Overloaded signatures which are not next to each other can lead to code which is hard to read and maintain.

Invalid:

(bar is declared in-between foo overloads)

type FooType = {
  foo(s: string): void;
  foo(n: number): void;
  bar(): void;
  foo(sn: string | number): void;
};

interface FooInterface {
  foo(s: string): void;
  foo(n: number): void;
  bar(): void;
  foo(sn: string | number): void;
}

class FooClass {
  foo(s: string): void;
  foo(n: number): void;
  bar(): void {}
  foo(sn: string | number): void {}
}

export function foo(s: string): void;
export function foo(n: number): void;
export function bar(): void {}
export function foo(sn: string | number): void {}

Valid:

(bar is declared after foo)

type FooType = {
  foo(s: string): void;
  foo(n: number): void;
  foo(sn: string | number): void;
  bar(): void;
};

interface FooInterface {
  foo(s: string): void;
  foo(n: number): void;
  foo(sn: string | number): void;
  bar(): void;
}

class FooClass {
  foo(s: string): void;
  foo(n: number): void;
  foo(sn: string | number): void {}
  bar(): void {}
}

export function foo(s: string): void;
export function foo(n: number): void;
export function foo(sn: string | number): void {}
export function bar(): void {}

Typescript directives reduce the effectiveness of the compiler, something which should only be done in exceptional circumstances. The reason why should be documented in a comment alongside the directive.

Invalid:

// @ts-expect-error
let a: number = "I am a string";

// @ts-ignore
let a: number = "I am a string";

// @ts-nocheck
let a: number = "I am a string";

Valid:

// @ts-expect-error: Temporary workaround (see ticket #422)
let a: number = "I am a string";

// @ts-ignore: Temporary workaround (see ticket #422)
let a: number = "I am a string";

// @ts-nocheck: Temporary workaround (see ticket #422)
let a: number = "I am a string";

of string the primitive) in addition to the non-explicit Function type and the misunderstood Object type.

There are very few situations where primitive wrapper objects are desired and far more often a mistake was made with the case of the primitive type. You also cannot assign a primitive wrapper object to a primitive leading to type issues down the line. For reference, the TypeScript handbook also says we shouldn't ever use these wrapper objects.

With Function, it is better to explicitly define the entire function signature rather than use the non-specific Function type which won't give you type safety with the function.

Finally, Object and {} means "any non-nullish value" rather than "any object type". Record<string, unknown> is a good choice for a meaning of "any object type". On the other hand object type means "any object type", but using Record<string, unknown> is much more preferable in this case too.

Invalid:

let a: Boolean;
let b: String;
let c: Number;
let d: Symbol;
let e: Function;
let f: Object;
let g: object;
let h: {};

Valid:

let a: boolean;
let b: string;
let c: number;
let d: symbol;
let e: () => number;
let f: Record<string, unknown>;
let g = Record<string, never>;

We sometimes have to suppress and ignore lint errors for some reasons. We can do so using ignore directives with rule names that should be ignored like so:

// deno-lint-ignore no-explicit-any no-unused-vars
const foo: any = 42;

This rule checks for the validity of the specified rule names (i.e. whether deno_lint provides the rule or not).

Invalid:

// typo
// deno-lint-ignore no-extra-sem
export const a = 42;

// unknown rule name
// deno-lint-ignore UNKNOWN_RULE_NAME
const b = "b";

Valid:

// deno-lint-ignore no-extra-semi
export const a = 42;

// deno-lint-ignore no-unused-vars
const b = "b";

Ignoring all rules can mask unexpected or future problems. Therefore you need to explicitly specify which rule(s) are to be ignored.

Invalid:

// deno-lint-ignore
export function duplicateArgumentsFn(a, b, a) {}

Valid:

// deno-lint-ignore no-dupe-args
export function duplicateArgumentsFn(a, b, a) {}

We sometimes have to suppress and ignore lint errors for some reasons and we can do so using ignore directives.

In some cases, however, like after refactoring, we may end up having ignore directives that are no longer necessary. Such superfluous ignore directives are likely to confuse future code readers, and to make matters worse, might hide future lint errors unintentionally. To prevent such situations, this rule detects unused, superfluous ignore directives.

Invalid:

// Actually this line is valid since `export` means "used",
// so this directive is superfluous
// deno-lint-ignore no-unused-vars
export const foo = 42;

Valid:

export const foo = 42;

Defined constructors of derived classes (e.g. class A extends B) must always call super(). Classes which extend non-constructors (e.g. class A extends null) must not have a constructor.

Invalid:

class A {}
class Z {
  constructor() {}
}

class B extends Z {
  constructor() {} // missing super() call
}
class C {
  constructor() {
    super(); // Syntax error
  }
}
class D extends null {
  constructor() {} // illegal constructor
}
class E extends null {
  constructor() { // illegal constructor
    super();
  }
}

Valid:

class A {}
class B extends A {}
class C extends A {
  constructor() {
    super();
  }
}
class D extends null {}

Incrementing for loop control variables in the wrong direction leads to infinite loops. This can occur through incorrect initialization, bad continuation step logic or wrong direction incrementing of the loop control variable.

Invalid:

// Infinite loop
for (let i = 0; i < 2; i--) {}

Valid:

for (let i = 0; i < 2; i++) {}

Getter functions return the value of a property. If the function returns no value then this contract is broken.

Invalid:

let foo = {
  get bar() {},
};

class Person {
  get name() {}
}

Valid:

let foo = {
  get bar() {
    return true;
  },
};

class Person {
  get name() {
    return "alice";
  }
}

Array construction is conventionally done via literal notation such as [] or [1, 2, 3]. Using the new Array() is discouraged as is new Array(1, 2, 3). There are two reasons for this. The first is that a single supplied argument defines the array length, while multiple arguments instead populate the array of no fixed size. This confusion is avoided when pre-populated arrays are only created using literal notation. The second argument to avoiding the Array constructor is that the Array global may be redefined.

