lateinit on primitives

# language-proposals

[xenoterracide]

lateinit on primitives

[Dico]

This would have to be implemented through boxing

[Dico]

Probably

[cbruegg]

Or a second field in the class. Disregarding the implementation difficulty, as a language user I've been annoyed by this several times.

[diesieben07]

var x: Int by Delegates.notNull()

[jw]

that does boxing and allocates an object for the delegate

[diesieben07]

Yes. And you can't do this without boxing.

[diesieben07]

There is no null for unboxed ints.

[jw]

Sure but "just" boxing is better than boxing and a delegate instantiation

[benleggiero]

What's wrong with boxing primitives? We do it all the time implicitly, if I remember correctly

[jw]

It puts allocation into a pathway where it shouldn't exist. It's why the JVM is working so hard to specialize generics over primitives.

[benleggiero]

I know the tradeoffs of boxing primitives, but sometimes it's necessary to get certain functionality, like Int?. I think lateinit is an acceptable place for that tradeoff, since the reason you'd want it on a primitive in the first place is because it's expensive to calculate, so adding a small overhead for boxing it would be negligible

[jw]

That's fine, but my comparison was against using the boxing+delegate vs. just boxing

[xenoterracide]

yeah, my alternative to using lateinit in this case is simply to make the primitive nullable thus boxing anyways, then I have to deref with !! seems like I’m doing the same thing, but it’s less consistent syntactically from everything else around it