Module resultCopy item path

Rust-like Result<T, E> type and error handling utils

I once had a fancy error object with TypeScript magic that tries to reduce allocation while maintaining Result-safety. It turns out that was slower than allocating plain objects for every return, because of how V8 optimizes things.

Don't even use isErr() helper functions to abstract. They are slower than directly property access in my testing.

Function that can fail

Instead of having functions throw, make it return instead.

// Instead of
function doSomethingCanFail() {
    if (Math.random() < 0.5) {
        return 42;
    }
    throw "oops";
}
// Do this
import type { Result } from "pure/result";

function doSomethingCanFail(): Result<number, string> {
    if (Math.random() < 0.5) {
        return { val: 42 };
    }
    return { err: "oops" };
}
Copy

This is similar to Rust:

fn do_something_can_fail() -> Result<u32, String> {
    if ... {
        return Ok(42);
    }

    Err("oops".to_string())
}
Copy

Calling function that can fail

The recommended pattern is

const x = doTheCall(); // x is Result<T, E>;
if (x.err) {
    // x.err is E, handle it
    return ...
}
// x.val is T
// ...
Copy

If your E type covers falsy values that are valid, use "err" in x instead of x.err. A well-known case is Result<T, unknown>. if(r.err) cannot narrow the else case to Ok, but if("err" in r) can.

A full example:

function getParam(name: string): Result<number, Error> {
    if (name === "a") {
        return { val: 13 };
    }
    if (name === "b") {
        return { val: 42 };
    }
    return { err: new Error("bad name") };
}

function multiplyFormat(
    name1: string,
    name2: string,
    prefix: string
): Result<string, Error> {
    const v1 = getParam(name1);
    if (v1.err) {
        console.error(v1.err);
        return v1;
    }
    const v2 = getParam(name1);
    if (v2.err) {
        console.error(v2.err);
        return v2;
    }

    const formatted = `${prefix}${v1.val * v2.val}`;
    return { val: formatted };
}
Copy

Interop with throwing functions

This library also has tryCatch to interop with throwing functions, and tryAsync for async functions.

import { tryCatch, tryAsync } from "pure/result";

// synchronous
const result1: Result<MyData, unknown> = tryCatch(() => JSON.parse<MyData>(...));
// or you can specify the error type:
const result2 = tryCatch<MyData, SyntaxError>(() => JSON.parse(...));

// asynchronous
async function doSomethingCanFail() {
    if (Math.random() < 0.5) {
        return 42;
    }
    throw "oops";
}
const result = await tryAsync<number, string>(() => doStuff);
Copy

Returning void

Use Void<E> as the return type if the function returns void on success

const x = doSomethingThatVoidsOnSuccess();
if (x.err) {
    return x;
}
// type of x is Record<string, never>, i.e. empty object
Copy

Why is there no match/map/mapErr, etc?

If you are thinking this is a great idea:

const result = foo(bar);
match(result,
    (okValue) => {
        // handle ok case
    },
    (errValue) => {
        // handle err case
    },
);
Copy

The vanilla if doesn't allocate the closures, and has less code, and you can control the flow properly inside the blocks with return/break/continue

const result = foo(bar);
if (result.err) {
    // handle err case
} else {
    // handle ok case
}
Copy

As for the other utility functions from Rust's Result type, they really only benefit because you can early return with ? AND those abstractions are zero-cost in Rust. Neither is true in JavaScript.

You can also easily write them yourself if you really want to.