The one exception to this rule is when creating a new array of fixed size, e.g. new Array(6). This is the conventional way to create arrays of fixed length.

Invalid:

// This is 4 elements, not a size 100 array of 3 elements
const a = new Array(100, 1, 2, 3);

const b = new Array(); // use [] instead

Valid:

const a = new Array(100);
const b = [];
const c = [1, 2, 3];

Promise constructors take an executor function as an argument with resolve and reject parameters that can be used to control the state of the created Promise. This function is allowed to be async but this is generally not a good idea for several reasons:

• If an async executor function throws an error, the error will be lost and won't cause the newly-constructed Promise to reject. This could make it difficult to debug and handle some errors.
• If an async Promise executor function is using await, then this is usually a sign that it is not actually necessary to use the new Promise constructor and the code can be restructured to avoid the use of a promise, or the scope of the new Promise constructor can be reduced, extracting the async code and changing it to be synchronous.

Invalid:

new Promise(async function (resolve, reject) {});
new Promise(async (resolve, reject) => {});

Valid:

new Promise(function (resolve, reject) {});
new Promise((resolve, reject) => {});

case or default clauses to be scoped with brackets.

Without brackets in the case or default block, the lexical declarations are visible to the entire switch block but only get initialized when they are assigned, which only happens if that case/default is reached. This can lead to unexpected errors. The solution is to ensure each case or default block is wrapped in brackets to scope limit the declarations.

Invalid:

switch (choice) {
  // `let`, `const`, `function` and `class` are scoped the entire switch statement here
  case 1:
    let a = "choice 1";
    break;
  case 2:
    const b = "choice 2";
    break;
  case 3:
    function f() {
      return "choice 3";
    }
    break;
  default:
    class C {}
}

Valid:

switch (choice) {
  // The following `case` and `default` clauses are wrapped into blocks using brackets
  case 1: {
    let a = "choice 1";
    break;
  }
  case 2: {
    const b = "choice 2";
    break;
  }
  case 3: {
    function f() {
      return "choice 3";
    }
    break;
  }
  default: {
    class C {}
  }
}

Declaring a class such as class A {}, creates a variable A. Like any variable this can be modified or reassigned. In most cases this is a mistake and not what was intended.

Invalid:

class A {}
A = 0; // reassigning the class variable itself

Valid:

class A {}
let c = new A();
c = 0; // reassigning the variable `c`

Comparing a value directly against negative may not work as expected as it will also pass for non-negative zero (i.e. 0 and +0). Explicit comparison with negative zero can be performed using Object.is.

Invalid:

if (x === -0) {}

Valid:

if (x === 0) {}

if (Object.is(x, -0)) {}

Use of the assignment operator within a conditional statement is often the result of mistyping the equality operator, ==. If an assignment within a conditional statement is required then this rule allows it by wrapping the assignment in parentheses.

Invalid:

let x;
if (x = 0) {
  let b = 1;
}

function setHeight(someNode) {
  do {
    someNode.height = "100px";
  } while (someNode = someNode.parentNode);
}

Valid:

let x;
if (x === 0) {
  let b = 1;
}

function setHeight(someNode) {
  do {
    someNode.height = "100px";
  } while ((someNode = someNode.parentNode));
}

Using a constant expression in a conditional test is often either a mistake or a temporary situation introduced during development and is not ready for production.

Invalid:

if (true) {}
if (2) {}
do {} while (x = 2); // infinite loop

Valid:

if (x) {}
if (x === 0) {}
do {} while (x === 2);

Control characters are invisible characters in the ASCII range of 0-31. It is uncommon to use these in a regular expression and more often it is a mistake in the regular expression.

Invalid:

// Examples using ASCII (31) Carriage Return (hex x0d)
const pattern1 = /\x0d/;
const pattern2 = /\u000d/;
const pattern3 = new RegExp("\\x0d");
const pattern4 = new RegExp("\\u000d");

Valid:

// Examples using ASCII (32) Space (hex x20)
const pattern1 = /\x20/;
const pattern2 = /\u0020/;
const pattern3 = new RegExp("\\x20");
const pattern4 = new RegExp("\\u0020");

debugger is a statement which is meant for stopping the javascript execution environment and start the debugger at the statement. Modern debuggers and tooling no longer need this statement and leaving it in can cause the execution of your code to stop in production.

Invalid:

function isLongString(x: string) {
  debugger;
  return x.length > 100;
}

Valid:

function isLongString(x: string) {
  return x.length > 100; // set breakpoint here instead
}

delete is used to remove a property from an object. Variables declared via var, let and const cannot be deleted (delete will return false). Setting strict mode on will raise a syntax error when attempting to delete a variable.

Invalid:

const a = 1;
let b = 2;
let c = 3;
delete a; // would return false
delete b; // would return false
delete c; // would return false

Valid:

let obj = {
  a: 1,
};
delete obj.a; // return true

The following APIs in Deno namespace are now marked as deprecated and will get removed from the namespace in the future.

IO APIs

• Deno.Buffer
• Deno.readAll
• Deno.readAllSync
• Deno.writeAll
• Deno.writeAllSync
• Deno.iter
• Deno.iterSync
• Deno.copy

The IO APIs are already available in std/io, so replace these deprecated ones with alternatives from std/io. For more detail, see the tracking issue.

Custom Inspector API

• Deno.customInspect

Deno.customInspect was deprecated in favor of Symbol.for("Deno.customInspect"). Replace the usages with this symbol expression. See deno#9294 for more details.

