For almost all of our code style rules, refer to the Airbnb JavaScript Style Guide.
When writing ES6 or React code, please also refer to the Airbnb React/JSX Style Guide.
We use Prettier to automatically style our code.
- You can run Prettier to fix the style on all files with
npm run prettier
- You can run Prettier in watch mode to fix the styles when they are saved with
npm run prettier-watch
There are a few things that we have customized for our tastes which will take precedence over Airbnb's guide.
-
When you have an event handler, do not prefix it with "on" or "handle". The method should be named for what it does, not what it handles. This promotes code reuse by minimizing assumptions that a method must be called in a certain fashion (eg. only as an event handler).
-
One exception for allowing the prefix of "on" is when it is used for callback
props
of a React component. Using it in this way helps to distinguish callbacks from public component methods.// Bad const onSubmitClick = () => { // Validate form items and submit form }; // Good const validateAndSubmit = () => { // Validate form items and submit form };
- Boolean props or variables must be prefixed with
should
oris
to make it clear that they areBoolean
. Useshould
when we are enabling or disabling some features andis
in most other cases.
// Bad
<SomeComponent showIcon />
// Good
<SomeComponent shouldShowIcon />
// Bad
const valid = props.something && props.somethingElse;
// Good
const isValid = props.something && props.somethingElse;
Any function declared in a library module should use the function myFunction
keyword rather than const myFunction
.
// Bad
const myFunction = () => {...};
export {
myFunction,
}
// Good
function myFunction() {
...
}
export {
myFunction,
}
Using arrow functions is the preferred way to write an anonymous function such as a callback method.
// Bad
_.map(someArray, function (item) {...});
// Good
_.map(someArray, (item) => {...});
Empty functions (noop) should be declare as arrow functions with no whitespace inside. Avoid _.noop()
// Bad
const callback = _.noop;
const callback = () => { };
// Good
const callback = () => {};
- Never use
var
- Use
const
when you are not reassigning a variable - Try to write your code in a way where the variable reassignment isn't necessary
- Use
let
only if there are no other options
// Bad
let array = [];
if (someCondition) {
array = ['addValue1'];
}
// Good
const array = [];
if (someCondition) {
array.push('addValue1');
}
We have standardized on using underscore.js methods for objects and collections instead of the native Array instance methods. This is mostly to maintain consistency, but there are some type safety features and conveniences that underscore methods provide us e.g. the ability to iterate over an object and the lack of a TypeError
thrown if a variable is undefined
.
// Bad
myArray.forEach(item => doSomething(item));
// Good
_.each(myArray, item => doSomething(item));
// Bad
const myArray = Object.keys(someObject).map(key => doSomething(someObject[key]));
// Good
const myArray = _.map(someObject, (value, key) => doSomething(value));
// Bad
myCollection.includes('item');
// Good
_.contains(myCollection, 'item');
// Bad
const modifiedArray = someArray.filter(filterFunc).map(mapFunc);
// Good
const modifiedArray = _.chain(someArray)
.filter(filterFunc)
.map(mapFunc)
.value();
Use lodashGet()
to safely access object properties and ||
to short circuit null or undefined values that are not guaranteed to exist in a consistent way throughout the codebase. In the rare case that you want to consider a falsy value as usable and the ||
operator prevents this then be explicit about this in your code and check for the type using an underscore method e.g. _.isBoolean(value)
or _.isEqual(0)
.
// Bad
const value = somePossiblyNullThing ?? 'default';
// Good
const value = somePossiblyNullThing || 'default';
// Bad
const value = someObject.possiblyUndefinedProperty?.nestedProperty || 'default';
// Bad
const value = (someObject && someObject.possiblyUndefinedProperty && someObject.possiblyUndefinedProperty.nestedProperty) || 'default';
// Good
const value = lodashGet(someObject, 'possiblyUndefinedProperty.nestedProperty', 'default');
-
Always document parameters and return values.
-
Optional parameters should be enclosed by
[]
e.g.@param {String} [optionalText]
. -
Document object parameters with separate lines e.g.
@param {Object} parameters
followed by@param {String} parameters.field
. -
If a parameter accepts more than one type use
*
to denote there is no single type. -
Use uppercase when referring to JS primitive values (e.g.
