Best Practices

Organizing Tests, Logging In, Controlling State

In February 2018 we gave a "Best Practices" conference talk at AssertJS. This video demonstrates how to approach writing fast, scalable tests.

https://www.youtube.com/watch?v=5XQOK0v_YRE

We have several Logging in recipes in our examples.

Selecting Elements

Every test you write will include selectors for elements. To save yourself a lot of headaches, you should write selectors that are resilient to changes.

Oftentimes we see users run into problems targeting their elements because:

  • Your application may use dynamic classes or ID's that change
  • Your selectors break from development changes to CSS styles or JS behavior

Luckily, it is possible to avoid both of these problems.

  1. Don't target elements based on CSS attributes such as: id, class, tag
  2. Don't target elements that may change their textContent
  3. Add data-* attributes to make it easier to target elements

How It Works:

Given a button that we want to interact with:

<button
  id="main"
  class="btn btn-large"
  name="submission"
  role="button"
  data-cy="submit"
>
  Submit
</button>

Let's investigate how we could target it:

SelectorRecommendedNotes
cy.get('button').click() NeverWorst - too generic, no context.
cy.get('.btn.btn-large').click() NeverBad. Coupled to styling. Highly subject to change.
cy.get('#main').click() SparinglyBetter. But still coupled to styling or JS event listeners.
cy.get('[name=submission]').click() SparinglyCoupled to the name attribute which has HTML semantics.
cy.contains('Submit').click() DependsMuch better. But still coupled to text content that may change.
cy.get('[data-cy=submit]').click() AlwaysBest. Isolated from all changes.

Targeting the element above by tag, class or id is very volatile and highly subject to change. You may swap out the element, you may refactor CSS and update ID's, or you may add or remove classes that affect the style of the element.

Instead, adding the data-cy attribute to the element gives us a targeted selector that's only used for testing.

The data-cy attribute will not change from CSS style or JS behavioral changes, meaning it's not coupled to the behavior or styling of an element.

Additionally, it makes it clear to everyone that this element is used directly by test code.

Real World Example

The Real World App (RWA) uses two useful custom commands for selecting elements for testing:

  • getBySel yields elements with a data-test attribute that match a specified selector.
  • getBySelLike yields elements with a data-test attribute that contains a specified selector.
// cypress/support/commands.ts

Cypress.Commands.add('getBySel', (selector, ...args) => {
  return cy.get(`[data-test=${selector}]`, ...args)
})

Cypress.Commands.add('getBySelLike', (selector, ...args) => {
  return cy.get(`[data-test*=${selector}]`, ...args)
})

Source: cypress/support/commands.ts

Text Content:

After reading the above rules you may be wondering:

If I should always use data attributes, then when should I use cy.contains()?

A rule of thumb is to ask yourself this:

If the content of the element changed would you want the test to fail?

  • If the answer is yes: then use cy.contains()
  • If the answer is no: then use a data attribute.

Example:

If we looked at the <html> of our button again...

<button id="main" class="btn btn-large" data-cy="submit">Submit</button>

The question is: how important is the Submit text content to your test? If the text changed from Submit to Save - would you want the test to fail?

If the answer is yes because the word Submit is critical and should not be changed - then use cy.contains() to target the element. This way, if it is changed, the test will fail.

If the answer is no because the text could be changed - then use cy.get() with data attributes. Changing the text to Save would then not cause a test failure.

Assigning Return Values

Many first time users look at Cypress code and think it runs synchronously.

We see new users commonly write code that looks like this:

// DONT DO THIS. IT DOES NOT WORK
// THE WAY YOU THINK IT DOES.
const a = cy.get('a')

cy.visit('https://example.cypress.io')

// nope, fails
a.first().click()

If you are new to Cypress and wanting to better understand how Commands work - please read our Introduction to Cypress guide.