Invalid:

// buffer
const a = Deno.Buffer();

// read
const b = await Deno.readAll(reader);
const c = Deno.readAllSync(reader);

// write
await Deno.writeAll(writer, data);
Deno.writeAllSync(writer, data);

// iter
for await (const x of Deno.iter(xs)) {}
for (const y of Deno.iterSync(ys)) {}

// copy
await Deno.copy(reader, writer);

// custom inspector
class A {
  [Deno.customInspect]() {
    return "This is A";
  }
}

Valid:

// buffer
import { Buffer } from "https://deno.land/std/io/buffer.ts";
const a = new Buffer();

// read
import { readAll, readAllSync } from "https://deno.land/std/io/util.ts";
const b = await readAll(reader);
const c = readAllSync(reader);

// write
import { writeAll, writeAllSync } from "https://deno.land/std/io/util.ts";
await writeAll(writer, data);
writeAllSync(writer, data);

// iter
import { iter, iterSync } from "https://deno.land/std/io/util.ts";
for await (const x of iter(xs)) {}
for (const y of iterSync(ys)) {}

// copy
import { copy } from "https://deno.land/std/io/util.ts";
await copy(reader, writer);

// custom inspector
class A {
  [Symbol.for("Deno.customInspect")]() {
    return "This is A";
  }
}

If you supply multiple arguments of the same name to a function, the last instance will shadow the preceding one(s). This is most likely an unintentional typo.

Invalid:

function withDupes(a, b, a) {
  console.log("I'm the value of the second a:", a);
}

Valid:

function withoutDupes(a, b, c) {
  console.log("I'm the value of the first (and only) a:", a);
}

Declaring a function of the same name twice in a class will cause the previous declaration(s) to be overwritten, causing unexpected behaviors.

Invalid:

class Foo {
  bar() {}
  bar() {}
}

Valid:

class Foo {
  bar() {}
  fizz() {}
}

When you reuse a condition in an if/else if statement, the duplicate condition will never be reached (without unusual side-effects) meaning this is almost always a bug.

Invalid:

if (a) {}
else if (b) {}
else if (a) {} // duplicate of condition above

if (a === 5) {}
else if (a === 6) {}
else if (a === 5) {} // duplicate of condition above

Valid:

if (a) {}
else if (b) {}
else if (c) {}

if (a === 5) {}
else if (a === 6) {}
else if (a === 7) {}

Setting the same key multiple times in an object literal will override other assignments to that key and can cause unexpected behaviour.

Invalid:

const foo = {
  bar: "baz",
  bar: "qux",
};

const foo = {
  "bar": "baz",
  bar: "qux",
};

const foo = {
  0x1: "baz",
  1: "qux",
};

Valid:

const foo = {
  bar: "baz",
  quxx: "qux",
};

When you reuse a case test expression in a switch statement, the duplicate case will never be reached meaning this is almost always a bug.

Invalid:

const someText = "a";
switch (someText) {
  case "a": // (1)
    break;
  case "b":
    break;
  case "a": // duplicate of (1)
    break;
  default:
    break;
}

Valid:

const someText = "a";
switch (someText) {
  case "a":
    break;
  case "b":
    break;
  case "c":
    break;
  default:
    break;
}

Empty block statements are legal but often represent that something was missed and can make code less readable. This rule ignores block statements that only contain comments. This rule also ignores empty constructors and function bodies (including arrow functions), which are covered by the no-empty-function rule.

Invalid:

if (foo) {}

while (foo) {}

switch (foo) {}

try {
  doSomething();
} catch (e) {
} finally {
}

Valid:

if (foo) {
  // empty
}

while (foo) {
  /* empty */
}

try {
  doSomething();
} catch (e) {
  // continue regardless of error
}

try {
  doSomething();
} finally {
  /* continue regardless of error */
}

Regular expression character classes are a series of characters in brackets, e.g. [abc]. if nothing is supplied in the brackets it will not match anything which is likely a typo or mistake.

Invalid:

/^abc[]/.test("abcdefg"); // false, as `d` does not match an empty character class
"abcdefg".match(/^abc[]/); // null

Valid:

// Without a character class
/^abc/.test("abcdefg"); // true
"abcdefg".match(/^abc/); // ["abc"]

// With a valid character class
/^abc[a-z]/.test("abcdefg"); // true
"abcdefg".match(/^abc[a-z]/); // ["abcd"]

An enum with no members serves no purpose. This rule will capture these situations as either unnecessary code or a mistaken empty implementation.

Invalid:

enum Foo {}

Valid:

enum Foo {
  ONE = "ONE",
}

An interface with no members serves no purpose. Either the interface extends another interface, in which case the supertype can be used, or it does not extend a supertype in which case it is the equivalent to an empty object. This rule will capture these situations as either unnecessary code or a mistaken empty implementation.

Invalid:

interface Foo {}
interface Foo extends Bar {}

Valid:

interface Foo {
  name: string;
}

interface Bar {
  age: number;
}

// Using an empty interface as a union type is allowed
interface Baz extends Foo, Bar {}

In destructuring, it is possible to use empty patterns such as {} or [] which have no effect, most likely not what the author intended.

Invalid:

// In these examples below, {} and [] are not object literals or empty arrays,
// but placeholders for destructured variable names
const {} = someObj;
const [] = someArray;
const {a: {}} = someObj;
const [a: []] = someArray;
function myFunc({}) {}
function myFunc([]) {}

Valid:

const { a } = someObj;
const [a] = someArray;

// Correct way to default destructured variable to object literal
const { a = {} } = someObj;

// Correct way to default destructured variable to empty array
const [a = []] = someArray;

function myFunc({ a }) {}
function myFunc({ a = {} }) {}
function myFunc([a]) {}
function myFunc([a = []]) {}

There is generally no good reason to reassign an exception parameter. Once reassigned the code from that point on has no reference to the error anymore.

Invalid:

try {
  someFunc();
} catch (e) {
  e = true;
  // can no longer access the thrown error
}

Valid:

try {
  someFunc();
} catch (e) {
  const anotherVar = true;
}

Use of the any type disables the type check system around that variable, defeating the purpose of Typescript which is to provide type safe code. Additionally, the use of any hinders code readability, since it is not immediately clear what type of value is being referenced. It is better to be explicit about all types. For a more type-safe alternative to any, use unknown if you are unable to choose a more specific type.

