Caution
@shopify/react-testing is deprecated.
Shopifolk, see Shopify/quilt-internal for information on the latest packages available for use internally.
A library for testing React components according to Shopify conventions.
yarn add @shopify/react-testingThis library allows you to test React components with a focus on type safety and testing based on a component’s external API. In order to keep the API small and easy-to-use, it will generally track to only the latest minor release of React.
A testcase using @shopify/react-testing tends to have more or less the following structure:
mountyour component with some props to get a wrapper- (optionally) mutate it using the wrapper's
setPropsor bytriggering callbacks on elements within it - make assertions based on the wrapper (often using the custom matchers)
This example uses jest as a test-runner.
import React from 'react';
import {mount} from '@shopify/react-testing';
import ClickCounter from './ClickCounter';
describe('<ClickCounter />', () => {
it('allows us to set props', () => {
const wrapper = mount(<ClickCounter defaultCount={0} />);
expect(wrapper.props.count).toBe(0);
expect(wrapper.text()).toBe('count: 0');
wrapper.setProps({defaultCount: 1});
expect(wrapper.props.count).toBe(1);
expect(wrapper.text()).toBe('count: 1');
});
it('triggers handlers', () => {
const wrapper = mount(<ClickCounter defaultCount={0} />);
wrapper.find('button').trigger('onClick');
wrapper.find('button').trigger('onClick');
expect(wrapper.text()).toBe('count: 2');
});
});This library ships with a few useful custom matchers for Jest. To include these matchers, import @shopify/react-testing/matchers in any file that is included as part of the setupFilesAfterEnv option passed to Jest.
import '@shopify/react-testing/matchers';
import {destroyAll} from '@shopify/react-testing';
afterEach(() => {
destroyAll();
});This will allow you to use matchers such as toContainReactText or toContainReactComponent on your tree.
import React from 'react';
import {mount} from '@shopify/react-testing';
import ClickCounter from './ClickCounter';
import LinkComponent from './LinkComponent';
describe('<ClickCounter />', () => {
it('renders a link to a cool website', () => {
const wrapper = mount(<ClickCounter defaultCount={0} />);
expect(wrapper).toContainReactComponent(LinkComponent, {
to: 'https://www.cool-website.com',
});
});
it('triggers handlers', () => {
const wrapper = mount(<ClickCounter defaultCount={0} />);
wrapper.find('button').trigger('onClick');
wrapper.find('button').trigger('onClick');
expect(wrapper).toContainReactText('count: 2');
});
});Mounts a component to the DOM and returns a Root instance. Note that for this to work, you must have a simulated browser environment, such as the jsdom environment that Jest uses.
createMount<MountOptions, Context, Async>(options: CreateMountOptions<MountOptions, Context, Async>): MountFunction
The mount function is powerful on its own, but applications will often want a more powerful version tailored to their application. A common example is app-wide context, where a set of context providers are generally assumed to be present for every component under test. It can also be useful for providing custom GraphQL infrastructure that enables easy testing of different API responses, such as the createGraphQL factory from @shopify/graphql-testing.
createMount enables this kind of customization by vending a custom mount function that will automatically wrap the component under test in an appropriate test wrapper. This custom mount function can do four things:
- Allow custom options to be passed as the second argument to mount, as specified by the
MountOptionsgeneric - Map passed options to an object containing all the relevant "context" (be it objects passed through react context providers, or other useful values for controlling the test harness)
- Use the resolved context to render react components around the element under test that use the context
- Perform some additional resolution after the component has mounted, including asynchronous behavior like resolving initial API results
These features are controlled by the generic type arguments to createMount, and the options detailed in the section below. Note that, no matter how many context providers or test wrapper you end up rendering your element within, all of the methods on the returned Root instance will still be scoped to within the tree actually under test.
Takes an object of options passed by a user of your custom mount (or an empty object), and should return an object containing the context you need for the test harness. If your Context type has non-optional keys, you must provide this option.
This function is called with the react element under test, the context created by context() (or an empty object), and the options passed by the user of your custom mount (or an empty object). This function must return a new react element, usually by wrapping the component in context providers.
Note:
rendercan be called multiple times for a given component. Yourrenderfunction (and any wrapping elements you put around the element under test) should be able to re-render from calling this function, ideally without unmounting the component under test.
This function allows you to perform additional logic after a component has been mounted. It gets called with a special Root instance that has one additional property: context, the object with the context you created in context() (or an empty object). You can use this hook to perform some additional resolution after the component has mounted, such as resolving all GraphQL.