Boolean
notbool
,Number
notint
, etc). -
When specifying a return value use
@returns
instead of@return
. If there is no return value do not include one in the doc. -
Avoid descriptions that don't add any additional information. Method descriptions should only be added when it's behavior is unclear.
-
Do not use block tags other than
@param
and@returns
(e.g.@memberof
,@constructor
, etc). -
Do not document default parameters. They are already documented by adding them to a declared function's arguments.
-
Do not use record types e.g.
{Object.<string, number>}
. -
Do not create
@typedef
to use in JSDocs. -
Do not use type unions e.g.
{(number|boolean)}
.
// Bad
/**
* Populates the shortcut modal
* @param {bool} shouldShowAdvancedShortcuts whether to show advanced shortcuts
* @return {*}
*/
function populateShortcutModal(shouldShowAdvancedShortcuts) {
}
// Good
/**
* @param {Boolean} shouldShowAdvancedShortcuts
* @returns {Boolean}
*/
function populateShortcutModal(shouldShowAdvancedShortcuts) {
}
We should avoid using object destructuring in situations where it reduces code clarity. Here are some general guidelines on destructuring.
General Guidelines
- Avoid object destructuring for a single variable that you only use once. It's clearer to use dot notation for accessing a single variable.
// Bad
const {data} = event.data;
// Good
const {name, accountID, email} = data;
React Components
Always use destructuring to get prop values. Destructuring is necessary to assign default values to props.
// Bad
function UserInfo(props) {
return (
<View>
<Text>Name: {props.name}</Text>
<Text>Email: {props.email}</Text>
</View>
}
UserInfo.defaultProps = {
name: 'anonymous';
}
// Good
function UserInfo({ name = 'anonymous', email }) {
return (
<View>
<Text>Name: {name}</Text>
<Text>Email: {email}</Text>
</View>
);
}
ES6 provides two ways to export a module from a file: named export
and default export
. Which variation to use depends on how the module will be used.
- If a file exports a single JS object (e.g. a React component, or an IIFE), then use
export default
- Files with multiple exports should always use named exports
- Files with a single method or variable export are OK to use named exports
- Mixing default and named exports in a single file is OK (e.g. in a self contained module), but should rarely be used
- All exports (both default and named) should happen at the bottom of the file
- Do not export individual features inline.
// Bad
export const something = 'nope';
export const somethingElse = 'stop';
// Good
const something = 'yep';
const somethingElse = 'go';
export {
something,
somethingElse,
};
Using the class
syntax is preferred wherever appropriate. Airbnb has clear guidelines in their JS style guide which promotes using the class syntax. Don't manipulate the prototype
directly. The class
syntax is generally considered more concise and easier to understand.
Classes have a default constructor if one is not specified. No need to write a constructor function that is empty or just delegates to a parent class.
// Bad
class Jedi {
constructor() {}
getName() {
return this.name;
}
}
// Bad
class Rey extends Jedi {
constructor(...args) {
super(...args);
}
}
// Good
class Rey extends Jedi {
constructor(...args) {
super(...args);
this.name = 'Rey';
}
}
JavaScript is always changing. We are excited whenever it does! However, we tend to take our time considering whether to adopt the latest and greatest language features. The main reason for this is consistency. We have a style guide so that we don't have to have endless conversations about how our code looks and can focus on how it runs.
So, if a new language feature isn't something we have agreed to support it's off the table. Sticking to just one way to do things reduces cognitive load in reviews and also makes sure our knowledge of language features progresses at the same pace. If a new language feature will cause considerable effort for everyone to adapt to or we're just not quite sold on the value of it yet we won't support it.
Here are a couple of things we would ask that you avoid to help maintain consistency in our codebase:
- Async/Await - Use the native
Promise
instead - Optional Chaining - Use
lodashGet()
to fetch a nested value instead - Null Coalescing Operator - Use
lodashGet()
or||
to set a default value for a possiblyundefined
ornull
variable
- Prop callbacks should be named for what has happened, not for what is going to happen. Components should never assume anything about how they will be used (that's the job of whatever is implementing it).