Invalid:

const someNumber: any = "two";
function foo(): any {
  return undefined;
}

Valid:

const someNumber: string = "two";
function foo(): undefined {
  return undefined;
}

In certain contexts, such as if, while or for statements, expressions are automatically coerced into a boolean. Therefore, techniques such as double negation (!!foo) or casting (Boolean(foo)) are unnecessary and produce the same result as without the negation or casting.

Invalid:

if (!!foo) {}
if (Boolean(foo)) {}
while (!!foo) {}
for (; Boolean(foo);) {}

Valid:

if (foo) {}
while (foo) {}
for (; foo;) {}

Non-null assertions are specified with an ! saying to the compiler that you know this value is not null. Specifying this operator more than once in a row, or in combination with the optional chaining operator (?) is confusing and unnecessary.

Invalid:

const foo: { str: string } | null = null;
const bar = foo!!.str;

function myFunc(bar: undefined | string) {
  return bar!!;
}
function anotherFunc(bar?: { str: string }) {
  return bar!?.str;
}

Valid:

const foo: { str: string } | null = null;
const bar = foo!.str;

function myFunc(bar: undefined | string) {
  return bar!;
}
function anotherFunc(bar?: { str: string }) {
  return bar?.str;
}

Disallows the use of unnecessary semi-colons

Extra (and unnecessary) semi-colons can cause confusion when reading the code as well as making the code less clean.

Invalid:

const x = 5;;

function foo() {};

Valid:

const x = 5;

function foo() {}

Case statements without a break will execute their body and then fallthrough to the next case or default block and execute this block as well. While this is sometimes intentional, many times the developer has forgotten to add a break statement, intending only for a single case statement to be executed. This rule enforces that you either end each case statement with a break statement or an explicit comment that fallthrough was intentional. The fallthrough comment must contain one of fallthrough, falls through or fall through.

Invalid:

switch (myVar) {
  case 1:
    console.log("1");

  case 2:
    console.log("2");
}
// If myVar = 1, outputs both `1` and `2`.  Was this intentional?

Valid:

switch (myVar) {
  case 1:
    console.log("1");
    break;

  case 2:
    console.log("2");
    break;
}
// If myVar = 1, outputs only `1`

switch (myVar) {
  case 1:
    console.log("1");
    /* falls through */
  case 2:
    console.log("2");
}
// If myVar = 1, intentionally outputs both `1` and `2`

Javascript allows for the reassignment of a function definition. This is generally a mistake on the developers part, or poor coding practice as code readability and maintainability will suffer.

Invalid:

function foo() {}
foo = bar;

const a = function baz() {
  baz = "now I'm a string";
};

myFunc = existingFunc;
function myFunc() {}

Valid:

function foo() {}
const someVar = foo;

const a = function baz() {
  const someStr = "now I'm a string";
};

const anotherFuncRef = existingFunc;

let myFuncVar = function () {};
myFuncVar = bar; // variable reassignment, not function re-declaration

In Javascript, String and Object for example are native objects. Like any object, they can be reassigned, but it is almost never wise to do so as this can lead to unexpected results and difficult to track down bugs.

Invalid:

Object = null;
undefined = true;
window = {};

ES module import bindings should be treated as read-only since modifying them during code execution will likely result in runtime errors. It also makes for poor code readability and difficult maintenance.

Invalid:

import defaultMod, { namedMod } from "./mod.js";
import * as modNameSpace from "./mod2.js";

defaultMod = 0;
namedMod = true;
modNameSpace.someExportedMember = "hello";
modNameSpace = {};

Valid:

import defaultMod, { namedMod } from "./mod.js";
import * as modNameSpace from "./mod2.js";

// properties of bound imports may be set
defaultMod.prop = 1;
namedMod.prop = true;
modNameSpace.someExportedMember.prop = "hello";

Variable initializations to JavaScript primitives (and null) are obvious in their type. Specifying their type can add additional verbosity to the code. For example, with const x: number = 5, specifying number is unnecessary as it is obvious that 5 is a number.

Invalid:

const a: bigint = 10n;
const b: bigint = BigInt(10);
const c: boolean = true;
const d: boolean = !0;
const e: number = 10;
const f: number = Number("1");
const g: number = Infinity;
const h: number = NaN;
const i: null = null;
const j: RegExp = /a/;
const k: RegExp = RegExp("a");
const l: RegExp = new RegExp("a");
const m: string = "str";
const n: string = `str`;
const o: string = String(1);
const p: symbol = Symbol("a");
const q: undefined = undefined;
const r: undefined = void someValue;

class Foo {
  prop: number = 5;
}

function fn(s: number = 5, t: boolean = true) {}

Valid:

const a = 10n;
const b = BigInt(10);
const c = true;
const d = !0;
const e = 10;
const f = Number("1");
const g = Infinity;
const h = NaN;
const i = null;
const j = /a/;
const k = RegExp("a");
const l = new RegExp("a");
const m = "str";
const n = `str`;
const o = String(1);
const p = Symbol("a");
const q = undefined;
const r = void someValue;

class Foo {
  prop = 5;
}

function fn(s = 5, t = true) {}

Function declarations in nested blocks can lead to less readable code and potentially unexpected results due to compatibility issues in different JavaScript runtimes. This does not apply to named or anonymous functions which are valid in a nested block context.

Variables declared with var in nested blocks can also lead to less readable code. Because these variables are hoisted to the module root, it is best to declare them there for clarity. Note that variables declared with let or const are block scoped and therefore this rule does not apply to them.

