Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Consider alternative type and syntax for Next #116

Open
sddamico opened this issue Sep 1, 2020 · 1 comment
Open

Consider alternative type and syntax for Next #116

sddamico opened this issue Sep 1, 2020 · 1 comment

Comments

@sddamico
Copy link

sddamico commented Sep 1, 2020

Is your feature request related to a problem? Please describe.

I brought this up in Slack but would like to continue the conversation here to open it to other folks that would like to participate.

The Next type is currently typealias'd to a Pair<Model, Effect<Msg>>. The usage of Pair comes from the desire to use the built in tuple types that the Kotlin language offers via generics and data class (Pair and Triple).

Unfortunately, unlike Elm's tuples which have a nice syntax for creation with parentheses (e.g. (1, 2, "a", "b")), Kotlin relies on constructors for its data classes with no short-hand replacement. To work around that, the language offers an infix function to to aid in Pair's creation.

This leads to idiomatic Oolong code in the Update/Init functions looking somewhat like this:

val update: Update<Msg, Model> = { msg, model ->
  model to none()
}

This syntax can be a little hard to decipher, especially as a novice reader of the code. It requires some knowledge of the Oolong type definitions to understand why this syntax is being used. The keyword to reads more like a mapping than a union (to me, at least), and I think there's an opportunity here to be more explicit with types and have a more readable syntax.

Describe the solution you'd like

Give an explicit type to Next and provide a more ergonomic infix operator for its creation to lean in to the Kotlin language features and tools.

Two steps to this solution:

  1. Redefine Next to data class Next<Model, Msg>(val model: Model, val effect: Effect<Msg>)
  2. Add a new infix function and: infix fun <Model, Msg> Model.and(effect: Effect<Msg>) = Next(this, effect)

This changes our above example to read like this:

val update: Update<Msg, Model> = { msg, model ->
    model and none()
}

Advantages:

  • the code reads as "return the updated model and these effects", which I think is a very nice improvement
  • extensions on Next will not pollute the Pair type in case there were extensions that were Oolong-specific
  • deconstructing the Next type will have named fields instead of first and second

Cons:

  • lose access to any extensions on Pair
  • potentially breaking change to the type system

note: I'd expect that we provide a to infix for compatibility initially, but have that deprecated

@Deprecated(
    message = "prefer `and` operator, `to` will be removed in a future update",
    replaceWith = ReplaceWith("this.and(effect)")
)
infix fun <Model, Msg> Model.to(effect: Effect<Msg>) = Next(this, effect)

The problem that this doesn't solve is the syntax for the optional Effect case, I think ideally you'd be able to have a syntax that supports returning just the model or returning the model and effects, e.g.

val update: Update<Msg, Model> = { msg, model ->
    model
}

Not sure what might be available for that solution (outside of a compiler plugin that allows for a more ergonomic tuple creation syntax)

Describe alternatives you've considered

  • with would be an acceptable alternative to and but it is already a reserved word in the Kotlin language
  • I opted against an infix operator (e.g. *) since that is not a standard Kotlin language feature and would likely be considered a bit obtuse in the way that to is today with the added negative of being less discoverable

Additional context

none

@pardom
Copy link
Collaborator

pardom commented Sep 5, 2020

I think a good first step while we consider alternatives, is to deprecate Next in favor of Pair. This is more in line with Elm using tuples and removes the unnecessary obfuscation. We may find that this clears up confusion about constructing the return type altogether. I've been experimenting with several alternatives, including a DSL.

I'll make a PR with this change so we can try it out.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
Development

No branches or pull requests

2 participants