is it desired behavior for reflection to be able t...

# announcements

[snowe]

is it desired behavior for reflection to be able to bypass kotlin null safety?

[Andreas Sinz]

What do you mean with bypass?

[snowe]

https://www.dropbox.com/s/blrrbx0gka6k3a9/Screenshot%202018-06-28%2014.38.35.png?dl=0

[snowe]

prints "Stupid Kotlin2!!!!"

[snowe]

and the object is non nullable.

[Andreas Sinz]

you are setting it to null with reflection?

[snowe]

well my coworker is, but yes.

[snowe]

I'm asking for him. I thought kotlin would verify this at runtime, but it seems not.

[Andreas Sinz]

thats expected, the JVM doesn't know anything about nullable/non-nullable types

[Andreas Sinz]

it does check it at the kotlin <-> java border, but not inside kotlin code

[snowe]

yeah, but kotlin generally inserts null checks. Why are we not receiving an NPE like if we used !!?

[snowe]

looking at the bytecode, kotlin generates a check for the !!

[Andreas Sinz]

imagine if kotlin would place null-checks everywhere inside the code, that's gonna decrease performance very heavily

[snowe]

I guess it would be crazy expensive to do that for every access of every object.

[snowe]

yeah.

[snowe]

but why not do it at the reflection boundary?

[snowe]

I actually don't know if he's using kotlin reflect, or java reflect... maybe that's why

[Andreas Sinz]

I don't know, you have to ask the Jetbrains guys

[snowe]

seems like he's using java reflection. I'll have him try it with kotlin-reflect.

[snowe]

thanks!

[mp]

reflection also lets you bypass final fields. Once you’re using reflection, it’s up to you to do the right thing

[Gerard Klijs]

I had a simular thing reading from a database. Kotlin asummed not null, but because the data was created from clojure, it was null, and gave a runtime error aborting the program.

[snowe]

Thanks. Yeah I assumed that was it, but had a coworker complaining 😉