Maybe someone here can shed more light on this? <h...

# library-development

[simon.vergauwen]

Maybe someone here can shed more light on this? https://kotlinlang.slack.com/archives/C7L3JB43G/p1678126163957399

[bod]

With it, a val can be initialized in the lambda. +1 runCatching should/could have used it (unless I'm missing something). Where do you see the unreachable code warning?

[simon.vergauwen]

It's actually not possible to add it for runCatching, or it results in a false positive. I know the benefits of using it, but we're getting false-negatives all over after adding it to Arrow. More context in Arrow.

[bod]

interesting. On my side I see the erroneous warning whether I add the contract or not 🤔

Scratch that, it seems to be flaky. Now I don't see it with or without the contract 😅

[simon.vergauwen]

Okay, that's really strange. For me it's consistent only with the EXACTLTY_ONCE contract. It's also been reported by other people (assuming using different versions of IDEA).

[bod]

but yeah in any case, looks like a bug 🙂

[simon.vergauwen]

Yes but in any case, that still poses the question to me why the Kotlin Std uses it so little 🤔 I.e. with the current version of kotlin.runCatching without EXACTLY_ONCE you cannot initialise val.

[simon.vergauwen]

Screenshot 2023-03-07 at 20.11.31.png

[bod]

I agree, it's odd - maybe runCatching existed prior to the contracts being introduced and they were just never added - but I'm just conjecturing

[franztesca]

Not only the false-negative on "Unreachable code", but there is a more serious issue:

val x: String
runCatching {
    if (1+1==2) throw Exception()
    x = "42"
}
println(x)

If runCatching had EXACTLY_ONCE, this code would compile with no problem, but x would not be initialized when the println statement is reached.
In general, if your function catches any throwable of the block, it is not safe to add EXACTLY_ONCE AFAIS.