Maybe Functions

The maybe function is a subtle monster that spreads it’s tentacles across the code-base. It’s alternating functionality of “does/doesn’t do something” makes code hard to understand, maintain and debug.

Example

Here’s a specific example, it’s a “maybe” function as it only returns the friends of a user, if the user is logged in. Basically it introduces a possible return null.

function maybeGetUser(): User | null {
  if (!loggedIn) {
    return null;
  }
  return fetchUser();
}

Notes:

• This is highly related to the “Null Object Pattern”, but I thought I would explain it from the perspective of functions.
• It’s also important to note here, that in the wild “maybe functions” might not be conveniently prefixed with “maybe”…

So what are the trade-off’s when using “maybe functions”?

Advantages

• Validation logic coupled with implementation
• Easy for the caller, can be called in many places

Disadvantages

• Caller has less control, they depend on validation hidden in “maybe function”
• Hides complexity but doesn’t actually remove it
• Things that depend on the output now need to handle the maybeness case

When a “maybe function” is implemented, it makes it far more likely that more “maybe functions” will be introduced to deal with all the null values. Here’s some pseudo code to show the basic problem:

// warning: not real react
function Page() {
  const bestFriends = maybeRenderBestFriends();
  return <>
    {bestFriends}
  </>;
}

function maybeRenderBestFriends() {
  const user = maybeGetUser();
  const friends = maybeGetFriends(user);
  const bestFriends = maybeFilterBestFriends(friends);
  return render(bestFriends);
}

function maybeGetUser(): User | null {
  if (!loggedIn) {
    return null;
  }
  return fetchUser();
}

function maybeGetFriends(user: User | null): Friend[] | null {
  if (!user) {
    return null;
  }
  return fetchUserFriends(user);
}

function maybeFilterBestFriends(friends: Friend[] | null): Friend[] | null {
  if (!friends) {
    return null;
  }
  return friends.filter(f => f.bestFriend);
}

You can see how this pattern maybe getting out of control… it does look clean in the consumer, with no control logic. However the null handling spreads to all new functions, and obscures what functions are actually doing… which is usually “nothing”, so why even call it?

How can we improve this?

Solution 1 - Lift the maybeness up

The easiest way to remove the “maybeness” from functions is to lift up the “maybe” logic into the consumer, this removes the possible null aspects of the function, which allows other functions to remove theirs too!

// warning: not real react
function Page() {
  const bestFriends = loggedIn
    ? renderBestFriends()
    : null;
  return <>
    {bestFriends}
  </>;
}

function renderBestFriends() {
  const user = getUser();
  const friends = getFriends(user);
  const bestFriends = filterBestFriends(friends);
  return render(bestFriends);
}

function getUser(): User {
  return fetchUser();
}

function getFriends(user: User): Friend[] {
  return fetchUserFriends(user);
}

function filterBestFriends(friends: Friend[]): Friend[] {
  return friends.filter(f => f.bestFriend);
}

This is much easier to understand, as each function does what it says and there’s much less “maybeness” to account for.

Solution 2 - Monads

It’s also possible to generalize and isolate the “maybeness” by using monads. In the below example, all null checks go through “runSafely”, which removes a lot of “maybeness” in remaining functions.

/** START MONAD */
class Maybe<T> {
  constructor(
    value: T | null
  )
  runSafely(fn: (val: T) => V): Maybe<V> {
    if (!this.value) {
      return new Maybe(null);
    }
    return new Maybe(fn(this.value));
  }
}
/** END MONAD */

function Page() {
  const bestFriends = maybeRenderBestFriends();
  return <>
    {bestFriends}
  </>;
}

function maybeRenderBestFriends() {
  const user = new Maybe(getUser());
  const friends = user.runSafely(getFriends);
  const bestFriends = friends.runSafely(filterBestFriends);
  return render(bestFriends);
}

function getUser(): User {
  if (!loggedIn) {
    return null
  }
  return fetchUser();
}

function getFriends(user: User): Friend[] {
  return fetchUser();
}

function filterBestFriends(friends: Friend[]): Friend[] {
  return friends.filter(f => f.bestFriend);
}

This might be appropriate for a lot of situations, but I believe the 1st solution should definitely be reached for first, as it doesn’t rely on another level of abstraction.

Conclusion

Maybe functions have always annoyed me personally and they can found in many codebases across any language that allows null. They seem to be trivial to add, but difficult to remove. But hopefully this illustrates the concern and ways to fix it.

“Functions should do something, not maybe do something…”

Ben Winding2024