no-unnecessary-type-parameters
Disallow type parameters that aren't used multiple times.
🔒Extending
"plugin:@typescript-eslint/ strict-type-checked "
in an ESLint configuration enables this rule.💡Some problems reported by this rule are manually fixable by editor suggestions .
💭This rule requires type information to run.
This rule forbids type parameters that aren't used multiple times in a function, method, or class definition.
Type parameters relate two types. If a type parameter is only used once, then it is not relating anything. It can usually be replaced with explicit types such as
unknown
.At best unnecessary type parameters make code harder to read. At worst they can be used to disguise unsafe type assertions.
warningThis rule was recently added, and has a surprising amount of hidden complexity compared to most of our rules. If you encounter unexpected behavior with it, please check closely the Limitations and FAQ sections below and our issue tracker . If you don't see your case covered, please reach out to us !
.eslintrc.cjsmodule.exports = {
"rules": {
"@typescript-eslint/no-unnecessary-type-parameters": "error"
}
};Try this rule in the playground ↗
Examples
- ❌ Incorrect
- ✅ Correct
Open in Playgroundfunction second<A, B>(a: A, b: B): B {
return b;
}
function parseJSON<T>(input: string): T {
return JSON.parse(input);
}
function printProperty<T, K extends keyof T>(obj: T, key: K) {
console.log(obj[key]);
}Open in Playgroundfunction second<B>(a: unknown, b: B): B {
return b;
}
function parseJSON(input: string): unknown {
return JSON.parse(input);
}
function printProperty<T>(obj: T, key: keyof T) {
console.log(obj[key]);
}
// T appears twice: in the type of arg and as the return type
function identity<T>(arg: T): T {
return arg;
}
// T appears twice: "keyof T" and in the inferred return type (T[K]).
// K appears twice: "key: K" and in the inferred return type (T[K]).
function getProperty<T, K extends keyof T>(obj: T, key: K) {
return obj[key];
}Limitations
Note that this rule allows any type parameter that is used multiple times, even if those uses are via a type argument. For example, the following
T
is used multiple times by virtue of being in anArray
, even though its name only appears once after declaration:declare function createStateHistory<T>(): T[];
This is because the type parameter
T
relates multiple methods in theT[]
together, making it used more than once.Therefore, this rule won't report on type parameters used as a type argument. That includes type arguments given to global types such as
Array
(including theT[]
shorthand and in tuples),Map
, andSet
.FAQ
The return type is only used as an input, so why isn't the rule reporting?
One common reason that this might be the case is when the return type is not specified explicitly. The rule uses uses type information to count implicit usages of the type parameter in the function signature, including in the inferred return type. For example, the following function...
function identity<T>(arg: T) {
return arg;
}...implicitly has a return type of
T
. Therefore, the type parameterT
is used twice, and the rule will not report this function.For other reasons the rule might not be reporting, be sure to check the Limitations section and other FAQs.
I'm using the type parameter inside the function, so why is the rule reporting?
You might be surprised to that the rule reports on a function like this:
function log<T extends string>(string1: T): void {
const string2: T = string1;
console.log(string2);
}After all, the type parameter
T
relates the inputstring1
and the local variablestring2
, right? However, this usage is unnecessary, since we can achieve the same results by replacing all usages of the type parameter with its constraint. That is to say, the function can always be rewritten as:function log(string1: string): void {
const string2: string = string1;
console.log(string2);
}Therefore, this rule only counts usages of a type parameter in the signature of a function, method, or class, but not in the implementation. See also #9735
Why am I getting TypeScript errors saying "Object literal may only specify known properties" after removing an unnecessary type parameter?
Suppose you have a situation like the following, which will trigger the rule to report.
interface SomeProperties {
foo: string;
}
// T is only used once, so the rule will report.
function serialize<T extends SomeProperties>(x: T): string {
return JSON.stringify(x);
}
serialize({ foo: 'bar', anotherProperty: 'baz' });If we remove the unnecessary type parameter, we'll get an error:
function serialize(x: SomeProperties): string {
return JSON.stringify(x);
}
// TS Error: Object literal may only specify known properties, and 'anotherProperty' does not exist in type 'SomeProperties'.
serialize({ foo: 'bar', anotherProperty: 'baz' });This is because TypeScript figures it's usually an error to explicitly provide excess properties in a location that expects a specific type. See the TypeScript handbook's section on excess property checks for further discussion.
To resolve this, you have two approaches to choose from.
If it doesn't make sense to accept excess properties in your function, you'll want to fix the errors at the call sites. Usually, you can simply remove any excess properties where the function is called.
Otherwise, if you do want your function to accept excess properties, you can modify the parameter type in order to allow excess properties explicitly by using an index signature :
interface SomeProperties {
foo: string;
// This allows any other properties.
// You may wish to make these types more specific according to your use case.
[key: PropertKey]: unknown;
}
function serialize(x: SomeProperties): string {
return JSON.stringify(x);
}
// No error!
serialize({ foo: 'bar', anotherProperty: 'baz' });Which solution is appropriate is a case-by-case decision, depending on the intended use case of your function.
I have a complex scenario that is reported by the rule, but I can't see how to remove the type parameter. What should I do?
Sometimes, you may be able to rewrite the code by reaching for some niche TypeScript features, such as the
NoInfer<T>
utility type (see #9751 ).But, quite possibly, you've hit an edge case where the type is being used in a subtle way that the rule doesn't account for. For example, the following arcane code is a way of testing whether two types are equal, and will be reported by the rule (see #9709 ):
type Compute<A> = A extends Function ? A : { [K in keyof A]: Compute<A[K]> };
type Equal<X, Y> =
(<T1>() => T1 extends Compute<X> ? 1 : 2) extends
(<T2>() => T2 extends Compute<Y> ? 1 : 2)
? true
: false;In this case, the function types created within the
Equal
type are never expected to be assigned to; they're just created for the purpose of type system manipulations. This usage is not what the rule is intended to analyze.Use eslint-disable comments as appropriate to suppress the rule in these kinds of cases.
Options
This rule is not configurable.
When Not To Use It
This rule will report on functions that use type parameters solely to test types, for example:
function assertType<T>(arg: T) {}
assertType<number>(123);
assertType<number>('abc');
// ~~~~~
// Argument of type 'string' is not assignable to parameter of type 'number'.If you're using this pattern then you'll want to disable this rule on files that test types.
Further Reading
TypeScript handbook: Type Parameters Should Appear Twice Effective TypeScript: The Golden Rule of Generics Related To
eslint-plugin-etc's no-misused-generics
wotan's no-misused-generics
DefinitelyTyped-tools' no-unnecessary-generics
Type checked lint rules are more powerful than traditional lint rules, but also require configuring type checked linting .
See Troubleshooting > Linting with Type Information > Performance if you experience performance degredations after enabling type checked rules.
Resources