If this option returns a Promise, the result of calling mount() will become a promise that resolves to the custom Root instance. Otherwise, it will synchronously return the Root instance. If you specify the Async generic argument as true, you must pass this option.
This function allows you to perform additional cleanup when a wrapper is destroyed. It gets called with a special Root instance that has one additional property: context, the object with the context you created in context() (or an empty object). You can use this hook to clean up any dangling promises or other operations when your test suite is being discarded.
We usually want to create a mocked version of the GraphQL infrastructure for our app to prevent relying on real API calls. We provide the @shopify/graphql-testing library to create a mock GraphQL source and Apollo client that uses it.
In our example mount, we want people to be able to pass a custom GraphQL instance. We want the initial GraphQL results to resolve, unless the user of mount specifies that GraphQL should not resolve until done manually. Finally, we want to expose this GraphQL instance on the returned wrapper for use to drive test results.
The custom mount for this situation would be built as demonstrated below.
import React from 'react';
import {ApolloProvider} from '@apollo/client';
import {createGraphQLFactory, GraphQL} from '@shopify/graphql-testing';
import {createMount} from '@shopify/react-testing';
// See graphql-testing docs for details
const createGraphQL = createGraphQLFactory();
// Here, we define the options a user can pass to mount. We need them to be able
// to pass two things: an optional GraphQL instance to drive the test, and an
// optional flag to skip initial GraphQL resolution.
interface Options {
graphQL?: GraphQL;
skipInitialGraphQL?: boolean;
}
// Next is the context. We only want to expose one thing as "context": The GraphQL
// instance driving the test.
interface Context {
graphQL: GraphQL;
}
// Now, we can create our custom mount function! Unfortunately, due to limitations in
// TypeScript, you usually need to pass all the generic arguments, including the last
// one, which specifies whether your `afterMount` is async or not.
export const mountWithGraphQL = createMount<Options, Context, true>({
// Step one: convert Options to Context
context({graphQL = createGraphQL()}) {
return {graphQL};
},
// Step two: use Context and Options to render the element under the test
// with the necessary providers
render(element, {graphQL}) {
return <ApolloProvider client={graphQL.client}>{element}</ApolloProvider>;
},
// Final step: if we need post-mount behavior, inject it in. If it returns
// a promise, like it does here, the final mount function will be async too.
async afterMount(root, {skipInitialGraphQL}) {
const {graphQL} = root.context;
// This makes it so any GraphQL resolution is wrapped in
// an act() block, which prevents setting state outside of
// act().
graphQL.wrap((perform) => root.act(perform));
if (skipInitialGraphQL) {
return;
}
// Here's the important bit: resolve the GraphQL so our first queries are
// in use for the component under test
await graphQL.resolveAll();
},
cleanup(root, {graphQL}) {
graphQL.client.stop();
},
});It is possible to extend a custom mount function with additional logic. This can help to provide more focused testing utilities for a section of the application that provides additional context to its subtree. Every function created by createMount has an extend method. This method has the same type parameters and options as createMount itself. When you create an extended mount function, your additional options are merged with the original mount’s options as follows:
- The resulting
mountfunction accepts the merged set of allowed options. - The root created by the resulting
mountfunction has acontextproperty that is the merged result of calling the original context and the extended context. - The
context()andrender()options you provide tomount.extend()will be called with the full, merged set of options. - The
render()option provided tomount.extend()is called first. The result of calling this function is then passed to the originalrender(). - The
afterMount()option provided tomount.extend()is called first. If it returns a promise, the resulting post-mount process will wait for it to resolve, and will then return the result of calling the originalafterMount(). If either the original options or the extended options return a promise fromafterMount, the resulting mount function will be asynchronous.
import {createMount} from '@shopify/react-testing';
interface Options {
pathname: string;
}
interface ExtendedOptions {
graphQLResult: object;
}
const mount = createMount<Options, Options>({
context: (options) => options,
render: (element, {pathname}) => (
<Router pathname={pathname}>{element}</Router>
),
});
const extendedMount = mount.extend<ExtendedOptions, ExtendedOptions>({
context: (options) => options,
render: (element, {graphQLResult}) => (
<GraphQLMock mock={graphQLResult}>{element}</GraphQLMock>
),
});
const mounted = extendedMount(<MyComponent />, {
pathname: '/',
graphQLResult: {},
});
// The final structure of this wrapper is:
// <Router><GraphQLMock><MyComponent /></GraphQLMock></Router>
//
// It also has a context field that merged the two `context()`
// results: typeof mounted.context === {pathname: string; graphQLResult: object}All mounted components are tracked in-memory. destroyAll() forcibly unmounts all mounted components and removes the DOM node used to house them. You should run this after each test that mounts a component (this is often done in a global afterEach hook).
