April 15, 2024

by Moshe Kolodny

On this publish, we’re excited to introduce SafeTest, a revolutionary library that provides a contemporary perspective on Finish-To-Finish (E2E) checks for web-based Person Interface (UI) functions.

Historically, UI checks have been performed via both unit testing or integration testing (additionally known as Finish-To-Finish (E2E) testing). Nevertheless, every of those strategies presents a novel trade-off: you must select between controlling the check fixture and setup, or controlling the check driver.

As an illustration, when utilizing react-testing-library, a unit testing answer, you keep full management over what to render and the way the underlying providers and imports ought to behave. Nevertheless, you lose the power to work together with an precise web page, which may result in a myriad of ache factors:

  • Issue in interacting with complicated UI components like <Dropdown /> parts.
  • Lack of ability to check CORS setup or GraphQL calls.
  • Lack of visibility into z-index points affecting click-ability of buttons.
  • Complicated and unintuitive authoring and debugging of checks.

Conversely, utilizing integration testing instruments like Cypress or Playwright offers management over the web page, however sacrifices the power to instrument the bootstrapping code for the app. These instruments function by remotely controlling a browser to go to a URL and work together with the web page. This method has its personal set of challenges:

  • Issue in making calls to another API endpoint with out implementing customized community layer API rewrite guidelines.
  • Lack of ability to make assertions on spies/mocks or execute code inside the app.
  • Testing one thing like darkish mode entails clicking the theme switcher or figuring out the localStorage mechanism to override.
  • Lack of ability to check segments of the app, for instance if a element is barely seen after clicking a button and ready for a 60 second timer to countdown, the check might want to run these actions and will likely be not less than a minute lengthy.

Recognizing these challenges, options like E2E Part Testing have emerged, with choices from Cypress and Playwright. Whereas these instruments try to rectify the shortcomings of conventional integration testing strategies, they produce other limitations on account of their structure. They begin a dev server with bootstrapping code to load the element and/or setup code you need, which limits their potential to deal with complicated enterprise functions which may have OAuth or a posh construct pipeline. Furthermore, updating TypeScript utilization might break your checks till the Cypress/Playwright group updates their runner.

SafeTest goals to deal with these points with a novel method to UI testing. The principle thought is to have a snippet of code in our application bootstrapping stage that injects hooks to run our tests (see the How Safetest Works sections for more information on what that is doing). Word that how this works has no measurable impression on the common utilization of your app since SafeTest leverages lazy loading to dynamically load the checks solely when working the checks (within the README instance, the checks aren’t within the manufacturing bundle in any respect). As soon as that’s in place, we are able to use Playwright to run common checks, thereby reaching the perfect browser management we would like for our checks.

This method additionally unlocks some thrilling options:

  • Deep linking to a particular check with no need to run a node check server.
  • Two-way communication between the browser and check (node) context.
  • Entry to all of the DX options that include Playwright (excluding those that include @playwright/check).
  • Video recording of checks, hint viewing, and pause web page performance for attempting out totally different web page selectors/actions.
  • Means to make assertions on spies within the browser in node, matching snapshot of the decision inside the browser.

SafeTest is designed to really feel acquainted to anybody who has performed UI checks earlier than, because it leverages the perfect elements of present options. Right here’s an instance of learn how to check a complete utility:

import  describe, it, count on  from 'safetest/jest';
import render from 'safetest/react';

describe('my app', () =>
it('masses the primary web page', async () =>
const web page = await render();

await count on(web page.getByText('Welcome to the app')).toBeVisible();
count on(await web page.screenshot()).toMatchImageSnapshot();
);
);

We are able to simply as simply check a particular element

import  describe, it, count on, browserMock  from 'safetest/jest';
import render from 'safetest/react';

describe('Header element', () =>
it('has a traditional mode', async () =>
const web page = await render(<Header />);

await count on(web page.getByText('Admin')).not.toBeVisible();
);

it('has an admin mode', async () =>
const web page = await render(<Header admin=true />);

await count on(web page.getByText('Admin')).toBeVisible();
);

it('calls the logout handler when signing out', async () =>
const spy = browserMock.fn();
const web page = await render(<Header handleLogout=fn />);

await web page.getByText('logout').click on();
count on(await spy).toHaveBeenCalledWith();
);
);

SafeTest makes use of React Context to permit for worth overrides throughout checks. For an instance of how this works, let’s assume now we have a fetchPeople perform utilized in a element:

import  useAsync  from 'react-use';
import fetchPerson from './api/individual';

export const Individuals: React.FC = () =>
const knowledge: folks, loading, error = useAsync(fetchPeople);

if (loading) return <Loader />;
if (error) return <ErrorPage error=error />;
return <Desk knowledge=knowledge rows=[...] />;

We are able to modify the Individuals element to make use of an Override:

 import  fetchPerson  from './api/individual';
+import createOverride from 'safetest/react';

+const FetchPerson = createOverride(fetchPerson);

export const Individuals: React.FC = () =>
+ const fetchPeople = FetchPerson.useValue();
const knowledge: folks, loading, error = useAsync(fetchPeople);

if (loading) return <Loader />;
if (error) return <ErrorPage error=error />;
return <Desk knowledge=knowledge rows=[...] />;

