Wrap async/sync result in the library

[keogami]

Its super common for async/sync operations to throw an error, but i don't see a constructor to conveniently wrap the operation in the Result. I propose a new constructor:

async function AsyncResultOf<T>(val: Promise<T>): Promise<Result<T, Error>> {
  try {
    return Ok(await val);
  } catch(err) {
    return Err(err);
  }
}

This allows us to go from the throw/catch error handling to the monadic error handling patterns const result = await AsyncResultOf(fetch(...))

I will go further to suggest an even more general constructor that translates both async and non-async fallible operations:

async function ResultOf<T>(val: () => Promise<T>): Promise<Result<T, Error>> {
  try {
    return Ok(await val());
  } catch(err) {
    return Err(err);
  }
}

This can of course be overloaded with the signature of AsyncResultOf.

(PS im trying to whip up a PR but the test keeps saying "Timed out while running tests" on my machine)

[keogami]

i changed results.ts interestingly.

[cawabunga]

I think your proposal is interesting. Have you considered using TaskEither from the fp-ts library? It could be more suitable for more complex scenarios. Additionally, if we're planning to support this constructor, we should also consider adding a wrapper for regular functions. Your syntax seems fitting for this purpose:

function resultOf<T, E = unknown>(val: () => T): Result<T, E> {
  try {
    return Ok(val());
  } catch(err) {
    return Err(err);
  }
}

// usage:
const result = resultOf(() => 10 / 0);

Also, I have modified the error type, as we cannot be certain that the thrown value is an error.

upd: I hadn't seen your PR https://github.com/sniptt-official/monads/pull/112 , when I sent the comment.

[keogami]

My bad, I should have linked this issue with my PR.

fp-ts's TaskEither is a good analog. However, Either is a more generalized form of Result as in Result makes it explicit that you will either get the result or the error.

And from non-async functions i meant regular functions, yes.

[danawoodman]

Would be cool to see this implemented :)