A Root object represents a mounted react tree. Most of the properties and methods it exposes are simply forwarded to the Element instance representing the top-level component you rendered:
- #children
- #descendants
- #props
- #isDOM
- #instance
- #domNode
- #domNodes
- #html()
- #text()
- #is()
- #prop()
- #find()
- #findAll()
- #findWhere()
- #findAllWhere()
- #trigger()
- #triggerKeypath()
This object also has a number of methods that only apply to the root of a component tree:
Re-mounts the component to the DOM. If the component is already mounted, this method will throw an error.
Unmounts the component from the DOM. If the component is not already mounted, this method will throw an error. This method can be useful for testing side effects that occur in componentWillUnmount or useEffect hooks.
Allows you to change a subset of the props specified when the component was originally mounted. This can be useful to test behaviour that is only caused by a change in props, such as getDerivedStateFromProps or its equivalent useRef/ useState hook version.
Performs an action in the context of a react act() block, then updates the internal representation of the react tree. You must use this whenever performing an action that will cause the react tree to set state and re-render, such as simulating event listeners being called. Failing to do so will print a warning, and the react tree will not be updated for subsequent calls to methods such as find().
function MyComponent() {
const [clicked, setClicked] = useState(false);
useEffect(() => {
const handler = () => setClicked(true);
document.body.addEventListener('click', handler);
return () => document.body.removeEventListener('click', handler);
}, [setClicked]);
return clicked ? <div>I’ve been clicked!</div> : <div>Nothing yet</div>;
}
const myComponent = mount(<MyComponent />);
// If you don’t do this, you’ll see a warning and the subsequent assertion
// will fail
myComponent.act(() => simulateClickOnBody());
expect(myComponent.text()).toContain('been clicked!');Unmounts the component and removes its associated DOM node. This method ensures that nothing leaks between tests. It is called on all un-destroyed Root objects when you call destroyAll()
Forces the root component to re-render. This can be necessary in some cases where globals change in a way that does not already cause a "natural" react update, but in general, this method should not be necessary.
The Element object represents a react element in the tree. This element can be a DOM node, custom react component, or one of the many "special" types react creates, such as context providers and consumers. The Element object also houses all of the methods that you will use to find rendered subcomponents (find and friends), get your react tree into the desired state (trigger), and ensure that state is correct (props).
It is important to understand that the Element object is only a snapshot representation of the react tree at one point in time. As soon as you use trigger to simulate calling a prop, or Root#act to commit an arbitrary update, the Element should be considered "stale" and discarded.
This getter returns the props for the component.
This getter returns the type of component. For DOM nodes, this will be a string representing the rendered DOM element. For custom react components, this will be the react component itself. For all other elements, this will be null.
This getter returns whether the element represents a DOM node.
This getter returns the instance associated with the component. Note: this property technically gives you access to fields like state and methods like setState, but doing so violates component boundaries and makes for bad tests. If you can avoid it, you should never use this getter. It should be seen only as an escape hatch when it is impossible to perform the update you need with props alone.
This getter returns an array of elements that represent the element children of this component in the react tree.
This getter returns an array of elements that represent all elements below this component in the react tree.
Returns all DOM nodes that are directly rendered by this component (that is, not rendered by descendant components).
Like domNodes, but expects only 1 or 0 DOM nodes to be direct children. If more than 1 DOM node is a child, this method throws an error. If no DOM nodes are children, this method returns null.
Returns the current value of the passed prop.
function MyComponent({name}: {name: string}) {
return <div>Hello, {name}!</div>;
}
function Wrapper() {
return <MyComponent name="Michelle" />;
}
const wrapper = mount(<Wrapper />);
expect(wrapper.find(MyComponent).prop('name')).toBe('Michelle');
// Will give you a type error
expect(wrapper.find(MyComponent).prop('firstName')).toBe('Uhh');Returns the text content of the component. This is the string of text you would receive from mapping over each DOM node rendered as a descendant of this component and taking its textContent.
Returns the HTML content of the component. This is the string of text you would receive from mapping over each DOM node rendered as a descendant of this component and taking its innerHTML.