If you're familiar with Cypress commands already, but find yourself using const, let, or var then you're typically trying to do one of two things:

  • You're trying to store and compare values such as text, classes, attributes.
  • You're trying to share values between tests and hooks like before and beforeEach.

For working with either of these patterns, please read our Variables and Aliases guide.

Visiting external sites

One of the first things many of our users attempt to do is involve 3rd party servers in their tests.

You may want to access 3rd party servers in several situations:

  1. Testing log in when your app uses another provider via OAuth.
  2. Verifying your server updates a 3rd party server.
  3. Checking your email to see if your server sent a "forgot password" email.

Initially you may be tempted to use cy.visit() or use Cypress to traverse to the 3rd party login window.

However, you should never use your UI or visit a 3rd party site when testing because:

  • It is incredibly time consuming and slows down your tests.
  • The 3rd party site may have changed or updated its content.
  • The 3rd party site may be having issues outside of your control.
  • The 3rd party site may detect you are testing via a script and block you.
  • The 3rd party site may be running A/B campaigns.

Let's look at a few strategies for dealing with these situations.

When logging in:

Many OAuth providers run A/B experiments, which means that their login screen is dynamically changing. This makes automated testing difficult.

Many OAuth providers also throttle the number of web requests you can make to them. For instance, if you try to test Google, Google will automatically detect that you are not a human and instead of giving you an OAuth login screen, they will make you fill out a captcha.

Additionally, testing through an OAuth provider is mutable - you will first need a real user on their service and then modifying anything on that user might affect other tests downstream.

Here are potential solutions to alleviate these problems:

  1. Stub out the OAuth provider and bypass using their UI altogether. You could trick your application into believing the OAuth provider has passed its token to your application.
  2. If you must get a real token you can use cy.request() and use the programmatic API that your OAuth provider provides. These APIs likely change more infrequently and you avoid problems like throttling and A/B campaigns.
  3. Instead of having your test code bypass OAuth, you could also ask your server for help. Perhaps all an OAuth token does is generate a user in your database. Oftentimes OAuth is only useful initially and your server establishes its own session with the client. If that is the case, use cy.request() to get the session directly from your server and bypass the provider altogether.

3rd party servers:

Sometimes actions that you take in your application may affect another 3rd party application. These situations are not that common, but it is possible. Imagine your application integrates with GitHub and by using your application you can change data inside of GitHub.

After running your test, instead of trying to cy.visit() GitHub, you can use cy.request() to programmatically interact with GitHub's APIs directly.

This avoids ever needing to touch the UI of another application.

Verifying sent emails:

Typically, when going through scenarios like user registration or forgotten passwords, your server schedules an email to be delivered.

  1. If your application is running locally and is sending the emails directly through an SMTP server, you can use a temporary local test SMTP server running inside Cypress Test Runner. Read the blog post "Testing HTML Emails using Cypress" for details.
  2. If your application is using a 3rd party email service, or you cannot stub the SMTP requests, you can use a test email inbox with an API access. Read the blog post "Full Testing of HTML Emails using SendGrid and Ethereal Accounts" for details.

Cypress can even load the received HTML email in its browser to verify the email's functionality and visual style:

The test finds and clicks the Confirm registration button
  1. In other cases, you should try using cy.request() command to query the an endpoint on your server that tells you what email has been queued or delivered. That would give you a programmatic way to know without involving the UI. Your server would have to expose this endpoint.
  2. You could also use cy.request() to a 3rd party email recipient server that exposes an API to read off emails. You will then need the proper authentication credentials, which your server could provide, or you could use environment variables. Some email services already provide Cypress plugins to access emails.

Having tests rely on the state of previous tests

You only need to do one thing to know whether you've coupled your tests incorrectly, or if one test is relying on the state of a previous one.

Change it to it.only on the test and refresh the browser.

If this test can run by itself and pass - congratulations you have written a good test.

If this is not the case, then you should refactor and change your approach.

How to solve this:

  • Move repeated code in previous tests to before or beforeEach hooks.
  • Combine multiple tests into one larger test.