// Bad
const propTypes = {
/** A callback to call when we want to save the form */
onSaveForm: PropTypes.func.isRequired,
};
// Good
const propTypes = {
/** A callback to call when the form has been submitted */
onFormSubmitted: PropTypes.func.isRequired,
};
- Do not use underscores when naming private methods.
- Add descriptions to all
propTypes
using a block comment above the definition. No need to document the types (that's whatpropTypes
is doing already), but add some context for each property so that other developers understand the intended use.
// Bad
const propTypes = {
currency: PropTypes.string.isRequired,
amount: PropTypes.number.isRequired,
isIgnored: PropTypes.bool.isRequired
};
// Bad
const propTypes = {
// The currency that the reward is in
currency: React.PropTypes.string.isRequired,
// The amount of reward
amount: React.PropTypes.number.isRequired,
// If the reward has been ignored or not
isIgnored: React.PropTypes.bool.isRequired
}
// Good
const propTypes = {
/** The currency that the reward is in */
currency: React.PropTypes.string.isRequired,
/** The amount of the reward */
amount: React.PropTypes.number.isRequired,
/** If the reward has not been ignored yet */
isIgnored: React.PropTypes.bool.isRequired
}
All propTypes
and defaultProps
must be defined at the top of the file in variables called propTypes
and defaultProps
.
These variables should then be assigned to the component at the bottom of the file.
MyComponent.propTypes = propTypes;
MyComponent.defaultProps = defaultProps;
export default MyComponent;
Any nested propTypes
e.g. that may appear in a PropTypes.shape({})
should also be documented.
// Bad
const propTypes = {
/** Session data */
session: PropTypes.shape({
authToken: PropTypes.string,
login: PropTypes.string,
}),
}
// Good
const propTypes = {
/** Session data */
session: PropTypes.shape({
/** Token used to authenticate the user */
authToken: PropTypes.string,
/** User email or phone number */
login: PropTypes.string,
}),
}
- Use inline ternary statements when rendering optional pieces of templates. Notice the white space and formatting of the ternary.
// Bad
{
const optionalTitle = props.title ? <div className="title">{props.title}</div> : null;
return (
<div>
{optionalTitle}
<div className="body">This is the body</div>
</div>
);
}
// Good
{
return (
<div>
{props.title
? <div className="title">{props.title}</div>
: null}
<div className="body">This is the body</div>
</div>
);
}
// Good
{
return (
<div>
{props.title
? <div className="title">{props.title}</div>
: <div className="title">Default Title</div>
}
<div className="body">This is the body</body>
</div>
);
}
In React Native, one must not attempt to falsey-check a string for an inline ternary. Even if it's in curly braces, React Native will try to render it as a <Text>
node and most likely throw an error about trying to render text outside of a <Text>
component. Use _.isEmpty()
instead.
// Bad! This will cause a breaking an error on native platforms
{
return (
<View>
{props.title
? <View style={styles.title}>{props.title}</View>
: null}
<View style={styles.body}>This is the body</View>
</View>
);
}
// Good
{
return (
<View>
{!_.isEmpty(props.title)
? <View style={styles.title}>{props.title}</View>
: null}
<View style={styles.body}>This is the body</View>
</View>
);
}
When writing a function component you must ALWAYS add a displayName
property and give it the same value as the name of the component (this is so it appears properly in the React dev tools)
function Avatar(props) {...};
Avatar.propTypes = propTypes;
Avatar.defaultProps = defaultProps;
Avatar.displayName = 'Avatar';
export default Avatar;
Stateless components vs Pure Components vs Class based components vs Render Props - When to use what?
Class components are DEPRECATED. Use function components and React hooks.
From React's documentation -
Props and composition give you all the flexibility you need to customize a component’s look and behavior in an explicit and safe way. Remember that components may accept arbitrary props, including primitive values, React elements, or functions. If you want to reuse non-UI functionality between components, we suggest extracting it into a separate JavaScript module. The components may import it and use that function, object, or a class, without extending it.
Use an HOC a.k.a. Higher order component if you find a use case where you need inheritance.
If several HOC need to be combined there is a compose()
utility. But we should not use this utility when there is only one HOC.