Returns a boolean indicating whether the component type matches the passed type. This function also serves as a type guard, so subsequent calls to values like props will be typed as the prop type of the passed component.
function MyComponent({name}: {name: string}) {
return <div>Hello, {name}!</div>;
}
function isMatch(element: Element<unknown>) {
// If we omitted element.is here, we would not know whether 'name' was a prop,
// so we would get a type error.
return element.is(MyComponent) && element.prop('name') === 'Chris';
}Finds a descendant component that matches type, where type is either a string or react component. If no matching element is found, null is returned. If a match is found, the returned Element will have the correct prop typing, which provides excellent type safety while navigating the react tree.
function MyComponent({name}: {name: string}) {
return <div>Hello, {name}!</div>;
}
function YourComponent() {
return <div>Goodbye, friend!</div>;
}
function Wrapper() {
return <MyComponent name="Michelle" />;
}
const wrapper = mount(<Wrapper />);
expect(wrapper.find(MyComponent)).not.toBeNull();
expect(wrapper.find(YourComponent)).toBe(null);You can optionally pass a second argument to this function, which is a set of props that will be used to further filter the matching elements. These props will be shallow compared to the props of each element.
function MyComponent({name}: {name: string}) {
return <div>Hello, {name}!</div>;
}
function YourComponent() {
return <div>Goodbye, friend!</div>;
}
function Wrapper() {
return (
<>
<MyComponent name="Michelle" />
<MyComponent name="Gord" />
</>
);
}
const wrapper = mount(<Wrapper />);
expect(wrapper.find(MyComponent, {name: 'Gord'})!.props).toMatchObject({
name: 'Gord',
});Like find(), but returns all matches as an array.
findWhere<Type = unknown>(predicate: (element: Element<unknown>) => boolean): Element<PropsForComponent<Type>> | null
Finds the first descendant component matching the passed function. The function is called with each Element from descendants until a match is found. If no match is found, null is returned.
findWhere accepts an optional generic argument that can be used to specify the type of the returned element. This argument is either a string or a React component, the same as the first argument on .find. If the generic argument is omitted then the returned element will have unknown props and thus calling .props and .trigger on it will cause type errors as those functions won't know what props are valid on your element:
function MyComponent({name}: {name: string}) {
return <div>Hello, {name}!</div>;
}
function Wrapper() {
return (
<>
<div id="Michelle" />
<MyComponent name="Gord" />
</>
);
}
const wrapper = mount(<Wrapper />);
const divElement = wrapper.findWhere<'div'>(
(node) => node.is('div') && node.prop('id').startsWith('M'),
);
const componentElement = wrapper.findWhere<typeof MyComponent>(
(node) => node.is(MyComponent) && node.prop('name').startsWith('G'),
);
expect(divElement.prop('id')).toBe('Michelle');
expect(componentElement.prop('name')).toBe('Gord'); findAllWhere<Type = unknown>(predicate: (element: Element<unknown>) => boolean): Element<PropsForComponent<Type>>[]
Like findWhere, but returns all matches as an array.
trigger<K extends FunctionKeys<Props>>(prop: K, ...args: Parameters<Props<K>>): ReturnType<Props<K>>
Simulates a function prop being called on your component. This is usually the key to effective tests: after you have mounted your component, you simulate a change in a subcomponent, and assert that the resulting react tree is in the expected shape. This method automatically uses Root#act when calling the prop, so updates will automatically be applied to the root component.
When you pass a key that is a prop on your component with a function type, this function will ensure that you pass arguments that are deeply partial versions of the types the prop expects. This allows you to, for example, pass an event object with only a few properties set to a button’s onClick prop. trigger returns whatever the result was of calling the prop.
function MyComponent({onClick}: {onClick(id: string): void}) {
return (
<button type="button" onClick={() => onClick(String(Math.random()))}>
Click me!
</button>
);
}
function Wrapper() {
const [id, setId] = React.useState<string>('');
return (
<>
<MyComponent onClick={setId} />
<div>Current id is: {id}</div>
</>
);
}
const wrapper = mount(<Wrapper />);
wrapper.find(MyComponent)!.trigger('onClick', 'some-id');
expect(wrapper.find('div')!.text()).toContain('some-id');Like trigger(), but allows you to provide a keypath referencing nested objects instead. Note that limitations in TypeScript prevent the same kind of type-safety as trigger guarantees.
function MyComponent({action}: {action: {onAction(): void; label: string}}) {
return (
<button type="button" onClick={action.onAction}>
{action.label}
</button>
);
}
const spy = jest.fn();
const myComponent = mount(
<MyComponent action={{label: 'Hi', onAction: spy}} />,
);
myComponent.triggerKeypath('action.onAction');
expect(spy).toHaveBeenCalled();Returns a text representation of either the root node, or any element within the mounted graph. debug() output can be tweaked using the options parameter.
