rationale.md

id	title
rationale	Rationale

Prettier is an opinionated code formatter. This document explains some of its choices.

What Prettier is concerned about

Correctness

The first requirement of Prettier is to output valid code that has the exact same behavior as before formatting. Please report any code where Prettier fails to follow these correctness rules — that’s a bug which needs to be fixed!

Strings

Double or single quotes? Prettier chooses the one which results in the fewest number of escapes. "It's gettin' better!", not 'It\'s gettin\' better!'. In case of a tie or the string not containing any quotes, Prettier defaults to double quotes (but that can be changed via the singleQuote option).

JSX has its own option for quotes: jsxSingleQuote. JSX takes its roots from HTML, where the dominant use of quotes for attributes is double quotes. Browser developer tools also follow this convention by always displaying HTML with double quotes, even if the source code uses single quotes. A separate option allows using single quotes for JS and double quotes for "HTML" (JSX).

Prettier maintains the way your string is escaped. For example, "🙂" won’t be formatted into "\uD83D\uDE42" and vice versa.

Empty lines

It turns out that empty lines are very hard to automatically generate. The approach that Prettier takes is to preserve empty lines the way they were in the original source code. There are two additional rules:

• Prettier collapses multiple blank lines into a single blank line.
• Empty lines at the start and end of blocks (and whole files) are removed. (Files always end with a single newline, though.)

Multi-line objects

By default, Prettier’s printing algorithm prints expressions on a single line if they fit. Objects are used for a lot of different things in JavaScript, though, and sometimes it really helps readability if they stay multiline. See object lists, nested configs, stylesheets and keyed methods, for example. We haven’t been able to find a good rule for all those cases, so Prettier instead keeps objects multiline if there’s a newline between the { and the first key in the original source code. A consequence of this is that long singleline objects are automatically expanded, but short multiline objects are never collapsed.

Tip: If you have a multiline object that you’d like to join up into a single line:

const user = {
  name: "John Doe",
  age: 30,
};

…all you need to do is remove the newline after {:

const user = {  name: "John Doe",
  age: 30
};

…and then run Prettier:

const user = { name: "John Doe", age: 30 };

And if you’d like to go multiline again, add in a newline after {:

const user = {
 name: "John Doe", age: 30 };

…and run Prettier:

const user = {
  name: "John Doe",
  age: 30,
};

♻️ A note on formatting reversibility

The semi-manual formatting for object literals is in fact a workaround, not a feature. It was implemented only because at the time a good heuristic wasn’t found and an urgent fix was needed. However, as a general strategy, Prettier avoids non-reversible formatting like that, so the team is still looking for heuristics that would allow either to remove this behavior completely or at least to reduce the number of situations where it’s applied.

What does reversible mean? Once an object literal becomes multiline, Prettier won’t collapse it back. If in Prettier-formatted code, we add a property to an object literal, run Prettier, then change our mind, remove the added property, and then run Prettier again, we might end up with a formatting not identical to the initial one. This useless change might even get included in a commit, which is exactly the kind of situation Prettier was created to prevent.

Decorators

Just like with objects, decorators are used for a lot of different things. Sometimes it makes sense to write decorators above the line they're decorating, sometimes it’s nicer if they're on the same line. We haven’t been able to find a good rule for this, so Prettier keeps your decorator positioned like you wrote them (if they fit on the line). This isn’t ideal, but a pragmatic solution to a difficult problem.

@Component({
  selector: "hero-button",
  template: `<button>{{ label }}</button>`,
})
class HeroButtonComponent {
  // These decorators were written inline and fit on the line so they stay
  // inline.
  @Output() change = new EventEmitter();
  @Input() label: string;

  // These were written multiline, so they stay multiline.
  @readonly
  @nonenumerable
  NODE_TYPE: 2;
}

There’s one exception: classes. We don’t think it ever makes sense to inline the decorators for them, so they are always moved to their own line.

// Before running Prettier:
@observer class OrderLine {
  @observable price: number = 0;
}

// After running Prettier:
@observer
class OrderLine {
  @observable price: number = 0;
}

Note: Prettier 1.14.x and older tried to automatically move your decorators, so if you've run an older Prettier version on your code you might need to manually join up some decorators here and there to avoid inconsistencies:

@observer
class OrderLine {
  @observable price: number = 0;
  @observable
  amount: number = 0;
}

One final thing: TC39 has not yet decided if decorators come before or after export. In the meantime, Prettier supports both:

@decorator export class Foo {}

export @decorator class Foo {}

Semicolons

This is about using the noSemi option.

Consider this piece of code:

if (shouldAddLines) {
  [-1, 1].forEach(delta => addLine(delta * 20))
}

While the above code works just fine without semicolons, Prettier actually turns it into:

if (shouldAddLines) {
  ;[-1, 1].forEach(delta => addLine(delta * 20))
}

This is to help you avoid mistakes. Imagine Prettier not inserting that semicolon and adding this line:

 if (shouldAddLines) {
+  console.log('Do we even get here??')
   [-1, 1].forEach(delta => addLine(delta * 20))
 }

Oops! The above actually means:

if (shouldAddLines) {
  console.log('Do we even get here??')[-1, 1].forEach(delta => addLine(delta * 20))
}

With a semicolon in front of that [ such issues never happen. It makes the line independent of other lines so you can move and add lines without having to think about ASI rules.

