True Myth / result / ap

Function: ap()

Call Signature

ap<A, B, E>(resultFn, result): Result<B, E>

Allows you to apply (thus ap) a value to a function without having to take either out of the context of their Results. This does mean that the transforming function is itself within a Result, which can be hard to grok at first but lets you do some very elegant things. For example, ap allows you to do this (using the method form, since nesting ap calls is awkward):

import { ap, ok, err } from 'true-myth/result';

const one = ok<number, string>(1);
const five = ok<number, string>(5);
const whoops = err<number, string>('oh no');

const add = (a: number) => (b: number) => a + b;
const resultAdd = ok<typeof add, string>(add);

resultAdd.ap(one).ap(five); // Ok(6)
resultAdd.ap(one).ap(whoops); // Err('oh no')
resultAdd.ap(whoops).ap(five) // Err('oh no')

Without ap, you'd need to do something like a nested match:

import { ok, err } from 'true-myth/result';

const one = ok<number, string>(1);
const five = ok<number, string>(5);
const whoops = err<number, string>('oh no');

one.match({
  Ok: n => five.match({
    Ok: o => ok<number, string>(n + o),
    Err: e => err<number, string>(e),
  }),
  Err: e  => err<number, string>(e),
}); // Ok(6)

one.match({
  Ok: n => whoops.match({
    Ok: o => ok<number, string>(n + o),
    Err: e => err<number, string>(e),
  }),
  Err: e  => err<number, string>(e),
}); // Err('oh no')

whoops.match({
  Ok: n => five.match({
    Ok: o => ok(n + o),
    Err: e => err(e),
  }),
  Err: e  => err(e),
}); // Err('oh no')

And this kind of thing comes up quite often once you're using Result to handle errors throughout your application.

For another example, imagine you need to compare the equality of two ImmutableJS data structures, where a === comparison won't work. With ap, that's as simple as this:

import { ok } from 'true-myth/result';
import { is as immutableIs, Set } from 'immutable';

const is = (first: unknown) =>  (second: unknown) =>
  immutableIs(first, second);

const x = ok(Set.of(1, 2, 3));
const y = ok(Set.of(2, 3, 4));

ok(is).ap(x).ap(y); // Ok(false)

Without ap, we're back to that gnarly nested match:

import Result, { ok, err } from 'true-myth/result';
import { is, Set } from 'immutable';

const x = ok(Set.of(1, 2, 3));
const y = ok(Set.of(2, 3, 4));

x.match({
  Ok: iX => y.match({
    Ok: iY => Result.of(is(iX, iY)),
    Err: (e) => ok(false),
  })
  Err: (e) => ok(false),
}); // Ok(false)

In summary: anywhere you have two Result instances and need to perform an operation that uses both of them, ap is your friend.

Two things to note, both regarding currying:

1. All functions passed to ap must be curried. That is, they must be of the form (for add) (a: number) => (b: number) => a + b, not the more usual (a: number, b: number) => a + b you see in JavaScript more generally.

   (Unfortunately, these do not currently work with lodash or Ramda's curry helper functions. A future update to the type definitions may make that work, but the intermediate types produced by those helpers and the more general function types expected by this function do not currently align.)

2. You will need to call ap as many times as there are arguments to the function you're dealing with. So in the case of this add3 function, which has the "arity" (function argument count) of 3 (a and b), you'll need to call ap twice: once for a, and once for b. To see why, let's look at what the result in each phase is:

   const add3 = (a: number) => (b: number) => (c: number) => a + b + c;
   
   const resultAdd = ok(add); // Ok((a: number) => (b: number) => (c: number) => a + b + c)
   const resultAdd1 = resultAdd.ap(ok(1)); // Ok((b: number) => (c: number) => 1 + b + c)
   const resultAdd1And2 = resultAdd1.ap(ok(2)) // Ok((c: number) => 1 + 2 + c)
   const final = maybeAdd1.ap(ok(3)); // Ok(4)

   So for toString, which just takes a single argument, you would only need to call ap once.

   const toStr = (v: { toString(): string }) => v.toString();
   ok(toStr).ap(12); // Ok("12")

One other scenario which doesn't come up quite as often but is conceivable is where you have something that may or may not actually construct a function for handling a specific Result scenario. In that case, you can wrap the possibly-present in ap and then wrap the values to apply to the function to in Result themselves.

Because Result often requires you to type out the full type parameterization on a regular basis, it's convenient to use TypeScript's typeof operator to write out the type of a curried function. For example, if you had a function that simply merged three strings, you might write it like this:

import Result from 'true-myth/result';
import { curry } from 'lodash';

const merge3Strs = (a: string, b: string, c: string) => string;
const curriedMerge = curry(merge3Strs);

const fn = Result.ok<typeof curriedMerge, string>(curriedMerge);

The alternative is writing out the full signature long-form:

const fn = Result.ok<(a: string) => (b: string) => (c: string) => string, string>(curriedMerge);

Aside: ap is not named apply because of the overlap with JavaScript's existing apply function – and although strictly speaking, there isn't any direct overlap (Result.apply and Function.prototype.apply don't intersect at all) it's useful to have a different name to avoid implying that they're the same.