Invalid:

if (someBool) {
  function doSomething() {}
}

function someFunc(someVal: number): void {
  if (someVal > 4) {
    var a = 10;
  }
}

Valid:

function doSomething() {}
if (someBool) {}

var a = 10;
function someFunc(someVal: number): void {
  var foo = true;
  if (someVal > 4) {
    let b = 10;
    const fn = function doSomethingElse() {};
  }
}

Specifying an invalid regular expression literal will result in a SyntaxError at compile time, however specifying an invalid regular expression string in the RegExp constructor will only be discovered at runtime.

Invalid:

const invalidRegExp = new RegExp(")");

Valid:

const goodRegExp = new RegExp(".");

Deno supports the triple-slash reference directives of types, path, lib, and no-default-lib. This lint rule checks if there is an invalid, badly-formed directive because it is most likely a mistake.

Additionally, note that only the types directive is allowed in JavaScript files. This directive is useful for telling the TypeScript compiler the location of a type definition file that corresponds to a certain JavaScript file. However, even in the Deno manual of the versions prior to v1.10 (e.g. v1.9.2), there was a wrong statement describing that one should use the path directive in such cases. Actually, the types directive should be used. See the latest manual for more detail. So this rule also detects the usage of the directive other than types in JavaScript files and suggests replacing it with the types directive.

Invalid:

JavaScript

/// <reference path="./mod.d.ts" />
/// <reference no-default-lib="true" />
/// <reference foo="bar" />

// ... the rest of the JavaScript ...

TypeScript

/// <reference foo="bar" />

// ... the rest of the TypeScript ...

Valid:

JavaScript

/// <reference types="./mod.d.ts" />
/// <reference lib="es2017.string" />

// ... the rest of the JavaScript ...

TypeScript

/// <reference types="./mod.d.ts" />
/// <reference path="./mod.d.ts" />
/// <reference lib="es2017.string" />
/// <reference no-default-lib="true" />

// ... the rest of the TypeScript ...

Non-space or non-tab whitespace characters can be very difficult to spot in your code as editors will often render them invisibly. These invisible characters can cause issues or unexpected behaviors. Sometimes these characters are added inadvertently through copy/paste or incorrect keyboard shortcuts.

The following characters are disallowed:

\u000B - Line Tabulation (\v) - <VT>
\u000C - Form Feed (\f) - <FF>
\u00A0 - No-Break Space - <NBSP>
\u0085 - Next Line
\u1680 - Ogham Space Mark
\u180E - Mongolian Vowel Separator - <MVS>
\ufeff - Zero Width No-Break Space - <BOM>
\u2000 - En Quad
\u2001 - Em Quad
\u2002 - En Space - <ENSP>
\u2003 - Em Space - <EMSP>
\u2004 - Tree-Per-Em
\u2005 - Four-Per-Em
\u2006 - Six-Per-Em
\u2007 - Figure Space
\u2008 - Punctuation Space - <PUNCSP>
\u2009 - Thin Space
\u200A - Hair Space
\u200B - Zero Width Space - <ZWSP>
\u2028 - Line Separator
\u2029 - Paragraph Separator
\u202F - Narrow No-Break Space
\u205f - Medium Mathematical Space
\u3000 - Ideographic Space

To fix this linting issue, replace instances of the above with regular spaces, tabs or new lines. If it's not obvious where the offending character(s) are try retyping the line from scratch.

Specifying a constructor for an interface or defining a new method for a class is incorrect and should be avoided.

Invalid:

class C {
  new(): C;
}

interface I {
  constructor(): void;
}

Valid:

class C {
  constructor() {}
}

interface I {
  new (): C;
}

namespace and module are both thought of as outdated keywords to organize the code. Instead, it is generally preferable to use ES2015 module syntax (e.g. import/export).

However, this rule still allows the use of these keywords in the following two cases:

• they are used for defining "ambient" namespaces along with declare keywords
• they are written in TypeScript's type definition files: .d.ts

Invalid:

// foo.ts
module mod {}
namespace ns {}

// bar.d.ts
// all usage of `module` and `namespace` keywords are allowed in `.d.ts`

Valid:

// foo.ts
declare global {}
declare module mod1 {}
declare module "mod2" {}
declare namespace ns {}

// bar.d.ts
module mod1 {}
namespace ns1 {}
declare global {}
declare module mod2 {}
declare module "mod3" {}
declare namespace ns2 {}

Symbols are created by being called as a function, but we sometimes call it with the new operator by mistake. This rule detects such wrong usage of the new operator.

Invalid:

const foo = new Symbol("foo");

Valid:

const foo = Symbol("foo");

function func(Symbol: typeof SomeClass) {
  // This `Symbol` is not built-in one
  const bar = new Symbol();
}

The following built-in objects should not be invoked like functions, even though they look like constructors:

• Math
• JSON
• Reflect
• Atomics

Calling these as functions would result in runtime errors. This rule statically prevents such wrong usage of them.

Invalid:

const math = Math();
const newMath = new Math();

const json = JSON();
const newJSON = new JSON();

const reflect = Reflect();
const newReflect = new Reflect();

const atomics = Atomics();
const newAtomics = new Atomics();

Valid:

const area = (radius: number): number => Math.PI * radius * radius;

const parsed = JSON.parse("{ foo: 42 }");

const x = Reflect.get({ x: 1, y: 2 }, "x");

const first = Atomics.load(foo, 0);

Octal numbers can be expressed via numeric literals with leading 0 like 042, but this expression often confuses programmers. That's why ECMAScript's strict mode throws SyntaxError for the expression.

Since ES2015, the other prefix 0o has been introduced as an alternative. This new one is always encouraged to use in today's code.

Invalid:

const a = 042;
const b = 7 + 042;

Valid:

const a = 0o42;
const b = 7 + 0o42;
const c = "042";

If objects are created via Object.create(null) they have no prototype specified. This can lead to runtime errors when you assume objects have properties from Object.prototype and attempt to call the following methods:

• hasOwnProperty
• isPrototypeOf
• propertyIsEnumerable

Instead, it's always encouraged to call these methods from Object.prototype explicitly.

