This will create a package.json file.

Next up, install our dependencies from npm:

npm i playwright @cucumber/cucumber typescript ts-node @types/node -D

Note the -D is short hand for --save-dev. We don't need separate @types dependencies for Cucumber and Playwright as they both come bundled with TypeScript definitions.

Now we have our project scaffolded we can setup TypeScript and Cucumber.js. Then we'll write our features and step definitions. But first we need to setup TypeScript.

Setup TypeScript

Run npx -p typescript tsc --init on the command line to generate a tsconfig.json. If you've never used npx, then no, it's not a spelling mistake. npx is a command that comes with npm 5.2 onwards that allows you to execute a command from an npm package, without having to install it globally first. The -p command is short for package and it's needed here because the command (tsc) is different from the package (typescript).

The generated tsconfig.json contains all the possible config options, with most of them commented out. It also contains a simple explanation of each one. So it's a good way to set up get started with TypeScript config.

Disable type checking and just do transpilation for ts-node. We can use a separate task just for type checking, so the transiplation is quicker when we're running the tests. Add the following to the tsconfig.json:

{
+ "ts-node": {
+ "transpileOnly": true
+ },
"compilerOptions": {
}

Now that we have TypeScript configured, we need to set up Cucumber.js.

Setup Cucumber.js

We'll set up cucumber to:

  • load ts-node/register to allow us to write step definitions in TypeScript
  • load a setup file (test.setup.ts) to configure Playwright
  • load step definitions from a step-definitions directory
  • run feature files from a features directory.

Add the following command to your package.json. This will allow us to run npm test on the command line:

{
+ "scripts": {
+ "test": "cucumber-js features/**/*.feature --require-module ts-node/register --require test.setup.ts --require step-definitions/**/*.ts"
+ }
}

Before we can run the command, we need to add the test setup file, step definitions and feature files. Let's start with the test setup.

Test setup file

We'll make use of two Cucumber.js features to set up our tests: Before, After, BeforeAll and AfterAll hooks and the 'World' instance. Scenario hooks allow you to run code before and after test runs and are a great place to intialise things. The World instance is, to quote the Cucumber.js docs:

an isolated context for each scenario, exposed to the hooks and steps as this.

In our case, the initialisation is the Playwright instances - the browser, context and page, the core concepts of Playwright. Browser instances in Playwright are expensive so we'll only create an instance once for all tests. We'll use a BeforeAll for this and an AfterAll to destory it. Contexts and pages in Playwright are cheap so we'll use a new one per test with a Before (and After) hook. We'll store the context and page on the World instance so they're available in each step definition as this.page and this.context.

But first, we need to create a TypeScript type that represents our World instance. We'll extend the built in World type from Cucumber. Create a types.ts file with the following:

import { World as CucumberWorld } from "@cucumber/cucumber";
import { BrowserContext, Page } from "playwright";
export interface OurWorld extends CucumberWorld {
context: BrowserContext;
page: Page;
}

Next, we need to setup Playwright. Create a test.setup.ts with the following. It's got comments throughout which should explain what it's doing:

// test.setup.ts
import { Before, BeforeAll, AfterAll, After } from "@cucumber/cucumber";
import { devices, chromium } from "playwright";
import { OurWorld } from "./types";
BeforeAll(async function () {
// Browsers are expensive in Playwright so only create 1
global.browser = await chromium.launch({
// Not headless so we can watch test runs
headless: false,
// Slow so we can see things happening
slowMo: 50,
});
});
AfterAll(async function () {
await global.browser.close();
});
// Create a new test context and page per scenario
Before(async function (this: OurWorld) {
const pixel2 = devices["Pixel 2"];
this.context = await global.browser.newContext({
viewport: pixel2.viewport,
userAgent: pixel2.userAgent,
});
this.page = await this.context.newPage();
});
// Cleanup after each scenario
After(async function (this: OurWorld) {
await this.page.close();
await this.context.close();
});

Step definitions can then access this.browser, this.context and this.page. This makes use of a TypeScript feature for defining the type of this in a function. But be warned: because we're using this in step definitions, we have to use proper functions rather than arrow functions. Arrow functions use what's known as 'lexical scoping', that is this uses normal variable rules to get the value for this based on where the function is defined.

Now that we've got a type definition for our World instance and have setup Playwright, we can write our step definitions.

Write a step definition

A step definition is a an expression in combination with a JavaScript (or in our case, TypeScript) function. The expression forms the text that will go in the feature file, and the function is what is called when that step executes. We'll use the function to control the browser using Playwright.

Create a step-definitions folder, and a file called homepage.ts. This file can actually be called anything with a ts extension: our Cucumber command uses a glob to load all step definitions: --require step-definitions/**/*.ts.

We'll write a simple use case of 3 step definitions with given, when, then. We'll use a cucumber expression, a regular expression and a string to show the different types. Again the following is commented to explain what's going on.

// step-definitions/homepage.ts
import { Given, When, Then } from "@cucumber/cucumber";
import assert from "assert";
import { OurWorld } from "../types";
// Using a cucumber expression
Given("I view {string}", async function (this: OurWorld, url: string) {
// Use the page instance from the World instance to navigate
await this.page.goto(`https://${url}`);
});
// Using a regular expression
When(/^I click '([^']*)'$/, async function (this: OurWorld, text: string) {
// Scroll to the link...
await this.page.$eval(`"${text}"`, (element) => {
element.scrollIntoView();
});
// ...then click it now it's within the viewport
await this.page.click(`"${text}"`);
});
Then("I expect to be on the accessibility page", async function (
this: OurWorld
) {
const heading1Text = (await this.page.textContent("h1")) as string;
assert.strictEqual(
trimExcessWhiteSpace(heading1Text),
"Accessibility statement"
);
});
// textContent includes whitespace, so use this method to trim
// See https://stackoverflow.com/a/42921059
const trimExcessWhiteSpace = (s: string) =>
s.replace(/[\n\r]+|[\s]{2,}/g, " ").trim();

Now we have our step definitions, we can now combine them into a feature file to write a test.

Write a feature

Our feature file will use the step definitions we've just created. We'll run a test against the gov.uk homepage, to assert that you can click the link in the footer to the accessibiltiy statement.

Create a file called homepage.feature in the features directory, and paste the following:

Feature: Random
A random feature using some Playwright stuff
Scenario: Govuk accessibility statement link
Given I view 'www.gov.uk'
When I click 'Accessibility statement'
Then I expect to be on the accessibility page

That's not the most exciting test in the world but hopefully it shows enough of how to use Cucumber.js.

Now it's all hooked up, run npm test on the command line and you should see something like:

1 scenario (1 passed) 3 steps (3 passed)

And that's it - automated tests running in the browser!

Next steps

I'm sure this example can be improved, after all it's a fairly basic setup. But 2 improvements jump out immediately:

  • separate our expressions from our functions in our step definitions. The convention is to use a support folder to house the different functions for re-use
  • use cucumber profiles to pass in different config, to target different browsers

I hope this tutorial has been useful - let me know @ediblecode on Twitter.