I think I've found a bug in Kotlin's compiler, but...

# getting-started

[Johann Pardanaud]

I think I've found a bug in Kotlin's compiler, but wanted to double-check with you. Here's a reproducible example:

class Wrapper<T>(val wrappedValue: T, parent: Wrapper<*>? = null)

fun <T : Any?, V> Wrapper<T>.wrapperOf(getter: KProperty1<T & Any, V>): Wrapper<V?> {
    return Wrapper(wrappedValue?.let { getter.get(it) }, this)
}

fun main() {
    // does not compile
    Wrapper("foo" as String?).wrapperOf(String::length)

    // but compiles with an intermediate variable
    val getter = String::length
    Wrapper("foo" as String?).wrapperOf(getter)
}

Link to the playground

If I comment the eleventh line, it compiles, despite using nearly the same code just below but with an intermediate value.

What do you think? Am I missing something?

[Johann Pardanaud]

It also works when writing:

Wrapper("foo" as String?).wrapperOf(String::length as KProperty1<String, Int>)

[Jacob]

Which do you think is correct behavior? Compiling or not?

[Youssef Shoaib [MOD]]

I'm not quite sure why, but just using a functional type instead of KProperty makes it work https://play.kotlinlang.org/#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

[Youssef Shoaib [MOD]]

This also works, where you separate it into a nullable and non-nullable variant. I think the issue is using T & Any.

[Johann Pardanaud]

@Jacob I expect all the code to compile

[Johann Pardanaud]

@Youssef Shoaib [MOD] Yes, it works with separate variants. But I'm not searching for an alternative, I just wonder if this is a bug. The more I dig, the more I think it is one.

[Johann Pardanaud]

reported: https://youtrack.jetbrains.com/issue/KT-59493