Invalid:

const a = foo.hasOwnProperty("bar");
const b = foo.isPrototypeOf("bar");
const c = foo.propertyIsEnumerable("bar");

Valid:

const a = Object.prototype.hasOwnProperty.call(foo, "bar");
const b = Object.prototype.isPrototypeOf.call(foo, "bar");
const c = Object.prototype.propertyIsEnumerable.call(foo, "bar");

JavaScript allows us to redeclare variables with the same name using var, but redeclaration should not be used since it can make variables hard to trace.

In addition, this lint rule disallows redeclaration using let or const as well, although ESLint allows. This is useful because we can notice a syntax error before actually running the code.

As for functions and parameters, JavaScript just treats these as runtime errors, throwing SyntaxError when being run. It's also beneficial to detect this sort of errors statically.

Invalid:

var a = 3;
var a = 10;

let b = 3;
let b = 10;

const c = 3;
const c = 10;

function d() {}
function d() {}

function e(arg: number) {
  var arg: number;
}

function f(arg: number, arg: string) {}

Valid:

var a = 3;
function f() {
  var a = 10;
}

if (foo) {
  let b = 2;
} else {
  let b = 3;
}

Multiple spaces in regular expression literals are generally hard to read when the regex gets complicated. Instead, it's better to use only one space character and specify how many times spaces should appear with the {n} syntax, for example:

// Multiple spaces in the regex literal are harder to understand how many
// spaces are expected to be matched
const re = /foo   bar/;

// Instead use `{n}` syntax for readability
const re = /foo {3}var/;

Invalid:

const re1 = /  /;
const re2 = /foo  bar/;
const re3 = / a b  c d /;
const re4 = /foo  {3}bar/;

const re5 = new RegExp("  ");
const re6 = new RegExp("foo  bar");
const re7 = new RegExp(" a b  c d ");
const re8 = new RegExp("foo  {3}bar");

Valid:

const re1 = /foo/;
const re2 = / /;
const re3 = / {3}/;
const re4 = / +/;
const re5 = / ?/;
const re6 = / */;

const re7 = new RegExp("foo");
const re8 = new RegExp(" ");
const re9 = new RegExp(" {3}");
const re10 = new RegExp(" +");
const re11 = new RegExp(" ?");
const re12 = new RegExp(" *");

Self assignments like a = a; have no effect at all. If there are self assignments in the code, most likely it means that the author is still in the process of refactoring and there's remaining work they have to do.

Invalid:

a = a;
[a] = [a];
[a, b] = [a, b];
[a, b] = [a, c];
[a, ...b] = [a, ...b];
a.b = a.b;

Valid:

let a = a;
a += a;
a = [a];
[a, b] = [b, a];
a.b = a.c;

Setters are supposed to be used for setting some value to the property, which means that returning a value from a setter makes no sense. In fact, returned values are ignored and cannot ever be used at all although returning a value from a setter produces no error. This is why static check for this mistake by the linter is quite beneficial.

Note that returning without a value is allowed; this is a useful technique to do early-return from a function.

Invalid:

const a = {
  set foo(x: number) {
    return "something";
  },
};

class B {
  private set foo(x: number) {
    return "something";
  }
}

const c = {
  set foo(x: boolean) {
    if (x) {
      return 42;
    }
  },
};

Valid:

// return without a value is allowed since it is used to do early-return
const a = {
  set foo(x: number) {
    if (x % 2 == 0) {
      return;
    }
  },
};

// not a setter, but a getter
class B {
  get foo() {
    return 42;
  }
}

// not a setter
const c = {
  set(x: number) {
    return "something";
  },
};

The following (a) properties of the global object, or (b) identifiers are "restricted" names in JavaScript:

• NaN
• Infinity
• undefined
• eval
• arguments

These names are NOT reserved in JavaScript, which means that nothing prevents one from assigning other values into them (i.e. shadowing). In other words, you are allowed to use, say, undefined as an identifier or variable name. (For more details see MDN)

function foo() {
  const undefined = "bar";
  console.log(undefined); // output: "bar"
}

Of course, shadowing like this most likely confuse other developers and should be avoided. This lint rule detects and warn them.

Invalid:

const undefined = 42;

function NaN() {}

function foo(Infinity) {}

const arguments = () => {};

try {
} catch (eval) {}

Valid:

// If not assigned a value, `undefined` may be shadowed
const undefined;

const Object = 42;

function foo(a: number, b: string) {}

try {
} catch (e) {}

In most cases, storing a reference to this in a variable could be avoided by using arrow functions properly, since they establish this based on the scope where the arrow function is defined.

Let's take a look at a concrete example:

const obj = {
  count: 0,
  doSomethingLater() {
    setTimeout(function () { // this function executes on the global scope; `this` evalutes to `globalThis`
      this.count++;
      console.log(this.count);
    }, 300);
  },
};

obj.doSomethingLater();
// `NaN` is printed, because the property `count` is not in the global scope.

In the above example, this in the function passed to setTimeout evaluates to globalThis, which results in the expected value 1 not being printed.

If you wanted to work around it without arrow functions, you would store a reference to this in another variable:

const obj = {
  count: 0,
  doSomethingLater() {
    const self = this; // store a reference to `this` in `self`
    setTimeout(function () {
      // use `self` instead of `this`
      self.count++;
      console.log(self.count);
    }, 300);
  },
};

obj.doSomethingLater();
// `1` is printed as expected

But in this case arrow functions come in handy. With arrow functions, the code becomes way clearer and easier to understand:

const obj = {
  count: 0,
  doSomethingLater() {
    setTimeout(() => { // pass an arrow function
      // `this` evaluates to `obj` here
      this.count++;
      console.log(this.count);
    }, 300);
  },
};

obj.doSomethingLater();
// `1` is printed as expected

This example is taken from MDN.