Let's imagine the following test that is filling out the form.

// an example of what NOT TO DO
describe('my form', () => {
  it('visits the form', () => {
    cy.visit('/users/new')
  })

  it('requires first name', () => {
    cy.get('#first').type('Johnny')
  })

  it('requires last name', () => {
    cy.get('#last').type('Appleseed')
  })

  it('can submit a valid form', () => {
    cy.get('form').submit()
  })
})

What's wrong with the above tests? They are all coupled together!

If you were to change it to it.only on any of the last three tests, they would fail. Each test requires the previous to run in a specific order in order to pass.

Here's 2 ways we can fix this:

1. Combine into one test

// a bit better
describe('my form', () => {
  it('can submit a valid form', () => {
    cy.visit('/users/new')

    cy.log('filling out first name') // if you really need this
    cy.get('#first').type('Johnny')

    cy.log('filling out last name') // if you really need this
    cy.get('#last').type('Appleseed')

    cy.log('submitting form') // if you really need this
    cy.get('form').submit()
  })
})

Now we can put an .only on this test and it will run successfully irrespective of any other test. The ideal Cypress workflow is writing and iterating on a single test at a time.

2. Run shared code before each test

describe('my form', () => {
  beforeEach(() => {
    cy.visit('/users/new')
    cy.get('#first').type('Johnny')
    cy.get('#last').type('Appleseed')
  })

  it('displays form validation', () => {
    cy.get('#first').clear() // clear out first name
    cy.get('form').submit()
    cy.get('#errors').should('contain', 'First name is required')
  })

  it('can submit a valid form', () => {
    cy.get('form').submit()
  })
})

This above example is ideal because now we are resetting the state between each test and ensuring nothing in previous tests leaks into subsequent ones.

We're also paving the way to make it less complicated to write multiple tests against the "default" state of the form. That way each test stays lean but each can be run independently and pass.

Creating "tiny" tests with a single assertion

We've seen many users writing this kind of code:

describe('my form', () => {
  before(() => {
    cy.visit('/users/new')
    cy.get('#first').type('johnny')
  })

  it('has validation attr', () => {
    cy.get('#first').should('have.attr', 'data-validation', 'required')
  })

  it('has active class', () => {
    cy.get('#first').should('have.class', 'active')
  })

  it('has formatted first name', () => {
    cy.get('#first').should('have.value', 'Johnny') // capitalized first letter
  })
})

While technically this runs fine - this is really excessive, and not performant.

Why you did this pattern in unit tests:

  • When assertions failed you relied on the test's title to know what failed
  • You were told that adding multiple assertions was bad and accepted this as truth
  • There was no performance penalty splitting up multiple tests because they run really fast

Why you shouldn't do this in Cypress:

  • Writing integration tests is not the same as unit tests
  • You will always know (and can visually see) which assertion failed in a large test
  • Cypress runs a series of async lifecycle events that reset state between tests
  • Resetting tests is much slower than adding more assertions

It is common for tests in Cypress to issue 30+ commands. Because nearly every command has a default assertion (and can therefore fail), even by limiting your assertions you're not saving yourself anything because any single command could implicitly fail.

How you should rewrite those tests:

describe('my form', () => {
  before(() => {
    cy.visit('/users/new')
  })

  it('validates and formats first name', () => {
    cy.get('#first')
      .type('johnny')
      .should('have.attr', 'data-validation', 'required')
      .and('have.class', 'active')
      .and('have.value', 'Johnny')
  })
})

Using after or afterEach hooks

We see many of our users adding code to an after or afterEach hook in order to clean up the state generated by the current test(s).

We most often see test code that looks like this:

describe('logged in user', () => {
  beforeEach(() => {
    cy.login()
  })

  afterEach(() => {
    cy.logout()
  })

  it('tests', ...)
  it('more', ...)
  it('things', ...)
})

Let's look at why this is not really necessary.