Now, in our check, we are able to override the response for this name:

const pending = new Promise(r =>  /* Do nothing */ );
const resolved = [{name: 'Foo', age: 23], title: 'Bar', age: 32]];
const error = new Error('Whoops');

describe('Individuals', () =>
it('has a loading state', async () =>
const web page = await render(
<FetchPerson.Override with=() => () => pending>
<Individuals />
</FetchPerson.Override>
);

await count on(web page.getByText('Loading')).toBeVisible();
);

it('has a loaded state', async () =>
const web page = await render(
<FetchPerson.Override with=() => async () => resolved>
<Individuals />
</FetchPerson.Override>
);

await count on(web page.getByText('Person: Foo, title: 23')).toBeVisible();
);

it('has an error state', async () =>
const web page = await render(
<FetchPerson.Override with=() => async () => throw error >
<Individuals />
</FetchPerson.Override>
);

await count on(web page.getByText('Error getting customers: "Whoops"')).toBeVisible();
);
);

The render perform additionally accepts a perform that will likely be handed the preliminary app element, permitting for the injection of any desired components wherever within the app:

it('has a folks loaded state', async () => 
const web page = await render(app =>
<FetchPerson.Override with=() => async () => resolved>
app
</FetchPerson.Override>
);
await count on(web page.getByText('Person: Foo, title: 23')).toBeVisible();
);

With overrides, we are able to write complicated check instances equivalent to guaranteeing a service technique which mixes API requests from /foo, /bar, and /baz, has the proper retry mechanism for simply the failed API requests and nonetheless maps the return worth accurately. So if /bar takes 3 makes an attempt to resolve the strategy will make a complete of 5 API calls.

Overrides aren’t restricted to simply API calls (since we are able to use additionally use page.route), we are able to additionally override particular app stage values like function flags or altering some static worth:

+const UseFlags = createOverride(useFlags);
export const Admin = () =>
+ const useFlags = UseFlags.useValue();
const isAdmin = useFlags();
if (!isAdmin) return <div>Permission error</div>;
// ...

+const Language = createOverride(navigator.language);
export const LanguageChanger = () =>
- const language = navigator.language;
+ const language = Language.useValue();
return <div>Present language is language </div>;

describe('Admin', () => {
it('works with admin flag', async () =>
const web page = await render(
<UseIsAdmin.Override with=oldHook =>
const oldFlags = oldHook();
return ...oldFlags, isAdmin: true ;
>
<MyComponent />
</UseIsAdmin.Override>
);

await count on(web page.getByText('Permission error')).not.toBeVisible();
);
});

describe('Language', () =>
it('shows', async () =>
const web page = await render(
<Language.Override with=previous => 'abc'>
<MyComponent />
</Language.Override>
);

await count on(web page.getByText('Present language is abc')).toBeVisible();
);
);

Overrides are a strong function of SafeTest and the examples right here solely scratch the floor. For extra info and examples, check with the Overrides section on the README.

SafeTest comes out of the field with highly effective reporting capabilities, equivalent to computerized linking of video replays, Playwright hint viewer, and even deep link directly to the mounted tested component. The SafeTest repo README hyperlinks to all of the example apps in addition to the reports

Image of SafeTest report showing a video of a test run

Many massive companies want a type of authentication to make use of the app. Sometimes, navigating to localhost:3000 simply leads to a perpetually loading web page. You must go to a special port, like localhost:8000, which has a proxy server to examine and/or inject auth credentials into underlying service calls. This limitation is among the predominant causes that Cypress/Playwright Part Assessments aren’t appropriate to be used at Netflix.

Nevertheless, there’s normally a service that may generate check customers whose credentials we are able to use to log in and work together with the applying. This facilitates creating a light-weight wrapper round SafeTest to routinely generate and assume that check consumer. As an illustration, right here’s mainly how we do it at Netflix:

import  setup  from 'safetest/setup';
import createTestUser, addCookies from 'netflix-test-helper';

sort Setup = Parameters<typeof setup>[0] &
extraUserOptions?: UserOptions;
;

export const setupNetflix = (choices: Setup) =>
setup(
...choices,
hooks: beforeNavigate: [async page => addCookies(page)] ,
);

beforeAll(async () =>
createTestUser(choices.extraUserOptions)
);
;

After setting this up, we merely import the above bundle rather than the place we might have used safetest/setup.

Whereas this publish targeted on how SafeTest works with React, it’s not restricted to simply React. SafeTest additionally works with Vue, Svelte, Angular, and even can run on NextJS or Gatsby. It additionally runs utilizing both Jest or Vitest based mostly on which check runner your scaffolding began you off with. The examples folder demonstrates learn how to use SafeTest with totally different tooling mixtures, and we encourage contributions so as to add extra instances.

At its core, SafeTest is an clever glue for a check runner, a UI library, and a browser runner. Although the commonest utilization at Netflix employs Jest/React/Playwright, it’s simple so as to add extra adapters for different choices.

SafeTest is a strong testing framework that’s being adopted inside Netflix. It permits for simple authoring of checks and offers complete stories when and the way any failures occurred, full with hyperlinks to view a playback video or manually run the check steps to see what broke. We’re excited to see the way it will revolutionize UI testing and sit up for your suggestions and contributions.