Await, do not make your E2E tests sleep

Sleeping: the worst practice in E2E testing and the concept of Deterministic Events.

I’m working on a big UI Testing Best Practices project on GitHub, I share this post to spread it and have direct feedback about one of the chapters.

When testing your UI, you define a sort of key points the app must pass through. Reaching these key points is an asynchronous process because, almost 100% of the times, your UI does not update synchronously. Those key points are called deterministic events, as known as something that you know that must happen.

It depends on the events you define and how your UI reaches them but, usually, there are some “long” waitings, like XHR requests, and some faster ones, like re-render updates.

The solution to the asynchronous updates seems handy: sleeping/pausing the test for a bunch of milliseconds, tenths of a second, or even seconds. It can make your test working because it gives the app the time to update itself and moving to the next deterministic event to be tested.

Consider that, except for specific and known waitings (like when you use setInterval or setTimeout), it’s totally unpredictable how much the sleeping time should be because it could depend on:

- the network state (for XHR requests)

- the total amount of available machine resources (CPU, RAM, etc.)

— a CI pipeline can limit them for example

— other apps can consume them on your local machine too

- the concurrence of other resource-consuming updates (canvas, etc.)

- a bunch of unpredictable game players like Service Workers, cache management, etc. that can make the UI update process faster or slower

Every fixed delay leads your test to be more brittle and increasing its duration. You are going to find a balance between false negatives — when the test fails because of a too low sleeping time — and exaggerate test duration. What about waiting just the right amount of time? The amount of time that makes the test as fast as possible!

Waitings fall in four main categories:

- page Load waitings: the first waiting to manage while testing your app, waiting for an event that allows you to understand that the page is interactive

- content waitings: waiting for DOM element that matches a selector

- XHR request waitings: waiting for an XHR request start or the corresponding response received

- custom waitings: waiting for everything strictly related to the app that does not fall into the above categories

Every UI testing tool manages waitings in different ways, sometimes automatically and sometimes manually. Below you can find some examples of implementing the listed waitings.

Page load waitings

Every E2E testing tool manages the page load waiting in a different way (in terms of what is waited before considering the page loaded).

Cypress

cy.visit('https://localhost:3000')

Puppeteer (and Playwright)

await page.goto('https://localhost:3000')

Selenium

driver.get('https://localhost:3000')
driver.wait(function() {
  return driver
    .executeScript('return document.readyState')
    .then(function(readyState) {
      return readyState === 'complete'
    })
})

TestCafé

fixture`Page load`.page`https://localhost:3000`

Content waitings

Take a look at the following examples to see how waiting for a DOM element could be implemented in the available tools.

Cypress

• waiting for an element:

// it waits up to 4 seconds by default
cy.get('#form-feedback')
// the timeout can be customized
cy.get('#form-feedback', {timeout: 5000})

• waiting for an element with specific content:

cy.get('#form-feedback').contains('Success')

Puppeteer (and Playwright)

• waiting for an element:

// it waits up to 30 seconds by default
await page.waitForSelector('#form-feedback')
// the timeout can be customized
await page.waitForSelector('#form-feedback', {timeout: 5000})

• waiting for an element with specific content:

await page.waitForFunction(selector => {
  const el = document.querySelector(selector)
  return el && el.innerText === 'Success'
}, {}, '#form-feedback')

Selenium

• waiting for an element:

driver.wait(until.elementLocated(By.id('#form-feedback')), 4000)

• waiting for an element with specific content:

const el = driver.wait(until.elementLocated(By.id('#form-feedback')), 4000)
wait.until(ExpectedConditions.textToBePresentInElement(el, 'Success'))

TestCafé

• waiting for an element:

// it waits up to 10 seconds by default
await Selector('#form-feedback')
// the timeout can be customized
await Selector('#form-feedback').with({timeout: 4000})

• waiting for an element with specific content:

await Selector('#form-feedback').withText('Success')

DOM Testing Library

• waiting for an element:

await findByTestId(document.body, 'form-feedback')

• waiting for an element with specific content:

const container = await findByTestId(document.body, 'form-feedback');
await findByText(container, 'Success')

XHR request waitings

Even if it’s a really important point, at the time of writing (July 2019) it seems that waiting for XHR requests and responses is not so common.

With exceptions for Cypress and Puppeteer, other tools/frameworks force you to look for something in the DOM that reflects the XHR result instead of looking for the XHR request itself.

You can read more about the topic:

? Selenium WebDriver: 5 Ways to Test AJAX Calls in Selenium WebDriver

? TestCafé: Wait Mechanism for XHR and Fetch Requests

? DOM Testing Library: await API

Cypress

• waiting for an XHR request/response

cy.server()
cy.route('https://dummy.restapiexample.com/api/v1/employees').as('employees')
cy.wait('@employees')
  .then(xhr => xhr.response.body)
  .then(body => {
    /* ... */
  })

Puppeteer (and Playwright)

• waiting for an XHR request:

await page.waitForRequest('https://dummy.restapiexample.com/api/v1/employees')

• waiting for an XHR response:

const response = await page.waitForResponse('https://dummy.restapiexample.com/api/v1/employees')
const body = response.json()

Custom waitings

The various UI testing tools/frameworks have built-in solutions to perform a lot of checks, but let’s concentrate on writing a custom waiting. Since UI testing is 100% asynchronous, a custom waiting should face recursive promises, a concept not so handy to manage at the beginning.

Luckily, there are some handy solutions and plugins to help us with that. Consider if we want to wait until a global variable (foo) is assigned with a particular value (bar): below you are going to find some examples.

Cypress

Thanks to the cypress-wait-until plugin you can do:

cy.waitUntil(() => cy.window().then(win => win.foo === 'bar'))

Puppeteer (and Playwright)

await page.waitForFunction('window.foo === "bar"')

Selenium

browser.executeAsyncScript(`
  window.setTimeout(function(){
    if(window.foo === "bar") {
      arguments[arguments.length - 1]();
    }
  }, 300);
`)

TestCafé

const waiting = ClientFunction(() => window.foo === 'bar')
await t.expect(waiting()).ok({ timeout: 5000 })

DOM Testing Library

await wait(() => global.foo === 'bar')

Some notes:

• unlike Cypress, Puppeteer, etc. DOM Testing Library is quite a different tool, that’s why the examples are not available for every single part.
• if there are better solutions or plugins to do the same, please let me know! I know Cypress, Puppeteer, and DOM Testing Library pretty well, but I can not say the same for Selenium and TestCafé.

Crossposting from Medium.

Hi ?? I’m Stefano Magni, I’m a passionate JavaScript developer and a Cypress Ambassador. I love creating high-quality products, testing and automating everything, learning and sharing my knowledge, helping people, speaking at conferences and facing new challenges. I work for Conio, an Italian Bitcoin startup.

You can find me on Twitter and GitHub. You can find all my recent contributions/talks etc. here.