Dangling state is your friend:

One of the best parts of Cypress is its emphasis on debuggability. Unlike other testing tools - when your tests end - you are left with your working application at the exact point where your test finished.

This is an excellent opportunity for you to use your application in the state the tests finished! This enables you to write partial tests that drive your application step by step, writing your test and application code at the same time.

We have built Cypress to support this use case. In fact, Cypress does not clean up its own internal state when the test ends. We want you to have dangling state at the end of the test! Things like stubs, spies, even routes are not removed at the end of the test. This means your application will behave identically while it is running Cypress commands or when you manually work with it after a test ends.

If you remove your application's state after each test, then you instantly lose the ability to use your application in this mode. Logging out at the end would always leave you with the same login page at the end of the test. In order to debug your application or write a partial test, you would always be left commenting out your custom cy.logout() command.

It's all downside with no upside:

For the moment, let's assume that for some reason your application desperately needs that last bit of after or afterEach code to run. Let's assume that if that code is not run - all is lost.

That is fine - but even if this is the case, it should not go in an after or afterEach hook. Why? So far we have been talking about logging out, but let's use a different example. Let's use the pattern of needing to reset your database.

The idea goes like this:

After each test I want to ensure the database is reset back to 0 records so when the next test runs, it is run with a clean state.

With that in mind you write something like this:

afterEach(() => {
  cy.resetDb()
})

Here is the problem: there is no guarantee that this code will run.

If, hypothetically, you have written this command because it has to run before the next test does, then the absolute worst place to put it is in an after or afterEach hook.

Why? Because if you refresh Cypress in the middle of the test - you will have built up partial state in the database, and your custom cy.resetDb() function will never get called.

If this state cleanup is truly required, then the next test will instantly fail. Why? Because resetting the state never happened when you refreshed Cypress.

State reset should go before each test:

The simplest solution here is to move your reset code to before the test runs.

Code put in a before or beforeEach hook will always run prior to the test - even if you refreshed Cypress in the middle of an existing one!

This is also a great opportunity to use root level hooks in mocha. A perfect place to put these is in the cypress/support/index.js file because it is always evaluated before any test code from your spec files.

Hooks you add to the root will always run on all suites!

// cypress/support/index.js

beforeEach(() => {
  // now this runs prior to every test
  // across all files no matter what
  cy.resetDb()
})

Is resetting the state necessary?

One final question you should ask yourself is - is resetting the state even necessary? Remember, Cypress already automatically clears localStorage, cookies, sessions, etc before each test. Make sure you are not trying to clean up state that is already cleaned up by Cypress automatically.

If the state you are trying to clean lives on the server - by all means, clean that state. You will need to run these types of routines! But if the state is related to your application currently under test - you likely do not even need to clear it.

The only times you ever need to clean up state, is if the operations that one test runs affects another test downstream. In only those cases do you need state cleanup.

Real World Example

The Real World App (RWA) resets and re-seeds its database via a custom Cypress task called db:seed in a beforeEach hook. This allows each test to start from a clean slate and a deterministic state. For example:

// cypress/tests/ui/auth.spec.ts

beforeEach(function () {
  cy.task('db:seed')
  // ...
})

Source: cypress/tests/ui/auth.spec.ts

The db:seed task is defined within the plugins file of the project, and in this case sends a request to a dedicated back end API of the app to appropriately re-seed the database.

// cypress/plugins/index.ts

on('task', {
  async 'db:seed'() {
    // Send request to backend API to re-seed database with test data
    const { data } = await axios.post(`${testDataApiEndpoint}/seed`)
    return data
  },
  //...
})

Source: cypress/plugins/index.ts

The same practice above can be used for any type of database (PostgreSQL, MongoDB, etc.). In this example, a request is sent to a back end API, but you could also interact directly with your database with direct queries, custom libraries, etc. If you already have non-JavaScript methods of handling or interacting with your database, you can use cy.exec, instead of cy.task, to execute any system command or script.

