i guess that's the real question: can i create an inline class whose value can be modified in place ...

# announcements

[bbaldino]

i guess that's the real question: can i create an inline class whose value can be modified in place to implement something like inlineClassInstance++ to increment its value

[karelpeeters]

inc never in-place in Kotlin, see https://kotlinlang.org/docs/reference/operator-overloading.html#increments-and-decrements

[bbaldino]

oh! so does that mean it's taken care of it behind the scenes though?

[bbaldino]

(trying it out now in a scratch)

[bbaldino]

i saw the thing about 'val' and assumed it'd need to be modified so wouldn't work, but seemed like a big limitation on inline classes

[karelpeeters]

Well in the end the operator ++ does behave as expected, it just always has to create a new instance, which for inline classes does't actually happen.

[bbaldino]

just tested it out and looks like it does work

[bbaldino]

awesome, thanks!

[bbaldino]

what about something like plusAssign ?

[karelpeeters]

plusAssign is in-place, if you haven't implemented that it reverts to plus and assigning at the callsite, so "out-place". Full details on the page I linked earlier.

[bbaldino]

looks like it's just kinda magic with plus

[bbaldino]

ok yeah

[bbaldino]

it'd be impossible to implement plusAssign for an inline class, right?

[karelpeeters]

Well the thing in the inline class can be mutable and you can mutate that, eg if it was a MutableList you could add to it.

[bbaldino]

oh ok, didn't know you could do an inline class with a more complex type like that

[bbaldino]

thanks for all the info @karelpeeters

[karelpeeters]

You're welcome!

[bbaldino]

inline classes look like they'll solve an annoying problem i've had for a long time. excited to try them out.

[karelpeeters]

What problem?

[bbaldino]

i need to model RTP sequence numbers which roll over at 2 bytes, so comparing and incrementing, etc. need to take that into account...but they're used a lot so never wanted the overhead of a full class

[karelpeeters]

Problem is that you can't put any constraints on the stored value, but maybe that's not a problem for you?

[bbaldino]

hmm...that is a good point, but probably not a deal breaker as it's no worse than what we have now (just using an Int)

[bbaldino]

can i not throw in init with an invalid value?

[karelpeeters]

There's no init for inline classes.

[bbaldino]

dang

[karelpeeters]

Because you couldn't force Java code to call that init.

[bbaldino]

it'd be great to be able to validate, but since it's no worse than current implementation in that regard it's not too bad

[hudsonb]

You could make the constructor private and provide a "constructor overload" (function matching class constructor case) that would perform that validation instead.

[karelpeeters]

No you can't make the constructor private for the same reason 🙂

[hudsonb]

You can with @Suppress("NON_PUBLIC_PRIMARY_CONSTRUCTOR_OF_INLINE_CLASS")

[karelpeeters]

Then you give up safety for Java interop, but that may be acceptable.

[hudsonb]

Yeah it obviously makes you unable to use new on the Java-side... but sometimes there is no Java-side

[karelpeeters]

Even worse, if you take an inline class as an argument Java code could pass illegal "instances".

[hudsonb]

Isnt that possible in either case? At least way youd have some safety Kotlin side, and as I said sometimes Java interop isnt a concern anyways

[karelpeeters]

Yeah you're right, this is a good idea in this case.

[bbaldino]

thanks @hudsonb i'll take a look at that!