Invalid:

const self = this;

function foo() {
  const self = this;
}

const bar = () => {
  const self = this;
};

Valid:

const self = "this";

const [foo] = this;

The access to this or super before calling super() in the constructor of derived classes leads to ReferenceError. To prevent it, this lint rule checks if there are accesses to this or super before calling super() in constructors.

Invalid:

class A extends B {
  constructor() {
    this.foo = 0;
    super();
  }
}

class C extends D {
  constructor() {
    super.foo();
    super();
  }
}

Valid:

class A extends B {
  constructor() {
    super();
    this.foo = 0;
  }
}

class C extends D {
  constructor() {
    super();
    super.foo();
  }
}

class E {
  constructor() {
    this.foo = 0;
  }
}

Because the control flow statements (return, throw, break and continue) unconditionally exit a block of code, any statements after them cannot be executed.

Invalid:

function foo() {
  return true;
  console.log("done");
}

function bar() {
  throw new Error("Oops!");
  console.log("done");
}

while (value) {
  break;
  console.log("done");
}

throw new Error("Oops!");
console.log("done");

function baz() {
  if (Math.random() < 0.5) {
    return;
  } else {
    throw new Error();
  }
  console.log("done");
}

for (;;) {}
console.log("done");

Valid

function foo() {
  return bar();
  function bar() {
    return 1;
  }
}

Use of the control flow statements (return, throw, break and continue) overrides the usage of any control flow statements that might have been used in the try or catch blocks, which is usually not the desired behaviour.

Invalid:

let foo = function () {
  try {
    return 1;
  } catch (err) {
    return 2;
  } finally {
    return 3;
  }
};

let foo = function () {
  try {
    return 1;
  } catch (err) {
    return 2;
  } finally {
    throw new Error();
  }
};

Valid:

let foo = function () {
  try {
    return 1;
  } catch (err) {
    return 2;
  } finally {
    console.log("hola!");
  }
};

operators.

! operators appearing in the left operand of the following operators will sometimes cause an unexpected behavior because of the operator precedence:

• in operator
• instanceof operator

For example, when developers write a code like !key in someObject, most likely they want it to behave just like !(key in someObject), but actually it behaves like (!key) in someObject. This lint rule warns such usage of ! operator so it will be less confusing.

Invalid:

if (!key in object) {}
if (!foo instanceof Foo) {}

Valid:

if (!(key in object)) {}
if (!(foo instanceof Foo)) {}
if ((!key) in object) {}
if ((!foo) instanceof Foo) {}

A label that is declared but never used is most likely developer's mistake. If that label is meant to be used, then write a code so that it will be used. Otherwise, remove the label.

Invalid:

LABEL1:
while (true) {
  console.log(42);
}

LABEL2:
for (let i = 0; i < 5; i++) {
  console.log(42);
}

LABEL3:
for (const x of xs) {
  console.log(x);
}

Valid:

LABEL1:
while (true) {
  console.log(42);
  break LABEL1;
}

LABEL2:
for (let i = 0; i < 5; i++) {
  console.log(42);
  continue LABEL2;
}

for (const x of xs) {
  console.log(x);
}

If there are variables that are declared but not used anywhere, it's most likely because of incomplete refactoring. This lint rule detects and warns such unused variables.

Variable a is considered to be "used" if any of the following conditions are satisfied:

• its value is read out, like console.log(a) or let otherVariable = a;
• it's called or constructed, like a() or new a()
• it's exported, like export const a = 42;

If a variable is just assigned to a value but never read out, then it's considered to be "not used".

let a;
a = 42;

// `a` is never read out

If you want to declare unused variables intentionally, prefix them with the underscore character _, like _a. This rule ignores variables that are prefixed with _.

Invalid:

const a = 0;

const b = 0; // this `b` is never used
function foo() {
  const b = 1; // this `b` is used
  console.log(b);
}
foo();

let c = 2;
c = 3;

// recursive function calls are not considered to be used, because only when `d`
// is called from outside the function body can we say that `d` is actually
// called after all.
function d() {
  d();
}

// `x` is never used
export function e(x: number): number {
  return 42;
}

const f = "unused variable";

Valid:

const a = 0;
console.log(a);

const b = 0;
function foo() {
  const b = 1;
  console.log(b);
}
foo();
console.log(b);

let c = 2;
c = 3;
console.log(c);

function d() {
  d();
}
d();

export function e(x: number): number {
  return x + 42;
}

export const f = "exported variable";

variables. Block scoped variables are defined using const and let keywords.

const and let keywords ensure the variables defined using these keywords are not accessible outside their block scope. On the other hand, variables defined using var keyword are only limited by their function scope.

Invalid:

var foo = "bar";

Valid:

const foo = 1;
let bar = 2;

In most situations, the global variable window works like globalThis. For example, you could call the fetch API like window.fetch(..) instead of fetch(..) or globalThis.fetch(..). In Web Workers, however, window is not available, but instead self, globalThis, or no prefix work fine. Therefore, for compatibility between Web Workers and other contexts, it's highly recommended to not access global properties via window.

Some APIs, including window.alert, window.location and window.history, are allowed to call with window because these APIs are not supported or have different meanings in Workers. In other words, this lint rule complains about the use of window only if it's completely replaceable with self, globalThis, or no prefix.

Invalid:

const a = await window.fetch("https://deno.land");

const b = window.Deno.metrics();

Valid:

const a1 = await fetch("https://deno.land");
const a2 = await globalThis.fetch("https://deno.land");
const a3 = await self.fetch("https://deno.land");

const b1 = Deno.metrics();
const b2 = globalThis.Deno.metrics();
const b3 = self.Deno.metrics();

// `alert` is allowed to call with `window` because it's not supported in Workers
window.alert("🍣");

// `location` is also allowed
window.location.host;

The with statement is discouraged as it may be the source of confusing bugs and compatibility issues. For more details, see with - JavaScript | MDN.