Type Parameters

A

A

B

B

E

E

Parameters

resultFn

Result<(a) => B, E>

result of a function from T to U

result

Result<A, E>

result of a T to apply to fn

Returns

Result<B, E>

Call Signature

ap<A, B, E>(resultFn): (result) => Result<B, E>

Allows you to apply (thus ap) a value to a function without having to take either out of the context of their Results. This does mean that the transforming function is itself within a Result, which can be hard to grok at first but lets you do some very elegant things. For example, ap allows you to do this (using the method form, since nesting ap calls is awkward):

import { ap, ok, err } from 'true-myth/result';

const one = ok<number, string>(1);
const five = ok<number, string>(5);
const whoops = err<number, string>('oh no');

const add = (a: number) => (b: number) => a + b;
const resultAdd = ok<typeof add, string>(add);

resultAdd.ap(one).ap(five); // Ok(6)
resultAdd.ap(one).ap(whoops); // Err('oh no')
resultAdd.ap(whoops).ap(five) // Err('oh no')

Without ap, you'd need to do something like a nested match:

import { ok, err } from 'true-myth/result';

const one = ok<number, string>(1);
const five = ok<number, string>(5);
const whoops = err<number, string>('oh no');

one.match({
  Ok: n => five.match({
    Ok: o => ok<number, string>(n + o),
    Err: e => err<number, string>(e),
  }),
  Err: e  => err<number, string>(e),
}); // Ok(6)

one.match({
  Ok: n => whoops.match({
    Ok: o => ok<number, string>(n + o),
    Err: e => err<number, string>(e),
  }),
  Err: e  => err<number, string>(e),
}); // Err('oh no')

whoops.match({
  Ok: n => five.match({
    Ok: o => ok(n + o),
    Err: e => err(e),
  }),
  Err: e  => err(e),
}); // Err('oh no')

And this kind of thing comes up quite often once you're using Result to handle errors throughout your application.

For another example, imagine you need to compare the equality of two ImmutableJS data structures, where a === comparison won't work. With ap, that's as simple as this:

import { ok } from 'true-myth/result';
import { is as immutableIs, Set } from 'immutable';

const is = (first: unknown) =>  (second: unknown) =>
  immutableIs(first, second);

const x = ok(Set.of(1, 2, 3));
const y = ok(Set.of(2, 3, 4));

ok(is).ap(x).ap(y); // Ok(false)

Without ap, we're back to that gnarly nested match:

import Result, { ok, err } from 'true-myth/result';
import { is, Set } from 'immutable';

const x = ok(Set.of(1, 2, 3));
const y = ok(Set.of(2, 3, 4));

x.match({
  Ok: iX => y.match({
    Ok: iY => Result.of(is(iX, iY)),
    Err: (e) => ok(false),
  })
  Err: (e) => ok(false),
}); // Ok(false)

In summary: anywhere you have two Result instances and need to perform an operation that uses both of them, ap is your friend.

Two things to note, both regarding currying:

1. All functions passed to ap must be curried. That is, they must be of the form (for add) (a: number) => (b: number) => a + b, not the more usual (a: number, b: number) => a + b you see in JavaScript more generally.

   (Unfortunately, these do not currently work with lodash or Ramda's curry helper functions. A future update to the type definitions may make that work, but the intermediate types produced by those helpers and the more general function types expected by this function do not currently align.)

2. You will need to call ap as many times as there are arguments to the function you're dealing with. So in the case of this add3 function, which has the "arity" (function argument count) of 3 (a and b), you'll need to call ap twice: once for a, and once for b. To see why, let's look at what the result in each phase is:

   const add3 = (a: number) => (b: number) => (c: number) => a + b + c;
   
   const resultAdd = ok(add); // Ok((a: number) => (b: number) => (c: number) => a + b + c)
   const resultAdd1 = resultAdd.ap(ok(1)); // Ok((b: number) => (c: number) => 1 + b + c)
   const resultAdd1And2 = resultAdd1.ap(ok(2)) // Ok((c: number) => 1 + 2 + c)
   const final = maybeAdd1.ap(ok(3)); // Ok(4)

   So for toString, which just takes a single argument, you would only need to call ap once.

   const toStr = (v: { toString(): string }) => v.toString();
   ok(toStr).ap(12); // Ok("12")

One other scenario which doesn't come up quite as often but is conceivable is where you have something that may or may not actually construct a function for handling a specific Result scenario. In that case, you can wrap the possibly-present in ap and then wrap the values to apply to the function to in Result themselves.

Because Result often requires you to type out the full type parameterization on a regular basis, it's convenient to use TypeScript's typeof operator to write out the type of a curried function. For example, if you had a function that simply merged three strings, you might write it like this:

import Result from 'true-myth/result';
import { curry } from 'lodash';

const merge3Strs = (a: string, b: string, c: string) => string;
const curriedMerge = curry(merge3Strs);

const fn = Result.ok<typeof curriedMerge, string>(curriedMerge);

The alternative is writing out the full signature long-form:

const fn = Result.ok<(a: string) => (b: string) => (c: string) => string, string>(curriedMerge);

Aside: ap is not named apply because of the overlap with JavaScript's existing apply function – and although strictly speaking, there isn't any direct overlap (Result.apply and Function.prototype.apply don't intersect at all) it's useful to have a different name to avoid implying that they're the same.

Type Parameters

A

A

B

B

E

E

Parameters

resultFn

Result<(a) => B, E>

result of a function from T to U

Returns

(result): Result<B, E>

Parameters

result

Result<A, E>

Returns

Result<B, E>