This practice is also common in standard which uses a semicolon-free style.

Print width

The printWidth option is more of a guideline to Prettier than a hard rule. It is not the upper allowed line length limit. It is a way to say to Prettier roughly how long you’d like lines to be. Prettier will make both shorter and longer lines, but generally strive to meet the specified print width.

There are some edge cases, such as really long string literals, regexps, comments and variable names, which cannot be broken across lines (without using code transforms which Prettier doesn’t do). Or if you nest your code 50 levels deep your lines are of course going to be mostly indentation :)

Apart from that, there are a few cases where Prettier intentionally exceeds the print width.

Imports

Prettier can break long import statements across several lines:

import {
  CollectionDashboard,
  DashboardPlaceholder,
} from "../components/collections/collection-dashboard/main";

The following example doesn’t fit within the print width, but Prettier prints it in a single line anyway:

import { CollectionDashboard } from "../components/collections/collection-dashboard/main";

This might be unexpected by some, but we do it this way since it was a common request to keep imports with single elements in a single line. The same applies for require calls.

Testing functions

Another common request was to keep lengthy test descriptions in one line, even if it gets too long. In such cases, wrapping the arguments to new lines doesn’t help much.

describe("NodeRegistry", () => {
  it("makes no request if there are no nodes to prefetch, even if the cache is stale", async () => {
    // The above line exceeds the print width but stayed on one line anyway.
  });
});

Prettier has special cases for common testing framework functions such as describe, it and test.

JSX

Prettier prints things a little differently compared to other JS when JSX is involved:

function greet(user) {
  return user
    ? `Welcome back, ${user.name}!`
    : "Greetings, traveler! Sign up today!";
}

function Greet({ user }) {
  return (
    <div>
      {user ? (
        <p>Welcome back, {user.name}!</p>
      ) : (
        <p>Greetings, traveler! Sign up today!</p>
      )}
    </div>
  );
}

There are two reasons.

First off, lots of people already wrapped their JSX in parentheses, especially in return statements. Prettier follows this common style.

Secondly, the alternate formatting makes it easier to edit the JSX. It is easy to leave a semicolon behind. As opposed to normal JS, a leftover semicolon in JSX can end up as plain text showing on your page.

<div>
  <p>Greetings, traveler! Sign up today!</p>; {/* <-- Oops! */}
</div>

Comments

When it comes to the content of comments, Prettier can’t do much really. Comments can contain everything from prose to commented out code and ASCII diagrams. Since they can contain anything, Prettier can’t know how to format or wrap them. So they are left as-is. The only exception to this are JSDoc-style comments (block comments where every line starts with a *), which Prettier can fix the indentation of.

Then there’s the question of where to put the comments. Turns out this is a really difficult problem. Prettier tries its best to keep your comments roughly where they were, but it’s no easy task because comments can be placed almost anywhere.

Generally, you get the best results when placing comments on their own lines, instead of at the end of lines. Prefer // eslint-disable-next-line over // eslint-disable-line.

Note that “magic comments” such as eslint-disable-next-line and $FlowFixMe might sometimes need to be manually moved due to Prettier breaking an expression into multiple lines.

Imagine this piece of code:

// eslint-disable-next-line no-eval
const result = safeToEval ? eval(input) : fallback(input);

Then you need to add another condition:

// eslint-disable-next-line no-eval
const result = safeToEval && settings.allowNativeEval ? eval(input) : fallback(input);

Prettier will turn the above into:

// eslint-disable-next-line no-eval
const result =
  safeToEval && settings.allowNativeEval ? eval(input) : fallback(input);

Which means that the eslint-disable-next-line comment is no longer effective. In this case you need to move the comment:

const result =
  // eslint-disable-next-line no-eval
  safeToEval && settings.allowNativeEval ? eval(input) : fallback(input);

If possible, prefer comments that operate on line ranges (e.g. eslint-disable and eslint-enable) or on the statement level (e.g. /* istanbul ignore next */), they are even safer. It’s possible to disallow using eslint-disable-line and eslint-disable-next-line comments using eslint-plugin-eslint-comments.

Disclaimer about non-standard syntax

Prettier is often able to recognize and format non-standard syntax such as ECMAScript early-stage proposals and Markdown syntax extensions not defined by any specification. The support for such syntax is considered best-effort and experimental. Incompatibilities may be introduced in any release and should not be viewed as breaking changes.

What Prettier is not concerned about

Prettier only prints code. It does not transform it. This is to limit the scope of Prettier. Let’s focus on the printing and do it really well!

Here are a few examples of things that are out of scope for Prettier:

• Turning single- or double-quoted strings into template literals or vice versa.
• Using + to break long string literals into parts that fit the print width.
• Adding/removing {} and return where they are optional.
• Turning ?: into if-else statements.
• Sorting/moving imports, object keys, class members, JSX keys, CSS properties or anything else. Apart from being a transform rather than just printing (as mentioned above), sorting is potentially unsafe because of side effects (for imports, as an example) and makes it difficult to verify the most important correctness goal.