Invalid:

with (someVar) {
  console.log("foo");
}

types or using type assertion.

When declaring a new variable of a primitive literal type, there are three ways:

1. adding an explicit type annotation
2. using normal type assertion (like as "foo", or <"foo">)
3. using const assertion (as const)

This lint rule suggests using const assertion because it will generally lead to a safer code. For more details about const assertion, see the official handbook.

Invalid:

let a: 2 = 2; // type annotation
let b = 2 as 2; // type assertion
let c = <2> 2; // type assertion
let d = { foo: 1 as 1 }; // type assertion

Valid:

let a = 2 as const;
let b = 2 as const;
let c = 2 as const;
let d = { foo: 1 as const };

let x = 2;
let y: string = "hello";
let z: number = someVariable;

Since ES2015, JavaScript supports let and const for declaring variables. If variables are declared with let, then they become mutable; we can set other values to them afterwards. Meanwhile, if declared with const, they are immutable; we cannot perform re-assignment to them.

In general, to make the codebase more robust, maintainable, and readable, it is highly recommended to use const instead of let wherever possible. The fewer mutable variables are, the easier it should be to keep track of the variable states while reading through the code, and thus it is less likely to write buggy code. So this lint rule checks if there are let variables that could potentially be declared with const instead.

Note that this rule does not check for var variables. Instead, the no-var rule is responsible for detecting and warning var variables.

Invalid:

let a = 0;

let b = 0;
someOperation(b);

// `const` could be used instead
for (let c in someObject) {}

// `const` could be used instead
for (let d of someArray) {}

// variable that is uninitialized at first and then assigned in the same scope is NOT allowed
// because we could simply write it like `const e = 2;` instead
let e;
e = 2;

Valid:

// uninitialized variable is allowed
let a;

let b = 0;
b += 1;

let c = 0;
c = 1;

// variable that is uninitialized at first and then assigned in the same scope _two or more times_ is allowed
// because we cannot represent it with `const`
let d;
d = 2;
d = 3;

const e = 0;

// `f` is mutated through `f++`
for (let f = 0; f < someArray.length; f++) {}

// variable that is initialized (or assigned) in another scope is allowed
let g;
function func1() {
  g = 42;
}

// conditionally initialized variable is allowed
let h;
if (trueOrFalse) {
  h = 0;
}

TypeScript module.

TypeScript supports the module keyword for organizing code, but this wording can lead to a confusion with the ECMAScript's module. Since TypeScript v1.5, it has provided us with the alternative keyword namespace, encouraging us to always use namespace instead whenever we write TypeScript these days. See TypeScript v1.5 release note for more details.

Invalid:

module modA {}

declare module modB {}

Valid:

namespace modA {}

// "ambient modules" are allowed
// https://www.typescriptlang.org/docs/handbook/modules.html#ambient-modules
declare module "modB";
declare module "modC" {}

In general, the primary reason to use async functions is to use await expressions inside. If an async function has no await expression, it is most likely an unintentional mistake.

Invalid:

async function f1() {
  doSomething();
}

const f2 = async () => {
  doSomething();
};

const f3 = async () => doSomething();

const obj = {
  async method() {
    doSomething();
  },
};

class MyClass {
  async method() {
    doSomething();
  }
}

Valid:

await asyncFunction();

function normalFunction() {
  doSomething();
}

async function f1() {
  await asyncFunction();
}

const f2 = async () => {
  await asyncFunction();
};

const f3 = async () => await asyncFunction();

async function f4() {
  for await (const num of asyncIterable) {
    console.log(num);
  }
}

// empty functions are valid
async function emptyFunction() {}
const emptyArrowFunction = async () => {};

// generators are also valid
async function* gen() {
  console.log(42);
}

JavaScript provides generator functions expressed as function*, where we can pause and later resume the function execution at the middle points. At these points we use the yield keyword. In other words, it makes no sense at all to create generator functions that contain no yield keyword, since such functions could be written as normal functions.

Invalid:

function* f1() {
  return "f1";
}

Valid:

function* f1() {
  yield "f1";
}

// generator function with empty body is allowed
function* f2() {}

function f3() {
  return "f3";
}

Because NaN is unique in JavaScript by not being equal to anything, including itself, the results of comparisons to NaN are confusing:

• NaN === NaN or NaN == NaN evaluate to false
• NaN !== NaN or NaN != NaN evaluate to true

Therefore, this rule makes you use the isNaN() or Number.isNaN() to judge the value is NaN or not.

Invalid:

if (foo == NaN) {
  // ...
}

if (foo != NaN) {
  // ...
}

switch (NaN) {
  case foo:
    // ...
}

switch (foo) {
  case NaN:
    // ...
}

Valid:

if (isNaN(foo)) {
  // ...
}

if (!isNaN(foo)) {
  // ...
}

When used with a value the typeof operator returns one of the following strings:

• "undefined"
• "object"
• "boolean"
• "number"
• "string"
• "function"
• "symbol"
• "bigint"

This rule disallows comparison with anything other than one of these string literals when using the typeof operator, as this likely represents a typing mistake in the string. The rule also disallows comparing the result of a typeof operation with any non-string literal value, such as undefined, which can represent an inadvertent use of a keyword instead of a string. This includes comparing against string variables even if they contain one of the above values as this cannot be guaranteed. An exception to this is comparing the results of two typeof operations as these are both guaranteed to return on of the above strings.

Invalid:

// typo
typeof foo === "strnig";
typeof foo == "undefimed";
typeof bar != "nunber";
typeof bar !== "fucntion";

// compare with non-string literals
typeof foo === undefined;
typeof bar == Object;
typeof baz === anotherVariable;
typeof foo == 5;

Valid:

typeof foo === "undefined";
typeof bar == "object";
typeof baz === "string";
typeof bar === typeof qux;