Too Many Maybes

I hang out in the Elm Slack channel from time to time, and people will often show up asking for help on their applications, ranging from questions about Elm's syntax to overall application design. One question that comes up pretty often is the right way to reduce the complexity of code like this:

type alias User = {
    username : Maybe String,
    email : Maybe String,
    name : Maybe String

renderUser : User -> Html
renderUser user =
  ul [] [
    li [] [ text <| Maybe.withDefault "" user.username ],
    li [] [ text <| Maybe.withDefault "" ],
    li [] [ text <| Maybe.withDefault "" ]

That's a lot of Maybe.withDefault calls, isn't it? I find that a lot of users new to Elm who aren't familiar with option types like Maybe tend to use it too much in a program. My question is this:

Does it make sense for a user to lack a username? To lack an e-mail?

Now, there are some Maybes that do belong here; for example, I can easily imagine an application where a human readable name is optional. The question is whether the domain you're modeling accommodates for it. Most websites require a user to have a username and an e-mail, so why should your Elm application have to deal with the notion of extracting from a Maybe something that always should be present?

The responses I've gotten to this is that maybe they're building a form so they need to store the values for the new user while the user is inputting them, or perhaps they're creating a User value from a JSON response, and the username and email keys may not exist.

In the case of the form, ask yourself: is a User object really the right abstraction you want to work with when processing a form? I think it would be better to use an alternative data structure (for instance, the humble Dict) to store your partial user values until the form is submitted, and then have a dictToUser : Dict String String -> Maybe User (or better yet, Dict String String -> Result FormError User) function perform the final conversion.

In the case of the JSON response, if your backend is providing you with incomplete information, I think that's an issue you should be dealing with local to the JSON-loading module, much like you'd deal with an invalid JSON document. The JSON-loading module can handle these unusual situations, rather then having the rest of your application shoulder the syntactic burden of a situation that probably shouldn't happen. This keeps exceptional behavior isolated to a single module in your application, which keeps your code cleaner and reduces cognitive load. I feel that it allows you to use Elm's type system to its full advantage!

If you're reading this and are new to Elm, hopefully I've helped give some perspective on using Maybe types in your application. Please let me know if I've helped, or if you think that I missed something!

Published on 2016-04-15