allPropsoverrides the default props filtering behaviour and instead includes all props in the output, by defaultclassName,aria-*, anddata-*props are omitted.depthdefines the number of children printed, by default all children are printed.verbositydefines the level of expansion that non-scalar props experience, the default value of1will expand objects one level deep
Typical usage should not require providing any options as the default verbosity and depth should be appropriate for the majority of inspections.
function ObjectText({data}: {data: {}}) {
return <span>{JSON.stringify(data)}</span>;
}
function Container({children}: PropsWithChildren<{}>) {
return children;
}
function MyComponent({onClick}: {onClick(id: string): void}) {
return (
<Container>
<button type="button" onClick={() => onClick(String(Math.random()))}>
<ObjectText data={{a: {very: {deep: {data: {object: 'with text'}}}}}} />
</button>
</Container>
);
}
const wrapper = mount(<MyComponent />);
// print the whole structure with one level of prop verbosity
console.log(wrapper.debug());
// print only the Container and button without any other children
console.log(wrapper.find(Container)!.debug({depth: 1}));
// find button by name and print all children with verbose props details
console.log(
wrapper
.findWhere((type) => type && type.name === 'button')!
.debug({verbosity: 9}),
);Checks whether a Root or Element object has specified props (asymmetric matchers like expect.objectContaining are fully supported). Strict type checking is enforced, so the props you pass must be a valid subset of the actual props for the component.
const myComponent = mount(<MyComponent />);
expect(myComponent.find('div')).toHaveReactProps({'aria-label': 'Hello world'});
expect(myComponent.find('div')).toHaveReactProps({
onClick: expect.any(Function),
});Like .toHaveReactProps(), but is not strictly typed. This makes it more suitable for asserting on data- attributes, which can’t be strongly typed.
const myComponent = mount(<MyComponent />);
expect(myComponent.find('div')).toHaveReactDataProps({
'data-message': 'Hello world',
});Asserts that at least one component matching type is in the descendants of the passed node. If the second argument is passed, this expectation will further filter the matches by components whose props are equal to the passed object (again, asymmetric matchers are fully supported).
const myComponent = mount(<MyComponent />);
expect(myComponent).toContainReactComponent('div', {
'aria-label': 'Hello world',
onClick: expect.any(Function),
});Asserts that a component matching type is in the descendants of the passed node a number of times. If the third argument is passed, this expectation will further filter the matches by components whose props are equal to the passed object (again, asymmetric matchers are fully supported). To assert that one component is or is not the descendant of the passed node use .toContainReactComponent or .not.toContainReactComponent.
const myComponent = mount(<MyComponent />);
expect(myComponent).toContainReactComponentTimes('div', 5, {
'aria-label': 'Hello world',
});Asserts that at least one context.Provider is in the descendants of the passed node. If the second argument is passed, this expectation will further filter the matches by providers whose value is equal to the passed object (again, asymmetric matchers are fully supported).
const MyContext = React.createContext({hello: 'world'});
const myComponent = mount(<MyComponent />);
expect(myComponent).toProvideReactContext(MyContext, {
hello: expect.any(String),
});Asserts that the rendered output of the component contains the passed string as text content (that is, the text is included in what you would get by calling textContent on all root DOM nodes rendered by the component).
const myComponent = mount(<MyComponent />);
expect(myComponent).toContainReactText('Hello world!');Asserts that the rendered output of the component contains the passed string as HTML (that is, the text is included in what you would get by calling outerHTML on all root DOM nodes rendered by the component).
const myComponent = mount(<MyComponent />);
expect(myComponent).toContainReactHtml('<span>Hello world!</span>');Why not use Enzyme instead?
Enzyme is a very popular testing library that heavily inspired the approach this library takes. However, our experience with Enzyme has not been ideal:
- It has frequently taken a long time to support new react features.
- It has a very large API surface area, much of which does not conform to Shopify’s testing conventions. For example, Enzyme provides APIs like
setStatewhich encourage reaching in to implementation details of your components. - Enzyme is unlikely to add features we use or need in a testing library, such as automatic unmounting and a built-in version
trigger().
Why not use react-testing-library instead?
While the premise of writing tests that mirror user actions is compelling, basing all tests off the raw DOM being produced becomes unmanageable for larger apps.
The React versions this library supports are spelled out via a peer dependency in the package.json
We currently provide support for Preact applications via a separate package. This may change in the future.