// Bad
export default compose(
withLocalize,
)(MyComponent);
// Good
export default compose(
withLocalize,
withWindowDimensions,
)(MyComponent);
// Good
export default withLocalize(MyComponent)
Note: If you find that none of these approaches work for you, please ask an Expensify engineer for guidance via Slack or GitHub.
React's documentation explains refs in detail. It's important to understand when to use them and how to use them to avoid bugs and hard to maintain code.
A common mistake with refs is using them to pass data back to a parent component higher up the chain. In most cases, you can try lifting state up to solve this.
There are several ways to use and declare refs and we prefer the callback method.
We love React and learning about all the new features that are regularly being added to the API. However, we try to keep our organization's usage of React limited to the most stable set of features that React offers. We do this mainly for consistency and so our engineers don't have to spend extra time trying to figure out how everything is working. That said, if you aren't sure if we have adopted something please ask us first.
In most cases, a custom hook is a better pattern to use than an HOC or Render Prop. They are easier to create, understand, use and document. However, there might still be a case for a HOC e.g. if you have a component that abstracts some conditional rendering logic.
Should I wrap all my inline functions with useCallback()
or move them out of the component if they have no dependencies?
The answer depends on whether you need a stable reference for the function. If there are no dependencies, you could move the function out of the component. If there are dependencies, you could use useCallback()
to ensure the reference updates only when the dependencies change. However, it's important to note that using useCallback()
may have a performance penalty, although the trade-off is still debated. You might choose to do nothing at all if there is no obvious performance downside to declaring a function inline. It's recommended to follow the guidance in the React documentation and add the optimization only if necessary. If it's not obvious why such an optimization (i.e. useCallback()
or useMemo()
) would be used, leave a code comment explaining the reasoning to aid reviewers and future contributors.
React saves the initial state once and ignores it on the next renders. However, if you pass the result of a function to useState()
or call a function directly e.g. useState(doExpensiveThings())
it will still run on every render. This can hurt performance depending on what work the function is doing. As an optimization, we can pass an initializer function instead of a value e.g. useState(doExpensiveThings)
or useState(() => doExpensiveThings())
.
The short answer is no. A longer answer is that sometimes we need to check not only that a dependency has changed, but how it has changed in order to run a side effect. For example, a prop had a value of an empty string on a previous render, but now is non-empty. The generally accepted practice is to store the "previous" value in a ref
so the comparison can be made in a useEffect()
call.
No! It is easy to confuse useCallback()
with a memoization helper like _.memoize()
or useMemo()
but they are really not the same at all. useCallback()
will return a cached function definition and will not save us any computational cost of running that function. So, if you are wrapping something in a useCallback()
and then calling it in the render then it is better to use useMemo()
to cache the actual result of calling that function and use it directly in the render.
A useEffect()
that does not include referenced props or state in its dependency array is usually a mistake as often we want effects to re-run when those dependencies change. However, there are some cases where we might actually only want to re-run the effect when only some of those dependencies change. We determined the best practice here should be to allow disabling the “next line” with a comment //eslint-disable-next-line react-hooks/exhaustive-deps
and an additional comment explanation so the next developer can understand why the rule was not used.
There are pros and cons of each, but ultimately we have standardized on using the function
keyword to align things more with modern React conventions. There are also some minor cognitive overhead benefits in that you don't need to think about adding and removing brackets when encountering an implicit return. The function
syntax also has the benefit of being able to be hoisted where arrow functions do not.
Our potentially larger collections of data (reports, policies, etc) are typically stored under collection keys. Collection keys let us group together individual keys vs. storing arrays with multiple objects. In general, do not add a new collection key if it can be avoided. There is most likely a more logical place to put the state. And failing to associate a state property with it's logical owner is something we consider to be an anti-pattern (unnecessary data structure adds complexity for no value).
For example, if you are storing a boolean value that could be associated with a report
object under a new collection key it is better to associate this information with the report itself and not create a new collection key.
Exception: There are some gotchas when working with complex nested array values in Onyx. So, this could be another valid reason to break a property off of it's parent object (e.g. reportActions
are easier to work with as a separate collection).
If you're not sure whether something should have a collection key reach out in #expensify-open-source
for additional feedback.