TypeScript's built-in typings are not perfect. ts-reset
makes them better.
Without ts-reset
:
- 🚨
.json
(infetch
) andJSON.parse
both returnany
- 🤦
.filter(Boolean)
doesn't behave how you expect - 😡
array.includes
often breaks on readonly arrays
ts-reset
smooths over these hard edges, just like a CSS reset does in the browser.
With ts-reset
:
- 👍
.json
(infetch
) andJSON.parse
both returnunknown
- ✅
.filter(Boolean)
behaves EXACTLY how you expect - 🥹
array.includes
is widened to be more ergonomic - 🚀 And several more changes!
// Import in a single file, then across your whole project...
import "@total-typescript/ts-reset";
// .filter just got smarter!
const filteredArray = [1, 2, undefined].filter(Boolean); // number[]
// Get rid of the any's in JSON.parse and fetch
const result = JSON.parse("{}"); // unknown
fetch("/")
.then((res) => res.json())
.then((json) => {
console.log(json); // unknown
});
-
Install:
npm i -D @total-typescript/ts-reset
-
Create a
reset.d.ts
file in your project with these contents:
// Do not add any other lines of code to this file!
import "@total-typescript/ts-reset";
- Enjoy improved typings across your entire project.
By importing from @total-typescript/ts-reset
, you're bundling all the recommended rules.
To only import the rules you want, you can import like so:
// Makes JSON.parse return unknown
import "@total-typescript/ts-reset/json-parse";
// Makes await fetch().then(res => res.json()) return unknown
import "@total-typescript/ts-reset/fetch";
For these imports to work, you'll need to ensure that, in your tsconfig.json
, module
is set to NodeNext
or Node16
.
Below is a full list of all the rules available.
ts-reset
is designed to be used in application code, not library code. Each rule you include will make changes to the global scope. That means that, simply by importing your library, your user will be unknowingly opting in to ts-reset
.
import "@total-typescript/ts-reset/json-parse";
JSON.parse
returning any
can cause nasty, subtle bugs. Frankly, any any
's can cause bugs because they disable typechecking on the values they describe.
// BEFORE
const result = JSON.parse("{}"); // any
By changing the result of JSON.parse
to unknown
, we're now forced to either validate the unknown
to ensure it's the correct type (perhaps using zod
), or cast it with as
.
// AFTER
import "@total-typescript/ts-reset/json-parse";
const result = JSON.parse("{}"); // unknown
import "@total-typescript/ts-reset/fetch";
Just like JSON.parse
, .json()
returning any
introduces unwanted any
's into your application code.
// BEFORE
fetch("/")
.then((res) => res.json())
.then((json) => {
console.log(json); // any
});
By forcing res.json
to return unknown
, we're encouraged to distrust its results, making us more likely to validate the results of fetch
.
// AFTER
import "@total-typescript/ts-reset/fetch";
fetch("/")
.then((res) => res.json())
.then((json) => {
console.log(json); // unknown
});
import "@total-typescript/ts-reset/filter-boolean";
The default behaviour of .filter
can feel pretty frustrating. Given the code below:
// BEFORE
const filteredArray = [1, 2, undefined].filter(Boolean); // (number | undefined)[]
It feels natural that TypeScript should understand that you've filtered out the undefined
from filteredArray
. You can make this work, but you need to mark it as a type predicate:
const filteredArray = [1, 2, undefined].filter((item): item is number => {
return !!item;
}); // number[]
Using .filter(Boolean)
is a really common shorthand for this. So, this rule makes it so .filter(Boolean)
acts like a type predicate on the array passed in, removing any falsy values from the array member.
// AFTER
import "@total-typescript/ts-reset/filter-boolean";
const filteredArray = [1, 2, undefined].filter(Boolean); // number[]
import "@total-typescript/ts-reset/array-includes";
This rule improves on TypeScript's default .includes
behaviour. Without this rule enabled, the argument passed to .includes
MUST be a member of the array it's being tested against.
// BEFORE
const users = ["matt", "sofia", "waqas"] as const;
// Argument of type '"bryan"' is not assignable to
// parameter of type '"matt" | "sofia" | "waqas"'.
users.includes("bryan");
This can often feel extremely awkward. But with the rule enabled, .includes
now takes a widened version of the literals in the const
array.
// AFTER
import "@total-typescript/ts-reset/array-includes";
const users = ["matt", "sofia", "waqas"] as const;
// .includes now takes a string as the first parameter
users.includes("bryan");
This means you can test non-members of the array safely.
import "@total-typescript/ts-reset/array-index-of";
Exactly the same behaviour of .includes
(explained above), but for .lastIndexOf
and .indexOf
.
import "@total-typescript/ts-reset/set-has";
Similar to .includes
, Set.has()
doesn't let you pass members that don't exist in the set:
// BEFORE
const userSet = new Set(["matt", "sofia", "waqas"] as const);
// Argument of type '"bryan"' is not assignable to
// parameter of type '"matt" | "sofia" | "waqas"'.
userSet.has("bryan");
With the rule enabled, Set
is much smarter:
// AFTER
import "@total-typescript/ts-reset/set-has";
const userSet = new Set(["matt", "sofia", "waqas"] as const);
// .has now takes a string as the argument!
userSet.has("bryan");
import "@total-typescript/ts-reset/map-has";
Similar to .includes
or Set.has()
, Map.has()
doesn't let you pass members that don't exist in the map's keys:
// BEFORE
const userMap = new Map([
["matt", 0],
["sofia", 1],
[2, "waqas"],
] as const);
// Argument of type '"bryan"' is not assignable to
// parameter of type '"matt" | "sofia" | "waqas"'.
userMap.has("bryan");
With the rule enabled, Map
follows the same semantics as Set
.
// AFTER
import "@total-typescript/ts-reset/map-has";
const userMap = new Map([
["matt", 0],
["sofia", 1],
[2, "waqas"],
] as const);
// .has now takes a string as the argument!
userMap.has("bryan");
import "@total-typescript/ts-reset/is-array";
When you're using Array.isArray
, you can introduce subtle any
's into your app's code.
// BEFORE
const validate = (input: unknown) => {
if (Array.isArray(input)) {
console.log(input); // any[]
}
};
With is-array
enabled, this check will now mark the value as unknown[]
:
// AFTER
import "@total-typescript/ts-reset/is-array";
const validate = (input: unknown) => {
if (Array.isArray(input)) {
console.log(input); // unknown[]
}
};
A common ask is to provide 'better' typings for Object.keys
, so that it returns Array<keyof T>
instead of Array<string>
. Same for Object.entries
. ts-reset
won't be including rules to change this.
TypeScript is a structural typing system. One of the effects of this is that TypeScript can't always guarantee that your object types don't contain excess properties:
type Func = () => {
id: string;
};
const func: Func = () => {
return {
id: "123",
// No error on an excess property!
name: "Hello!",
};
};
So, the only reasonable type for Object.keys
to return is Array<string>
.
A common request is for ts-reset
to add type arguments to functions like JSON.parse
:
const str = JSON.parse<string>('"hello"');
console.log(str); // string
This appears to improve the DX by giving you autocomplete on the thing that gets returned from JSON.parse
.
However, we argue that this is a lie to the compiler and so, unsafe.
JSON.parse
and fetch
represent validation boundaries - places where unknown data can enter your application code.
If you really know what data is coming back from a JSON.parse
, then an as
assertion feels like the right call:
const str = JSON.parse('"hello"') as string;
console.log(str); // string
This provides the types you intend and also signals to the developer that this is slightly unsafe.