Unnecessary Waiting

In Cypress, you almost never need to use cy.wait() for an arbitrary amount of time. If you are finding yourself doing this, there is likely a much simpler way.

Let's imagine the following examples:

Unnecessary wait for cy.request()

Waiting here is unnecessary since the cy.request() command will not resolve until it receives a response from your server. Adding the wait here only adds 5 seconds after the cy.request() has already resolved.

cy.request('http://localhost:8080/db/seed')
cy.wait(5000) // <--- this is unnecessary

Unnecessary wait for cy.visit()

Waiting for this is unnecessary because the cy.visit() resolves once the page fires its load event. By that time all of your assets have been loaded including javascript, stylesheets, and html.

cy.visit('http://localhost/8080')
cy.wait(5000) // <--- this is unnecessary

Unnecessary wait for cy.get()

Waiting for the cy.get() below is unnecessary because cy.get() automatically retries until the table's tr has a length of 2.

Whenever commands have an assertion they will not resolve until their associated assertions pass. This enables you to describe the state of your application without having to worry about when it gets there.

cy.intercept('GET', '/users', [{ name: 'Maggy' }, { name: 'Joan' }])
cy.get('#fetch').click()
cy.wait(4000) // <--- this is unnecessary
cy.get('table tr').should('have.length', 2)

Alternatively a better solution to this problem is by waiting explicitly for an aliased route.

cy.intercept('GET', '/users', [{ name: 'Maggy' }, { name: 'Joan' }]).as(
  'getUsers'
)
cy.get('#fetch').click()
cy.wait('@getUsers') // <--- wait explicitly for this route to finish
cy.get('table tr').should('have.length', 2)

Web Servers

We do NOT recommend trying to start your back end web server from within Cypress.

Any command run by cy.exec() or cy.task() has to exit eventually. Otherwise, Cypress will not continue running any other commands.

Trying to start a web server from cy.exec() or cy.task() causes all kinds of problems because:

  • You have to background the process
  • You lose access to it via terminal
  • You don't have access to its stdout or logs
  • Every time your tests run, you'd have to work out the complexity around starting an already running web server.
  • You would likely encounter constant port conflicts

Why can't I shut down the process in an after hook?

Because there is no guarantee that code running in an after will always run.

While working in the Cypress Test Runner you can always restart / refresh while in the middle of a test. When that happens, code in an after won't execute.

What should I do then?

Start your web server before running Cypress and kill it after it completes.

Are you trying to run in CI?

We have examples showing you how to start and stop your web server.

Setting a global baseUrl

Adding a baseUrl in your configuration allows you to omit passing the baseUrl to commands like cy.visit() and cy.request(). Cypress assumes this is the url you want to use.

Adding a baseUrl can also save some time during the initial startup of your Cypress tests.

When you start running your tests, Cypress does not know the url of the app you plan to test. So, Cypress initially opens on https://localhost + a random port.

Without baseUrl set, Cypress loads main window in localhost + random port

Url address shows localhost:53927/__/#tests/integration/organizations/list_spec.coffee

As soon as it encounters a cy.visit(), Cypress then switches to the url of the main window to the url specified in your visit. This can result in a 'flash' or 'reload' when your tests first start.

By setting the baseUrl, you can avoid this reload altogether. Cypress will load the main window in the baseUrl you specified as soon as your tests start.

Configuration file (cypress.json by default)

{
  "baseUrl": "http://localhost:8484"
}

With baseUrl set, Cypress loads main window in baseUrl

Url address bar shows localhost:8484/__tests/integration/organizations/list_spec.coffee

Having a baseUrl set gives you the added bonus of seeing an error if your server is not running during cypress open at the specified baseUrl.

Test Runner with warning about how Cypress could not verify server set as the baseUrl is running

We also display an error if your server is not running at the specified baseUrl during cypress run after several retries.

The terminal warns and retries when the url at your baseUrl is not running