I want to write a function that takes an intersect...

# announcements

[glenkpeterson]

I want to write a function that takes an intersection type, but I don't know the syntax (or if it's possible in Kotlin). Here's what I'd imagine using a single ampersand like so:

// Error here ---v
fun <T : Slugged & BaseModel> suggestSlug(
    tempSlug: String,
    clazz: Class<T>,
): String

I want to make sure the class passed to this function implements Slugged (to be sure it has a .slug property) and extends the abstract class BaseModel (to be sure it has an .id property and is a database @Entity class - I'm using Ebean).

I found this, which shows the set-theory, but doesn't show the syntax I need (I apologize if I am misunderstanding any of this):
https://kotlinlang.org/spec/type-system.html#intersection-types

Should I just make an abstract SluggedBaseModel ? I'd rather not because abstract vars annotated @Column need to be initialized and this should be a non-null required constructor parameter for each entity bean that uses it.

[Zach Klippenstein (he/him) [MOD]]

Kotlin doesn’t support expressing intersection types in code. It can handle intersection types that are inferred, but you can’t actually express such a type yourself unfortunately.

[CLOVIS]

fun <T> suggestSlug(
  tempSlug: String,
  clazz: Class<T>,
): String where T : Slugged, T : BaseModel

https://kotlinlang.org/docs/generics.html

[CLOVIS]

It only works if you know the exact interfaces you want as parent though, you (very sadly) can't write where T : S, T : Q (but in your case it seems you know them so it should work fine)

[glenkpeterson]

Perfect. Thanks @CLOVIS! The most direct link I found is https://kotlinlang.org/docs/generics.html#upper-bounds

A simple | or & would be prettier (especially if you could use it in a typealias), but at least the feature is in the language.

[glenkpeterson]

Of course, calling this function from Java now requires a cast to (Class) and a @SuppressWarnings({"rawtypes", "unchecked"}) but I can live with that.

[rnett]

Denotable intersection and union types are on the team's radar, but it's not going to happen before the frontend re-write at a minimum (~ a year? no firm timeline). See https://youtrack.jetbrains.com/issue/KT-13108, it was mentioned in the 1.5 event too

[CLOVIS]

I'm surprised it requires anything from the Java side, my understanding was that it's a direct equivalent of Java's <T : A & B>

Sadly it's not powerful enough for what I want to do with it, hopefully the team with rework on that soon